[From the U.S. Government Printing Office, www.gpo.gov]
NOAA Technical Report NMFS 123 April 1995 NOAA Coastal Change Analysis Pr 'ogram (C-CAP): Guidance for Regional Implementation J. E. Dobson E. A. Bright R. L. Ferguson D. W. Field L. L. Wood K. D. Haddad H. Iredale III J. R. Jensen V. V. Klemas R J. Orth J. P. Thomas SH1 1 U.S. Department of Commerce .A44672 no. 123 c.2 U.S. Department of Commerce NOAA Ronald H. Brown Secretary National Oceanic and Atmospheric AdministrationTechni*cal D. James Baker Under Secretary for Oceans and Atmosphere Reports NMFS National Marine Fisheries Service Technical Reports of the Fisheg Bulletin Rolland A. Schmitten Assistant Administrator for Scientific Editor Fisheries Dr. Ronald W. Hardy Northwest Fisheries Science Center National Marine Fisheries Service, NOAA 2725 Mondake Boulevard East Seattle, Washington 98112-2097 oxgv Or FQS, Editorial Conunittee Dr. Andrew E. Dizon National Marine Fisheries Service Dr. Linda L. Jones National Marine Fisheries Service 15 Dr. Richard D. Methot National Marine Fisheries Service r s o Dr. Theodore W. Pietsch University of Washington Dr. Joseph E. Powers National Marine Fisheries Service Dr. Tim D. Smith National Marine Fisheries Service Managing Editor James W. Orr The NOAA Technical Report NMFS National Marine Fisheries Service (ISSN 0892-8908) series is published by Scientific Publications Office the Scientific Publications Office, National 7600 Sand Point Way NE, BIN C15700 Marine Fisheries Service, NOAA, 7600 Seattle, Washington 98115-0070 Sand Point Way NE, Seattle, WA 98115- 0070. Although the contents have not been copyrighted and may be reprinted entire- ly, reference to tile source is appreciated. The NOAA Technical Report NMFS series of the F4@ Bulletin carries peer- The Secretary of Commerce lias deter- reviewed, lengthy original research reports, taxonomic keys, species synopses, flora mined that the publication of this series is and fauna studies, and data intensive reports on investigations in fishery science, necessary in the transaction of the public engineering, and economics, The series was established in 1983 to replace two business required by law of this Depart- subcategories of the Technical Report series: "Special Scientific Report- ment. Use of funds for printing of this series lias been approved by die Director Fisheries" and "Circular." Copies of the NOAA Technical Report NMFS are avail- of the Office of Management and Budget. able free in limited numbers to government agencies, both federal and state. For sale by the U.S. Department of They are also available in exchange for other scientific and technical publications Commerce, National Technical Informa- in the marine sciences. individual copies may be obtained from the U.S. Depart- tion Service, 5285 Port Royal Road, ment of Commerce, National Technical Information Service, 5285 Port Royal Springfield, VA 22161. Road, Springfield, VA 22161. NOAA Technical Report NMFS 123 A Technical Report of the Fishe?y Bulletin NOAA Coastal Change Analysis Program (C-CAP): Guidance for Regional Implementation J. E. Dobson and E. A. Bright Oak Ridge National Laborato?y Oak Ridge, TN 3 7831 R. L. Fe'rguson, D. W. Field, and L. L. Wood NOAAINAYS Beaufort Laboratoiy Beaufort, NC 28516 K. D. Haddad Florida Dept. of Environmental Protection, Marine Research Institute St. Petersburg, FL 33701 H. Iredale III NOAAINESDIS National Oceanog7-aphic Data Center Washington, DC 20235 J. R. Jensen Dept. of Geography University of South Carolina Columbia, SC 29208 V. V. Klemas College of Marine Studies, University of Delaware Newark, DE 19716 R. J. Orth Virginia Institute of Marine Science College of William and Ma7y Gloucester Point, VA 23062 J. P. Thomas NOAAINMFS Office of Protected Resources Silver Spring, MD 20910 April 1995 U.S. Department of Commerce Seattle, Washington Contents Preface ........................................................................ vii Executive Summary ............................................................... ix Chapter 1. Introduction The Coastal Region Management Problem .......................................... I The NOAA Coastal Change Analysis Program (C-CAP) Solution ........................ I C-CAP's National Scope and Regional Implementation ................................ 3 Change Detection Every One to Five Years .......................................... 4 The Need for Standardization and Guidelines ....................................... 4 General Steps Required to Conduct Regional C-CAP Projects ............................ 4 Chapter 2. 7`he C-C4P Coastal Land-Cover Classification System Introduction ................................................................... 7 Superclasses of the C-CAP System ................................................. 7 Uplands .................................................................... 7 Wetlands .................................................................... 9 Water and Submerged Land .................................................... 10 Chapter 3. Monitoring Uplands and Wetlands Using Satellite Remote Sensor Data Remote Sensing System Considerations ............................................. I I Temporal Resolution .......................................................... I I Spatial Resolution and Look Angle .............................................. 11 Spectral Resolution ........................................................... 11 Radiometric Resolution ........................................................ 12 The Preferred C-CAP Satellite Sensor System ...................................... 12 Important Environmental Characteristics ........................................... 13 Atmospheric Conditions ....................................................... 13 Soil Moisture Conditions ....................................................... 13 Vegetation Phenological Cycle Characteristics ..................................... 14 Effects of Tidal Stage on Image Classification ...................................... 14 Image Processing Data to Inventory Upland and Wetland Change ....................... 14 Rectification of Multiple-Date Remote Sensor Data ................................. 14 Radiometric Normalization of Multiple-Date Images ................................ 15 Selecting the Appropriate Change Detection Algorithm ............................... 16 Change Detection Using Write Function Memory Insertion .......................... 16 Multiple-Date Composite Image Change Detection ................................. 17 image Algebra Change Detection ............................................... 17 Postclassification Comparison Change Detection ................................... 19 iii Multiple-Date Change Detection Using a Binary Change Mask Applied to Tb_1 or Tbl ..... 20 Multiple-Date Change Detection Using Ancillary Data Source as Tb .................... 25 Manual On-screen Digitization of Change ......................................... 26 Selecting Appropriate Classification Algorithms ...................................... 28 Supervised and Unsupervised Image Classification Logic ............................ 29 Selection of Training and Verification Samples for Supervised and Unsupervised Classification .............................................................. 31 Use of Collateral Data in Image Classification ...................................... 32 Cartographic Portrayal of Classification and Change Detection Maps .................... 33 The Concept of the Minimum Measurement Unit .................................. 33 Analog (Hardcopy) Cartographic Products ........................................ 33 CWapter 4. Monitoring Submerged Land Using Aerial Photography C-Cap Focus on Aerial Photography of Submersed Rooted Vascular Plants (SRV) .......... 35 Ancillary Categories of Submersed Habitat ........................................ 35 Ancillary Technologies for Collection of Submersed Habitat Data ..................... 35 Aerial Photography of SRV ....................................................... 36 Film ........................................................................ 36 Metric Photography and Photographic Scale ...................................... 36 Flightlines, Reconnaissance Flights, and Photographic Overlap ....................... 37 Environmental Considerations ................................................... 37 Photointerpretation of SRV ........................................................ 38 Field Surveys ................................................................... 39 Species and Habitat at Randomly Selected Stations ................................. 39 Signature Verification and Supplemental Spatial Data ............................... 40 Base Maps and Registration of Habitat Polygons ..................................... 40 Planimetric Base Maps ......................................................... 40 Transfer of Polygons to the Map Coordinate Projection System ....................... 41 Digitization of Habitat Polygons ................................................... 41 Change Detection with Aerial Photographic Data .................................... 42 Recent Photography .......................................................... 42 Historical Photography ........................................................ 42 Change Detection of Seagrass Habitat in North Carolina ............................ 44 Chapter 5. Spacial Data Quality Assurance and Control Lineage ....................................................................... 45 Positional Accuracy and Precision ................................................. 45 Generalization Versus Error ...................................................... 46 Attribute Accuracy and Precision .................................................. 47 Logical Consistency ............................................................. 47 Completeness .................................................................. 47 Temporal Accuracy and Precision ................................................. 47 iv Fitness for Use ............................................................... 47 Recommended Accuracy Assessment Test ......................................... 48 Sample Selection and Field Mapping ............................................. 48 Accuracy Assessment for Individual Date Classification of Upland and Wetland Habitat Data ....................................................... 48 Accuracy Assessment for Individual Date Classification of Water and Submerged Land Data ....................................................... 49 Accuracy Assessment for Land-Cover Change Data ................................. 49 Comparison and Statistical Analysis .............................................. 49 Chapter 6. P@-oduct Availability Digital Product ................................................................. 51 Description and Availability .................................................... 51 Digital Product Redistribution Restrictions ........................................ 51 Digital Product Liability Disclaimer .............................................. 51 Digital Product Format and Contents ............................................ 51 Digital Product Medium ........................................................ 52 Digital Product Cost .......................................................... 52 Digital Product Ancillary Documentation ......................................... 52 Hardcopy Products ............................................................. 52 Upland and Wetland Habitats .................................................. 52 Submersed Habitats ........................................................... 52 Chapter 7. Users and Information Needs Chapter 8. Regional Participation Purpose ....................................................................... 55 Regional Project Summaries ...................................................... 55 St. Croix River Estuary (Border of Maine and New Brunswick, Canada) ................ 55 Coastal Massachusetts ......................................................... 55 Universities of Connecticut and Rhode Island ..................................... 55 University of Delaware ......................................................... 55 Oak Ridge National Laboratory ................................................. 56 Virginia Institute of Marine Sciences ............................................. 56 North Carolina State University ................................................. 56 Beaufort Laboratory, National Marine Fisheries Service ............................. 57 University of South Carolina .................................................... 57 State of Florida, Department of Environmental Protection ........................... 57 Texas Parks and Wildlife Department ............................................ 58 Columbia River, Tillamook Bay, and Willapa Bay (Oregon and Washington) ............ 58 Hubbard Glacier and Russell Fjord, Alaska ........................................ 58 V Acknowledgments ................................................................. 59 Literature Cited ................................................................. 61 Appendices Appendix Table 1 - U.S. Geological Survey Land-Cover Classification Scheme for Remote Sensor Data ................................................. 65 Appendix Table 2 - U.S. Fish and Wildlife Service Wetland Classification System .......... 67 Appendix 3 - C-CAP Wetland Classification Scheme Definitions ....................... 69 Appendix 4 - C-CAP Workshops ................................................. 79 Appendix 5 - C-CAP Protocol Development Research ................................ 81 Appendix 6 - Workshop Participants .............................................. 83 Vi Preface This unique document represents a first attempt to edly will occur, we intend to update this document develop guidelines that will allow researchers and re- periodically. These updates, however, require time to source managers alike to quantitatively monitor changes publish, so anyone planning to use these guidelines that are occurring in the abundance of emergent and should contact either the senior author or program submergent wetlands'and adjacent uplands in coastal manager to obtain drafts of any revised chapters that regions. Such information is essential in order to effec- have not yet been published. tively relate changes in coastal land use to changes in Finally, I would like to express my appreciation to the the productivity of estuaries and coastal waters on a authors for their fine effort and to Dr. Don Scavia, regional scale. Director of NOAA's Coastal Ocean Program, for his This is a document that was developed from the support, both financial and moral, during the develop- input of approximately 200 research scientists and re- ment of this document. I believe we have made a signifi- source managers that attended five regional workshops cant step in addressing an important coastal issue. and several topical interagency meetings. Thus, we be- lieve it represents a general consensus of how to ap- Ford A Cross proach the issue of quantifying land-cover and wetland Manager, C-CAP change in coastal regions. Because improvement in National Marine Fisheries Service/NOAA existing technologies and in our understanding of how Beaufort Laboratory to measure habitat change on a regional scale undoubt- Beaufort, NC 28516 Vii Executive Summary The Coastal Change Analysis Program I (C-CAP) is de- Aggregations to larger areas (representing habitats, wild- veloping a nationally standardized database on land- life refuges, or management districts) will be provided cover and habitat change in the coastal regions of the on a case-by@case basis. Ongoing C-CAP research will United States. C-CAP is part of the Estuarine Habitat continue to explore techniques for remote determina- Program (EHP) of NOAA's Coastal Ocean Program tion of biomass, productivity, and functional status of (COP). C-CAP inventories coastal submersed habitats, wetlands and will evaluate new technologies (e.g. re- wetland habitats, and adjacent uplands and monitors mote sensor systems, global positioning systems, image changes in these habitats on a one- to five-year cycle. processing algorithms) as they become available. Se- This type of information and frequency of detection lected hardcopy land-cover change maps will be pro- are required to improve scientific understanding of the duced at local (1:24,000) to regional scales (1:500,000) linkages of coastal and submersed wetland habitats with for distribution. Digital land-cover change data will be adjacent uplands and with the distribution, abundance, provided to users for the cost of reproduction. and health of living marine resources. The monitoring Much of the guidance contained in this document cycle will vary according to the rate and magnitude of was developed through a series of professional work- change in each geographic region. Satellite imagery shops and interagency meetings that focused on a) (primarily Landsat Thematic Mapper), aerial photo- coastal wetlands and uplands; b) coastal submersed graphy, and field data are interpreted, classified, ana- habitat including aquatic beds; c) user needs; d) re- lyzed, and integrated with other digital data in a geo- gional issues; e) classification schemes; f) change detec- graphic information system (GIS). The resulting land- tion techniques; and g) data quality. Invited partici- cover change databases are disseminated in digital form pants included technical and regional experts and rep- for use by anyone wishing to conduct geographic analy- resentatives of key State and Federal organizations. sis in the completed regions. Coastal habitat managers and researchers were given C-CAP spatial information on coastal change will be an opportunity for review and comment. input to EHP conceptual and predictive models to sup- This document summarizes C-CAP protocols and pro- port coastal resource policy planning and analysis. G cedures that are to be used by scientists throughout the CAP products will include 1) spatially registered digital United States to develop consistent and reliable coastal databases and images, 2) tabular summaries by state, change information for input to the GCAP nationwide county, and hydrologic unit, and 3) documentation. database. It also provides useful guidelines for contribu- tors working on related projects. It is considered a work- Formerly known as the "Coast Watch Change Analysis Project." ing document subject to periodic review and revision. ix aiapter I Introduction The Coastal Region Management submersed habitat may be a sensitive integrator of over- Problem all water quality and potential for change in fisheries productivity. Submersed rooted vascular aquatic beds The conterminous United States lost 53 percent of its define habitat critical for the support of many recre- wetlands to agricultural, residential, and commercial ational and sport fisheries (Ferguson et al., 1980; Zieman, land use from the 1780's to 1980's (Dahl, 1990). Oil 1982; Phillips, 1984; Thayer et al., 1984; Zieman and spills occurring throughout the world continue to dev- Zieman, 1989; Klemas et al., 1993). Changes in up- astate coastal wetlands Uensen et al., 1990; Narumalani lands, wetlands, and submersed habitats can be rapid and et al., 1993). Sea level has risen approximately 130 m in pervasive. Hence, effective management requires frequent the past 17,500 years. More abundant "greenhouse" monitoring of coastal regions (at least twice per decade). gases in the atmosphere may be increasing the Earth's It has long been suspected that a crucial factor in the average temperature (Clarke and Primus, 1990) and observed decline of fisheries in most coastal regions is may, yet again, accelerate the global sea level rise, even- the declining quantity and quality of habitat. Land- tually inundating much of today's coastal wetlands (Lee cover change is a direct measure of quantitative habitat et al., 1992). Unfortunately, current projections for loss or gain. For many marine fisheries the habitats (i.e. U.S. population growth in coastal regions suggest accel- land covers) of greatest importance are saltmarsh and erating losses of wetlands and adjacent habitats, as waste scagrass. Other fisheries, such as those for salmon, de- loads and competition for limited space and resources pend on a variety of habitats that may include upland as increase (U.S. Congress, 1989). Coastal wetlands and well. Land-cover change is also a direct measure of submersed habitats are being destroyed by erosion, increases or decreases in sources of pollution, sedimen- dredge and fill, impoundments, toxic pollutants, tation, and other factors that determine habitat quality. eutrophication, and (for submersed habitats) excessive Increases in developed land, for example, are accompa- turbidity and sedimentation. Many marine finfish and nied by land disturbance that increases erosion and shellfish depend on these coastal habitats for their sur- sedimentation and by hydrologic alteration that in- vival. Salt marsh grasses, mangroves, macroalgae, and creases runoff. Similarly, cultivated land is associated submersed grasses and forbs are essential as nourish- with fertilizer and pesticide use that ultimately affects ment and animal habitat. Continued loss of these wet- the marine environment. Hence, land-cover change is lands may lead to the collapse of coastal ecosystems and linked to habitat quantity and quality. associated fisheries. Documentation of the loss or gain of coastal wetlands is needed for their conservation and effective managment of marine fisheries (Haddad and The NOAA Coastal Change Analysis Program Ekberg, 1987; Haddad and McGarry, 1989; Kiraly et al., (C-CAP) Solution 1990; Kean et al.'). Submersed grasses and forbs include seagrasses that For these reasons, the National Oceanic and Atmo- require high salinity and other species of submersed spheric Administration (NOAA) Coastal Ocean Pro- rooted vascular plants (SRV) that tolerate or require gram initiated the Coastal Change Analysis Program low salinity water. Submersed grasses and forbs may be (C-CAP), a cooperative interagency, State, and Federal crucial indicators of water quality and overall health of effort to detect coastal upland and wetland land cover coastal ecosystems (Dennison et al., 1993). Submersed and submersed vegetation and to monitor change in vegetation has the additional requirement of living at the coastal region of the United States (Cross and Tho- photic depths and therefore is particularly sensitive to mas, 1992; Haddad, 1992). The project uses digital water clarity (Kenworthy and Haunert, 1991). Change remote sensor data, in situ measurement in conjunc- (increase or decrease in areal extent, movement, con- tion with global positioning systems (GPS), and geo- solidation or fragmentation, or qualitative change) in graphic information system (GIS) technology to moni- Kean, T. H., C. Campbell, B. Gardner, and W. K. Reilly. 1988. tor changes in coastal wetland habitats and adjacent Protecting America's wetlands: an action agenda. Final Report of uplands. Landsat multispectral scanner (MSS) data, the National Wetlands Policy Forum. The Conservation Founda- Landsat Thematic Mapper (TM) data, and SPOT high tion, Washington, D.C. resolution visible (HRV) data have been used success- 2 NOAA Technical Report NMFS 123: Dobson et aL: Coastal Change Analysis Program fully to detect major categories of wetlands (Haddad local, State, and Federal wetland management strate- and Harris, 1985; Jensen et al., 1993b; Lade et al.2). gies, and construction of predictive models. C-CAP di- However, they have not been used previously to map or rectly supports NOAA's responsibilities in estuarine and monitor wetlands for regional or national coverage. marine science, monitoring, and management as legis- The use of satellite imagery for mapping wetlands pro- lated in the Fish and Wildlife Coordination Act; the vides a number of advantages over conventional aerial Magnuson Fishery Conservation and Management Act; photographs including timeliness, synopticity, and re- the Coastal Zone Management Act; the Clean Water duced costs. While aerial photography may be appro- Act; the Marine Protection, Research, and Sanctuaries priate for high resolution cartography, satellite imagery Act; and the National Environmental Policy Act. Land- is better suited and less costly for rapid, repeated obser- cover change data are essential to the implementation vations over broad regions (Haddad and Harris, 1985; of a "No Net Loss" wetlands policy. Bartlett, 1987; Klemas and Hardisky, 1987; Ferguson et A large community of managers, scientists, and users al., 1993). Although the program will stress the use of were involved in developing a C-CAP protocol at the satellite imagery, particularly for coastal wetlands and national level. Guidance in this document was derived adjacent uplands, aerial photography or a combination from a series of professional workshops and interagency of photography and satellite imagery (TM or SPOT) working group meetings which focused on will be used for mapping SRV (Orth and Moore, 1983) and certain other habitats, as suggested by Patterson 0user needs (1986) and Lade et al. (1988). A methodology to pho- 0upland, wetland, and water classification schemes tographically observe, analyze, and display spatial change eregional boundary issues in habitat defined by the presence of SRV was a prereq- 0cartographic datum and data structures uisite to a nationwide change detection effort (Thomas 0selection of appropriate satellite imagery and aerial and Ferguson, 1990). photography The C-CAP nationally standardized database will be *field work and field verification methods used to monitor land-cover and habitat change in the esatellite remote sensing of coastal wetlands and coastal regions of the United States (Thomas and uplands Ferguson, 1990; Thomas et al. 1991) and to improve ephoto interpretation of coastal submersed habitat, understanding of coastal uplands, wetlands (e.g. salt including seagrasses marshes), and submersed habitats (e.g. seagrass) and 0calibration among regions and scenes their linkages with the distribution, abundance, and 0classification and change detection algorithms health of living marine resources. Coastal regions of 9geographic information processing and analysis the U.S. will be monitored every one to five years de- 0regional ecological modeling pending on the anticipated rate and magnitude of *quality assurance and control change in each region and the availability of suitable *product availability and distribution remote sensing and in situ measurements. This moni- *research issues toring cycle will provide feedback to habitat managers on the success or failure of habitat management poli- Approximately 40 scientists and environmental man- cies and programs. Frequent feedback to managers will agers attended each major regional workshop held in enhance the continued integrity or recovery of coastal the Southeast, Northeast, Pacific Coast, and Great Lakes ecosystems and the attendant productivity and health regions; about 200 individuals participated in all work- of fish and other living marine resources at minimal shops and special meetings. The community of users cost. In addition, the geographical database will allow and providers of coastal habitat information were given managers and scientists to evaluate and, ultimately, to an opportunity for review and comment. A detailed list predict cumulative direct and indirect effects of coastal of workshops is provided in Appendix 4. development on wetland habitats and living marine Although C-CAP is national in scope, it is based on resources. Initially, C-CAP products will document cur- procedures also applicable at local and regional levels. rent land-cover distribution and change that have oc- Much of the content of this document is based on curred in the recent past. The database, as it increases C-CAP sponsored. research conducted at the regional with each subsequent monitoring cycle, will be an in- level. For example, Klemas et al. (1993) of the College valuable baseline resource for research, evaluation of of Marine Studies at the University of Delaware devel- oped the "C-CAP Coastal Land Cover Classification Sys- 2 Lade, P. K., D. Case, J. French, and H. Reed. 1988. Delineation and tem" by investigating existing upland and wetland clas- classification of submerged aquatic vegetation using SPOT satellite sification systems and then synthesizing a new system multispectral digital data. Final report to the Maryland Dept. Natu- ral Resources, Tidewater Administration, Coastal Resources Div., that is practical at the regional level. Dobson and Bright Annapolis, MD. (1991, 1992, and 1993) of the Oak Ridge National Chapter 1: Introduction 3 Laboratory (ORNL) developed a regional prototype to proach is to designate 1) standard coverage limits for inventory uplands and wetlands in the Chesapeake Bay general application and 2) extended coverage limits region. Jensen et al. (1993a) evaluated various change for regions with special needs. Standard coverage will detection algorithms for inland and coastal wetland utilize biological and other geographical boundaries environments near Charleston, S. C. Ferguson et al. appropriate to the needs of specified C-GAP users iden- (1993) developed a regional prototype to inventory tified through the protocol workshops. Extended cov- SRV in North Carolina based on protocols developed erage will be defined for each regional project in col- by the Beaufort Laboratory, Southeast Fisheries Sci- laboration with states and other regional organiza- ence Center, National Marine Fisheries Service (NMFS). tions. NOAA will make every effort to identify and Khorram et al. (1992) investigated methods of seamlessly accommodate research, conservation, management, and integrating multiple-region C-CAP databases. the needs of other interests that rely on wetland maps The C-CAP protocol continues to evolve and im- and data. Regional projects will be designed to identify prove. For example, projects underway in 1993 include special needs that may require extended coverage and analysis of the effects of tidal stage on remote-sensing to suggest sources of funds to support the additional classification, change detection accuracy assessment, cost of extended coverage. refined techniques for classification of forested wet- The estuarine drainage area (EDA), defined by lands, and advanced change detection techniques (Ap- NOAA's National Ocean Service (NOS) as the "land pendix 5). Research continues on functional health and water component of an entire watershed that most indicators (e.g. biomass, productivity), plant stress (e.g. directly affects an estuary," is an appropriate standard mangrove freeze), new data-collection instruments, and coverage area for C-CAP. For the purposes of this pro- regional ecological modeling. Thus, C-CAP will con- gram, all U.S. coasts are or will be defined as part of an tinue to have a strong research and development com- EDA. The boundary of each EDA basin is defined to be ponent to improve and refine its operational techniques. consistent with U.S. Geological Survey (USGS) hydro- logic units and codes. The estuarine drainage boundary as defined by NOS National Scope and Regional Implementation is considered a standard inland boundary for of C-CAP C-CAP regional projects. Regional analysts may employ C-CAP protocols upstream, but C-CAP funding is not No single Federal or State organization will collect all intended for coverage beyond the EDA. However, the information residing in the C-CAP database. In- C-CAP funding may be used to purchase satellite scenes stead, regional inventories will be completed by re- that extend beyond the EDA if they are necessary to gional experts following C-CAP guidelines. Therefore, cover the coastal region. Functional definitions, such as it is important to define the logic used to specify a C- "limits of tidal influence," may be employed in response CAP region. First, regional boundaries must coincide to local situations justified by local user communities with the following NOAA/NMFS regions: and local or regional experts on a coastal region-by- region or estuary-by-estuary basis. Regional analysts Northeast -Virginia through Maine, including the should be aware of local, State, and Federal rights and Great Lakes responsibilities and should seek intergovernmental and Southeast - Texas through North Carolina, U.S. Vir- interagency cooperation. Because C-CAP interests in- gin Islands, and Puerto Rico clude the effects of eutrophication due to development Northwest - Oregon, Washington, and Alaska of uplands, information from outside the EDA may be Southwest - California, Hawaii, Midway Islands, Wake, justified in high order streams that extend beyond the Guam, Mariana Islands, American Samoa, coastal region. In this case, the point where the river Johnston Atoll, Trust Territory of the Pa- enters the region will be defined as a point source for cific Islands, Baker and Howland Islands, inputs. and Jarvis Island. The offshore boundary of each region is defined as the seaward extent of wetlands, seagrass, coral, or other Coastal regions may be further subdivided, as appropri- submersed habitat detectable using remote sensing sys- ate, on the basis of State and other administrative bound- tems. The functional definition of limits of detection aries or ecoregions as defined, for example, by Omernik normally will be based on satellite and aerial sensors (1987). and will vary within and among regions. Both the limits The boundary should encompass coastal watersheds of detection and the actual bathymetric range of SRV plus offshore coral reefs, algae, and seagrass beds in the are based on light attenuation and, thus, will not be a photic zone. In keeping with the goals of C-CAP and consistent bathymetric contour even within a single anticipated funding constraints, the recommended ap- region. 4 NOAA Technical Report NMFS 123: Dobson et a].: Coastal Change Analysis Program Overlap of regions, consistent with TM scene bound- (e.g. urban analysis, forest inventory) often are not ary overlap, is preferred so that analysts may calibrate suitable for use in C-CAP. Aquatic beds, and even coastal results from neighboring regions. A healthy exchange wetlands, may not be identifiable on aerial photographs between neighboring regional analysts could reconcile obtained for other purposes. differences, not only in the area of overlap, but also in signature identification across both regions. Each re- gional project team will be responsible for calibrating The Need for Standardization and the relationship between remotely sensed spectral in- Guidelines formation and other information such as field mea- surements of biomass and photosynthetic rates. Histori- C-CAP desires to create a synoptic, digital database of cally, such measurements have focused on relatively few coastal wetland and upland land cover by class for a of the many species, habitats, and land-cover types of base time period and to identify change between the significance in the coastal region. Analysts should also base period and other time periods. The use of satellite ensure that protocols originally developed for north- remote sensing to inventory uplands and wetlands, con- ern temperate latitudes are modified sufficiently to serve ventional aerial photography to inventory submerged well in tropical areas of the southern United States, lands, and GIS to analyze the data are important ele- Caribbean, and Pacific Ocean, and in the Arctic areas ments of the C-CAP methodology. However, the goal of of Alaska. It will be necessary, for example, to use differ- completing an accurate change detection product over- ent methods and sensors for coral reefs than for wet- rides any given technical consideration. Therefore, lands. Similarly, the identification of Arctic muskeg timely high-quality information from aerial photographs, may require different methods and sensors from those topographic maps, field experience, or other sources used to identify temperate, herbaceous wetland. may be used to prepare C-CAP products if appropriate guidelines are followed. By standardizing procedures at the national level, Change Detection Every One to Five Years this document will benefit not only C-CAP but also coastal management research conducted by other State The frequency of change detection is a crucial issue. and Federal agencies. C-CAP desires to facilitate the For most regions in the United States, the base year exchange of standardized data among programs, de- (referred to as Tb or Date I in the diagrams) should be crease duplication, and improve the quality and utility the most recent year for which acceptable satellite im- of decision support for wetlands policy, management, agery for uplands and wetlands or aerial photographs and research activities. All data accepted for inclusion for submersed habitat can be obtained, and for which and eventual distribution in the C-CAP database must sufficient in situ information is available to conduct an adhere to the protocol described in this manual. The error evaluation. Exceptions may occur in regions where protocol is designed to allow flexibility in the use of cloud cover is a perennial problem or where other elements of the classification scheme and in the choice considerations favor aerial photographs over satellite of remote sensor data, classification and change detec- imagery. The choice of the second date of imagery tion procedures, and other key elements that vary re- (Date b-I or b+I) may be more flexible. It may be gionally. However, potential users must adhere to the desirable to choose a date one to five years earlier than protocol in order to maintain high quality information the base period to capture recent changes in coastal in the C-CAP database. Coastal land-cover change data- habitats. Plans should then be made for another change bases derived independently from C-CAP will be con- analysis no later than five years after the base time. sidered for dissemination as C-CAP products if originat- However, plans may be altered abruptly when natural ing organizations can document compliance with or human-induced events, such as hurricanes and oil C-CAP protocol and data quality standards. spills, occur. Five years is the recommended frequency of change detection for most regions, but shorter periods may be General Steps Required to Conduct Regional necessary in regions undergoing rapid economic devel- C-CAP Projects opment or affected by catastrophic events. Longer peri- ods may be necessary where funds are limited or where The general steps required to conduct regional C-CAP change is exceptionally slow. Regional analysts are ad- change detection projects using satellite remotely sensed vised to evaluate rates of change and explicitly recom- data are summarized in Table 1. This document is mend the base year and change period as a part of each organized according to these specific requirements and, regional project proposal. Unfortunately, remotely in certain instances, provides step-by-step instructions sensed data obtained specifically for other purposes to be used when conducting regional projects. One of Chapter 1: Introduction 5 the first requirements of regional participants is to scheme was suitable for all C-CAP requirements. There- precisely identify land-cover classes of interest to be fore, great effort went into the development of the G monitored and eventually placed in the C-CAP change CAP Coastal Land-Cover Classification System, which detection database. This must be performed in con- can be used to inventory uplands and wetlands by using junction with an appropriate classification scheme. Un- satellite remote sensor data and to inventory SRV by fortunately, no existing standardized classification using metric aerial photography. Table I General steps required to conduct regional G-CAP change detection projects to extract upland and wetland information using satellite remote sensing systems. Each major step is listed in the order to be accomplished. 1. State the regional change detection problem b. Preprocess the multiple-date remotely sensed data a. Define the region 1) Geometric rectification b. Specify frequency of change detection (I to 5 yr) 2) Rachometric correction (or normalization) c. Identify classes of the C-CAP Coastal Land-Cover c. Select appropriate change detection algorithm from Classification System the three C-CAP alternatives d. Apply appropriate image classification logic if necessary 2. Consider significant factors when performing change 1) Supervised detection 2) Unsupervised a. Remote sensing system considerations 3) Hybrid 1) Temporal resolution e. Perform change detection using CIS algorithms 2) Spatial resolution 1) Highlight selected classes using change detection 3) Spectral resolution matrix 4) Radiometric resolution 2) Generate change map products 5) The preferred C-CAP remote sensing system 3) Compute change statistics b. Environmental considerations 1) Atmospheric conditions 4. Conduct quality assurance and control 2) Soil moisture conditions a. Assess spatial data quality 3) Vegetation pherrological cycle characteristics b. Assess statistical accuracy of 4) Tidal stage 1) Individual date classification 2) Change detection products 3. Conduct image processing of remote sensor data to extract upland and wetland information 5. Distribute C-CAP Results a. Acquire appropriate change detection data a. Digital products 1) In situ and collateral data b. Analog (hardcopy) products 2) Remotely sensed data a) Base year (Time b) b) Subsequent year(s) (Time b-1 or b+I) Chapter 2 The C-CAP Coastal Land-Cover Classification System Introduction categories (Dobson, 1993a). C-GAP focuses on land cover and its relationship to other functional compo- It is essential that the coastal land-cover information nents of landscape (Dobson, 1993b). Definitions of the stored in the G-GAP database be taxonomically correct pertinent terms are as follows: and consistent with coastal wetland information de- rived from other agencies. The G-GAP Coastal Land- *land cover-vegetation, soils, rocks, water (in its vari- Cover Classification System (Table 2) includes three ous forms), and constructed materials covering the Level I superclasses (Klemas et al., 1993): land surface, physically present and visible. aland use-economic and cultural activities, permit- LO-Upland, ted or not, that are practiced at a place which may or 2.0-Wetland, and may not be manifested as visible land-cover features. 3.0-Water and Submerged Land. For example, forestry land use may be visibly mani- fested as forest land cover, but recreational land use These superclasses are subdivided into classes and sub- may occur in many different types of land cover, classes at Levels II and III, respectively. While the cat- often without visible evidence of recreational use. egories Wetland and Water and Submerged Land con- landscape-the zone of interaction and convergence stitute the primary habitats of interest to NOAA, Up- of the atmosphere, the hydrosphere, and the solid lands are also included because they influence adjacent earth. Its vertical bounds are determined by the fre- wetlands and water bodies. The classification system is quency and extent of interactions pertinent to a given hierarchical, reflects ecological relationships, and fo- field of inquiry. Horizontally, landscape may be di- cuses on land-cover classes that can be discriminated vided into areal units defined by physical or cultural primarily from satellite remote sensor data. It was features pertinent to a field of inquiry. adapted and designed to be compatible with other nationally standardized classification systems, especially While all categories of the CCAP classification system can be represented as two-dimensional features at the � the U.S. Geological Survey (USGS) "Land Use and mapping scale of 1:24,000, some features may be mapped Land Cover Classification System For Use with Re- as lines (e.g. a Marine/Estuarine Rocky Shore) or points mote Sensor Data" (Anderson et al., 1976; USGS, (e.g. unique landmarks). Most linear and point features 1992; Appendix Table 1), will be obtained from nonsatellite sources of information � the U.S. Fish and Wildlife Service (USFWS) "Classifi- (e.g. aerial photography or in situ measurement using cation of Wetlands and Deepwater Habitats of the GPS). Those classes and subclasses that are required by C- United States" (Cowardin et al., 1979; Wilen, 1990; CAP and which each regional C-CAP project will include Appendix Table 2), and in its database are underlined in Table 2. The underlined � the U.S. Environmental Protection Agency (USEPA) classes, with the exception of aquatic beds, can generally Environmental Monitoring and Assessment Program be detected by satellite remote sensors, particularly when (EMAP) classification system. supported by surface in situ measurement Dedicated workshops on the C-CAP classification sys- tem and productive discussions and reviews with repre- Superclasses of the C-CAP System sentatives from each of these major agencies resulted in a classification system that is in harmony with other Uplands major U.S. land-cover databases. The C-CAP Coastal Land-Cover Classification System includes upland, wet- The Upland superclass consists of seven subclasses land, submerged land, and water in a single, compre- (Table 2): Developed Land, Cultivated Land, Grass- hensive scheme. An attempt has been made to identify land, Woody Land, Bare Land, Tundra, and Snow/Ice. land-cover classes that can be derived primarily through Upland classes are adapted from Level I classes in the remote sensing and that are important indicators of USGS Land-Use and Land-Cover Classification System ecosystem change. Modifications were necessary to rec- (Anderson et al., 1976; USGS, 1992; Appendix Table oncile inconsistencies between Anderson et al. (1976) 1). Detailed definitions of all C-CAP classes and sub- and Cowardin et al. (1979) and to remove all land-use classes in Table 1 are found in Appendix 3. 7 8 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program Table 2 C-CAP Coastal Land-Cover Classification System (Modified from Klemas et al., 1993). C-CAP is committed to include the underlined classes in the land cover change databases. 1.0 Upland 2.41 Deciduous 2.411 Forest 1.1 Developed Land 2.412 Scrub/shrub 2.413 Dead 1.11 High Intensity 2.42 Evergreen 1.12 Low Intensity 2.421 Forest 2.422 Scrub/Shrub 1.2 Cultivated Land 2.423 Dead 2.43 Mixed 1.21 Orchards/ Groves/Nurseries 2.431 Forest 1.22 Vines/Bushes 2.432 Scrub/shrub 1.23 Cropland 2/433 Dead 1.3Grassland 2.5 Riverine Unconsolidated Shore (Beach, Flat, Bar) 1.31 Unmanaged 2.51 Cobble-Gravel 1.32 Managed 2.52 Sand 2.53 Mud/Organic 1.4 Woody Land 2.6 LacusLrine Unconsolidated Shore (Beach, Flat, Bar) 1.41 Deciduous 1.411 Forest 2.61 Cobble-Gravel 1.412 Scrub/Shrub 2.62 Sand 1.42 Evergreen 2.63 Mud/Organic 1.421 Forest 1.422 Scrub/Shrub 2.7 Palustrine Unconsolidated Shore (Beach, Flat, Bar) 1.43 Mixed 1.431 Forest 2.71 Cobble-Gravel 1.432 Scrub/Shrub 2.72 Sand 2.73 Mud/Organic 1.5 Bare Land 2.8 Palustrine Emergent Wetland (Persistent) 1.6Tundra 1.7 Snow/Ice 2.9 Palustrine Woody Wetland 1.71 Perennial Snow/Ice 2.91 Deciduous 1.72 Glaciers 2.911 Forest 2.912 Scrub/shrub 2.913 Dead 2.0 Wetland 2.92 Evergreen 2.921 Forest 2.1 Marine/Estuarine Rocky Shore 2.922 Scrub/shrub 2.11 Bedrock 2.923 Dead 2.12 Rubble 2.93 Mixed 2.931 Forest 2.2 Marine/Estuarine Unconsolidated Shore 2.932 Scrub/shrub (Beach, Flat, Bar) 2.933 Dead 2.21 Cobble-gravel 3.0 Water and Submerged Land 2.22 Sand 2.23 Mud/Organic 3.1 Water 2.3 Marine /Estuarine Emergent Wetland 3.11 Marine/Estuarine 2.31 Haline (Salt Marsh) 3.12 Riverine 2.32 Mixohaline (Brackish Marsh) 3.13 Lacustrine (Basin > 20 acres) 3.14 Palustrine (Basin < 20 acres) 2.4 Estuarine Wooft Wetland Chapter 2: C-CAP Coastal Land-Cover Classification System 9 Table 2 (continued) 3.2 Marine/Estuarine Reef 3.41 Rooted Vascular/Algal/Aquatic Moss 3.42 Floating Vascular 3.3 Marine /Estuarine Aquatic Bed 3.5 Lacustrine Aquatic Bed (Basin > 20 acres) 3.31 Algal (e.g., kelp) 3.32 Rooted Vascular (e.g., seagrass) 3.51 Rooted Vascular/Algal/Aquatic Moss 3.321 (High Salinity (2!5 ppt; Mesohaline, 3.52 Floating Vascular Polyhaline, Euhaline, Hyperhaline) 3.322 Low Salinity (< 5 ppt; Oligohaline, Fresh) 3.6 Palustrine Aquatic Bed (Basin < 20 acres) 3.4 Riverine Aquatic Bed 3.61 Rooted Vascular/Algal/Aquatic Moss 3.62 Floating Vascular Developed Land (derived from the Anderson et al. ing or for growing and harvesting hay and straw for [1976] Urban or Built-Up class) characterizes con- animal feed. structed surfaces composed of concrete, asphalt, roof- Woody Land includes nonagricultural trees and ing, and other building materials with or without veg- shrubs. The category alleviates the problem of separat- etation. This class has been divided into two subclasses ing various sizes of trees and shrubs using satellite remote based on the amount of constructed surface relative to sensor data but allows a height-based separation if high the amount of vegetated surface present. High-Inten- resolution aerial photographs are available. The class may sity Developed Land contains little or no vegetation. be partitioned into three subclasses: Deciduous, Evergreen, This subclass includes heavily built-up urban centers as and Mixed. These three subclasses generally can be dis- well as large constructed surfaces in suburban and rural criminated with satellite remote-sensing systems. areas. Large buildings (such as multiple-family hous- Bare Land (derived from Barren Land of Anderson ing, hangars, and large barns), interstate highways, and et al. [1976]) is composed of bare soil, rock, sand, silt, runways typically fall into this subclass. Low-Intensity gravel, or other earthen material with little or no veg- Developed Land contains substantial amounts of con- etation. Anderson et al.'s Barren Land was defined as structed surface mixed with substantial amounts of veg- having limited ability to support life; C-CAP's Bare Land etated surface. Small buildings (such as single family is defined by the absence of vegetation without regard housing, farm outbuildings, and sheds), streets, roads, to inherent ability to support life. Vegetation, if present, and cemeteries with associated grasses and trees typi- is more widely spaced and scrubby than that in the cally fall into this subclass. vegetated classes. Unusual conditions such as a heavy Cultivated Land ("Agricultural Land" of Anderson et rainfall may occasionally result in growth of a short- al. [1976]) includes herbaceous (cropland) and woody lived, luxuriant plant cover. Wet, nonvegetated, exposed (orchards, nurseries, vineyards, etc.) cultivated lands. Sea- lands are included in the Wetland categories. Bare Land sonal spectral signatures, geometric field patterns, and may be bare temporarily because of human activities. The road network patterns may help identify this land-cover transition from Woody Land, Grassland, or Cultivated type. Always associated with agricultural land use, culti- Land to Developed Land, for example, usually involves vated land is used for the production of food and fiber. a Bare Land phase. Developed Land also may have tempo- Grassland differs from "Rangeland" of Anderson et rary waste and tailing piles. Woody Land may be clearcut, al. (1976) by excluding shrub-brushlands. Unmanaged producing a temporary Bare Land phase. When it may be Grasslands are dominated by naturally occurring grasses inferred from the data that the lack of vegetation is due to and forbs which are not fertilized, cut, tilled, or planted an annual cycle of cultivation (e.g. plowing), the land is regularly. Managed Grasslands are maintained by hu- not included in the Bare Land class. Land temporarily man activity such as fertilization and irrigation, are without vegetative cover because of cropping or tillage is distinguished by enhanced biomass productivity, and classified as Cultivated Land, not Bare Land. can be recognized through vegetative indices based on spectral characteristics. Examples of such areas include lawns, golf courses, forest or shrub areas converted to Wedands grassland, or areas of permanent grassland with altered species composition. This category includes managed Wetlands are lands where saturation with water is the pastures and pastures with vegetation that grows vigor- dominant factor determining soil development and the ously as fallow. Managed Grasslands are used for graz- types of plant and animal communities living in the soil 10 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program and on its surface (Cowardin et al., 1979). A character- sensors. These wetlands are classified as Riverine Water istic feature shared by all wetlands is soil or substrate and Lacustrine Water. that is at least periodically saturated with or covered by water. The upland limit of wetlands is designated as 1) the boundary between land with predominantly hydro- Water and Submerged Land phytic cover and land with predominantly mesophytic or xerophytic cover; 2) the boundary between soil that All areas of open water with <30% cover of trees, shrubs, is predominantly hydric and soil that is predominantly persistent emergent plants, emergent mosses, or lichens nonhydric; or 3) in the case of wetlands without vegeta- are assigned to the superclass Water and Submerged I-and, tion or soil, the boundary between land that is flooded whether the area is considered wetland or deepwater or saturated at some time during the growing season habitat under the Cowardin et al. (1979) classification. each year and land that is not (Cowardin et al., 1979). The Water class includes Cowardin et al.'s (1979) Most wetlands are vegetated and found on soil. classes Rock Bottom and Unconsolidated Bottom, and Wetland in the C-CAP Coastal Land-Cover Classifica- Nonpersistent Emergent Wetlands, as well as Reefs and tion System (Table 2) includes all areas considered Aquatic Beds that are not identified as such. Most wetland by Cowardin et al. (1979) except for bottoms, C-CAP products will display water as a single class. How- reefs, aquatic beds, and nonpersistent emergent wet- ever, the major systems (Marine/ Estuarine, Riverine, lands. The class subdivision was adopted primarily from Lacustrine, Palustrine) are ecologically different from the Cowardin et al. system, shown in Appendix Table 2. one another, and for this reason, the C-CAP system At Level 11, C-CAP incorporates certain Cowardin et al. identifies the four systems as Level III subclasses: 3.11- classes (e.g. Rocky Shore, Unconsolidated Shore, Emer- Marine/Estuarine Water, 3.12-Riverine Water, 3.13- gent Wedand) or grouped Cowardin et al. classes (e.g. Lacustrine Water, and 3.14-Palustrine Water. While Woody Wetland may be further divided into Scrub- C-CAP does not require these subclasses, the option is Shrub and Forested categories) in combination with Cow- provided to participants who may have such data avail- ardin et al. systems (i.e. Marine, Estuarine, Riverine, Lacus- able from ancillary sources. Having the water subclasses trine, Palustrine). Thus, a typical Level II class in the also makes the C-CA-P scheme more compatible with C-CAP system might be Palusuine Woody Wetland. the Cowardin et al. (1979) system. The subclass 3.11- Marine and Estuarine Rocky Shores (Cowardin et al., Marine/Estuarine Water includes Bottoms and unde- 1979) were combined into a single class, Marine/Estua- tected Reefs and Aquatic Beds. The subclasses 3.12- rine Rocky Shore. The same logic was used to produce Riverine Water, 3.13-Lacustrine Water, and 3.14- Marine/Estuarine Unconsolidated Shore. Palustrine Water include Bottoms and undetected Salinity exhibits a horizontal gradient in coastal es- Aquatic Beds as well as Nonpersistent Emergent Wet- tuarine marshes. This is evident not only through the lands. Palustrine waterbodies, defined as covering <20 direct measurement of salinity but in the horizontal acres, are smaller than Lacustrine waterbodies. distribution of marsh plants (Daiber, 1986). Therefore, C-CAP combined Marine and Estuarine Reefs and the Estuarine Emergent Wetland class is partitioned Aquatic Beds into two classes: Marine/Estuarine Reefs into Haline (Salt) and Mixohaline (Brackish) Marshes. and Marine/Estuarine Aquatic Beds. Marine/ Estuarine For both subclasses, the C-CAP classification system Aquatic Beds includes the subclass Rooted Vascular, uses the Cowardin et al. (1979) definitions. Mixohaline which is subdivided into High Salinity Q!5 ppt) and salinity ranges from 0.5 ppt to 30 ppt, and Haline salin- Low Salinity (<5 ppt). The @!5 ppt salinity ievel separates ity is @!30 ppt. Within a marsh, plant zonation is usually seagrasses from submersed grasses and forbs that toler- quite evident. Along the Atlantic coast of North America ate or require low salinity. Both types of plants define the pioneer plant on regularly flooded mudflats is aquatic beds, submersed habitats that are important to saltmarsh cordgrass, Spa7tina alternijlora, which often the C-CAP project. High Salinity includes mesohaline, appears in pure stands. In more elevated areas that are polyhaline, euhaline, and hyperhaline salinity catego- flooded less frequently, saltmeadow hay, Spailina Pat- ries of Cowardin et al. (1979). Low Salinity includes ens, often dominates. The upland interfaces are bor- oligohaline and fresh categories (<5 ppt salinity), dered by marsh elder, Ivafrutescens, and groundsel tree, With the noted exceptions, most of the Wetland and Bacchafis halimifolia. Thus, salt marshes may be subdi- Water classes have definitions similar to those contained vided further into High Marsh and Low Marsh, but this in Cowardin et al. (1979) so that data can be inter- distinction is not required in C-CAP regional projects. changed with other programs, such as the USFWS Na- The C-CAP Coastal Land-Cover Classification System tional Wetlands Inventory (NWI) program, which is based does not attempt to identify freshwater nonpersistent on the Cowardin et al. (1979) classification system. De- emergent wetlands because they are invisible during tailed definitions of all superclasses, classes, and subclasses much of the year and are difficult to detect by remote shown in Table 2 are provided in Appendix 3. Chapter 3 Monitoting Uplands and Wetlands Using Satellite Remote Semor Data Successful remote-sensing change detection of uplands Geometric rectification algorithms Uensen, 1986; Novak, and wetlands in coastal regions requires careful atten- 1992) are used to register the images to a standard map tion to 1) sensor systems, 2) environmental characteris- projection (Universal Transverse Mercator [UTM] for tics, and 3) geodetic control. Failure to understand the most U.S. projects). Rectification should result in the impact of the various parameters on the change detec- two images having a root mean square error (RMSE) of tion process can lead to inaccurate results. Ideally, the !!@ �0.5 pixel. RMSE @! �0.5 pixel may result in the identi- remotely sensed data used to perform C-CAP change fication of spurious areas of change between the two detection are acquired by a remote sensor system that datasets. See "Rectification of Multiple-date Remote holds the following factors constant: temporal, spatial Sensor Data" for a summary of C-CAP image rectifica- (and look angle), spectral, and radiometric. It is in- tion requirements. structive to review each of these parameters and iden- It is possible to perform change detection using data tify why they have a significant impact on the success of collected by two different sensor systems with different C-CAP remote-sensing change detection projects. Table IFOV's, e.g. Landsat TM data (30 x 30 in) for date I and 3 summarizes the characteristics of some of the most SPOT HRV data (20 x 20 in) for date 2. In such cases, it important satellite remote-sensing systems. is necessary to decide upon a representative minimum mapping unit (e.g. 20 x 20 in) and then resample both datasets to this uniform pixel size. This does not present Remote-Sensing System Considerations 2, significant problem as long as one remembers that the information content of the resampled data can Temporal Resolution never be greater than the IFOV of the original sensor Two important temporal resolutions should be held system (i.e. even though the Landsat TM data are constant when performing coastal change detection resampled to 20 x 20 in pixels, the information was still using multiple dates of remotely sensed data. First, the acquired at 30 x 30 in resolution, and one should not data should be obtained from a sensor system which expect to be able to extract additional spatial detail in acquires data at approximately the same time of day the dataset). (e.g. Landsat TM data are acquired before 0945 h for Some remote-sensing systems like SPOT collect data most of the conterminous United States). This elimi- at off-nadir look angles as much as �20' (Table 3), i.e. nates diurnal sun angle effects which can cause anoma- the sensors obtain data of an area on the ground from lous differences in the reflectance properties of re- an "oblique" vantage point. Two images with signifi- motely sensed objects. Second, whenever possible it is cantly different look angles can cause problems when desirable to use remotely sensed data acquired on anni- used for change detection purposes. For example, con- versary dates (e.g. 1 October 1988 versus I October sider a maple forest consisting of very large, randomly 1993). Using anniversary date imagery removes sea- spaced trees. A SPOT image acquired at 0* off-nadir will sonal sun angle differences that can make change de- look directly down upon the "top" of the canopy. Con- tection difficult and unreliable Uensen et al., 1993a). versely, a SPOT image acquired at 20' off-nadir will Usually, precise anniversary date imagery is not avail- record reflectance information from the "side" of the able. The determination of acceptable near-anniversary canopy. Differences in reflectance from the two datasets dates then depends on local and regional factors such as can cause spurious change detection results. There- phenological cycles and annual climatic regimes. fore, the data used in a remote-sensing digital change detection should be acquired with approximately the same look angle whenever possible. Spatial Resolution and Look Angle Accurate spatial registration of at least two images is Spectral Resolution essential for digital change detection. Ideally, the re- motely sensed data are acquired by a sensor system that A fundamental assumption of digital change detection collects data with the same instantaneous-field-of-view is that there should exist a difference in the spectral (IFOV) on each date. For example, Landsat TM data response of a pixel on two dates if the biophysical collected at 30 x 30 in spatial resolution (Table 3) on materials within the IFOV have changed between dates. two dates are relatively easy to register to one another. Ideally, the spectral resolution of the remote sensor 12 NOAA Technical Report NMFS 123: Dobson et at.: Coastal Change Analysis Program Table 3 Selected satellite remote-sensing system characteristics; abbreviations: MSS=multispectral scanner; TM=thematic mapper. Spectral resolution Spatial resolution Temporal Radiometric Remote sensor system (gm) (m) resolution (d) resolution (bits) Landsat MSS 1-5 Band 1 (0.50-0.60) 80 x 80 18 81 Band 2 (0.60-0.70) 80 x 80 18 8 Band 3 (0.70-0.80) 80 x 80 18 8 Band 4 (0.80 -1.1) 80 x 80 18 8 Landsat TM 4-6 Band 1 (0.45-0.52) 30 x 30 16 8 Band 2 (0.52-0.60) 30 x 30 16 8 Band 3 (0.63-0.69) 30 x 30 16 8 Band 4 (0.76-0.90) 30 x 30 16 8 Band 5 (1.55-1.75) 30 x 30 16 8 Band 7 (2.08-2.35) 30 x 30 16 8 Band 6 (10.4-12.5) 120 x 120 16 8 Landsat TM ro, pAN2 Band 8 (0.5-0.90) 15 x 15 16 8 SPOT HRV, XS Band 1 (0.50-0.59) 20 x 20 pointable 8 Band 2 (0.61-0.68) 20 x 20 pointable 8 Band 3 (0.79-0.89) 20 x 20 pointable 8 SPOT HRV, PAN Pan (0.51-0.73) lox 10 pointable 8 1 Landsat MSS I and 2 collected data in 7 bits. 2 The panchromatic (PAN) band was found on Landsat 6, which was lost during a launch mishap. system is sufficient to record reflected radiant flux in values ranging from 0 to 255 (Table 3). Ideally, the spectral regions that best capture the most descriptive sensor systems collect the data at the same radiometric spectral attributes of the object. Unfortunately, differ- precision on both dates. When the radiometric resolu- ent sensor systems do not record energy in exactly the tion of data acquired by one system (e.g. MSS I with 7- same portions of the electromagnetic spectrum, i.e. bit data) are compared with data acquired by a higher bandwidths (Table 3). For example, Landsat MSS radiometric resolution instrument (e.g. TM with 8-bit records energy in four relatively broad bands, SPOT data) then the lower resolution data (e.g. 7-bit) should HRV sensors record in three relatively coarse multi- be "decompressed" to 8-bit data for change detection spectral bands and one panchromatic band, and TM purposes. However, the precision of decompressed records in six relatively narrow optical bands and one brightness values can never be better than the original, broad thermal band (Table 3). Ideally, the same sensor uncompressed. data. system is used to acquire imagery on multiple dates. When this is not possible, the analyst should select bands which approximate one another. For example, The Preferred C-CAP Satellite Sensor System SPOT bands I (green), 2 (red), and 3 (near-infrared) can be used successfully with TM bands 2 (green), 3 TM is currently the primary sensor recommended for (red), and 4 (near-infrared) or MSS bands I (green), 2 GCAP image acquisition and change analysis for all (red), and 4 (near-infrared). Many of the change detec- land cover except aquatic beds. Although its spatial reso- tion algorithms to be discussed do not function well lution is not as good as that of a SPOT satellite or aircraft when bands from one sensor system do not match MSS image, a TM image is generally less expensive to those of another sensor system. For example, using TM acquire and process for large-area coverage. Compared band I (blue) with either SPOT or MSS data is not wise. with SPOT imagery, TM has better spectral resolution and specific spectral bands that are more applicable to wet- lands delineation (bands 5 and 7). In addition, TM is Radiometric Resolution preferred over SPOT because TM has collected data for a longer time (since 1982, as opposed to SPOT since 1986) Converting satellite remote sensor data from analog to and because many TM scenes of U.S. coastal regions were digital usually results in 8-bit brightness values with systematically collected on a routine basis. Chapter 3: Monitoring Uplands and Wetlands 13 There are advantages and disadvantages to using other thin layer of haze can alter spectral signatures in satel- sensors. Aircraft multispectral scanners are more ex- lite images enough to create the false impression of pensive and complex to use over large regions Uensen spectral change between two dates. Obviously, 0% cloud et al., 1987). However, good algorithms are now avail- cover is preferred for satellite imagery and aerial pho- able for georeferencing, and in certain cases (e.g. when tography. At the upper limit, cloud cover >20% is usu- higher spectral or spatial resolution is needed and when ally unacceptable. In addition, clouds not only obscure unfavorable climatic conditions for satellite sensors ex- terrain but the cloud shadow also causes major image ist) aircraft sensors may be optimum. The SPOT sensor classification problems. Any area obscured by clouds or has a greater temporal coverage because the satellite affected by cloud shadow will filter through the entire can collect data off-nadir. However, if off-nadir SPOT change detection process, severely limiting the utility of imagery is used for C-CAP change analyses, the data the final change detection product. Therefore, regional must be normalized to compensate for different look analysts must use good professional judgment to evalu- angles that may preclude pixel-to-pixel spectral-change ate such factors as the criticality of the specific locations analysis. Nevertheless, SPOT imagery may be a reason- affected by cloud cover and shadow and the availability able alternative in certain areas because of cloud cover of timely surrogate data for those areas obscured (e.g. or other impediments to TM data availability. perhaps substituting aerial photography interpretation C-CAP remains flexible to. take. advantage of new for a critical area). Even when the stated cloud cover is sensors and other technologies that become operational 0%, it is advisable to "browse" the proposed image on during the lifetime of the program. Regional partici- microfiche at the National Cartographic Information pants should work with the C-CAP program coordina- Center in each State to confirm that the cloud cover tors to ensure that the sensor selection meets the fol- estimate is correct. lowing C-CAP requirements: Assuming no cloud cover, the use of anniversary dates helps to ensure general, seasonal agreement between � Standard radiometrically corrected TM data are re- the atmospheric conditions on the two dates. However, quired, and geocoded (georeferenced) data are op- if dramatic differences exist in the atmospheric condi- tional. If geocoded data are selected, the coordinate tions present on the n dates of imagery to be used in the system should be UTM. change detection process, it may be necessary to re- � Regional participants must collaborate with C-CAP move the atmospheric attenuation in the imagery. Two managers to ensure that the exchange medium and alternatives are available. First, sophisticated atmo- its format will be amenable to the processing capabili- spheric transmission models can be used to correct the ties of the participants. remote-sensor data if substantial in situ data are avail- � C-CAP normally will purchase and archive the raw able on the day of the overflights. Second, an alterna- data in collaboration with the regional image pro- tive empirical method may be used to remove atmo- cessing center. In cases where the regional partici- spheric effects. A detailed description of one empirical pants already have usable raw imagery or are making method of image-to-image normalization is found in their own purchases, formal agreements between "Radiometric Normalization of Multiple-Date Images. C-CAP managers and participants must address ven- dor licensing and other legal requirements as well as C-CAP archiving and quality-control protocol. Soil Moisture Conditions Important Environmental Characteristics Ideally, the soil moisture conditions should be identical for the n dates of imagery used in a change detection Failure to understand the impact of various environ- project. Extremely wet or dry conditions on one date mental characteristics on the remote-sensing change can cause serious change detection problems. There- detection process can also lead to inaccurate C-CAP fore, when selecting the remotely sensed data to be results. When performing change detection it is desir- used for change detection it is very important not only able to hold environmental variables as constant as to look for anniversary dates but also to review precipi- possible. Specific environmental variables and their tation records to determine how much rain or snow fell potential impacts are described below. in the days and weeks prior to data collection. When soil moisture differences between dates are significant for only certain parts of the study area (perhaps due to Atinospheric Conditions a local thunderstorm), it may be necessary to stratify (eliminate) those affected areas and perform a sepa- There should be no clouds, haze, or extreme humidity rate analysis that can be added back in the final stages on the days remote-sensing data are collected. Even a of the project. 14 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program Vegetation Phenological Cycle Characteristics acceptable scenes may be quite small. In some regions it may be necessary to seek alternative data such as Vegetation grows according to seasonal and annual SPOT satellite data, aerial photographs, or other land- phenological cycles. Obtaining near-anniversary images cover databases. For most regions, mean low tide (MLT) greatly minimizes the effects of wetland seasonal phe- or lower will be preferred, one or two feet above MLT nological differences that may cause spurious change will be acceptable, and three feet or more will be unac- to be detected in the imagery. One must also be careful ceptable (Jensen et al., 1993a). Ideally, tides for aerial about two other factors when dealing with upland sea- photographic surveys of submersed habitat should ap- sonal agricultural crops. First, many monoculture crops proach low tide as predicted in NOS tide tables, but (e.g. corn) normally are planted at approximately the optimal visualization of the subtidal bottom depends same time of year. A month lag in planting date be- on water clarity as well as depth. Two of the 1993 C-CAP tween fields having the same crop can cause serious protocol development projects focus on improving the change detection error. Second, many monoculture C-CAP protocol for tidal effects (see Appendix 5). crops are composed of different species (or strains) of the same crop, which can cause the crop to reflect energy differently on multiple dates of anniversary im- Image Processing Data to Inventory Upland agery. These observations suggest that the analyst must and Wetland Change know the biophysical characteristics of the vegetation as well as the cultural land-tenure practices in the study With the classification scheme developed and the ap- area so that imagery which meets most of these charac- propriate remote-sensor data selected, it is possible to teristics can be selected for change detection. process the data to extract upland and wetland change The choice of image date is best determined by mu- information. This involves geometric and radiometric tual agreement among remote-sensing specialists, bi- correction, selection of an appropriate change detec- ologists, ecologists, and local experts. The selection of tion algorithm, classification if necessary, creation of the acceptable window of acquisition will be made inde- change detection products, and error evaluation (Table pendently by participants in each region. No single 1). A separate section (Chapter 4) describes the extrac- season will serve for all areas because of substantial tion of information on SRV because aerial photography latitudinal variation extending from temperate to tropi- and significantly different photogrammetric techniques cal regions. For example, coastal marshes in the mid- must be utilized. Atlantic region are best inventoried fromJune through October while submersed habitats in southern Florida may be best inventoried in November. Even within Rectification of Multiple-Date Remote Sensor regions, some cover types will be more easily distin- Data guished in different seasons. For example, in the Carib- bean, estuarine seagrasses can be best detected in early Georeferencing (spatial registration of a remotely sensed January, yet marine seagrasses can be best detected in image to a standard map projection) is a necessary step May or June. Technically, these vegetation patterns in digital change detection and cartographic represen- should be monitored at optimal times throughout the tation. The following C-CAP recommendations should year, but cost limitations usually limit the analyst to a be followed when rectifying the base image to a stan- single date. dard basemap: eCreocoded base TM images can be purchased if pre- Effects of Tidal Stage on Image Classification ferred by regional analysts. However, participants should be aware that some analysts have reported Tidal stage is a crucial factor in satellite image scene undocumented variations in commercial products that selection and the timing of aerial surveys. Ideally, tides can lead to poor registration in certain regions, espe- should be constant between time periods, but this would cially where local relief requires substantial terrain rule out synoptic satellite sensors since tidal stages are correction. Additional registration may be necessary not synchronized within a region or even within a single to achieve the C-CAP standard precision of RMSE image. Alternatively, analysts should avoid selecting the �0.5 pixel. Therefore, it is recommended that each highest tides and should take into account the tide regional project perform its own base image-to-map stages occurring throughout each scene. Tidal effect rectification by using data that is radiometrically cor- varies greatly among regions. In the Northwest, for rected but not geocoded. example, when all of the temporal, atmospheric, and *Ground control points (GCP's) used to compute rec- tidal criteria are taken into account, the number of tification transformation coefficients should be rela- Chapter 3: Monitoring Uplands and Wetlands 15 tively static features in the landscape (e.g. road inter- of reducing or better controlling co-registration er- sections) or should be based on new GPS measure- ror among images. Selection and consistency of con- ments taken in the field. When GCP's are digitized trol points and rectification algorithms are important from USGS 7.5' (1:24,000) maps, analysts should use to the success of this technique. Cubic convolution the marginal information and available updates to algorithms normally yield the most precise spatial fit, improve the location of control points. GCP's should but cubic convolution and bilinear interpolation al- be extracted from mylar copies of the USGS maps gorithms suffer from the disadvantage of averaging whenever possible to minimize system-produced digi- pixel brightness values. Nearest-neighbor algorithms tizing error. Traditional paper maps expand and con- are spatially less precise, but they offer the advantage tract with changes in relative humidity and should of retaining pixel brightness values through the pro- not be used for digitizing GCP's. cesses of rectification and registration. �C-CAP recommends the use of the current NAD '83 national datum. Unfortunately, most existing map series are based on the NAD '27 datum. NAD '27 will be Radiometric Normalization of Multiple-Date acceptable on a region-by@region basis until published Images maps based on NAD '83 are universally available. �In all but the flattest coastal regions, terrain correc- The use of remotely sensed data to classify coastal and tion of imagery may be necessary to reduce image upland land cover on individual dates is contingent distortion caused by local relief. upon there being a robust relationship between re- �The required coordinate system is UTM. If another motely sensing brightness values (BV's) and actual sur- coordinate system is used (e.g. state plane), it is the face conditions. However, factors such as sun angle, responsibility of the regional analyst to provide com- Earth/Sun distance, detector calibration differences plete documentation and conversion equations. between the various sensor systems, atmospheric condi- �It is the responsibility of the regional analyst to un- tion, and sun/target/ sensor geometry (phase angle) derstand (or seek advice concerning) the variety of will also affect pixel brightness value. Differences in rectification-resampling algorithms (e.g. bilinear in- direct beam solar radiation due to variation in sun terpolation, nearest neighbor, cubic convolution) and angle and Earth/sun distance can be calculated accu- their impact on the data. Nearest-neighbor resampling rately, as can variation in pixel BV's due to detector is recommended. calibration differences between sensor systems. Remov- ing atmospheric and phase-angle effects requires infor- Rectification of an earlier date (Tb._,) or later date mation about the gaseous and aerosol composition of (Tb,j) to the base image (Tb) can be accomplished in the atmosphere and the bidirectional reflectance char- several ways. The primary concern is to accomplish the acteristics of elements within the scene. However, at- most exact co-registration of pixels from each time mospheric and bidirectional reflectance information period and thus reduce a potentially significant source of are rarely available for historical remotely sensed data. error in change analysis (Lunetta et al., 1991). The follow- Also, some analysts may not have the necessary exper- ing are minimum recommendations and requirements: tise to perform a theoretically based atmospheric path radiance correction on remotely sensed data. Hence, it �Geocoded and terrain-corrected TM data can be or- is suggested that a relatively straightforward "empirical dered from commercial vendors. Two separate im- scene normalization" be employed to match the detec- ages can be overlaid according to like coordinates, tor calibration, astronomic, atmospheric, and phase- but this technique may introduce error if prior angle conditions present in a reference scene. geocoding was not precisely the same in both images. Image normalization reduces pixel BV variation The regional analyst has no control in this process, caused by nonsurface factors, so variations in pixel BV's but if high precision is accomplished by the vendor, between dates can be related to actual changes in sur- the analyst can significantly reduce image processing face conditions. Normalization enables the use of im- effort at the regional facility. age analysis logic developed for a base-year scene to be �The regional analyst can geocode the image to UTM applied to other scenes. This can be accomplished us- coordinates as was done with the base image. If this ing techniques pioneered by the U.S. Bureau of Land technique is adopted, it is important to use the iden- Management (Eckhardt et al., 1990). Image normaliza- tical GCP's and resampling algorithm that were used tion is achieved by developing simple regression equa- to rectify the base image. tions between the brightness values of "normalization �For multiple images, the preferred technique is to targets" present in Tb and the scene to be normalized rectify nongeocoded images directly to the geocoded (e.g. T,._j or T,,,j). Normalization targets are assumed to base image. This technique may have the advantage be constant reflectors, therefore any changes in their 16 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Prograin brightness values are attributed to detector calibration, Selecting the Appropriate Change Detection astronomic, atmospheric, and phase-angle differences. Algorithm Once these variations are removed, changes in BV may be related to changes in surface conditions. C-CAP is the first Federal program to state as a primary Acceptance criteria for potential "normalization tar- goal the monitoring of coastal habitat change using gets" (Eckhardt et al., 1990) are as follows: satellite technology (Cross and Thomas, 1992). The implementation and continuing evolution of the pro- � Targets must be at approximately the same elevation gram is based on the fact that improved cartographic, as the land cover of primary interest within the scene. digital image processing, and photointerpretation meth- Most aerosols in the atmosphere occur <1000 m above ods must be developed for a program of this geographic ground level (AGL). Selecting a mountain-top nor- coverage, spatial resolution, and temporal frequency malization target, thus, would be of little use in esti- (nationwide, 30 x 30 m pixel, every one to five years). mating atmospheric conditions near sea level. Al- Initial implementation of C-CAP will require a blend of though C-CAP projects are on the coast, many re- traditional and innovative approaches to change analysis. gions include areas of substantial local relief. Because the program has adopted a digital format, with � Targets should contain only minimal amounts of veg- TM as a primary sensor, new techniques in processing can etation. Vegetation spectral reflectance can change be easily incorporated into future iterations. over time because of environmental stresses and plant The selection of an appropriate change detection phenology. Good targets include bare soil fields and algorithm is very important Uensen, 1986; Dobson and deep, nonturbid water bodies. Bright, 1991, 1992, and 1993; Jensen et al., 1993a). � Targets must be on relatively flat terrain so that incre- First, it will have a direct impact on the type of image mental changes in sun angle between dates will have classification to be performed (if any). Second, it will the same proportional increase or decrease in direct dictate whether important "from-to" information can be beam sunlight for all normalization targets. extracted from the imagery. C-CAP requires that from-to � Normalization targets should have approximately the information be readily available in digital form suitable same texture over time. Changing textural patterns for geographic analysis and for producing maps and tabu- indicate variability within the target, which could mean lar summaries. At least seven change detection algorithms that the reflectance of the target as a whole may not are commonly used by the remote-sensing community: be constant over time. For example, a mottled pat- tern on what had previously been a uniformly gray, 1. Change Detection Using Write Function Memory In- dry lake bed indicates changing surface moisture con- sertion-Example: Kittredge and Fort Moultrie, S.C. ditions, which would eliminate the dry lake bed from 2. Multiple-Date Composite Image Change Detection- consideration as a normalization target. No example provided. 3. Image Algebra Change Detection (Band Differencing The mean BV's of the Tb targets are regressed against or Band Ratioing) -No example provided. the mean BV's of the T,,_i or Tb.,, targets for the n bands 4.Postclassification Comparison Change Detection- used in the classification of the remote sensor data (e.g. Example: Fort Moultrie, S.C. TM bands 2, 3, and 4). The slope and yintercept of the 5. Multiple-Date Change Detection Using a Binary Mask n equations are then used to normalize the Tb+1 or Tb._1 Applied to Tb.-l-Example: Chesapeake Bay, Md. Landsat TM data to the Tb Landsat TM data. Each regres- 6. Multiple-Date Change Detection Using Ancillary Data sion model contains an additive component (y-inter- Source as TC---No example provided. cept) that corrects for the difference in atmospheric 7. Manual On-Screen Digitization of Change-No ex- path radiance between dates and contains a multiplica- ample provided. tive term (slope) that corrects for the difference in detector calibration, sun angle, Earth/Sun distance, It is instructive to review these alternatives, identify atmospheric attenuation, and phase angle between dates. those acceptable to C-CAP, and provide specific ex- It is customary first to normalize the remote-sensor amples where appropriate. data and then perform image rectification (using near- est-neighbor resampling if image classification is to take place). These data are then ready for individual date Change Detection Using Write Function Memory classification or the application of various multi-image Insertion change detection algorithms. Most studies that attempt to monitor biophysical properties such as vegetation It is possible to insert individual bands of remotely biomass, chlorophyll absorption, and health require sensed data into specific write function memory banks atmospheric correction. (red, green, and/or blue) in the digital image process- Chapter 3: Monitoring Uplands and Wetlands 17 ing system (Fig. 1) to visually identify change in the imagery Uensen et al., 1993b). For Multi-Date Visual Change Detection example, consider two Landsat TM scenes of Using Write-Function Memory Insertion the Fort Moultrie quadrangle near Charles- ton, SC, obtained on I I November 1982 and Date 1 band n Red image plane 19 December 1988. Band I of the 1982 image Green image plane was placed in the green image plane; band I Date 2 band n Blue image plane of the 1988 image, in the red image plane; Date 3 band n and no image, in the blue image plane (Fig. 2). All areas that did not change between the Advantages: Disadvantages: two dates are depicted in shades of yellow visual examination of 2 or 3 , non-quantitative les (i.e. in additive color theory, equal intensiti years of non-specific change - no 'from-to'change class information of green and red make yellow). The graphic depicts numerous changes, including Figure I � beach and sand bar accretion (red) and Diagram of Multiple-Date Change Detection using Write Function Memory erosion (green), insertion Uensen, 1994). � new urban development (red), and � changes in tidal stage between dates (green 1987 and 1988; Eastman and Fulk, 1993). This results and red). in the computation of eigenvalues and factor loadings that are used to produce a new, uncorrelated PCA Advantages of this technique include the possibility image dataset. Usually, several of the new bands of of looking at two and even three dates of remotely information are directly related to change. The diffi- sensed imagery at one time, as demonstrated byjensen culty arises when trying to interpret and label each et al. (1993b). Unfortunately, the technique does not component image. Nevertheless, the method is valu- produce a classified land-cover database for either date able and is used frequently. and, thus, does not provide quantitative information The advantage of the techniques is that only a single on the amount of area changing from one land-cover classification is required. Unfortunately, it is often diffi- category to another. Nevertheless, it is an excellent cult to label the change classes, and no from-to change analog method for quickly and qualitatively assessing class information is available, the amount of change in a region, which might help to select one of the more rigorous change detection tech- niques to be discussed. Image Algebra Change Detection It is possible to simply identify the amount of change Multiple-Date Composite Image Change Detection between two images by band ratioing or image differencing the same band in two images that have Numerous researchers have rectified multiple dates of previously been rectified to a common basemap. Image remotely sensed imagery (e.g. selected bands of two differencing involves subtracting the imagery of one TM scenes of the same region) and placed them in a date from that of another (Fig. 4). The subtraction single dataset (Fig. 3). This composite dataset can be results in positive and negative values in areas of radi- analyzed in a number of ways to extract change infor- ance change and zero values in areas of no-change in a mation. First, a traditional classification using all n bands new "change image." In an 8-bit (28) analysis with pixel (six in the example in Fig. 3) may be performed. Unsu- values ranging from 0 to 255, the potential range of pervised classification techniques will result in the cre- difference values is -255 to 255. The results are nor- ation of "change" and "no-change" clusters. The analyst mally transformed into positive values by adding a con- must then label the clusters accordingly. stant, c (usually 255). The operation is expressed math- Other researchers have used principle component ematically as analysis (PCA) to detect change Uensen, 1986). Again, the method involves registering two (or more) dates of Dijk @ BV (1) - BV.. k(2) + c remotely sensed data to the same planimetric basemap ijk as described earlier and then placing them in the same where dataset. A PCA based on variance-covariance matrices D-h = change pixel value, or a standardized PCA based on an analysis of correla- Br.-k(l) = brightness value at Tb, tion matrices is then performed (Fung and LeDrew, B@-k(2) = brightness value at Tb-, or Tbl, 18 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program c = a constant (e.g. 255), distributed around the mean and pixels of change are i = line number, found in the tails of the distribution. Band ratioing j = column number, and involves exactly the same logic except a ratio is com- k = a single band (e.g. TM band 4). puted between Tb and Tb.,, or Tb.-jand the pixels that did not change have a value of "l " in the change image. The "change image" produced using image differ- A critical element of both image-differencing and encing usually yields a BV distribution approximately band-ratioing change detection is deciding where to Gaussian in nature, where pixels of no BV change are place the threshold boundaries between "change" and A"T A At- ?Mv lot 4k Y JP1 v -1A 7 oil 40 Figure 2 Example of Multiple-Date Change Detection using Write Function Memory Insertion using two dates of Landsat Thematic Mapper imagery of Fort Moultrie, S. C. Red image plane = TM band 3, 19 Dec 1988; green image plane = TM band 3, 9 Nov 1982; blue image plane = blank. Chapter 3: Monitoring Uplands and Wetlands 19 no-change" pixels displayed in the histogram Multi-Date Composite Change Detection of the change image Uensen, 1986). Often, a standard deviation from the mean is selected 2 :1 3Rectified Thematic and tested empirically. Conversely, most ana- Date 1 4 Mapper bands lysts prefer to experiment empirically, plac- 0 ing the threshold at various locations in the 3 tails of the distribution until a realistic amount Date 2 -its; 4 of change is encountered. Thus, the amount _9@1 of change selected and eventually "recoded" or for display is often subjective and must be based on familiarity with the study area. There 6 Principal are also analytical methods that can be used Traditional Components to select the most appropriate thresholds. Classification Unfortunately, image differencing simply identifies those areas that may have changed Advantages: Disadvantages: and provides no information on the nature of requires single classification - difficult to label change classes the change, i.e. no from-to information. Nev- - no 'from-to' change classes available ertheless, the technique is valuable when used in conjunction with other techniques such as Figure 3 the multiple-date change detection using a 4 1_ Diagram of Multiple-Date Composite Image Change Detection (Jensen, 1994). binary change mask to be discussed in " u tiple-Date Change Detection Using a Binary Change Mask Applied to Tb_1 or Tb,,." Image Algebra Change Detection 2 Postdassification Comparison Change 3- Rectified Thematic Detection Date 1 4 Mapper bands The most commonly used quantitative method of change detection is postclassification com- 2 Rectified Thematic parison Uensen, 1986; Jensen et al., 1993a) 3 - and may be used in regional C-CAP projects Date 2 4 Mapper bands under certain conditions. It requires rectifi- cation and classification of each of the re- Composite motely sensed images (Fig. 5). These two maps Dataset are then compared on a pixel-by@pixel basis by using a "change detection matrix" to be discussed. Unfortunately, every error in the Image differenced or individual date classification map will also be band ratioed image present in the final change detection map (Rutchey and Velcheck, 1993). Therefore, it is imperative that the individual classification Recoded to produce binary maps used in the postclassification change 'Change/No-change' Mask detection method be extremely accurate (Augenstein et al., 1991; Price et al., 1992). To demonstrate the postclassification com- parison change detection method, consider Advantages: Disadvantages: the Kittredge (40 river miles inland from efficient method of identifying - no 'from-to'change classes available pixels which have changed - requires careful selection of the Charleston, S.C.) and Fort Moultrie, S.C. study in brightness value between dates 'change/no-change' threshold areas (Fig. 6) Uensen et al., 1993a). Nine classes of land cover were inventoried on each Figure 4 date (Fig. 7). The 1982 and 1988 classifica- Diagram of Image Algebra Change Detection Uensen, 1994). tion maps were then compared on a pixel-by- pixel basis using an n x n GIS "matrix" algorithm whose ues from 1 to 81. The analyst then selected specific logic is shown in Figure 8. This resulted in the creation from-to classes for emphasis. Only a select number of V- of "change images maps" consisting of brightness val- the 72 (W-n) possible off-diagonal from-to land-cover 20 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program change classes summarized in the change matrix (Fig. Postclassification comparison change detection is 8) were selected to produce the change detection maps widely used and easy to understand. When conducted (Fig. 9). For example, all pixels which changed from by skilled image analysts it represents a viable tech- any land cover in 1982 to Developed Land in 1988 were nique for the creation of C-CAP change detection prod- color coded red (RGB=255, 0, 0) by selecting the ap- ucts. Advantages include the detailed from-to informa- propriate from-to cells in the change detection matrix tion and the classification map for each year. Unfortu- (10, 19, 28, 37, 46, 55, 64, and 73). Note that the change nately, the accuracy of change detection is heavily de- classes are draped over a TM band-4 image of the study pendent on the accuracy of the two separate classifica- area to facilitate orientation. Similarly, all pixels in tions. Postclassification comparison is not recommended 1982 that changed to Estuarine Unconsolidated Shore for C-CAP regional projects except under special cir- by 19 December 1988 (cells 9,18, 27, 36,45, 54,63, and cumstances, such as when different sensors are involved 72) were depicted in yellow (RGB=255, 255, 0). If de- or when two separate organizations are classifying the sired, the analyst could highlight very specific changes, same region at different times. such as all pixels that changed from Developed Land to Estuarine Emergent Wetland (cell 5 in the matrix), by assigning a unique color look-up table value (not Multiple-date Change Detection Using a Binary shown). A color-coded version of the change detection Change Mask Applied to Tb--i or Tb-,, matrix can be used as an effective from-to change de- tection map legend Uensen and Narumalani, 1992). This method of change detection is highly recom- mended for C-CAP regional projects. First, the analyst selects the base image, T.. Date 2 may be an earlier image Tb._1 or a later image Multi-Date Change Detection Using Tlw* A traditional classification of Tb is per- Post-Classification Comparison formed by using rectified remote sensor data. Next, one of the bands (e.g. band 3 in Figure 2 Rectified Thematic 10) from both dates of imagery are placed in a " Ma Date 3 new dataset. The two band dataset is then ana- 4 pper bands lyzed by using various image algebra functions (e.g. band ratioing, image differencing, princi- classification map of Date 1 pal components analysis) to produce a new image file. The analyst usually selects a thresh- old value to identify spectral change and no- change pixels in the new image as discussed in 2 "Image Algebra Change Detection." The spec- e Date 2 3 ctified Thematic tral change image is then recoded into a binary 4 Mapper bands mask file, consisting of pixels with spectral change between the two dates, and these are viewed as Classification map of Date 2 candidate pixels for categorical change. Great care must be exercised when creating the change/no-change binary mask (Dobson and Classification map of Date 1 Bright, 1993; Jensen et al., 1993a). The change mask is then overlaid onto Tb-i or Tb,., of the analysis and only those pixels which were de- Change map produced using tected as having changed are classified in Tb._1 or 'change detection matrix' logic Tbl* A traditional postclassification comparison applied to Date 1 and Date 2 can then be applied to yield from-to change classification maps information. Hence, many pixels wi th sufficient change to be included in the mask of candidate Advantages: Disadvantages: pixels may not qualify as categorical land-cover - provides 'from- to' change * dependent on accuracy of individual change. class information date classifications Dobson and Bright (1991, 1992, and 1993) * next base year is already # requires two separate classifications completed used this change detection methodology to inventory change in the area surrounding the Figure 5 Chesapeake Bay using TM imagery obtained ;S7 Diagr-am of Postclassification Comparison Change Detection (Jensen, 1994). on 9 September 1984 and 3 November 1988 Chapter 3: Monitoring Uplands and Wetlands 21 7 ... ....... % % a Kittredge, S.C. 11/09/82 12/19/88 XA 11, '411 W0. 4e@ C d Fort Moultrie, S.C. 11/09/82 Scale 12/19/88 Landsat Thematic Mapper Data 22@ Meters Bands 4,3,2 = RGB 5000 0 5000 Figure 6 Rectified Landsat Thematic Mapper data: (a and b) obtained for the Kittredge, S. C., 75 quadrangle C-CAP study area, 9 Nov 1982 and 19 Dec 1988 (Jensen et al., 1993a) (c and d) Obtained for the Fort Moultrie, S. C., 75 quadrangle study area, 9 Nov 1982 and 12 Dec 1988. 22 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program 40 % 2% Kittredge, S. C. 11/09/82 12/19/88 _JEw 41 C d Fort Moultrie, S. C. 11/09/82 12/19/88 Legend Developed Land r___j Estuarine Unconsolidated Shore Grassland Cultivated Land Woody Land Riverine Aquatic Bed Palustrine Woody Wetland Water Estuarine Emergent Wetland Figure 7 Multiple-date land-cover classification maps: (a and b) Kittredge, S. C., study area, produced from 9 Nov 1982 and 19 Dec 1988 Landsat TM data. (c and d) Fort Moultrie, S. C., study area, produced from 9 Nov 1982 and 19 Dec 1988 Landsat TM data Uensen et al., 1993a). Chapter 3: Monitoring Uplands and Wetlands 23 V) "a 1988 "ZI a CIO CI E r_ "From -To" ca Change Detection Legend TO: S@ r_ > ii From: U 0 W a. W Developed Land Cultivated Land MEESE: 1982 Grassland E MESS -m 27 36 Woody Land ME MEE 1 45 Estuarine Emergent Wetland VMS ME` Riverine Aquatic Beds M 14 63 Palustrine Woody Wetland no Water MEESE Estuarine Uncolsolidated Bottom NEESE. @3'1 Color look-up table values in Red 255 255 255 255 255 255 0 0 255 change detection map -iw-Green 0 255 255 255 163 0255 0 255 Blue 0 255 255 255 0 255 255 255 0 No change in landcover between dates, and not selected for display Change in land cover between dates, F@ but not selected for display New Developed Land (cells 10, 19, 28,37, 46,55,64,73) shown in red (RGB=255,0,0) New Estuarine Unconsolidated Shore (cells 9,18,27,36,45,54,63,72) shown in yellow (RGB=255,255,O) Figure 8 Change detection matrix. The basic elements of a change detection matrix may be used to select specific "from-to" classes for display in a "postclassification comparison" change detec- tion map. There arc (n2 - n) off-diagonal possible change classes which may be displayed in the change detection map (72 in this example) although some may be highly unlikely. The colored off-diagonal cells in this diagram were used to produce the change maps in Figure 9. For example, any pixel in the 1982 map that changed to Developed Land by 1988 is red (RGB=255,0,0). Any pixel that changed into Estuarine Unconsolidated Shore by 1988 is yellow (RGB=255,255,O). Individual cells can be color coded in the change map to identify very specific "from-to" changes (Jensen et al., 1993a). _7@ 4 lip j 1 dr A 0 A Kittredge, S.C. Fort Moul Changes in Land Cover 1982 to 1988 Scale Developed Land =5MMEwe Palustrine Woody Wetland 5 60@0 0 50 Estuarine Emergent Wetland Estuarine Unconsolidated Shore Riverine Aquatic Bed water Figure 9 Change detection maps of the Kittredge and Fort Moultrie, S. C., study areas derived from analysis of 11 Nov 1982 and 19 Dec 1988 Lands The nature of the change classes selected for display are summarized in Figure 8. The change information is overlaid onto the Landsat image of each date for orientation purposes Uensen et al., 1993a). Chapter 3: Monitoring Uplands and Wetlands 25 imagery and only those pixels which were detected as Tb-,, image is required. It may also be possible to update having changed were classified in the earlier image. A the NWI map (Tb) with more current wetland informa- from-to matrix similar to the one shown in Figure 9 was tion (this would be done using a GIS "dominate" func- then used to produce a change map of the region (Fig. tion and the new wetland information found in the Tb_1 15). Summary statistics for the region are found in or Tbl classification). The disadvantage is that the NWI Table 4. This process may be repeated with a later scene data must be digitized and generalized to be compat- to determine successive change. ible with the C-CAP Coastal Land-Cover Classification This method may reduce change detection errors System, then converted from vector to raster format to (omission and commission) and provides detailed be compatible with the raster remote-sensor data. Any from-to change class information. The technique re- manual digitization and subsequent conversion intro- duces effort by allowing analysts to focus on the small duces error into the database which may not be accept- amount of area that has changed between dates. In able (Lunettaetal., 1991). most regional projects, the amount of actual change over one to five years is probably no greater than 10% of the total area. The method Multi-Date Change Detection Using A is complex, requiring a number of steps, and Binary Change Mask Applied to Date 2 the final outcome is dependent on the quality of the change/no-change binary mask used in 2 the analysis. A conservative threshold may ex- Date 1 3 Rectified Thematic 4 Map clude real change while a liberal threshold may per bands create problems similar to those of the post- classification comparison technique (See Traditional classification "Postclassification Comparison Change Detec- of Date 1 tion.") 3 Date 1 band 3 Multiple-date Change Detection Using @F`_ @3 Date2ba,d3 AncWary Data Source as Tb Image algebra to identify Sometimes a land-cover data source may be change pixels, e.g. ratio of A"- "! used in place of a traditional remote-sensing multidate band 3 data. image in the change detection process. For Create change pixel mask example, the NWI is inventorying all wetlands I in the United States at the 1:24,000 scale. Date 2 Mask out change pixels 2 Some of these data have been digitized. In- 3in Date 2 imagery and stead of using a remotely sensed image as Tb classify in the analysis, it is possible to substitute the Classification map of Date 2 digital NWI map of the region (Fig. 16). In this case, the NWI map would be "recoded" to be compatible with the C-CAP Coastal Land- Classification map of Date I Cover Classification System (Table 2). This should not be difficult since the two systems are highly compatible. Next, Tb.-i or Tb-,, of rm Post-Classification Comparison Change Detection the analysis is classified and then compared on a pixel-by-pixel basis with Tb information. or Update Date 1 map with Traditional from-to information can then be Date 2 change information derived. As with any other postclassification using GIS dominate function comparison, the accuracy of the change data- base is dependent on the accuracy of both Advantages: Disadvantages: input databases (C-CAP and NWI) may reduce change detection - requires a number of steps Advantages of the method include the use errors (omission and comission) - dependent on quality of'change/ - provides 'from- to' change no-change' binary mask of a well-known, trusted data source (NWI) class information and the possible reduction of errors of omis- sion and commission. Detailed from-to infor- Figure 10 mation may be obtained by using this method. Diagram of Multiple-Date Change Detection Using a Binary Change Also, only a single classification of the Tb_1 or Mask Applied to Date 2 Uensen, 1994). 26 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program el '_4 Ap 4r !L 4@ P, 'A V -4, V, M Path 14 Row 34 9121184 Metomkin Inlet AM tw Figure I I Rectified Thematic Mapper imagery of the Metomkin Inlet area obtained on 21 Sep 1984 (Dobson and Bright, 1992). Manual On-Screen Digitization of Change basemap and compared to identify change. Digitized high resolution aerial photographs displayed on a CRT Considerable amounts of high resolution remote sen- screen can be interpreted easily using standard photo sor data are now available (e.g. SPOT 10 X 10 m, the interpretation techniques based on size, shape, shadow, aircraft mounted Calibrated Airborne Spectrographic texture, etc. (Ryerson, 1989). Therefore, it is becoming Imager [CASI] of the National Aerial Photography Pro- increasingly common for analysts to interpret visually gram [NAPP]). These data can be rectified and used as both dates of aerial photographs (or other type of planimetric basemaps or orthophotomaps. Often aerial remote-sensor data) on the screen, annotate the impor- photographs are scanned (digitized) at high resolu- tant features using heads-up on-screen digitizing, and tions into digital image files (Light, 1993). These pho- compare the various images to detect change (Cowen tographic datasets can then be registered to a common et al., 1991; Cheng et al., 1992; Lacy, 1992; Wang et al., Chapter 3: Monitoring Uplands and Wetlands 27 IVA AW- @41 TM Path 14 Row 1. 11/03188 Metomkin Inlet .-Tn- -A Figure 12 Rectified Thematic Mapper imagery of the Metomkin Inlet area obtained on 3 Nov 1988 (Dobson and Bright, 1992). 1992; Westmoreland and Stow, 1992). The process is to interferences from aquatic as well as atmospheric especially easy when 1) both digitized photographs (or sources. As with other new technologies, demonstra- images) are displayed on the CRT side by side, and 2) tion of the appropriateness of interpretation of scanned they are topologically linked through object-oriented photographs will be a critical step in expanding the G programming so that a polygon drawn around a feature CAP Protocol (Also see "Accuracy Assessment for Indi- on one photograph will also be drawn around the same vidual Date Classification of Water and Submersed Habi- feature on the other photograph. Scanning aerial pho- tat Data"). The manual on-screen approach is recom- tographs unavoidably reduces the spatial and spectral mended as a useful adjunct to other change detection resolution of source data. This loss may be significant in methods. Its principle drawback is the time required to photographs of submerged features, which are subject cover large regions in such a labor-intensive fashion. 28 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program E 4** c, V V N x e@, i > V V,_ 44 N, C-CAP 63 MrTMM - - @7 g 11/03/88 -v- Developed-High d eveloped-Low 5=1, - Al Cropland Grassland Deciduous Forest A S nI 1@ I '- Evergreen Forest J J @ J , I ; , Mixea Forest Scrub/Shrub Palustrine Forest Estuarine Marsh Palus rine Marsh t r1dal Flats _14 Barren Land Water 4 7-1 J_rq \-71 Figure 13 Classification map of 3 Nov 1988 Landsat Thematic Mapper imagery of the Metomkin Inlet area (Dobson and Bright, 1992). Selecting Appropriate Classification The previous section indicated that these three of Algorithms the seven most commonly used change detection algo- rithms are acceptable for C-CAP regional projects: C-CAP requires that the classification procedures used as part of the change detection process be approved 6 Postclassification Comparison and documented. Classification algorithms used in each e Change Detection Using a Binary Change Mask Ap- region will be selected based on the capabilities and needs plied to Tb._1 or T., of the regional participants. C-CAP assumes that the re- e Change Detection Using Ancillary Data Source as Tb. gional participants are experienced in image processing and mapping. If not, C-CAP will attempt to provide funda- Each of these requires a complete pixel-by-pixel classifi- mental technical assistance on a case-by-case basis. cation of one date of imagery and, at least, a partial Chapter 3: Monitoring Uplands and Wetlands 29 _41.4 - 410.4c V4 1b .0e 11b 4@ 41 41 V It W. Spectral Change Mask File 'AW P4 01. Figure 14 Binary "change/no-change mask" produced by image differencing TM bands 3, 4, and 5 of each date (Dobson and Bright, 1992). classification of an additional date. Hence, it is instruc- will have sufficient expertise to assess the advantages of tive to review the C-CAP-approved image classification alternative classification algorithms and to recognize logic which may be used in the regional projects. when human pattern recognition and other types of intervention are necessary. In practice, it may be neces- sary to employ a suite of algorithms including both Supervised and Unsupervised Image supervised and unsupervised statistical pattern recogni- Classification Logic tion approaches. Currently, maximurn-likelihood clas- sifiers often serve as a good first step, but new statistical The primary reason for employing digital image classi- approaches are being developed and implemented on a fication algorithms is to reduce human labor and im- routine basis Uensen et al., 1987; Hodgson and Plews, prove consistency. It is expected that regional analysts 1989; Foody et al., 1992). It is important for analysts to 30 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program remain flexible with regard to procedures and algorithms. pixels are assigned to the nearest class in n-dimensional Standard supervised and unsupervised classification feature space, or passed to a maximum-likelihood clas- techniques have been available for more than 20 years sification algorithm that assigns an unknown pixel to and are well documented in texts byjensen (1986) and the class in which it has the highest probability of being Campbell (1987). In a supervised classification, the a member. Great care must be exercised when selecting analyst "trains" the classifier by extracting mean and training samples (Mausel et al., 1990). covariance statistics for known phenomena in a single In an unsupervised classification, the computer is date of remotely sensed data (Gong and Howarth, 1990). allowed to query the multispectral properties of the These statistical patterns are then passed to a mini- scene by using user-specified criteria and to identify x mum-distance-to-means algorithm in which unknown mutually exclusive clusters in n-dimensional feature 1-4 -1-4 41 W 1% 2tAI V to Y Fi gure 15 WIA6 44 7 %,.a _.Rat Lcj@@ MAP A map showing selected C-CAP change classes derived from analysis of the 21 Sep 1984 and 3 Nov 1988 Landsat TM data of the Metomkin Inlet area (Dobson and Bright, 1992). Chapter 3: Monitoring Uplands and Wetlands 31 Multi-Date Change Detection Using An Ancillary Data Source as Date 1 Ancillary data source e.g., Dat 1 National Wetlands inventory Map ...............2Rectified Thematic Date 2 -3 4 Mapper bands Classification map of Date 2 Classification map of Date 1 Perform Post-Classification Comparison Change Detection or Update Date 1 NWI map with Date 2 change information using GIS dominate function Advantages: Disadvantages: � may reduce change detection dependent on quality of ancillary errors (omission and cornission) information � provides 'from-to'change class information � requires a single classification Figure 16 Diagram of Multiple-Date Change Detection Using Ancillary Data Source as Date I Uensen, 1994). space (Chuvieco and Congalton, 1988). The analyst of five training sites per land-cover class be collected. must then convert (label) the x spectral clusters into This creates a representative training set when per- information classes such as those found in the C-CAP forming supervised classification and makes labeling Coastal Land-Cover Classification System. Training sites clusters much easier in an unsupervised classification. visited in the field and identifiable in the digital imag- In addition to the image analysts, the field team should ery are also indispensable when labeling clusters in an contain specialists in ecology, biology, forestry, geogra- unsupervised classification. The following sections dis- phy, statistics, and other pertinent fields, such as agronomy. cuss C-CAP guidelines for collecting training and verifi- Field samples should be stratified by land-cover type and cation samples. by various physical factors such as slope, elevation, vegeta- tion density, species mix, season, and latitude. The po- lygonal boundary of all field sites should be measured Selection of Training and Verification Samples using GPS whenever possible, and the locational, tempo- for Supervised and Unsupervised Classification ral, and categorical information should be archived. The collection of field training sites often requires Only training sites that were actually visited on the multiple visits to the field. Some of the field sites may be ground by experienced professionals should be selected used to train a classifier or label a cluster while a certain for extracting the multispectral statistical "signature" of proportion of the field sample sites should be held back a specific class when performing a supervised or unsu- to be used for classification error assessment, which will pervised classification. It is suggested that a minimum be discussed. 32 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program Table 4 Statistical summary of areal change (in ha) by land-cover class for the Metomkin Inlet area shown in Figures 12-16. Read across each row to find which categories the 1988 totals came from. Read down each column to find which categories the 1984 totals changed to. Bold numbers along the diagonal indicate the area that did not change from 1985 to 1988. uncon. = unconsolidated. 1984 Classification Developed Grassland/ Forest Scrub/ Palustrine Estuarine Palustrine Water/ Bare land cultivated land shrub forest emergent emergent uncon. shore land TOTAL 1988 Classification Developed land 1,158 85 8 0 0 4 0 1 0 1,256 Grassland/ cultivated 0 219341 562 0 0 2 0 0 17 21,922 Forest land 0 165 189915 0 0 1 0 0 0 19,081 Scrub/shrub 0 240 562 854 0 2 0 0 0 1,658 Palusuine forest 0 20 9 0 787, 0 0 0 0 816 Estuarine emergent 0 26 9 0 0 11,587 0 13 8 11,643 Palustrine emergent 0 0 0 0 0 0 0 0 0 0 Water/uncon. shore 0 4 0 0 0 2 0 37,172 144 37,322 Bare land 0 19 0 0 0 23 0 124 507 673 TOTAL 1,158 21,900 20,065 854 787 11,621 0 37,310 676 94,371 The following materials are indispensable to a suc- is preferred rather than transparent-overlay techniques, cessful field exercise: which are cumbersome and difficult to use under field conditions. � Imagery geocorrected to a standard map projection C-CAP investigators have assembled and tested a field � Topographic maps at 1:24,000 or the largest available station based on a color laptop computer with commer- scale cial software. At present the software supports visualiza- � Global Positioning System (GPS) tion of raster imagery (e.g. satellite data, digital � Aerial photographs orthophotographs, scanned aerial photographs) and vector databases (e.g. TIGER road networks, NWI wet- it is advisable to perform, at least, a cursory classifica- lands). A version of the software soon to be available tion before initiating fieldwork. In this case, both raw from commercial vendors will allow realtime input of and classified data should be taken to the field. The GPS coordinates. It will then be possible to follow field primary function of the cursory classification is to guide movements directly on the image and map data. The field workers in targeting the covers and signatures that software also allows for completion of field forms on are most difficult and confusing. Keep in mind that the screen in the field. Preliminary tests are encouraging, vast majority of all cover will be easy to identify on the but the field station is not fully operational at this time. ground and on the imagery. Efficient use of field time One shortcoming, for example, is the poor performance provides for quick verification of easy cover types and of active matrix color screens in sunlight. maximum attention to difficult, unusual, and ecologi- cally critical cover types. Field investigators should anticipate the need to know Use of Collateral Data in Image Classification not only the geodetic coordinates of training sites but also the layout of the road network that will provide The overriding goal is to produce accurate individual access. It is advisable to imbed roadway information date classifications and accurate change detection data- into the raw imagery. This can be done using the Bu- bases. Any information or operation that enhances data reau of the Census Topologically Integrated Geographic quality is generally encouraged. C-CAP does not en- Encoding and Referencing (TIGER) files. Imbedding dorse the notion that the use of collateral data in a Chapter 3: Monitoring Uplands and Wetlands 33 remote-sensing project is "hedging." Instead, the objec- CAP regional projects, the input data will be 30 x 30 in tive is to use collateral data innovatively to improve the pixel data recorded by a Landsat TM sensor. The mini- accuracy of the C-CAP database. mum measurement unit, however, combines the ability There are many potential sources of collateral data (e.g. sensor limitations) and effort (e.g. field verifica- including soil maps, NOAA coastlines (T-sheets), tim- tion) required to measure a category with the spatial ber surveys, USGS digital line graphs, and digital eleva- precision and accuracy necessary to accomplish the tion models (for elevation, slope, and aspect). These intended use of the data. Each land-cover category can be incorporated by masking, filtering, probability could potentially have a different minimum measure- weighting, or including in the signature file (Ryerson, ment unit based on the size of individual parcels and 1989; Baker et al., 1991). Depending on the impor- the distinctiveness of the signature. Thus, the mini- tance of each category, analysts may use certain catego- mum measurement unit differs from a traditional mini- ries to overrule others Uensen et al., 1993a). mum mapping unit, which by definition imposes a pre- The NWI is an especially valuable collateral database determined polygon (or pixel) size for all land-cover that may be of value when classifying wetlands. Re- categories (for example, a rule that all parcels of one gional analysts should incorporate NWI data to the hectare or larger will be mapped). This traditional ap- maximum extent possible. NWI data are recognized as proach is acceptable for manual mapping using analog the most authoritative and complete source of wetlands aerial photographs but is difficult to apply to raster land-cover data (Wilen, 1990). However, NWI maps are imagery. Regional analysts will be responsible for defin- not temporally synchronized in each region and are ing minimum measurement units, which will generally not in a digital format for many regions. An approach be larger than a single pixel but no larger than three based on complementary use of NWI and imagery will pixel dimensions on the short axis. be an asset to both C-CAP and NWI. At a minimum, Regardless of the minimum measurement unit, NWI maps, digital data, or both should be used to change analysis will be conducted pixel by pixel. C-CAP define training samples, to check intermediate results, protocol requires that the inherent resolution of the and to aid in the final verification of the wetlands raw data must be retained throughout the classification portion of the G-GAP maps. NWI digital data may be and change-analysis processes. Aggregation and filter- used as a probability filter in the classification process. ing of pixels should occur only in regard to carto- In this approach, C-CAP recommends an "innocent graphic presentation of the completed change detec- until proven guilty" attitude toward the NWI data. In tion database. other words, the NWI category is considered correct for Regardless of the techniques employed, the final da- a given pixel area for each time period, unless spectral tabase should be capable of representing land-cover by signatures or collateral data suggest that the NWI cat- class for the base time, land cover by class for each egory is incorrect or a land-cover change has occurred. earlier or later time, and land-cover change by class for Even if the NWI data were 100% correct at the time of each change period. The final database should contain NWI mapping, overriding by spectral data would be the full change matrix (all "from" and "to" categories) necessary to detect change over time. Ultimately in for each change period. turn, the G-CAP change detection database can assist NWI managers in determining the need for NWI updates. Analog (Hardcopy) Cartographic Products Hardcopy maps of the final database are not specifically Cartographic Portrayal of Classification and required by C-CAP, but they are certain to be useful Change Detection Maps when presenting results. Often it is useful to produce a smaller scale regional map (usually requiring some pixel C-CAP products must meet stringent cartographic stan- aggregation) that gives an impression of the scope of dards. The following sections discuss the minimum the effort and to produce several larger scale maps at measurement unit and its proper use when aggregating full resolution that demonstrate the level of detail and change information. Formats of classification maps and highlight notable findings. All maps should come directly change maps must satisfy C-CAP criteria whenever from the final database complying with CCAP protocols, hardcopy maps are produced. but overlaying or imbedding ancillary data, such as DLG and TIGER data, is encouraged with proper notation. If the statistical summary of changes is present on a The Concept of the Minimum Measurement Unit map, C-CAP recommends that the numbers included in it always be calculated for the area shown on the The minimum measurement unit is a measure of both map. It is not acceptable to associate the summary of the precision and accuracy of input data. For most C- changes for one area (larger or smaller) with a map of 34 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program another. The statistical summaries of the change detec- trix numbers change with aggregation for the same tion matrix must always be calculated from the data- territory. base at full resolution, rather than from the aggregated Technically, the minimum cartographic presentation data of the plot file. It is not advisable to allow the is 1) a map for the base time, 2) a map showing gains by numerical count of class area to float with the level of class, and 3) a map showing losses by class. A full classi- cartographic aggregation. Unless all counts are based fication for the earlier or later (nonbase) time may be on the full resolution database, some classes composed useful, but it is not essential to present the matrix of of small features may disappear at higher levels of possible changes. Examples of some of these products aggregation. Map readers may become confused if ma- are found in Figures 6-13. Chapter 4 Monitoring Submerged Land Using Aerial Photography GCAP Focus on Aerial Photography of Ancillary Technologies for Collecting Submersed Rooted Vascular Plants (SRV) Submersed Habitat Data Photic submerged land can support submersed rooted Some successes have been reported with satellite imag- vascular plants (SRV) (including salt-requiring sea- ery and a number of other technologies in monitoring grasses and oligolialine and freshwater tolerant grasses photic submerged land. Presently, these technologies and forbs), macroalgae, and coral reefs (see "Water supplement, and eventually may replace, aerial photog- and Submerged Land" and Appendix 3). The G-CAP raphy for change detection in SRV. Some of them are Coastal Land-Gover Classification (see Table 1) identi- briefly mentioned here. fies Marine/Estuarine Aquatic Beds, specifically SRV, Satellite imagery has some advantages and disadvan- of primary importance to be inventoried and placed in tages compared with photography. Satellite data gener- the GCAP database (Klemas et a]., 1993). Many of the ally have greater spectral resolution than aerial photog- steps discussed in Chapter 3 to monitor uplands and raphy but lesser spatial resolution. Satellite imagery is wetlands are pertinent to monitor SRV. However, there already in a digital format whereas information derived are significant differences which cannot be ignored (Table from aerial photography must eventually be digitized 5). Important considerations include the following: to be quantitatively analyzed. Landsat and SPOT data have been successfully used to inventory some � mapping SRV is primarily a photogrammetric task, macroalgae such as the giant kelp, Macrocystis pyiifera, rather than a satellite task, requiring an entirely dif- along southern California shorelines Uensen et al., 1980; ferent sensor system (aircraft, camera filter, and film); Augenstein et al., 1991). In clear shallow tropical waters � aerial photography is normally not radiometrically with highly reflective substrate, Landsat imagery may (except for color balance between photographs) or discriminate sandy from coral reef or seagrass areas geometrically corrected; (Luczkovich et al., 1993) or provide an estimate of � time of day, sensor altitude, and flightline placement biomass for unispecific beds of Thalassia testudinum are very flexible, unlike fixed orbit satellite sensor (Armstrong, 1993). In the turbid estuaries of the east- systems; ern United States, Landsat and SPOT imagery can be � numerous environmental conditions must be consid- used to detect some (e.g. large, dense, shallow) but not ered (sea state, water clarity, water depth, low altitude all of the SRV that is visible in the best aerial photogra- atmospheric conditions) to optimize photography; phy. Aerial photography is, in fact, often used as "ground and truth" when interpreting satellite imagery. Because of � aerial photographs are in analog format. the fixed orbital paths of satellites, it is only fortuitous when a satellite image is acquired under optimum con- These differences are so significant that it is instructive ditions to inventory SRV (see "Environmental Consid- to focus on aerial photography of SRV. erations"). For these reasons, aerial photography is the C-CAP imagery of choice for comprehensive mapping and change detection (Ferguson and Wood, 1990; Tho- Ancillary Categories of Submersed Habitat mas and Ferguson, 1990; Orth et al., 1991; Ferguson et al., 1992 and 1993). Photo interpretation supported by Other types of submersed habitat classified by C-CAP surface-level signature verification and species identifi- can be monitored with guidelines similar to those pre- cation is qualitatively and spatially more reliable for sented here for SRV. At a minimum, regional coopera- SRV than are satellite-based methods. tors are requested to map and conduct change analysis Several technologies may provide valuable supple- for SRV. Increasing the number of habitat types to be mental data to aerial photographic detection of habitat included in the study will be based on local or regional change. These include closed circuit television (CCTV) interest and support for the effort. For example, a on an airplane, small boat, or remotely operated ve- comprehensive mapping of SRV, macroalgae, and coral hicle (ROV), side-scan and down-looking sonar, new reefs is underway in the Florida Keys (see "State of satellite sensors, airborne spectral scanners and digital Florida, Department of Environmental Protection"). video scanners, and digitized photography. Such new 35 36 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program Table 5 technology will be incorporated into the C-CAP guide- General steps required to conduct regional C-CAP lines as it is demonstrated to meet qualitative, quantita- change detection projects to extract water and sub- tive, resolution, and geographic positioning standards. merged land information using aerial photography. At present, CCTV is effective for surveillance applica- Each major step is listed in the order to be accomplished. tions but georeferencing and rectification fall short of metric quality photography. Airborne multispectral scan- 1. State the regional change detection problem ners and digital cameras are technologies with applica- a. Define the region tions in the demonstration stage of development. b. Specify frequency of change detection (I to 5 yr) Direct mapping of habitat borders can be performed c. Identify classes of the C-CAP Coastal Land-Cover with differentially correctable GPS instrumentation Classification System when the perimeter of that habitat can be visually ob- served or detected with the aid of instruments in the 2. Consider significant factors when performing change field. Differentially corrected GPS can provide posi- detection tions of surface level data at an accuracy suitable to a. Remote sensing system considerations supplement or assess the accuracy of aerial photographic 1) Spatial resolution and scale 2) Flightline considerations data. With differential correction, single position fixes 3) Spectral resolution and film/filter combination with GPS are accurate to a circular error probable 4) Temporal resolution and diurnal sun angle (CEP)of �5 m 50% of the time. The methodology for 5) The preferred C-CAP aerial photography system using GPS in accuracy assessment and monitoring of b. Environmental considerations SRV is a current research topic funded by C-CAP. 1) Atmospheric conditions 2) Turbidity conditions 3) Vegetation phenological cycle characteristics 4) Tidal stage Aerial Photography of SRV 5) Surface roughness and sun glint conditions Film 3. Interpret aerial photographs to extract water and submerged land information The recommended film for aerial photography of SRV a. Acquire appropriate change detection data is Aerocolor 2445 color-negative film. Second choices 1) in situ surface level verification and basemaps are Aerochrome 2448 color-reversal and Aerographic 2) Aerial photography 2405 black-and-white negative film. A haze filter should a) Base year (Td always be used to minimize the degrading effect of haze b) Subsequent year(s) (T., or T,,) on photographic images. We do not recommend infra- b. Preprocess the multiple-date photography 1) Radiometric correction (color balance) red film for delineating SRV. In our experience in 2) Optically register photography to planimetric North Carolina with tandem cameras, Aerochrome 2443 basemap false-color infrared film was much less effective than c. Select appropriate change detection algorithm color film at recording benthic features in shallow, (usually postclassification comparison) moderately turbid water. True color film gives more d. Image analysis 1) Monoscopic (interpretation of single photos or information than black-and-white or infrared film, is orthophotographs) critical for initial mapping attempts in new or unfamil- 2) Stereoscopic (analog or analytical) iar areas, and may permit identification of species in e. Transfer polygons to planimetric basemap some tropical areas. Color negative film also appears to f. Digitize polygons be better than color reversal or black-and-white film for g. Perform change detection using GIS algorithms 1) Highlight selected classes using change detection identification of habitat under moderately turbid or matrix hazy conditions. Color transparency prints are dimen- 2) Generate change map products sionally stable and are most amenable to illuminating 3) Compute change statistics dark areas of the photograph under magnification. 4. Conduct quality assurance and control Paper prints are not as dimensionally stable as transpar- a. Assess spatial data quality encies (i.e. paper prints are subject to stretching and b. Assess statistical accuracy of shrinking) but they are more resistant than transparen- 1) Individual date classification cies to damage from handling when used for field work. 2) Change detection products 5. Distribute C-CAP results Metric Photography and Photographic Scale a. Digital products b. Analog (hardcopy) products Metric-quality aerial photographs (:53' of tilt off-nadir and including camera calibration data) are essential Chapter 4: Monitoring Submerged Lands 37 and should be acquired with a protocol similar to that 1:24,000 (1 inch = 2,000 ft), a standard 9 X 9 inch aerial employed by NOAA's Photogrammetry Branch (1980) photograph has a coverage of 18,000 x 18,000 ft. Large to produce the highest quality data possible. The need areas (relative to coverage of a single photograph) of for rectification of photography is minimized by pre- open water require parallel flightlines and bridging of cise control of aircraft altitude and orientation relative the large-scale photography to control points with the to the vertical during photography and by interpreta- small-scale photography, construction of towers, etc., tion in stereo. Photography should be obtained at a to supplement horizontal control features or inflight scale appropriate to the areal extent of habitat, local GPS positioning of photographic centers. water conditions, type of habitat being studied, and Overlap of photographs includes endlap of adjacent resolution requirements for the resultant data. Scale is photographs along a flightline and sidelap of photo- a compromise among resolution of signatures, cover- graphs along parallel flightlines. Sixty percent endlap age of habitat, inclusion of land features sufficient for allows stereoscopic interpretation, facilitates interpre- horizontal control, and cost. Photographic scale should tation from the most central region of the photographs, normally range from 1:12,000 to 1:24,000. For exten- and compensates for loss of coverage due to sun glint in sive areas of high and variable turbidity such as Chesa- the photographs. (Sun glint is the image of the sun peake Bay and eastern North Carolina, 1:24,000 or reflected off the surface of the water. See "Sun Angle.") 1:20,000 scale photographs may be adequate when the Sidelap of 30% ensures contiguous coverage of adja- water is clear. For chronically turbid estuarine or brack- cent flightlines and produces a block of aerial photo- ish water areas, 1:12,000 or larger scale photographs graphs that may be subjected to photogrammetric obtained at times of minimal turbidity may be required bundle adjustment if necessary. for acceptable visualization of submerged features. Small-scale photography may be necessary to bridge habitat delineated in larger scale photographs to local Environmental Considerations horizontal control points on adjacent land features that are not included in the larger scale photographs. GPS Knowledge of the study area that is important to a onboard the airplane for positioning photographic cen- successful project includes the plant species compris- ters during exposure may reduce this limitation of larger ing SRV; morphology and phenology of these plants; scale photography. For extensive areas of relatively clear depth range and location of known habitat; locations water, such as the Florida Keys, a scale of 1:48,000 may with water depth potentially suitable for habitat, types be sufficient and cost effective. This is a current C-CAP and locations of benthic features that may confuse photo research topic (see "State of Florida, Department of interpretation of SRV; seasonality of turbidity, weather, Environmental Protection"). and haze; daily patterns in wind speed and direction; and progression of sun angle through the day. Primary and secondary seasonal windows and the day and time Flightlines, Reconnaissance Flights, and Photo- to conduct photography are selected to optimize the graphic Overlap visibility of habitat in the photography. Surface waters in different locations and at different times of the year Flightlines are planned with reference to aeronautical will be more or less sensitive to turbidity from local and nautical charts to include all areas known to have, runoff, plankton blooms, local resuspension of sedi- or which potentially could have, SRV. The efficiency of ment, and surface waves, Seasonal and daily trends for photographic missions can be optimized by minimizing haze, cloud cover, wind direction, wind duration, and the number of flightlines and by contingency planning. wind velocity should be included in planning for pho- Some airspace is restricted for military or other use, for tography. The decisions of when to have the aircraft example, and is indicated on aeronautical charts. Nau- arrive at the study area (within the seasonal window) tical charts provide bathymetric data useful for desig- and when to collect photography are based on NOS nating potential habitat areas when combined with lo- tide tables, local knowledge of factors affecting water cal knowledge of the depth of vegetated bottoms. Re- clarity and depth, observation of recent weather pat- connaissance flights can provide valuable perspective terns (precipitation, wind direction, and wind speed), on SRV distribution if timed to optimize visualization of and water clarity. The final decision to photograph shallow bottoms (see "Environmental Considerations"). includes observations from the air based on the pilot's Ideally, each photograph in a flightline records cul- estimate of haze, cloud cover, and overall visibility. tural and shoreline features required to register the Primary and secondary photographic windows should image to the base map, about 1/3 of the exposure. This be one or two months duration to ensure optimal con- permits correction of photographic scale and orienta- ditions for photography. For single day missions it may tion to the external reference system. At a scale of be possible to have the plane and flightcrew fly to the 38 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program study area on the day of photography. In our experi- floating debris should not be visible from the air or in the ence in North Carolina, staging of the plane and flight photographs. For some areas, ocean swell can be an im- crews to the study area several times for several days was portant consideration and should not exceed 3 ft. required to complete missions involving more than one day of actual photography. SunAngle Sun angle affects the illumination of benthic features, sun glint, and shadows from tall shoreline Phenology-The best time of year to acquire photogra- features in the photographs. A sun angle of 20-25* is phy is during the season of maximum biomass or flow- optimal to record benthic features (Keller, 1978). A ering of dominant species, considering the phenologic sun angle of 15-30' is recommended by C-CAP. This overlap for the entire community. This is June for the interval maximizes the time for photography consider- SRV of the Pacific Northwest and Atlantic Northeast, ing both the illumination of submerged features and April and May for eelgrass in eastern North Carolina, sun glint. Sun angles above 15* illuminate the bottom and September for most of the other species of SRV in sufficiently for photographic purposes. Sun glint also the eastern United States. increases with sun angle but precludes visualization of benthic features where it occurs in the photograph. As Clouds and Haze It is best to have no clouds and sun angle increases, sun glint also increases and moves minimal haze. Thin broken clouds or thin overcast from the edge toward the center of the photograph. above the plane may be acceptable when these are Loss of coverage due to sun glint at sun angles of up to determined by visualization from the air neither to cast about 30' is compensated (to ensure monoscopic cov- shadows nor adversely affect illumination of the study erage, at a minimum) by the recommended endlap of area. Haze reduces illumination and clarity of the im- 60% (see "Flightlines, Reconnaissance Flights, and Pho- age of benthic features being recorded in the photo- tographic Overlap"). Eighty percent endlap will im- graph. Cooperators are referred to the "Aerial Photog- prove coverage when high sun angles cannot be avoided. raphers Clear Day Map," U.S. Department of Com- Photography at sun angles above 30' is not recom- merce, Environmental Data Service. mended. Sun glint is minimized when the sun and land are on the same side of the plane because sun glint Turbidity-Aerial photography should be conducted does not occur on land. Shadows from tall objects on when turbidity is low. Care should be exercised in areas shore such as trees, however, can preclude visualization adjacent to sources of suspended sediment and nutri- of benthic features and may be a factor when the land ents. Data collection should be avoided during seasonal and sun are on the same side of the plane. phytoplankton blooms or immediately following heavy rains or persistent strong winds. Potential days for pho- tography are those during the photographic window Photointerpretation of SRV when high water clarity is expected, based on local experience, recent weather patterns, and surface level Habitat defined by the presence of SRV can be inter- observation. The flightcrew should confirm water clar- preted from metric-quality aerial photographs exposed ity from the air on the day of photography. as recommended in the previous sections. The accurate identification of SRV in aerial photographs requires Tides-Generally, aerial photography should be col- visual evaluation of the fundamental elements of image lected within �2 hours of the lowest tide predicted by interpretation (tone, color, contrast, texture, shadow, etc.). NOS tide tables, although factors affecting water depth It also requires extensive experience at ground level in the and water clarity should be considered simultaneously. study area; the photographic images of habitat and In general, extreme low tide, which may be -0.5 to -1.0 m nonhabitat features vary in ways which cannot readily be or more around the U.S. coast is preferred, if compat- modeled, described, or communicated. Training for a ible with other constraints. The significant "lag" in the habitat change analysis effort includes literature research; tidal stage of some estuaries should be considered for discussions with local ecologists and biologists; site visits data acquisition. on foot, swimming (snorkel or scuba), or small boat; overflights in a small plane; and examination of historical aerial photographs of the area. Training of photo inter- Wind and Surface Waves-No wind and no waves is best preters is active throughout the life of the project. for aerial photography. Low wind (<10 mph) may be SRV are best observed by using stereo pairs of photo- acceptable. The direction, persistence, fetch (the distance graphs and high quality stereoscopic instruments (e.g. that wind can blow unobstructed over water), and recent Wild, AVIOPRET, APT2, stereoscopes) - Polygons are wind events should be taken into account. Breaking waves traced on overlays fixed to each photograph. To be and associated turbidity, white caps, lines of bubbles, and delineated as habitat, recognizable and verified signa- Chapter 4: Monitoring Submerged Lands 39 tures of SRV must be present in the photographs. SRV SRV with unrecognized signatures due to poor pho- (and other benthic features) in a given area will present tographic conditions cannot be mapped as habitat un- a variety of signatures depending upon the species less the area is rephotographed or additional sources of present, bottom sediment, depth, season, haze, clouds, data are incorporated into the database. When photo water clarity and surface disturbances, and sun angle at interpretation is difficult or not possible, the preferred the time of photography. option is to rephotograph the area under better condi- The designation of a given area as SRV is a function tions. Although desirable, this may not be possible. of the minimum detection unit, the minimum map- Even under the best photographic conditions, delinea- ping unit, and the proximity of the area to other SRV. tion of all or part of some habitat polygons may require Assuming a photographic scale of 1:24,000, high qual- additional effort in regard to surface level verification or ity optics, high resolution film, and ideal conditions direct inclusion of surface level data. Polygon borders (e.g. dense clusters of large vigorous shoots growing on derived from surface level data must be so designated in light-colored sediment in shallow, clear, calm water), it the lineage database for "truth in labeling" requirements is usually possible to have a minimum detection unit of (see "Digital Product"). Suitable surface level positioning approximately I in. All detected SRV that appear to be techniques include GPS or more traditional survey posi- in a continuum with adjacent SRV in an area exceeding tioning techniques that can be demonstrated to provide 0.03 ha will be mapped as a single. polygon. The mini- the positional accuracy required by C-CAP. mum mapping unit is the smallest area to be mapped as Within a polygon of SRV, the extent of bottom cover- habitat. At the C-CAP map scale of 1:24,000, the mini- age by shoots of SRV and the pattern of distribution of mum mapping unit is 0.03 ha for SRV (i.e. a diameter the shoots or bed form (e.g. circular, doughnut-shaped, of about 0.8 mm on the map represents a diameter of irregular patches, or continuous cover of SRV) reflects about 20 in or an area of about 0.03 ha on the ground). the interaction of biotic, physical, and anthropogenic Therefore, isolated groups of shoots with a diameter of factors. Coverage and bed form can be estimated from less than 20 in may be detected but not mapped as aerial photographs but is not a requirement of C-CAP. habitat. The presence of SRV signature in the photo- An example of a coverage index is an adaptation of the graph defines habitat if 1) the total area exceeds 0.03 ha; crown density scale originally developed to categorize 2) no unvegetated discontinuities, such as dredged or coverage by trees crowns in forests (Orth et al., 1991). natural channels, partition the distribution into spatial However, coverage indices and bed-form identifications units less than 0.03 ha; and 3) unvegetated areas be- are affected by factors such as water depth and brightness tween plants are not large relative to the minimum of bottom sediments. The degree of contrast between mapping unit. Unfortunately, not all areas of SRV can shoots and exposed sediment and the clarity of the photo- be detected when photographic conditions are less than graphic image determines the minimum detection unit of ideal. Because of the constraint of the minimum map- features within SRV. Comparison of habitats with differ- ping unit and the possibility of suboptimal photogra- ent depths, water clarity, or substrate brightness, there- phy, delineations of SRV will tend to be conservative. fore, is problematic. Analysis of change over time at a The degree of underestimation depends upon the at- given location may be useful but requires consistent pho- mospheric and hydrographic conditions at the time of tographic conditions and field verification. Changes in photography, the experience of the photo interpreter, coverage or bed form over time in a given location may and the nature of the subject area. indicate changing conditions in that habitat polygon or Optimizing conditions for photography will mini- disturbances, such as scarring by boat propellers. mize underestimation of SRV, particularly in areas that Some data including species, biomass, productivity, are intrinsically more difficult to interpret. Where habi- functional status, and health of SRV may not be inter- tat edges are clearly distinct in superior-quality photog- pretable from the aerial photographs. Species identifi- raphy, they may also be detected in inferior-quality cation is not possible from aerial photography in tem- photography (e.g. high biomass of SRV along a clear perate areas such as North Carolina and the Chesa- water channel with a steep bank of light-colored sedi- peake Bay. In some tropical areas, species distributions ment). In other cases where the edges are not clearly and photographic signatures may be sufficiently dis- distinct in superior-quality photography, they are likely tinct to discriminate by species. to remain undetected in inferior photography (e.g. low biomass of SRV growing on a shallow depth gradient of Field Surveys deep, turbid water over dark-colored sediment). The deep-water edge of habitat often will be difficult to Species and Habitat at Randomly Selected Stations delineate. This edge may also be at high risk for loss due to degradation in water quality that limits the illumination of Once selected by stratified random sampling of poten- the bottom with photosynthetically active radiation. tial habitat, stations are observed for SRV species and 40 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program the presence or absence of aquatic beds during the sary, to collect surface level data for inclusion in the same season and preferably within one year of the spatial database. Surface level data intended to aug- photography. Stations are stratified by water depth and ment photo interpreted data require differentially cor- water body. Water depth determines if sampling can be rected GPS positioning to a CEP of �5 m. accomplished by wading, snorkeling, or scuba diving. Clear water with a bottom depth of @!2.5 m or somewhat shallower turbid water may require scuba. Stratification Base Maps and Registration of Habitat permits flexibility in sampling intensity and effort (sam- Polygons pling by scuba requires special training and resources and takes about twice the time per station). Bathymetry and Accurate and up-to-date planimetric base maps of coastal reference coordinates in NOAA nautical charts of the land features are essential for georeferencing (estab- study area facilitate selection and positioning of stations. lishing of geographic location) and scaling polygons of Navigation to stations is with GPS. The spatial density of habitat interpreted from aerial photographs. C-CAP points is adjusted according to the resources and scale of recommends 1) use of the most accurate and up-to- the project (e.g. an average of 1.5 to 2.5 nmi from station date base map available for the study area and 2) use of to station in North Carolina). Great care is taken to in- the most cost-effective technology to apply local hori- clude all locations of potential habitat in the surface level zontal control to interpreted data by registration of the survey. SRV are limited to water depths less than about 2 photographs to base maps. The base map and the regis- m at mean lower low water (MLLW) for Chesapeake Bay. tration technology may vary regionally. A similar depth limit was determined for that habitat in eastern North Carolina. To determine that depth in North Carolina, potential habitat was sampled to water depths of Planimetric Base Maps 10 ft MLLW (Ferguson and Wood, 1990,1994). SRV are not known to occur seaward of the barrier islands in The accuracy of the base map used for local horizontal North Carolina. In sharp contrast, the maximum depth control places a limit on the accuracy of the C-CAP for SRV is 9 m off the northwest coast of Florida. product. The two base maps broadly available are NOAA The presence or absence of aquatic beds and species shoreline and USGS 7.5' topographic maps. NOS pro- of SRV are determined within an area equal to the duces highly accurate shoreline maps based on tide- minimum mapping unit and centered around the nomi- coordinated and fully rectified photography (Swanson, nal station location. If SRV are present, visual observa- 1949; Ellis, 1978; Slamma, 1980; NOAA Photogramme- tions of the number, size, and distribution of groups of try Branch, 1989; Crowell et al., 1991). When available plant shoots are recorded. These data are translated and current, NOAA shoreline and coastal data should into an assessment of the presence or absence of an be used for C-CAP projects (e.g. Ferguson et al., 1991). aquatic bed at the station considering the spatial distri- These data, available in graphic and digital form, are bution of SRV relative to the minimum detection and products of the NOAA Coastal Mapping Program and mapping units. The goal is to assess presence data in a are available from NOS. Shoreline data are produced manner relevant to photo interpretation (see "Photo from tide-coordinated photographic data and ground interpretation of SRV"). Ancillary data recorded are level survey data by the Photogrammetry Branch of water depth, salinity, water clarity, latitude and longi- NOS and meet or exceed national map accuracy stan- tude, and descriptions of benthic sediment 'algae, ani- dards. Horizontal ground control meets or exceeds mals or animal shells, boulders, etc. A GPS position fix third-order class I specifications found in the geodetic is taken to be differentially corrected (postprocessing) control standards (Federal Geodetic Control Commit- to a CEP of �5 m. If the station data are not required to tee, 1984). The Coastal Mapping Project of the Photo- verify photo interpretation (see below), they can be grammetry Branch provides data that depict the delin- used to estimate the accuracy of the habitat data (see eation of the mean high water line, the limit of emer- "Recommended Accuracy Assessment Test"). gent vegetation (apparent shoreline) and/or cultural shoreline, and in some areas, e.g. North Carolina, the approximate MLLW line. NOS shoreline data are a Signature Verification and Supplemental Spatial data source for NOAA nautical charts and USGS topo- Data graphic maps. Coverage of the U.S. coastline is not com- plete, however, and for some areas the data may be dated. Locations selected from the photographs are observed In some locations, USGS 75 topographic maps may be during the same season and within one year of the the only base maps available at a scale of 1:24,000. These photographic mission. The purpose of this survey is to maps delineate the high tide line and cultural features resolve uncertainties in the photographs and, if neces- and may meet CCAP requirements. In many instances, Chapter 4: Monitoring Submerged Lands 41 however, these maps are out of date and temporal changes The three-dimensional stereo model of the aerial pho- in shorelines may cause problems in the application of tographs is leveled and scaled in the analytical plotter local horizontal control to compile the habitat polygons (AP) and the interpreter views a three-dimensional land- (Ferguson et al., 1989; Ferguson and Wood, 1990). This scape during photo interpretation. All polygonal inter- can reduce the positional and scaling accuracy of habitat pretations are automatically stored in digital xy coordi- data which is critical for change analysis (see "Recent nates in their proper planimetric position during photo Photography"). Care should be taken to determine the interpretation (Welch et al., 1992), avoiding any error effective date of coastal features in these maps. Updates of that might arise during information transfer in meth- these maps generally include cultural but not natural ods I and 2 discussed above. The polygon data are changes in shoreline. Coverage of the coastal United States registered and digitized without the errors that are is almost comprehensive, but dated. In some coastal areas, associated with transfer in a zoom transfer scope or by 1:24,000 scale orthophotoquads, have been published as hand digitization. Unfortunately, analytical stereo- an alternative to topographic maps. Orthophotoquads at plotters are expensive and their use requires special a scale of 1:24,000 are unsatisfactory for compilation from training. Some additional expense to locate x, y, and z aerial photography in remote areas. Orthophotoquads do control points may be necessary to successfully level the not have delineated shorelines, which may be needed block of aerial photography. Recent advances in soft- when the preferred cultural features are insufficient to copy photogrammetry allow analytical stereoplotter register the photograph to the map base. functions to be accomplished using UNIX type worksta- tions and image processing software (e.g. ERDAS ORTHO-max). Therefore, this alternative will become Transfer of Polygons to the Map Coordinate more affordable and attractive in the future. Projection System An adaptation of the third approach is being tested by NOAA and the State of Florida. Photo interpretation Polygons of habitat interpreted from aerial photographs is done as in approach 1. Registration and digitization are mapped into a standard map projection coordinate of the interpreted habitat polygons is completed in the system. The UTM projection is recommended. C-CAP AP. Due to the high expense of AP and the specializa- protocol allows the polygons interpreted from aerial tion of AP technicians, this option may be feasible for photography to be transferred onto planimetrically ac- processing data from SRV interpreters who do not have curate basemaps using three approaches: direct access to or training on an AP. 1) Stereoscopically interpret the photographs and optically scale the polygons and photographic image to fit planimetric horizontal control in the basemap with a Digitization of Habitat Polygons zoom transfer scope. This is the least expensive and often the most reliable approach. Habitat delineations Habitat polygons that have been transferred to the drawn at the photographic scale through stereo view- planimetric base map according to procedure I or 2 ing under magnification are transferred using camera above require digitization to be incorporated into the lucida principles from the photographic overlay di- GCAP spatial data base. Normally, digitization is ac- rectly onto the planimetric basemap. complished using a digitizing tablet. Polygons are digi- 2) Process the aerial photographs into planimetrically tized with a digitizing table in point mode. The overlays accurate orthophotographs, and interpret and directly are labeled according to the base map. Compilations trace habitat polygons onto the planimetric base map. are checked for clear delineation and cartographic ac- interpretation of the orthophotographs is performed ceptability of line work, existence of and consistency in using monoscopic airphoto interpretation techniques. feature attributes, and adequacy of horizontal control The orthophotographs must be at the same scale as the points. Compilations are checked along neat lines to base map or the images must be enlarged or reduced to confirm edgeline match and label match for polygons the map scale. This approach applies orthophotographic extending over adjoining maps. Any inconsistencies rectification (Thrower and Jensen, 1976), which cor- are brought to the attention of the map author. rects relief displacement in the original photographs Compilations are affixed to a digitizing table for and ensures planimetric mapping results in the data- georeferencing and data entry. The accuracy of the base. Some loss of detail may occur since the ortho- reference points, the four corners of the neat line, and photography is a generation away from the original no less than four internal tick marks on the overlays are aerial photography. The process is expensive but accu- checked to ensure that control points are within �0.02 racy is improved in areas with substantial vertical relief. inches. This translates to �40 ft or �12.2 in from its 3) Delineate and simultaneously rectify and digitize stated location. If tolerance is exceeded on any one habitat polygons by using an analytical stereo plotter. point, new control points are selected, digitized, and 42 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program reevaluated until all points test within tolerance. Infor- different years are independently interpreted, verified, mation regarding the georeferencing error for each and compiled to the base map. In this case indepen- control point is recorded on a documentation form. In dence does not mean different photo interpreters, com- addition the technician records other information about pilers, and field personnel but rather an avoidance of the overlay manuscript such as scale, size, media type, side-by-side comparison of the data until after classifica- source map information, and author. tion is complete. Postclassification change detection Polygons are digitized with the cartographic style and can be accomplished graphically, or polygons may be accuracy that is represented on the source manuscript. A digitized and compared by using a geographic informa- technician performs digitizing and data processing to map tion system to detect spatial displacement and to quan- completion, including matching edgelines, preparing ini- tify change. Although simple in concept, the statistics tial check plots, and reviewing, editing, and preparing of change analysis are not well understood. Development final check plots. All linework and labeling are reviewed of consensus for statistical evaluation of qualitative or using check plots produced at the source map scale. Each spatial change is a subject of ongoing C-CAP research. arc is checked for acceptance on a light table with the final As an expedient to postclassification change detec- check plot overlaid on the source map. Digitized linework tion, photographs from different years are compared should conceal original linework with exceptions for dif- directly or with mapped polygons. By using such com- ference in line thickness, differences in media, and subtle parisons, areas where change may have occurred can be differences of horizontal control on the source map and identified rapidly but subjectively. Determining what con- in digital files. Unacceptable data is flagged, edited, and stitutes significant change and how to objectively quantify reviewed prior to acceptance into the digital database. A the degree of change remain to be accomplished. data layer specification form is completed for formal docu- mentation at the conclusion of all digitizing. Scan digitizing may be an acceptable alternative to Historical Photography hand digitizing and could be applied at one of two stages: 1) when polygons are positioned on overlays of The earliest metric-quality aerial photographs were ac- base maps or 2) when polygons are interpreted from quired in about 1939. Prior to 1960, virtually all aerial individual photographs. Large format scanners would photographs were black and white. Incomplete cover- be required to scan an entire map, approximately age, lack of coordination with tide, lack of camera 19 x 23 inch, in one pass. A standard desktop scanner, calibration data, inappropriate scale, sun angle, and 8.5 X 11 inch, could scan the overlay from a single 9 x 9 inappropriate time of year, or poor quality for visualiza- inch photograph. In the latter case, geopositioning tion of benthic features often make these photographs might be accomplished digitally without the use of the unacceptable for a C-CAP change analysis. Interpreta- zoom transfer scope. In either case, the digital product tion of historical photographs is likely to proceed with would have to meet the same positional tolerances de- limited or no concurrent surface level information for scribed above for data entered with the digitizing tablet. signature verification and should be attempted only by interpreters with extensive experience in the study area. Unless historical photography meets the C-CAP require- Change Detection WiLdi Aerial Photographic ments listed in "Aerial Photography of SRW and is sup- Data ported by surface level data as discussed in "Field Surveys," the historical presence or absence of SRV at a given loca- The C-CAP objective of site-specific change detection tion may remain an open question. Some but not all SRV places greater emphasis on accuracy and precision of can be identified in less than optimal photography and be spatial data than required in one-time inventories or confirmed in the literature or in the memory of local regional summaries of change. Methodology for moni- residents. A visible signature for "bare bottom" or another toring site-specific change on a statewide or regional nonhabitat signature is required to interpret absence of scale is a recent development (Ferguson and Wood, habitat at the time of photography. As a result, documen- 1990; Orth ct al., 1991; Ferguson et al., 1993). Quantita- tation of loss may be more likely than documentation of tive historical data, with possible exceptions in Chesa- gain of SRV with historical photography. peake Bay or spatially limited study sites, does not exist. Historical photographs may contain limited but valu- able information on presence of submersed habitat other than SRV. Canopies of the giant kelp, Macrocystis Recent Photography pyfifera, for example, are readily discernible in color infrared (IR) photography because they have very high C-CAP recommends post-classification change detec- IR reflectance against a background of water that has tion for SRV. Photographs taken in the same season of no reflectance. The case of photo interpretation of Chapter 4: Monitoring Submerged Lands 43 some macroalgae allows historical photography to be names of USGS 7.5' quadrangles that locate the area of used to identify this and perhaps other types of habitat. interest. The ESIC office conducts a microfiche or com- The most complete and general. (but not compre- puter-based search. Information produced includes lati- hensive) source for historical photography is the Earth tude and longitude, emulsion, scale, month, year, source Resources Observation Systems (EROS) Data Center in of the photography, cloud cover, camera, and frame Sioux Falls, South Dakota EROS records can be searched numbers. Sources of this photography are USGS, Na- through the Earth Sciences Information Center (ESIC) of tional Aeronautics and Space Administration (NASA), the USGS. The searcher must supply coordinates or the USFWS, Agricultural Stabilization and Conservation Ser- I.b + Aw @THROLINA lk@ +34 YAREA ig ? I. 34'474 N 41 0 ATLANTIC 0 C E A N G 41 + 34-43' BROWNS ISLAND 1166 0 MARCH APRIL 'V sAv HABrrAT SPA-nAL SCALE 100ACRE@ 100HECTARES 34* 37 + 3@ 0 1 2 3 4 5 MILES 0 1 2 3 4 5 KILOMETERS Figure 17 Seagrass habitat in Back Sound and southern Core Sound in 1985, gray, and in 1988, green. Pure gray indicates habitat present in 1985 but not in 1988. Light green indicates habitat in 1988 but not in 1985. The overlap of gray and green indicates the presence of habitat in both 1985 and 1988. (A) Head of the Hole, an area where seagrass habitat decreased due to mechanical harvesting of clams. (B) Spoil deposition island from which uncontained spoil was released into the water and buried seagrass habitat (Ferguson et al., 1992, 1993). 44 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program vice (ASCS), Soil Conservation Service (SCS), U.S. Forest ter surface. The aerial photographs were interpreted ste- Service (USES), and some private companies. All histori- reoscopically and the polygons were transferred to plani- cal photography identified in an ESIC search is reproduc- metric NOS shoreline maps with a zoom transfer scope. A ible. Substantial collections of historical photography may graphical postclassification overlay approach was used to also be found in other Federal or State agencies, universi- visually identify changes between years (Fig. 17). A gray ties, or private companies. These collections of photogra- tone in the chart indicates habitat present in 1985 but not phy may be available for reproduction, distribution, loan, in 1988. Light green indicates habitat in 1988 but not in or examination. Federal sources include U.S. Army Corps 1985. The overlay of gray and green indicates the pres- of Engineers (COE), NOAA, National Archives for pre- ence of seagrass habitat in both 1985 and 1988. 1956 photos, and the Smithsonian Institution. Summary statistics, obtained via automated geomet- ric analysis of digitized video images of individual poly- gons (pixel size <0.03 ha) revealed that seagrass habitat Change Detection of Seagrass Habitat in North is a major resource in the study area, comprising about Carolina 35% of the subtidal land. Total area of habitat changed less than 6%, from 7,030 ha in 1985 to 6,637 ha in 1988. Ferguson et al. (1993) followed the C-CAP guidelines Polygons along the mainland and Harkers Island tended in Back Sound and southern Core Sound in North to be linear and close to shore. Large broad areas of Carolina. That study demonstrated the feasibility of seagrass habitat were present in the subtidal shallows monitoring spatial change in SRV using C-CAP guide- east of Browns Island, north of Shackleford Banks, and lines for large-scale metric aerial photography, photo west of Core Banks. The total number of habitat poly- interpretation, geographic positioning, and postclass- gons was similar in the two years, 151 in 1985 and 149 in ification change detection techniques. Aerochrome MS 1988. Reliability of detected change was conducted by 2448 color-reversal film was exposed in March 1985 at reinspection of the photography and is summarized in 1:20,000 and 1: 12,000 scales. Aerocolor 2445 color-nega- Ferguson et al. (1993). Some areas of detected change tive film was exposed at 1:24,000 scale in April 1988. All were confirmed by surface-level observations and two of aerial photography was obtained by the NOS Photo- these were associated with known anthropogenic dis- grammetry Branch. The photography was coordinated turbances. Some areas of detected change were con- with low tide and sun angle and was collected with firmed but could not be associated with potential causes. minimal haze, no clouds below the aircraft, and no Still others could not be confirmed, which may have visible shadows from high clouds. Water was essentially been the result of variable quality in the photography. free of white caps and clear enough for identification In a continuation of this study, the study area was re- of vegetated and shallow unvegetated bottoms. Epi- photographed in 1992, selected polygons were mapped sodic wind, haze, local turbidity, and airborne pollen with GPS at surface level during the 1992 photographic often precluded photography for one or more days. window, and surface-level verification of signature was The sun angle during photography ranged from 15 to completed in 1993. Data for all three years, 1985, 1988, 30'. This sun angle localized sun glare to one edge of the and 1992, will be digitized and change and positional photography while presenting illumination below the wa- accuracy will be assessed in a GIS. Chapter 5 Spatial Data Quality Assurance and Control Quality assurance and control (QA/QC) from data ac- measures of accuracy, field-based reference data are quisition through final database compilation are the exclusively preferred over other data sources, including responsibility of each regional project team. Accep- aerial photographs. tance of the final database into the C-CA-P archive and dissemination system are contingent upon the demon- stration that the project has complied with the manda- Lineage tory requirements stated in this document. C-CAP standards of data quality are based on authori- The sources, scales, or resolutions, and dates of materi- tative references (Goodchild and Kemp, 1990; Chris- als involved in the preparation of all regional C-CAP man, 1991; Congalton, 1991; Lunetta et al., 1991; NIST, databases must be documented (Lunetta et al., 1991), 1992). These documents recommend that producers of including data document e satellite images or aerial photographs used in the � Lineage-A record of the type of data sources and analysis, the operations involved in the creation of a database. 0 aerial photographs (including oblique photographs) � Positional accuracy and precision-The closeness of used as an aid in training or field verification if the locational information (in xy coordinates) to the photographs directly influenced the identification of true position. land-cover types for significant portions of a given � Attribute accuracy and precision-The closeness of area, attribute values to their true values. 0 collateral information such as NWI data or soils maps � Logical consistency-The adherence of internal data if the information directly influenced the identifica- structures to established conventions or stated rules. tion for significant portions of a given area, � Completeness-The degree to which the data ex- 0 planimetric basemaps, haust the universe of possible items. 0 state and county land-cover inventories or other sur- face level data, and C-CAP has added to this list 0 sources and techniques of georeferencing, especially for submerged land and other land where identifi- *Temporal accuracy and precision-The time over able features are sparse. which source materials were acquired and observa- tions were made. Positional Accuracy and Precision Users are responsible for determining Positional accuracy is concerned with the accuracy of Fitness for use-The degree to which the data quality the geometric placement of points, lines, and polygon characteristics of each database and its components boundaries. In land-cover databases, polygons are de- collectively suit an intended application. rived either from raster spectral data representing dis- crete pixels or from closed polygons delineating the The GCAP protocol also distinguishes between edges of spectral signatures in photographs. In the first case, the placement of polygon boundaries depends on � Accuracy-The closeness of results, computations, or a) selection of spectral signatures for class boundaries estimates to true values (or values accepted to be and b) registration of pixel locations. The second case true), and generally applies to C-CAP SRV projects in which the � Precision-The number of decimal places or signifi- primary intent is to delineate limits between presence cant digits in a measurement. and absence of habitat classes. In this application, poly- gons of class I tend to occur as discrete objects in a The accuracy of the resulting land-cover database for large polygon of class 0 that has specified boundaries each time period and for change between time periods landward and unspecified boundaries seaward. In addi- is a crucial measure of the success of C-CAP. Several tion, one or more polygons of class 0 may be included different types of accuracy are involved, and some of within a polygon of class 1. The placement of polygon them are difficult to measure. For rigorous statistical boundaries depends on a) limits of signatures attrib- 45 46 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program uted to the habitat class and b) registration of horizon- tween aerial photography for the base map and for the tal control points present in the base map and visible in submersed habitat. A positional accuracy that meets the photography. In both cases, signature selection di- national map accuracy standards is assumed for sub- rectly affects attribute distribution which, in turn, af- mersed habitat data. At the compilation scale of 1:24,000 fects the size and shape of polygons. This effect is most this amounts to �13.3 m on the ground, close to the �15 common at polygon edges but may occur throughout meter precision of Landsat TM data for uplands and the polygon, for example, as internal voids or as cir- wetlands. cumscribed polygons of different classes. The selection of spectral signatures for class boundaries is similar to the task of generalization that cartographers have tradi- Generalization Versus Error tionally faced in deciding where to draw boundaries between land-cover features. It is a tribute to the power of modern information For most remote-sensing applications, positional ac- technologies that what we used to call generalization, curacy on the order of �1-2 pixels has not been a major we now call error. With analog maps it has always been concern. Neither has positional accuracy for photo- necessary to use human judgment in deciding, for ex- graphic delineations of submersed habitat been a ma- ample, precisely where a forest becomes a field. In jor concern or a subject of independent verification. reality most forests have some grass, and most fields For a single time period, positional errors may not have some trees or shrubs. In natural circumstances the greatly affect the aggregate area of each land-cover boundary is not a precise line but rather a "fuzzy" zone type. Positional errors may be difficult to detect even of highly variable width in which the predominant land when a specific polygon in the field is visited. For G cover grades from one class to another. Scale and reso- CAP, however, positional accuracy is a crucial concern lution are crucial determinants of such boundaries. In (Ferguson et al., 1992 and 1993). The change database an analog map, scale limits the feasibility of drawing the amounts to a comparison that will conspicuously record densities and convolutions of lines that would be neces- positional errors of one or more pixel dimensions in sary to represent each patch of forest or each individual the satellite imagery and errors in excess of about 10 m tree. Conceptually there will always be unrepresented in the photographic images used to delineate submersed boundaries because, in the modern sense of fractals, a habitat. This compounds the problem of recognizing nearly infinite number of convolutions are possible. real changes, which also tend to concentrate at polygon Digital systems are capable of representing a much edges and class boundaries. larger portion of all possible boundaries, but there are The registration of pixel locations is a purely geomet- practical limitations affecting digital systems as well. In ric problem which has been greatly improved with re- current technology the most often encountered limita- cent advances in sensors, GPS, and image processing tion is the established resolution of satellite sensors. systems. Many vendors claim a positional accuracy of @Vhile the terms "error" and "accuracy" are frequently +0.5 pixel root mean square error (RMSE) for commer- used in regard to generalized boundaries, conceptually cial image processing systems and a CEP of 3-5 rn for the "accurate" boundary can only be determined on GPS. Selective availability (SA, the intentional distor- the basis of a highly specific set of criteria that goes far tion of GPS signals for military security purposes) re- beyond what can actually be implemented for large duces CPS precision to a CEP of 40 m when SA is in areas. Land-cover phenomena are prime examples of operation. Differential readings by multiple receivers fuzzy sets. This fuzzy characteristic is explicitly recog- can improve the quality of positional data, even when nized in the procedures of image processing (for ex- SA is active, to a CEP of <5 m. C-CAP regional analysts ample, the use of maximum-likelihood statistics), but should verify vendor claims to their own satisfaction the remote-sensing community traditionally has pre- based on sources of higher precision. Unless stated oth- sumed that a "right" answer or "ground truth" can be erwise, a geometric registration of �0.5 pixel RMSE will be determined if the analyst can get close enough to see assumed for all GGAP regional databases (�15 m if Landsat the polygon and its boundary in the field or on the TM data are used). photograph, Yet different investigators "see" different For submerged land, the registration of polygon edges land covers, a problem that is especially troublesome is a function of the metric quality of photographs, meth- when the area is large enough to require multiple teams. ods used to transfer the information to a planimetric In reality, land-cover phenomena are fuzzy sets whether map projection, and quality of the digitization per- viewed directly in the field or through remote sensors. formed. Positional accuracy is therefore subject to the Fuzziness persists because each class is defined, not by a accuracy of the base map including deviations not only discrete boundary, but by factors that grade from one between the source photography and the base map but class to another-spatially, temporally, categorically, and also actual changes in the study area in the time be- observationally. Also, classification and accuracy assess- Chapter 5: Spatial Data Quality 47 ment procedures are not always implemented in a timely (Congalton, 1991). Generally, these procedures serve manner but often months or years after the image is well for current time periods and for relatively small collected or analyzed. study areas. Past time periods, however, cannot be field Generalization also occurs within delineated poly- verified. Conventional procedures also are difficult to gons whether derived from satellite or photographic apply to large areas. Accuracy assessment of large change images. In both cases a finite limit for signature detec- databases is currently infeasible due to the combina- tion and mapping exists. Minimum detection units are tion of past time period, large area, and the excessive one pixel for spectral scanner data and about one meter number of "from" and "to" classes. for high altitude photographic images. Elimination of "salt and pepper" and preservation of reasonable accu- racy for perimeter or areal estimates requires a mini- Logical Consistency mum mapping unit of 4 pixels or about 0.4 ha. At a compilation scale of 1:24,000 the smallest polygon that Tests for logical consistency should indicate that all row can be traced from a photograph is about 20 m in and column positions in the selected latitude/longi- diameter or an effective area on the ground of about tude window contain data. Conversion, integration, and 0.03 ha. Realistically, the goal for improving generaliza- registration with vector files should indicate that all tion should be to strive for consistency more than positions are consistent with earth coordinates. Attribute "accuracy." files must be logically consistent. For example, when Reference data for accuracy assessment must have a examining the change matrix for logical consistency, resolution and reliability that meet or exceed those of very few pixels should change from the urban category the C-CAP remotely sensed data. The reliability, incltid- to any other category or from water to any category ing attribute and positional accuracy, must be demon- other than bare ground or marsh. The range of appro- strated prior to its qualification as reference data for C- priate tests is left to the judgment and experience of CAP. Reference data, including surface level observa- regional analysts. All attribute classes should be mutu- tions, must be evaluated in accordance with C-CAP's ally exclusive. The criteria cannot be met if land-use minimum detection unit and minimum mapping unit classes are included along with land-cover classes. for the remote data and with the classification system used to categorize the habitat. The presence of a char- acteristic species or natural or cultural feature may or Completeness may not, in itself, establish an area as a particular type of habitat. A number of questions need to be answered The classification scheme should be comprehensive, to conclude the appropriate category of land cover to containing all anticipated land covers. The C-CAP assign based on the reference data: Does a characteris- Coastal Land-Cover Classification System is intended to tic species or feature meet the minimum detection unit provide complete coverage, but regional analysts may of the remote sensor? What other characteristic species find special land covers that are not included. It is the or features also are present within the minimum map- responsibility of regional project personnel to ensure ping unit? and What conclusion can be drawn from the that all categories are included and that all pixels are reference data as to the C-CAP category for a given assigned a category. Regional analysts may use their location based upon data generalized to the minimum discretion in deciding at what classification system level detection and minimum mapping units? (0 to 3) they wish to classify. The level need not be the same for all branches of the classification scheme. Attribute Accuracy and Precision Temporal Accuracy and Precision Attribute accuracy is a measure of the probability that the land-cover type for any given polygon is properly Regional analysts should document the time of data identified according to the land-cover scheme. For ex- collection for the primary input data to at least the ample, the identification of a substantial polygon of precision of year, day, and hour. "High-Intensity Developed" land as "Deciduous Woody Wetland" is a clear instance of categorical error. If 15% of all sample polygons for this class are misclassified as Fitness for Use "Deciduous Woody Wetland" and other categories, the categorical accuracy for the "High-Intensity Developed" C-CAP workshops have involved many discussions with class is 85%. The remote-sensing literature is replete potential users and have devoted a great deal of effort with procedures for measuring attribute accuracy to field verification and other types of verification. 48 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program C-CAP is confident that the databases resulting from cedures, and manage the data production process to compliance with this document will be of sufficient meet DQO's. C-CAP has conducted three workshops quality to support most policy and management activi- on accuracy assessment and sponsored two protocol ties as well as some regulatory, enforcement, and re- development projects in the hope of devising new pro- search activities. The spatial precision and attribute cedures that will work for accuracy assessment of large accuracy are not sufficient for enforcement of indi- land-cover change databases. vidual small permits, but they may be useful in evaluat- Nevertheless, the following material identifies sound ing cumulative impacts in the vicinity of a permit site or procedures that may be used to obtain unbiased field for evaluating individual sites larger than the minimum information which, in turn, may be statistically evalu- mapping unit. In the southeast region of the United ated to perform an accuracy assessment for a single States, a vast majority of the total area of coastal salt time period. This is a blind field test in which the field marsh or seagrass habitat that is potentially subject to mapping personnel will not see the C-CAP Land-Cover direct loss, according to permits submitted, would be and Land-Cover-Change Maps until all mapping has detectable in the C-CAP data (Rivera et al., 1992). Data- been completed. bases will also be of value in many applications, such as Since the field mapping personnel may be unfamil- land-use planning, unrelated to the C-CAP mission. iar with C-CAP, it is advised that they be required to Ultimately, however, only the user can make the deci- submit a memorandum stating the design of the field sion regarding fitness for use. mapping implementation. Early in the effort, regional analysts should review the design, in collaboration with NOAA, and reach agreement with the field mapping Recommended Accuracy Assessment Test personnel regarding final implementation. Field per- sonnel should be provided copies of the land-cover The recommended accuracy assessment for C-CAP re- classification scheme and should be trained in its use. gional databases is a test based on comparison with The field personnel will be responsible for ensuring independent field samples. Independence should be the positional accuracy and precision of each sample guaranteed through the use of personnel who are not site and each land-cover class boundary within each familiar with and do not have access to the results of the site. Field personnel will be responsible for determin- land-cover classification (Congalton, 1991). ing physical accessibility and obtaining permission for legal access to the sample sites. An early determination will be made regarding who Sample Selection and Field Mapping is responsible for acquiring the best available aerial photographs, topographic maps, and other collateral Regional analysts are responsible for selecting unbi- data to ensure an accurate mapping of each sample site ased, statistically meaningful area samples for field veri- for each time period. These materials will assist in map- fication in the accuracy assessment process. ping land cover and land-cover change for each site at 1:24,000 scale. Positional accuracy shall comply with national map accuracy standards. The determination Accuracy Assessment for Individual Date of class type will be based primarily on field observa- Classification of Upland and Wetland Habitat tion. The determination of class areas and boundaries Data will be based primarily on aerial photographs. The final results for each sample polygon will be provided in a Accepted procedures in the remote-sensing, carto- digital form. graphic, and geographic literature assess 1) the posi- After completing field mapping, regional analysts tional accuracy of identifiable, stable features and 2) will compare the generated map for each sample site the categorical accuracy at the interior of class poly- with the C-CAP map for the same site. All discrepancies gons. Unfortunately, the methods often neglect the will be referred back to the field mapping personnel for fuzzy nature of land cover-categorically (e.g. the class a final check. The regional analyst may request a special boundary between grass and marsh), spatially (e.g. the examination and may accompany the field mapping polygon boundary between water and marsh), tempo- personnel for a final reconciliation of any discrepan- rally, and observationally. Given these limitations, it is cies for which field error is suspected. not feasible at this time to provide a quantitative esti- The regional analysts will compile the results of all mate of accuracy with every C-CAP regional database. A sample polygon comparisons and conduct a statistical reasonable alternative is to establish data quality objec- analysis. The results of this analysis will be provided to tives (DQO) designed to serve expected uses, establish the field mapping personnel for review and comment. and consistently implement a set of protocols and pro- At this point the field mapping personnel may also see Chapter 5: Spatial Data Quality 49 the C-CAP land-cover and land-cover-change maps for Accuracy Assessment for Land-Cover Change the sample quadrangles. Data The field mapping personnel will provide a brief documentation of the field mapping task for inclusion The methodological difficulties of accuracy assessment in the final accuracy assessment report to be prepared for the final change map are significantly greater than by the regional analysts. The field mapping personnel those for a single, current database. Remote-sensing, must be given an opportunity to comment on the re- cartographic, and geographic literature provide no guid- sults of the final statistical analysis if they so choose. ance on techniques for assessing the accuracy of a change detection map (Lunetta et al., 199 1; Jensen and Narumalani, 1992). Even for a single-time database, Accuracy Assessment for Individual Date existing procedures are ineffective for past land cover Classification of Water and Submerged Land since the recommended "source of higher accuracy" Data cannot include actual field verification. Change detec- tion databases compound the difficulty because they Accuracy assessment for submersed habitat is similar to always include a past time period and a large number of that for emergent and upland habitat but it should be "from" and "to" categories (potentially the square of noted that data for submersed habitat is intrinsically the number of categories for each time). This large vector, not raster. Positional accuracy of polygon bor- change matrix can make accuracy assessment more ders and attribute accuracy of a point location can both expensive than the original classification and change be assessed. Habitat polygons or areas of potential habi- detection effort. Furthermore, if the distribution of tat should be stratified by class and region (water body) error is thoroughly depicted by class and position, the and randomly selected. Additional sample locations accuracy database may be as large as the thematic data- from potential habitat sites (i.e. sites of suitable depth base itself. Even worse, both the distribution of error but apparently devoid of habitat) should be randomly and the distribution of actual change tend to concen- selected. Verification locations should be identified by trate on the same circumstances (for example, polygon latitude and longitude coordinates and visited in the edges and transitional classes, such as marsh and field with GPS navigation. The nature of the habitat, if palustrine forest). C-CAP is sponsoring workshops to present, should be documented by inspection or sam- develop improved methods for assessing the accuracy pling if necessary and the position of the sample or of change databases and maps. observation recorded to a CEP of <5 in. The entire perimeter of a small polygon or a section, e.g. 0.5 kin of the perimeter of a large polygon, should be positioned Comparison and Statistical Analysis by differentially corrected GPS at a point spacing of 3 to 20 in depending upon the degree of curvature in the The C-CAP land-cover database and the field-mapped perimeter. Differential GPS provides CEP of <5 in for verification database should be compared and mea- single position fixes. C-CAP projects found that single sured to determine differences in attributes for the point differential GPS position fixes did not exceed 10 base time period and for change that can be recog- in (Ferguson, R. L.,J. A. Scope, and L. L. Wood, unpub- nized in the field. The measures obtained from this lished data). Habitats at the minimum mapping limit, comparison are numerical differences relating to the i.e. with diameters on the order of 20 in, therefore, sample sites only. It will then be necessary to employ should be located and delineated with multiple posi- statistical algorithms to determine what the differences tion estimates. GPS manufacturers recommend collec- reveal about the accuracy of the entire regional data- tion of multiple position fixes for a time period of base. These algorithms should be designed to estimate about 4 min to achieve CEP on the order of I or 2 in the attribute accur-acy and positional accuracy of the change with differentially collected GPS. Multiple position fixes database. The necessary algorithms are not currently avail- obtained at strategic points around the perimeter of able in the remote-sensing, cartographic, and geographic the smallest mapped polygons would be required to literature (Congalton et al., 199 I;Jensen and Narumalani, ensure mapping the polygon rather than generating a 1992). GCAP ftinded two protocol development projects scattered pattern of points. in an attempt to remedy this deficiency. 50 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program Chapter 6 Product Availability Digital Product acknowledge NOAA as the source of the product when- ever data are redistributed and must provide an ac- Description and Availability counting to NOAA stating who received copies of the database. If the redistributed data are modified, an Regional databases generated by C-CAP participants accompanying disclaimer must acknowledge NOAA as will be provided to the C-CAP project director in accor- the source of the original data, must state the nature of dance with procedures specified in research funding the modifications, and must relieve NOAA of responsi- proposals (RFP), statements of work (SOW), funding bility for the modified data. documents, memoranda of understanding (MOU), or other applicable documents under which each regional Liability Disclaimer-The user of C-CAP data will hold project is authorized and conducted. The purpose of the U.S. Government and its agencies, officers, and this transfer is to place each regional database into a employees harmless against any and all liability, includ- central archive from which all data will be made avail- ing costs and expenses, which may result from or arise able to the public. It will be the responsibility of the out of any use of the data. regional participants to document and certify that the data have been prepared in accordance with C-CAP protocols. C-CAP may conduct additional data quality Digital Product Format and Contents and accuracy assessment tests before final submission to the archive. The data should adhere to the Spatial The goal is to exchange the digital products in the Data Transfer Standard (SDTS) proposed by the Fed- Federal Spatial Data Transfer Standard (SDTS) format eral Geographic Data Committee and adopted as a for raster data. Until the SDTS raster standard is avail- Federal Information Processing Standard (FIPS) (NIST, able, the initial data products may not adhere to the 1992). Commercial implementations are not currently final standard. available but will be marketed by software vendors in the near future. At a minimum, the standard should be Product Identifiers and Characteristics-Each data considered a near-term goal with one or more de facto transmittal from NODC to the user will be accompa- standards-such as DLG, ARC, DXF (geometry only), nied by documentation provided by the data producers and ERDAS-accepted in the interim. Lineage, quality, stating the following: and format information should be transmitted with the data disseminated to users. 0Geographical coverage in UTM coordinates The digital product for each region will be a change aUTM zone number matrix of land cover by class for coastal submersed 0Computer and operating systems used to create the habitats, emergent coastal wetlands, and adjacent up- file lands. T'he only regional database currently completed 0Precision of the computer system (e.g. 16-bit, 32-bit) and available to the public is the Chesapeake Bay Land- eSoftware used to create the file (e.g. ERDAS Imagine Cover Change Database for 1984 and 1988-89. 8.2, ARC-Info 7.0) Digital products are available from 0Type of file (ASCII, binary, ERDAS.IMG, ARC-Info coverage) National Oceanographic Data Center (NODC) 0Description and format of header file 1825 Connecticut Avenue, NW *Data record format Washington, DC 20235 0Number of classes (202) 606-4549 0Class names 0Number of pixels by class and by file, including the When more C-CAP regions have been completed, an null class (i.e. no data in pixel) on-line electronic catalog will be created for users to browse. The header file for each database will repeat the quan- titative portion of this information. Digital Product Redistribution Restrictions-The prod- uct file will contain statements defining the responsibil- Product Data Quality-The documentation will describe ity of the user in regard to C-CAP data. The user must the lineage, date and source of data (i.e. instrument, 51 52 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program and platform), resolution, positional accuracy and pre- Digital Product Ancillary Documentation cision, attribute accuracy and precision, logical consis- tency, completeness, and temporal accuracy and preci- General Protocol-A copy of the "C-CAP Guidance for sion of the data being transferred. Regional Implementation" for the version used to pro- duce the product will be available from the NODC in Guidance Version-The product file will contain a field digital form for the cost of reproduction. indicating the "C-CAP Guidance for Regional Imple- mentation" version used to produce the image product. Specific Digital Products Documentation-Regional analysts may provide ancillary documentation for dis- Transfer Verification Parametem--Each C-CAP prod- semination by NODC if both the documentation and uct will contain unique verification parameters for the the corresponding database are provided to NODC in a image raster data and a confirmation algorithm that standard digital format. can be applied to the image values. The algorithm tests whether the database received by the user is equivalent to the original. If an image has been damaged or modi- fied, application of the algorithm will produce results Hardcopy Products different from the master values in the original data file maintained at the NODC. The occurrence of each class Upland and Wetland Habitats value (including the no data class) can be tabulated and compared with the original summary statistics. Hardcopy maps of uplands and wetlands for selected areas will be produced for informational purposes, pri- Derived Data and Quality-The data product may in- marily to illustrate database content. At present there clude derived data, such as tabular summaries of land are no plans to publish hardcopy maps for general sale cover and accuracy assessments for specified areas (e.g. and distribution to the public. Requests for informa- counties, watersheds, wildlife management areas). Data tional maps will be considered on a case-by-case basis. are defined as "derived data" if they cannot be used to Individuals and organizations should make their re- reconstitute the C-CAP data at the pixel level. quests in writing to Digital Data Values--The data values in raster format are Dr. Ford A Cross, Director numerical values representing the land-cover categories Beaufort Laboratory described in this document (see "The CCAP Coastal Land- NOAA/National Marine Fisheries Service Cover Classification System"). A lookup table or other 10 1 Pivers Island Road accompanying statement will define the relationship be- Beaufort, NC 28516-9722 tween the stored values and the land-cover categories. Organizations wishing to serve as value-added vendors of hardcopy products derived from C-CAP data should Digital Product Medium write to this address. Digital data products are available on 9-track magnetic tapes and CD-ROM. As the completed coverage ex- Submersed Habitats pands, these data may be available on other magnetic and optical media. Hardcopy maps are routinely produced as part of the submersed habitat change analysis because the tech- niques are currently based on aerial photographic in- Digital Product Cost terpretation in analog form. A limited number of publi- cation-quality maps are reproduced at the completion The organization conducting each regional project will of each regional task. Individuals and organizations receive one copy of the final database as distributed by may request copies on a "first come, first served" basis NODC at no cost. All other users will be charged the by writing to Dr. Ford A Cross at the address listed standard NODC reproduction fee. above. Chapter 7 Users and Information Needs Table 6 presents a matrix developed by participants in from their own regional perspective and submit their the regional concerns breakout group at the C-CAP modifications to C-CAP. This will enable C-CAP to gen- Rhode Island Workshop (see Appendix 4). The matrix erate matrices for each region or a single national matches potential uses with C-CAP products and indi- matrix that will help ensure that C-CAP products meet cates the relative value of the product according to use. the broadest range of user needs possible. Interested parties are encouraged to modify this table Table 6 The potential utility of C-CAP coastal land-cover information; abbreviations: H=high; M=moderate; L=Iow. Potential uses Map data Digital data Tables Physical boundaries Error estimation Technical review M M L L L Decisions M M L L L Modeling L H H H H Interstate coordination H H H H H Enforcement L L L L L Hazard response H H L L L Policy L L H H H Management and planning H H H H H Education H H M M H Citizens H L L L L Commerce H H L L H Research H H M M H 53 54 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program Oiapter 8 Regional Participation Purpose Universities of Connecticut and Rhode Island NOAA C-CAP will endeavor to cooperate with all ongo- Faculty members of the Universities of Connecticut ing wetlands mapping and change-detection programs and Rhode Island worked cooperatively to examine at the Federal, State, and regional levels. Priorities for several issues concerning coastal land-cover classifica- NOAA funding allocation will be tion and change detection in the Northeast (Hurd et al., 1992). In the first part of the project, detailed GIS 1) biogeographic diversity, data on coastal wetlands in Rhode Island derived from 2) joint funding efforts, and aerial photography were used to establish coastal wet- 3) existing field-based studies. lands signatures for input to a digital classification of Landsat TM imagery. This work is crucial in assessing Other considerations will include: the extent to which an existing coastal wetlands dataset (e.g. NWI digital data) can be used to establish a classi- 1) areas of rapid development, fication for a larger TM dataset. Other areas of impor- 2) areas disturbed by major storms or other natural tance to C-CAP include assessments of 1) classification events, and approaches best suited to characterize wetlands in south- 3) areas disturbed by hazardous technologies (e.g. oil ern New England; 2) techniques for monitoring coastal spills). wetlands change in the Northeast using several change detection techniques to look at TM imagery from the same location for 1988 and 1982; and 3) multistate, Regional Project Summaries multi-institutional collaboration in southern New England. St. Croix River Estuary (Border of Maine and New Brunswick, Canada) University of Delaware This is a cooperative effort involving the U.S. Fish and University of Delaware faculty members at the Center Wildlife Service (USFWS), the Gulf of Maine Program, for Remote Sensing played a lead role in developing and Environment Canada. A change detection analysis the interagency land-cover classification system used by was performed using TM imagery from 1985 and 1992. C-CAP (Klemas et al., 1993). The system was developed The image processing and change detection analysis during joint meetings with representatives from key was performed at Oak Ridge National Library (ORNL). government agencies including NOAA, USGS, USEPA, Five field verification exercises were carried out in con- USFWS, and COE. junction with USFWS personnel. The C-CAP change Currently, University of Delaware faculty members detection product has been completed and submitted are developing remote-sensing and field techniques for to NODC. measuring indicators of wetland condition and func- tional health over large wetland areas. An overview of wetland health assessment techniques has been pre- Coastal Massachusetts pared, with special emphasis on wetland condition and functional health indicators that can be monitored with This is a cooperative submerged land effort involving remote sensors (Patience and Klemas, 1993). The over- the Massachusetts Department of Environmental Pro- view report contains a comprehensive literature search tection (DEP) Wetlands Conservancy Program. A pilot and chapters describing the techniques and their sta- project focusing on training Massachusetts personnel tus. A joint study has been initiated with investigators in current SRV mapping techniques and adapting the from Louisiana State University and USEPA to work on C-CAP protocol for use in Massachusetts was conducted impaired and healthy pilot test sites in Louisiana in the spring and summer of 1993. Photo interpreta- marshes. Field data, including measures of biomass, tion and mapping are being performed by DEP person- soils, hydrology, chemistry, biology, and light reflec- nel with technical assistance from the NMFS Beaufort tance, are being correlated with Landsat TM imagery to Laboratory. The SRV polygons will be added to wetland assess biomass and stress indicators over large areas data on coastal orthophoto maps. with the help of modified models and techniques devel- 55 56 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program oped during previous studies. The data derived from areas were prepared to illustrate static land cover for these investigations are crucial to C-CAP for early detec- 1984 and 1988/1989 and land-cover change between tion of functional change in habitat. these dates. The final database was delivered to NODC and is available on CD-ROM for purchase by the public. This analysis was conducted on graphics workstations Oak Ridge National Laboratory employing ERDAS image processing software, ERDAS raster GIS software, and Oak Ridge Geographics GIS The C-CAP prototype and first regional project was software. Processing and verification techniques were conducted in the Chesapeake Bay region by ORNL similar to those employed in the initial MSS/TM analy- (Dobson and Bright, 1991, 1992, and 1993). In the first sis. Investigators revisited the area in the spring and phase, a land-cover classification was completed for a summer of 1991 and participated in the Maryland field four-scene area using MSS data for 24-25 October 1978, verification workshop (See Appendix 4). Finally, the and change detection was completed for portions of a database was modified to accommodate the new C-CAP scene in the vicinity of Metomkin Inlet, VA, using MSS land-cover classification scheme'and to incorporate sug- data for 12 September 1974, MSS data for 24 October gestions and corrections resulting from the Maryland 1978, and TM data for 18 November 1982. The Virginia workshop. The protocols developed for the Chesapeake Institute of Marine Sciences (VIMS) was contracted to Bay project have been incorporated into the C-CAP assist in field verification and training-sample identifi- protocols. Thus the final C-CAP Chesapeake Bay Land- cation. The results of this prototype served as a proof- Cover Change Database complies with this document. of-principle for large-area change analysis, and the meth- ods and techniques served as the basis for the draft protocol presented at the first protocol development Virginia Institute of Marine Sciences workshop. This initial prototype and proof-of-principle was con- The Virginia Institute of Marine Sciences has been ducted at the Oak Ridge Geographics Laboratory. All conducting photographic mapping of submersed veg- cartographic and geographic information processing etation in the entire Chesapeake Bay beginning in 1978 was conducted by ORNL personnel using Oak Ridge and annually since 1984 (Orth et al., 1990 and 1991). Geographics software. Tentative land-cover classes were Although not funded by C-CAP, this important work is determined on the basis-of supervised training samples considered a regional C-CAP project because of the in areas of known land cover. The tentative classes were voluntary collaboration among principal investigators. checked with information available from other sources Methodology for the Chesapeake Bay project was a such as 1:24,000 USGS topographic maps and wetlands starting point for the C-CAP protocol. Data from Chesa- inventories (NWI and county marsh inventories). In- peake Bay have been provided to C-CAP to attempt to vestigators visited the area on 4-6 November 1985 for overlay it with the land-cover data for Chesapeake Bay field verification of the tentative classes and for identifi- generated by ORNL. Historically, Chesapeake Bay has cation of additional training samples in the Wacha- suffered a dramatic decline in SRV and associated fish- preague, Metomkin Inlet, and Saxis areas of Virginia. eries. From 1984 to 1990, however, SRV habitat in- Land-cover classes were determined through iterative creased from 15,400 to 24,313 ha. refinement of supervised training samples. Investiga- tors visited the area in August 1986 for field verification of final land-cover classes in the York River estuary of North Carolina State University Virginia and the Tangier Island and Blackwater Na- tional Wildlife Refuge areas of Maryland. Finally, the A land-cover classification project was conducted by the entire dataset was compared digitally on a cell-by-cell Computer Graphics Center at North Carolina State basis to land-cover data from the USGS Land Use Data University (NCSU) prior to the University's involve- Analysis (LUDA) database in order to resolve certain ment in C-CAP. Coincidentally, the four-scene area ana- classes. lyzed by NCSU was contiguous with the four-scene area In the second phase, the change detection was ex- analyzed by ORNL in the Chesapeake Bay project. Scene tended to cover the full four-scene area by using TM dates are contemporaneous with the 1988 Chesapeake data for 27 August 1984, 21 September 1984, 3 Novem- Bay scenes. Faculty members of NCSU cooperated with ber 1988, and 10 October 1989. The final product ORNL research staff to investigate the potential for consisted of a classified land-cover change matrix data- merging portions of these two independently conducted base for the entire Chesapeake Bay area. Regional maps land-cover classifications based on TM digital data. The at 1:500,000 scale and numerous local area maps cover- goal was to merge the project areas and form a seamless ing individual USGS 1:100,000 and 1:24,000 quadrangle regional land-cover classification from the Chesapeake Chapter 8: Regional Participation 57 Bay to Dare County, N.C. One of the major problems tent with current photographs, on stable media. The investigated was the development of a classification topographic maps are virtually complete for North Caro- scheme adaptable to both areas. This research was a lina (with the exception of Currituck Sound) but are crucial test of the C-CAP concept of regional compat- out of date (mid-forties photography with occasional ibility among neighboring databases developed by dif- photo-revision for cultural features dated in the seven- ferent organizations. ties or eighties). NOAA shoreline manuscripts are based on 1988 or more recent photography but are not com- plete for North Carolina. Necessary photographs for Beaufort Laboratory, National Marine Fisheries construction of manuscripts to complete shorelines in Service North Carolina were obtained by NOAA Photogram- metry Branch in 1988-92. The project in North Carolina is researching protocol Habitat polygons were coordinate digitized by State for conducting and verifying change detection in SRV, of North Carolina personnel and incorporated into a including seagrasses and low-salinity-tolerant grasses and statewide ARC-Info database referenced to the State forbs. Simultaneously the project is completing the first Plane Coordinate System. comprehensive inventory of such habitat in North Three two-color charts of seagrass habitat at a scale of Carolina. The project wasjointly funded by the Albemarle 1:36,000 and measuring about 3 x 4 ft were published Pamlico Program of EPA's National Estuary Program. and are available at no cost. Aerial photography to delineate SRV was first com- missioned in 1985 (Bogue, Back, and southern Core Sounds). The rest of the aerial photography for all University of South Carolina areas of potential SRV between Bogue Inlet and the Virginia border were taken between 1988 and 1992. All University of South Carolina faculty members performed photography was subcontracted to the NOAA Photo- a detailed investigation of the geographic area cen- grammetry Unit and acquired at scales of 1:12,000, tered on two 7.5' U.S. Geological Survey quadrangles 1:20,000, 1:24,000, or 1:50,000. The smallest scale pho- (quads) along South Carolina's coastal plain. These tography provided a bridge between parallel flightlines quads, representative of many other quads in coastal (at 1:24,000) in eastern Pamlico Sound, where minor South Carolina, provide an opportunity to examine two dimensions of some habitat areas exceeded 3 nmi. very different wetland communities. One quad is di- All aerial photography from 1985 through 1991 was rectly on the coast and contains extensive Spartina interpreted and most was compiled on base maps. The alterniWora marsh, developed and undeveloped beach interpretation was supported by extensive systematic front, and a mature maritime forest. The other quad is and directed sampling throughout the study area. At 40 river miles inland and contains significant inland the time of photography, stations were selected by freshwater wetlands with extensive bottoraland hard- statified random sampling, visited, and sampled for woods. The project identified optimum parameters for species of submersed plants and ancillary data (sedi- conducting accurate coastal change detection includ- ment particle size and organic content, water depth, ing, but not limited to, 1) an optimum wetlands classifi- salinity, temperature, and Secchi depth and the pres- cation scheme; 2) an optimum type of remotely sensed ence of exposed peat deposits, shells, algae, or debris data; 3) optimum digital image processing pattern rec- which might confuse signature identification). All loca- ognition algorithms for C-CAP land-cover classification; tions of known and potential habitat, water <6 ft MLLW 4) the applicability and utility of including ancillary on nautical charts, were sampled by positioning a rect- data (e.g. NWI digital data) in the classification pro- angular matrix of points over the nautical chart. Station cess; 5) optimum change detection algorithm logic; positions approximately two scaled nautical miles apart and 6) detailed error evaluation. Results were reported were extracted from the chart and visited with the aid injensen et al. (1993a). of LORAN C, now, preferably, GPS. After receipt and preliminary interpretation of the photographs, field surveys were conducted to verify the range of habitat State of Florida, Department of Environmental signatures and confirm false signatures. Protection Photographs initially interpreted monoscopically are now interpreted stereoscopically. Polygons are traced State of Florida personnel are mapping submersed habi- on stable film at the photograph scale, rectified, and tat in the Florida Keys, Florida Bay, Biscayne Bay, and transferred to base maps with a zoom transfer scope. Tampa Bay. Photography was conducted in the winter Base maps are NOAA shoreline manuscripts, if avail- of 1991 and 1992 by the NOAA Photogrammetry Branch. able, or USGS 7.5' topographic series maps, if consis- The effort in the Keys is cooperative with NOAA's Ma- 58 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis program rine Sanctuary Program. The Keys were photographed CREST and its cooperators in conjunction with ORNL. at 1:48,000 because of cost considerations and will dem- Several field verification exercises have been performed, onstrate resolution of signatures of submersed habitat and a final change detection project is expected in fall at a scale smaller than that acceptable with the current 1994. This information should be useful to a variety of C-CAP protocol. The motion-compensating camera used managers that are presently dealing with severely stressed in this case should enhance resolution. Photography in salmon stocks throughout the study area. Florida Bay is being interpreted and ground verified in fiscal year 1993. C-CAP is partially funding the cost of photography and interpretation. The Hubbard Glacier and Russell Fjord, Alaska This is a cooperative effort involving NMFS's Auke Bay Texas Parks and Wildlife Department Laboratory. A 1986 image is the only image available at this time that meets C-CA-P cloud cover specifications. This agency is currently processing TM imagery, as per The implications of the future movements of the the steps outlined in this document, for the entire Hubbard Glacier make this project unique. During 1986 Texas coast with technical assistance from ORNL. Two the Hubbard Glacier blocked off the mouth of the scenes in the Galveston Bay area for 2 December 1988 Russell Fjord and created the world's largest glacier- have already been classified, and a change detection formed lake. Within months, rising water levels caused analysis was performed, comparing a November 1992 the glacier to burst, restoring tidal flow to the Fjord. scene with the southernmost of the 1988 scenes. Classi- Glacier experts predict that there is a 90% chance that fication has been aided by an abundance of ground the Hubbard Glacier will block off the mouth of the reference data as well as digital NWI data that are Russell Fjord again within the next 10 years. Because available for most of the Texas coast. the portion of the glacier that will block off the Fjord is bigger than the one in 1986, it is predicted that the glacier will not burst. This would cause the rising waters Columbia River, Tillamook Bay, and Willapa Bay to exit the Fiord at the end opposite the glacier, flowing (Oregon and Washington) into Old Situk Creek. This may significantly affect a very important salmon fishery, crucial to the inhabit- This is a cooperative effort involving cooperating agen- ants of nearby Yakutat, Alaska. C-CAP is presently look- cies within the Columbia River Estuary Study Taskforce ing for another image to perform a change analysis and (CREST), NMFS's Point Adams Field Station, Ham- provide more baseline information for future change- mond, Oreg., and Washington State personnel. Imag- detection activities, should the glacier again close off ery for September 1989 and 1992 has been obtained Russell Fjord. The 1986 image has been processed by and a change-detection analysis is being performed by ORNL and the data has been submitted to NODC. Acknowledgments Many individuals contributed to this effort at regional poration; Richard Kleckner, Kathy Lins, Keven Roth, C-CAP workshops, at meetings, and through private and Peg Rawson, U.S. Geological Survey; Mark Laustrup, communications. We wish to thank all of them, espe- National Biological Survey; Doug Norton, U.S. Envi- cially the following who have helped us continuously ronmental Protection Agency; Robert Peplies, East Ten- over several years to improve this classification system: nessee State University; and Warren Pulich, Texas De- Michael DeMers, New Mexico State University; Francis partment of Parks and Wildlife. Brent Moll of Oak Golet, University of Rhode Island; Steven Hoffer, Ridge National Laboratory assisted in the technical re- Lockheed Engineering and Sciences Co.; Michael vision of a preliminary draft. The Chesapeake bay pro- Hodgson, Oak Ridge National Laboratory; Jimmy totype project was funded in part by NOAA's Chesa- Johnston, National Biological Survey; Bill Wilen, U.S. peake Bay Program. The development of this docu- Fish and Wildlife Service; Donley Kisner, Bionetics Cor- ment was funded by NOAA's Coastal Ocean Program. 59 60 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program Literature Cited 63 Lacy, R. detection of areal changes on Marco Island, Florida: appli- 1992. South Carolina finds economical way to update digital cation of digital image processing and remote sensing road data. GIS World 5:58-60. techniques. U.S. Fish Wild. Serv., Biol. Rep. 86(10):1-87. Lee, J. K., R. A. Park, and P. W. Mausel. Phillips, R. C. 1992. Application of geoprocessing and simulation modeling 1974. Temperate grass flats. InH.T. Odum,B.J. Copeland, to estimate impacts of sea level rise on the northeast coast of and E. A. McMahan (eds.), Coastal ecological systems of the Florida. Photogramm. Eng. Remote Sens. 58:1579-1586. United States, Vol. 2, p. 244--314. The Conservation Foun- Light, D. dation, Washington, D.C. 1993. The National Aerial Photography Program as a geo- 1984. The ecology of eelgrass meadows in the Pacific North- graphic information system resource. Photogramm. Eng. west: a community profile. U.S. Fish and Wildlife Service, Remote Sens. 59:61-65. FWS/OBS-84/24, 85 p. Luczkovich,J.J., T. W. Wagncrj L. Michalek, and R. W. Stoffle. Price, K. P., D. A. Pyke, and L. Mendes. 1993. Discrimination of coral reefs, seagrass meadows, and 1992. Shrub diebackin a semiarid ecosystem: the integration sand bottom types from space: a Dominican Republic case of remote sensing and GIS for detecting vegetation study. Photogramm. Eng. Remote Sens. 59:385-389. change. Photogramm. Eng. Remote Sensing 58:455-463. Lunetta, R. S., R. C. Congalton, L. K. Fenstermakerj. R. Jensen, K. Riveraj A., A. Magerjr., R. L. Ferguson, D. W. Field, and F. A Cross. C. McGwire, and L. R. Tinney. 1992. Verification of submerged aquatic vegetation alterations 1991. Remote sensing and geographic information system associated with Corps of Engineers permit requests. Pro- data integration: error sources and research issues. Photo- ceedings of the first thematic conference on remote sensing gramm. Eng. Remote Sens. 57:677-687. for marine and coastal environments, New Orleans, LA, 15- Mausel, P. W., W. J. Kramber, and J. K. Lee.' 17june 1992, p. 781-794. 1990. Optimum band selection for supervised classification Rutchey, K., and L. Velcheck. of multispectral data. Photogramm. Eng. Remote Sens. 1943. Development of an everglades vegetation map using a 56:55-60. SPOT image and the Global Positioning System. Photo- Narumalani, S.j R. Jensen, M. Hayes, J. Michel, T. Montello, gramm. Eng. Remote Sens. 150:767-775. and J. Robinson. Ryerson, R. 1993. Using remote sensing to assess habitat in the Saudi 1989. Image interpretation concerns for the 1990's and les- Arabian Gulf before the Gulf War oil spill. Geo Info Sys- sons from the past. Photogramm. Eng. Remote Sens. tems 3(6):33-41. 55:1427-1430. NIST (National Institute of Standards and Technology). Slamma, C. C. 1992. Spatial data transfer standard. FIPS Publ. 173, Na- 1980. Manual of photogrammetry, 4th ed. American Soci- tional Technical Information Service, Springfield, VA. ety of PhotogrammeLry, Falls Church, VA, 1056 p. NOAA Photogrammetry Branch. Swanson, L. W. 1980. Standard operating procedures for Flight Operations 1949. Topographical manual, part 2, photogrammetry. U.S. Group of the Photographic Operations Section. U.S. Dep. Dep. Commer., Spec. Publ. No. 249, 570 p. Commer., NOAA, Rockville, MD. Thayer, G. W., W.J. Kenworthy, and M. S. Fonseca. 1989. Coastal mapping program operations manual. Sec. 3, 1984. The ecology of eelgrass meadows of the Atlantic coast: QualiLyprogram. U.S. Dep. Commer., NOAA, Rockville, MD. a community profile. U.S. Fish Wildl. Serv. FWS/OBS-84/ Novak, K. 02, 147 p. 1992. Rectification of digital imagery. Photogramm. Eng. Thomasj P., and R. L. Ferguson. Remote Sens. 58:339-344. 1990. National Oceanic and Atmospheric Administration's Odum, E. P. habitat mapping under the Coastal Ocean Program. In S. 1971. Fundamentals of ecology. W. B. Saunders, Philadel- J. Kiraly, F. A Cross, and J. D. Buffington (eds.), Federal phia, PA. coastal wetland mapping programs, p. 27-37. U.S. Fish Omernikj M. Wild. Serv. Biol. Rep. 90(18). 1987. Ecoregions of the United States. EPA Environmental Thomas, J. P., R. L. Ferguson, J. E. Dobson, and F. A Cross. Research Lab., EPA/600/D-87, Corvallis, OR. 1991. NOAA's CoastWatch: Change Analysis Program. Coas- Orth, R. J., and K. A. Moore: tal Wetlands Coastal Zone '9 1 Conference, Long Beach, CA, 1983. Submerged vascular plants: techniques for analyzing their July 1991, p. 259-267. distribution and abundance. J. Mar. Tech. Soc. 17:38-52. Thrower, N.J. W., and J. R. Jensen. Orth, R. J., K. A. Moore, and J. E. Nowak. 1976. The orthophoto and orthophotomap: characteristics, de- 1990. Monitoring seagrass distribution and abundance pat- velopment and application. Am. Cartographer 3(l):39-56. terns: a case study from the Chesapeake Bay. InS.J.Kiraly, U. S. Congress. F. A Cross, and J. D. Buffington (eds.), Federal coastal 1989. Coastal waters in jeopardy: reversing the decline and wetland mapping programs, p. 111-123. U.S. Fish Wild. protecting America's coastal resources. Oversight Report of Serv. Biol. Rep. 90(18). the Committee on Merchaat Marine and Fisheries, Serial Orth, R. J., R. L. Ferguson, and K. D. Haddad. 100-E. U.S. Govt. Print. Off., Washington, D.C., 47 p. 1991. Monitoring seagrass distribution and abundance USGS (U. S. Geological Survey. National Mapping Program). patterns. Coastal Wetlands Coastal Zone '91 Conference, Long 1992. Standards for digital line graphs for land use and land Beach, CA, American Society of Civil Engineers, p. 281-300. cover, technical instructions. Referral STO-1-2. Patience, N., and V. V. Klemas. Wang, J., P. M. Treitz, and P. J. Howarth. 1993. Wetland functional health assessment using remote 1992. Road network detection from SPOT imagery for updat- sensing and other techniques: literature search. U.S. Dep. ing geographical information systems in the rural-urban Commer., NOAA Tech. Memo. NMFS-SEFSC-319. fringe. Intj Geogr. Inform. Syst. 6(2):141-157. Patterson, S. G. Welch, R., M. Remillard, andj. Alberts. 1986. Mangrove community boundary interpretation and 1992. Integration of GPS, remote sensing, and GIS techniques 64 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program for coastal resource management. Photogramm. Eng. Re- (eds.), Federal coastal wetland mapping programs, p. 9- mote Sens. 58:1571-1578. 19. U.S. Fish Wild. Serv. Biol. Rep. 90 (18). Westmoreland, S., and D. A. Stow. Ziemanj C. 1992. Category identification of changed land-use polygons 1982. The ecology of the seagrasses of south Florida: a commu- in an integrated image processing/geographic information nityprofile. U.S. Fish Wildl. Sm., FWS/OBS-82/25,185 p. system. Photogramm. Eng. Remote Sens. 58:1593-1599. Ziemanj C., and R. T. Zieman. Wilen, B. 0. 1989. The ecology of the seagrass meadows of the west coast 1990. The U.S. Fish and Wildlife Service's National Wetlands of Florida: a community profile. U.S. Fish Wildl. Serv. Inventory. In S. J. Kiraly, F. A. Cross and J. D. Buffington Biol. Rep. 85 (7.25). 185 p. Laffature Cited Anderson, J. R., E. E. Hardy, J. T. Roach, and R. E. Witmer. Crowell, M., S. P. Leatherman, and Y11. K. Buckley. 1976. A land use and land cover classification system for use 1991. Historical shoreline change: error analysis and map- with remote sensor data. U.S. Geol. Surv. Prof. Pap., 964p. ping accuracy. J. Coastal Res. 7:839-852. Armstrong, R. A. Dahl, T. W. 1993. Remote sensing of submerged vegetation canopies for 1990. Wetlands losses in the United States 1780's to biomass estimation. Int.J. Remote Sens. 14:621-627. 1980's. U.S. Fish and Wildlife Service, Washington, D.C. Augenstein, E. W., D. A. Stow, and A. S. Hope. Daiber, F. C. 1991. Evaluation of SPOT HRV-XS data for kelp resource 1986. Conservation of tidal marshes. Van Nostrand inventories. Photogramm. Eng. Remote Sens. 57:501-509. Reinhold, New York. Baker, J. R., S. A. Briggs, V. Gordon, A. R. Jones, J. J. Settle, Dennison, W. C., R.J. Orth, K. A. Moore,J. C. Stevenson, V. Carter, J. R. Townshend, and B. K. Wyatt. P. W. Bergstrom, and R. A. Batiuk. 1991. Advances in classification for land cover mapping using 1993. Assessing water quality with submerged aquatic vegeta- SPOTHRVimagery. Int.j. Remote Sens. 12:1071-1085. tion: habitat requirements as barometers of Chesapeake Bay Bartlett, D. S. health. Bioscience 43(2):86-94. 1987. Remote sensing of tidal wetlands. In V. V. Klemas, J. Dobson, J. E. P. Thomas, and J. B, Zaitzeff (eds.), Remote sensing of 1993a. Land cover, land use differences distinct. GIS World estuaries, proceedings of a workshop, p. 145-156. U.S. Dep. 6(2):20-22. Commer., NOAA, Washington, D.C. 1993b. A conceptual framework for integrating remote sens- Bormann, F. H., and G. E. Likens. ing, GIS, and geography. Photogramm.Eng. Remote Sens. 1969. The watershed-ecosystem concept and studies of nutri- 59:1491-1496. enL cycles. In G. M. VanDyne (ed.), The ecosystem con- Dobson, J. E., and E. A. Bright. ceptin natural resource management, p. 49-67. Academic 1991. CoastWatch-detecting change in coastal wetlands. Press, New York. Geo. Info. Systems (Jan.), p. 36-40. Campbell, J. B. 1992. CoastWatch Change Analysis Program (C-CAP) Chesa- 1987. Introduction to remote sensing. Guilford Press, peake Bay regional project. In Global change and education, New York. Vol.l,p.109-110. American Society for Photograrnmetry and Caspers, H. Remote Sensing, ASPRS/ACSM 92 Tech. Pap., Bethesda, MD. 1967. Estuaries: analysis of definitions and biological 1993. Large-area change analysis: the CoastWatch Change considerations. InG. H. Lauff (ed.), Estuaries, p. 6-8. Am. Analysis Project (C-CAP). Proceedings of The Pecora 12 Assoc. Adv. Sci. Publ. 83. Conference (In Press). Cheng, T. D., G. L. Angelici, R. E. Slye, and M. Ma. Eastman, J. R., and M. Fulk. 1992. Interactive boundary delineation of agricultural lands 1993. Long sequence time series evaluation using standard- using graphics workstations. Photogramm. Eng. Remote ized principal components. Photogramm. Eng. Remote Sens. 58:1439-1443. Sens. 59:991-996. Chrisman, N. R. Eckhardt, D. W., J. P. Verdin, and G. R. Lyford. 1991. The error component in spatial data. InD.J.Maguire, 1990. Automated update of an irrigated lands GIS using SPOT M. F. Goodchild, and D. W. Rhind (eds.), Geographical infor- HRV imagery. Photograrnm. Eng. Remote Sens. 56: mation systems: principles and applications. Wiley,NewYork. 1515-1522. Chuvieco, E., and R. G. Congalton. Ellis, M. 1988. Using cluster analysis to improve the selection of train- 1978. Coastal mapping handbook. U.S. Government Print- ing statistics in classifying remotely sensed data. Photo- ing Office, Washington, D.C. gramm. Eng. Remote Sens. 54:1275-1281. Federal Geodetic Control Committee. Clarke, J. A., and J. A. Primus. 1984. Standards and specifications for geodetic control 1990. Sea-level changes resulting from future retreat of ice networks. NOAA, National Geodetic Information Branch, sheets: an effect of C02 warming of the climate. In M. J. N/CG17X2, Sect. 2, p. 1. Tooley, and I. Sherman (eds.), Sea-level changes, p. 356- Ferguson, R. L., and L. L. Wood. 370. Blackwell, Oxford. 1990. Mapping submerged aquatic vegetation in North Caro- Congalton, R. G. lina with conventional aerial photography. In S. J. Kiraly, 1991. A review of assessing the accuracy of classifications of F. A Cross, and J. D. Buffington (eds.), Federal coastal remotely sensed data. Remote Sens. of Environ. 37:35-46. wetland mapping programs, p. 125-133. U.S. Fish Wild. Cowardin, L. M., V. Carter, F. C. Golet, and E. T. LaRoe. Serv. Biol. Rep. 90(18). 1979. Classification of wetlands and deepwater habitats of the 1994. Rooted vascular beds in the Albemarle-Pamlico estua- United States. U.S. Fish and Wildl. Serv. FWS/OBS-79/31. rine system. EPA Albemarle-Pamlico Study, North Carolina Cowen, D. J., J. R. Jensen, and J. Halls. Dept. of Environment, Health, and Natural Resources, Rep. 1991. Maintenance of TIGER files using remotely sensed No. 94-02, 122 p. data. Am. Soc. Photogramm. Remote Sens. 4:31-40. Ferguson, R. L., G. W. Thayer, and IF. R. Rice. Cross, F. A, and J. P. Thomas. 1980. Marine primary producers. In F. J. Vernberg and 1992. CoastWatch Change Analysis Program (C-CAP): an over- W. Vernberg (eds.), Functional adaptation of marine organ- view Chesapeake Bay regional project. In Global change isms, p. 9-69. Academic Press, New York. and education, Vol, 1, p. 57. American Society for Photo- Ferguson, R. L., J. A. Rivera, and L. L. Wood. grammetry and Remote Sensing, ASPRS/ACSM 92 Tech. 1989. Seagrasses in Southern Core Sound, NC. NOAA Pap., Bethesda, MD. Coastal Ocean Program Submerged Aquatic Vegetation 61 62 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program Study [chart]. U.S. Dep. Commer., SE Fish. Sci. Center, Jensen, J. R. Beaufort Lab., Beaufort, NC. 1986. Introductory digital image processing: a remote sens- Ferguson, R. L., L. L. Wood, and B. T. Pawlak. ing perspective. Prentice-Hall, Englewood Cliffs, NJ. 1991. SAV habitat from Drum Inlet to Ocracoke Inlet, 1994. Introductory digital image processing: a remote sensing NC. NOAA Coastal Ocean Program Submerged Aquatic perspective, 2nd ed. Prenfice-Hall, Englewood Cliffs, NJ. Vegetation Study [chart]. U.S. Dep. Commer., SE Fish. Jensen,J. R., and S. Narumalani. Sci. Cent., Beaufort Lab., Beaufort, NC. 1992. Improved remote sensing and GIS reliability diagrams, 1992. SAV habitat in 1985 and 1988: Cape Lookout to Drum image genealogy diagrams, and thematic map legends to Inlet, NC. NOAA Coastal Ocean Program Submerged enhance communication. Int. Remote Sens. 6(B6): Aquatic Vegetation Study [chart]. U.S. Dep. Commer., SE 125-132. Fish. Sci. Cent., Beaufort Lab., Beaufort, NC. Jensen,J. R., L. Tinney, andj. Estes. Ferguson, R. L., L. L. Wood, and D. B. Graham. 1980. Remote sensing techniques for kelp surveys. Photo- 1992. Detection of change in submerged coastal habitat. In gramm. Eng. Remote Sens. 46:743-755. Global change and education, Vol. 1, p. 70-79. American Jensen,J. R., E. W. Ramsey, H. E. Mackey, E. Christensen, Society for Photogrammetry and Remote Sensing, ASPRS/ and R. Sharitz. ACSM 92 Tech. Pap., Bethesda, MD. 1987. Inland wetland change detection using aircraft MSS 1993. Monitoring spatial change in seagrass habitat with aerial data. Photogramm. Eng. Remote Sens. 53:521-529. photography. Photogramm. Eng. Remote Sens. 59:1033-1038. Jensen,J. R., E. W. Ramsey,J. M. Holmes,J. E. Michel, B. Savitsky, Foody, G. M., N. A. Campbell, N. M. Trodd, and T. F. Wood. and B. A. Davis. 1992. Derivation and applications of probabilistic measures of 1990. Environmental sensitivity index (ESI) mapping for oil class membership from the maximum-likelihood classifi- spills using remote sensing and geographic information cation. Photogramm. Eng. Remote Sens. 58:1335-1341. system technology. Int. J. Geographical Information Sys- Fung, T., and E. LeDrew. tems 4:181-201. 1987. Application of principal components analysis for change Jensen,J. R, D.J. Cowen, 0. Weatherbee,j. D. Althausen, S. detection. Photogramm. Eng. Remote Sensing 53:1649-1658. Narumalani, B.J. Kjerfve, P. Talwani, and R. Lacy. 1988. The determination of optimal threshold levels for 1992. Change detection algorithm evaluation: CoastWatch change detection using various accuracy indices. Photo- protocol development in South Carolina. In Global change gramm. Eng. Remote Sens. 54:1449-1454. and education, Vol. 1, p. 120-129. American Society for Gong, P., and P.J. Howarth. Photogrammetry and Remote Sensing, ASPRS/ACSM 92 1990. An assessment of some factors influencing multispec- Tech. Pap., Bethesda, MD. tral land-cover classification. Photogramm. Eng. Remote Jensen, J. R., D.J. Cowen, S. Narumalani,J. D. Althausen, Sens. 56:597-603. and 0. Weatherbee. Goodchild, M. F., and K. K_ Kemp (eds.). 1993a. An evaluation of CoastWatch change detection proto- 1990. Technical issues in GIS: NCGIA core curriculum, 3 col in South Carolina. Photogramm. Eng. Remote Sens. vols. National Center for Geographic Information and 59:1039-1046. Analysis, Santa Barbara, CA. Jensen, J. R., S. Narumalani, 0. Weatherbee, and H. E. Mackey. Haddad, K. D. 1993b. Measurement of seasonal and yearly cattail and 1992. CoastWatch Change Analysis Program (C-CAP) remote waterlily changes using multiclate SPOT panchromatic sensing and GIS protocols. In Global change and educa- data. Photogramm. Eng. Remote Sens. 59:519-525. tion,Vol.l,p.58-69. American Society for Photogramme- Keller, M. try and Remote Sensing, ASPRS/ACSM 92 Tech. Pap., 1978. A study of applied photogrammetric bathymetry in the Bethesda, MD. National Oceanic and Atmospheric Administration. In Haddad, K. D., and D. R. Ekberg. Coastal mapping symposium proceedings, p. 45-54. Ameri- 1987. Potential of Landsat TM imagery for assessing the na- can Society of Photogrammetry, Falls Church, VA. tional status and trends of coastal wetlands. In 0. Magoon, Kenworthy, W. J., and D. E. Harmert (eds.). H. Converse, D. Miner, L. Tobin, D. Clark, and G. Dormurat 1991. The light requirements of seagrasses. U.S. Dep. (eds.), Coastal Zone '87, Vol. 5, p. 5182-5195. American Commer., NOAA Tech. Memo. NMFS-SEFSC-287. Society of Civil Engineers, New York. Khorram, S.,J. A. Brockhaus, and H. Cheshire. Haddad, K. D., and B. A. Harris. 1992. Merger of independently derived wetland inventory 1985. Use of remote sensing to assess estuarine habitats. In data bases. In Global change and education, Vol. 1, p. 0. Magoon, H. Converse, D. Miner, D. Clark, and L. Tobin 111-119. American Society for Photogrammetry and Re- (eds.), Coastal Zone '85, American Society of Civil Engi- mote Sensing, ASPRS/ACSM 92 Tech. Pap., Bethesda, MD. neers, New York. Kiraly S. J., F. A Cross, and J. D. Buffington (eds.). Haddad, K. D., and G. McGarry. 1990. Overview and recommendations. In S. J. Kiraly, F. A 1989. Basin-wide management: a remote sensing/GIS Cross, and J. D. Buffington (eds.), Federal coastal wetland approach. InO. Magoon, H. Converse, D. Miner, L. Tobin, mapping programs, p. 1-7. U.S. Fish Wild. Serv. Biol. Rep. and D. Clark (eds.), Coastal Zone '89, Vol. 2, p. 1822-1836. 90(18). American Society of Civil Engineers, New York. Klernas, V.,J. E. Dobson, R. L. Ferguson, and K_ D. Haddad. Hodgson, M. E., and R. W. Plews. 1993. A coastal land cover classification system for the NOAA 1989. N-dimensional display of cluster means in feature CoastWatch Change Analysis Project, J. Coastal Res. 9: space. Photogramm. Eng. Remote Sens. 55:613-619. 862-872. Hurd, J. D., Civco, D. L., LaBash, C. L., and P. V. August. Klemas, V., and M. A. Hardisky. 1992. Coastal wetland mapping and change detection in the 1987. Remote sensing of estuaries: an overview. InV.Klemas, northeastern United States. In Global change and education, J. P. Thomas, and J. B. Zaitzeff (eds.), Remote sensing of Vol. 1, p. 130-139. American Society for Photogrammetry and estuaries. p. 91-120. U.S. Dep. Commer., NOAA, Wash- Remote Sensing, ASPRS/ACSM 92 Tech. Pap., Bethesda, MD. ington, D. C. Appendix 1 U.S. Geological Survey Land-Cover Classification Scheinefor Remote Sensor Data Appendix Table I Summary of Level I and Level II elements of the U.S. Geological Survey "Land Use and Land Cover Classification System for Use with Remote Sensor Data" (Anderson et al., 1976; USGS, 1992). Level Land-use and land-cover class I Urban or Built-Up Land 11 Residential 12 Commercial and Services 13 Industrial 14 Transportation, Communications, and Utilities 15 Industrial and Commercial Complexes 16 Mixed Urban or Built-up 17 Other Urban or Built-up Land 2 Agricultural Land 21 Cropland and Pasture 22 Orchards, Groves, Vineyards, Nurseries, and Ornamental Horticultural Areas 23 Confined Feeding Operations 24 Other Agricultural Land 3 Rangeland 31 Herbaceous Rangeland 32 Shrub-Brushland Rangeland 33 Mixed Rangeland 4 Forest Land 41 Deciduous Forest Land 42 Evergreen Forest Land 43 Mixed Forest Land 5 Water 51 Streams and Canals 52 Lakes 53 Reservoirs 54 Bays and Estuaries 6 Wedand 61 Forested Wettand 61 Nonforested Wetland 7 Barren Land 71 Dry Salt Flats 72 Beaches 73 Sandy Areas Other than Beaches 74 Bare Exposed Rock 75 Strip Mines, Quarries, and Gravel Pits 76 Transitional Areas 77 Mixed Barren Land 8 Tundra 81 Shrub and Brush Tundra 82 Herbaceous Tundra 83 Bare Ground Tundra 84 Wet Tundra 85 Mixed Tundra 9 Perennial Snow or lee 91 Perennial Snowfields 92 Glaciers 65 66 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program - Appendix 2 U.S. Fish and Wildlife Smdce Ketland Classification Schente Appendix Table 2 Summary of the classification hierarchy of wetlands and deepwater habitats, showing systems, subsystems, and classes of the U.S. Fish and Wildlife Service "Classification of Wetlands and Deepwater Habitats of the United States" (Cowardin et al., 1979). System Subsystem Class Rock Bottom Subtidal F-_ Unconsolidated Bottom Aquatic Bed Reef Marine Aquatic Bed - Interlical F-_ Reef Rocky Shore Unconsolidated shore Rock Bottom - Subtidal F-_ Unconsolidated Shore Aquatic Bed Reef Estuarine Aquatic Bed Reef Streambed - Intertidal Rocky Shore Unconsolidated Shore Emergent Welland Scrub-Shrub Welland Forested Wetland Rock Bottom Unconsolidated Bottom Aquatic Bed Tidal Streambed CO Rocky Shore < Unconsolidated Shore Emergent Welland Rock Bottom Lu Unconsolidated Bottom Lu Lower Perennial Aquatic Bed Rocky Shore z < Riverine Unconsolidated Shore 0 Emergent Wetland 0 z :5 Rock Bottom Lu Unc nsolidated Bottom 0 Upper Perennial Aquatic Bed Rocky Shore Unconsolidated Shore Intermittent Streambed Rock Bottom Limnetic Unconsolidated Bottom Aquatic Bed Lacustrine Rock Bottom Unconsolidated Bottom Littoral Aquatic Bed Rocky Shore Unconsolidated Shore Emergent Wetland Rock Bottom Unconsolidated Bottom Aquatic Bed palustrine Unconsolidated Shore Moss-Lichen Welland Emergent Welland Scrub-Shrub Wetland Forested Welland 67 68 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program - Appendix 3 C-CAP Coastal Land-Cover aassification System Definitions The C-CAP Coastal Land-Cover Classification System, de- areas in which a significant land area is covered by concrete scribed in Chapter 2 (Table 2), was developed to meet C-CAP and asphalt or other constructed materials. Vegetation, if requirements (KIemas et al., 1993). It is intended to be com- present, occupies <20% of the landscape. Examples of such patible with other classification systems to facilitate the ex- areas include apartment buildings, skyscrapers, shopping cen- change of data among related programs, especially USGS, ters, factories, industrial complexes, large barns, airport run- NWI, and EPA's EMAP. Those classes underlined in Table 2 ways, and interstate highways. arc of greatest importance to the C-CAP program and most can be detected by satellite sensors such as TM and SPOT. 1.12-Low Intensity (Mixed Pixels)-Low-In tensity Developed Land includes areas with a mixture of constructed materials (e.g. roofing, metal, concrete, asphalt) and vegetation or Categories of the C-CAP Classification System other cover. Constructed materials account for 50-79% of total area. These areas commonly include single-family hous- The system starts with three superclasses: 1.0-Uplands, 2.0- ing areas, especially in suburban neighborhoods, but may in- Wetlands, and 3.0-Water and Submerged Land. These super- clude scattered surfaces associated with all types of land use. As classes are subdivided into classes and subclasses at the sec- the percentage of constructed material cover decreases, this ond and third levels, respectively. Most of the classes and category grades into Cultivated, Grassland, Woody, and other subclasses in the C-CAP system are taken from Anderson et land-cover classes. A large building surrounded by several acres al. (1976), Cowardin et al. (1979), and USGS (1992). How- of grass, for example, might appear as one or more pixels of ever, a few definitions have been modified to resolve conflicts High-Intensity Developed Land, one or more pixels of Low- between the Anderson et al. and Cowardin et al. categories, Intensity Developed Land, and many pixels of Grassland. and some finer categories have been added (e.g. High-Inten- sity Developed Land and Low-Intensity Developed Land) - 1.2-Cultivated Land 1.0-Upland Agricultural Land in the Anderson et al. (1976) classification system was defined as The superclass 1.0-Upland is divided into seven classes: 1.1- Developed Land, 1.2-Cultivated Land, 1.3-Grassland, 1.4- land used primarily for production of food and Woody Land, 1.5-Bare Land, 1.6-Tundra, and 1.7-Snow/Ice. fiber. On high-altitude imagery, the chief indications of agricultural activity will be distinctive geometric field and road patterns on the landscape and the traces pro- 1.1 -Developed Land duced by livestock or mechanized equipment." This class is composed of areas of intensive anthropogenic C-CAP renamed this class "Cultivated Land" to emphasize use. Much of the land is covered by structures and impervi- land cover rather than land use. This category contains areas ous surfaces. Anderson et al. (1976) called these areas "Ur- that have been planted, tilled, or harvested. Pastures and ban or Built-up Land" although the definition clearly in- hayfields that are in a state of tilling or planting are also cluded suburban and rural areas: included. Otherwise, pasture or hayfield with well-established grasses are placed in the Grassland category (3.0). The Culti- "Included in this category are cities; towns; villages; strip vated Land class is divided into three subclasses: 1.21-Orchards/ developments along highways; transportation, power, and Groves/Nurserics, 1.22-Vincs/Bushes, and 1.23-Cropland. communications facilities; and areas such as those occu- pied by mills, shopping centers, industrial and commer- 1.21-Orchards/Groves/Nurseri,es-This category includes cial complexes, and institutions that may, in some in- woody-stemmed crops that are dominated by single-stemmed, stances, be isolated from urban areas." woody vegetation that is unbranched 0.6 to 0.9 m (2 to 3 ft) above the ground, having a height >3 m (10 ft). Some ex- To clarify this apparent contradiction, C-CAP specifies all amples of the crops included are apple and cherry orchards, constructed surfaces regardless of land use. Developed Lands and palm date groves. Anderson et al. (1976) states are divided into two Level 11 groups: 1.11-High Intensity and 1.12-Low Intensity. "Orchards and groves produce the various fruits and nut crops. Tree nurseries that provide seedlings for planta- 1.1 I-High Intensity (Solid Cover)-High-Intensity Developed tion forestry also are included." Land includes heavily built-up urban centers and large con- structed surfaces in suburban and rural areas with a variety of Isolated fruit trees and other orchards substantially smaller different land uses. The High-Intensity category contains than the areal unit of observation are not included. Pine 69 70 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program plantations are not included in this class; they are assigned to 1.31-Unmanaged-The Unmanaged, Herbaceous, Grasslands the Forest, Evergreen, Woody category (1.421). category refers to herbaceous cover that is allowed to grow naturally and is not fertilized, cut, or tilled and planted 1. 22-Vines/Bushes-Vines/ Bushes refers to areas of multiple regularly. This category includes, but is not limited to, the stemmed, woody-stemmed crops that are shrubs <3 rn (10 ft) Anderson et al. (1976) "Herbaceous Rangeland" category: in height. Examples of crops included in this category are blueberries, grapes, and other vines and bushes producing "the tall grass (or true prairie), short grass, bunch grass various fruit or nut crops (Anderson et al., 1976). This group or palouse grass, and desert grass regions ... Bunch grass has a different spectral signature than other Cultivated Land and desert grass are found in many locations, represent- groups because of the size and spatial configuration of the ing transitional situations to desert shrub. Typical occur- vines, shrubs, and bushes. rences of grasslands include such species as the various bluestems (Andropogon), grama grasses (Bouteloua), wheat- 1.23-Cropland-This class of Cultivated Land refers to any grasses (Agropyron), needle-grasses (Stipa), and fescues crop type that is planted on a regular basis. Crops may be (Festuca). This category also includes the palmetto prai- planted annually in the same field year after year or on a rie areas of south-central Florida, which consist mainly rotating schedule. Anderson et al. (1976) states of dense stands of medium length and tall grasses such as wiregrass (Aristida stricta) and saw palmettos (Seronoa "The several components of Cropland now used for agri- ripens), interspersed occasional palms (Sabal palmetto), cultural statistics include: cropland harvested, including and shrubs." bush fruits; cultivated summer-fallow and idle cropland, land on which crop failure occurs; and cropland in soil- Unmanaged grasslands are found throughout the United improvement grasses and legumes." States, often as a transitional phase in the regrowth of aban- doned Cropland, clearcut Woody Land, or land affected by C-CAP has modified this category to emphasize the instanta- natural disturbance, neous state of the land at the time of observation. Hence, for example, Cultivated Land in a five-year rotation scheme will 1.32-Managed-These grasslands are maintained by human be categorized as Cropland for the four years the land is activity and include lawns, golf courses, pastures, hayfields, tilled and as Grassland for the one year the land is fallow and and other areas of grassland in which seeding, fertilization, covered by grasses. A fallow period of several years may result or irrigation enhance biomass productivity. This category in a transition from Cropland to Grassland to Scrub/Shrub may contain vegetation that grows as fallow if vigorous growth and back to Cropland. persists due to the residual effects of management practices Nurseries and horticultural areas (which include floriculture, in the nonfallow state. seed, and sod areas) used perennially for those purposes are included in this category if woody-stemmed plants are not grown. Greenhouses normally fall in the Developed Land category. 1.4-Woody Land The Woody Land class includes any species with an aerial 1.3-Grassland stem that persists for more than one season. The class is divided into three subclasses: 1.41-Deciduous, 1.42-Evergreen, The Grassland category includes lands covered by natural and 1.43-Mixed. and managed herbaceous cover. Historically, grassland has been defined as land where the potential natural vegetation 1.4 I-Deciduous-The Deciduous Woody subclass includes all is predominantly grasses, grasslike plants, and forbs, and forest and shrub areas having a predominance of trees and where natural herbivory was an important influence in its shrubs that lose their leaves or needles at the end of the frost- pre-civilization state. Anderson et al. (1976) state free season or at the beginning of a dry season. Areas in this category are composed of greater than two-thirds deciduous "Some grasslands have been or may be seeded to intro- trees and shrubs. The Deciduous Woody category can be duce or domesticate plant species. The Grassland (Her- divided into two groups: 1.41 I-Forest and 1.412-Scrub/Shrub. baceous) category contains both managed and 1.41 I-Forest-Deciduous Forest includes areas dominated unmanaged or natural herbaceous cover. The Grassland by single stemmed, woody vegetation unbranched 2-3 ft above (Herbaceous) category can be found in every state in the the ground having a height @!6 in (20 ft). Forest Deciduous United States along with Canada and Mexico." Woody areas have a tree-crown areal density (crown closure percentage) of @!10 percent, are stocked with trees capable of The C-CAP category includes lands with herbaceous cover at the producing timber or other wood products, and exert an time of observation regardless of origin or potential. Pastures, influence on the climate or water regime. In most parts of the hayfields, and natural rangelands are included. Also included United States, these would be the hardwoods such as oak, are lawns and other managed grassy areas such as parks, ce- Quercus, maple, Acer, or hickory, Caiya, and the "soft" hard- eteries, golf courses, road fights-of-way, and other herbaceous- woods, such as aspen, Populus tremuloides. Tropical hardwoods covered, landscaped areas. The Grassland class is divided into are included in the Evergreen Forest (Woody Land) category two subclasses: 1.31-Unmanaged and 1.32-Managed. (1.421). Deciduous forest types characteristic of Wetland, Appendix 3: C-CAP Land-Cover Classification Definitions 71 such as tupelo, Nyssa, or cottonwoods, Populus deltoides, are 1.43-Mixed-The Mixed Wood), class includes all forest and not included in this category. shrub areas where both evergreen and deciduous trees and 1.412-Scrub/Shrub-Deciduous Scrub/Shrub includes all shrubs grow and neither predominates. When evergreen and areas having a predominance of shrub that lose their leaves deciduous species each respectively occupy @:33 % of an area, or needles at the end of the frost-free season or at the begin- the land is classified as Mixed Woody. The Mixed Woody ning of the dry season (Anderson et al., 1976). This category category is subdivided into two additional categories: 1.431- contains vegetation that is <6 in (20 ft) in height. The species Forest and 1.432-Scrub/Shrub. include true shrubs, young trees, and trees or shrubs that are 1.431 Forest-This class includes all forested areas where small or stunted because of environmental conditions. True both evergreen and deciduous trees are growing and neither shrubs are those woody@stemmed species that exhibit several predominate. erect, spreading, or prostrate stems and a general bushy 1.432 Scrub/Shrub-This class includes all shrub areas appearance. Shrub Lands may represent a successional stage where both evergreen and deciduous shrubs are growing and leading to forests or they may be relatively stable communi- neither predominate. ties. Forest regrowth composed of young trees <6 in tall is also included in this category. 1.5-Bare Land 1.42-Evergreen-The Evergreen Woody subclass contains for- eIsts and shrubs that do not lose their leaves or needles at the The Bare Land class, modified from "Barren Land" in Ander- end of a frost-free season or at the beginning of a dry season. son et al. (1976) is composed of bare rock, sand, silt, gravel, The Evergreen Woody category is subdivided into two addi- or other earthen material with little or no vegetation regard- tional categories: 1.421-Forest and 1.422-Scrub/Shrub. less of its inherent ability to support life. Vegetation, if present, 1.421-Forest-Evergreen Forest includes areas in which is more widely spaced and scrubby than that in the vegetated @!67% of the trees remain green throughout the year. Both categories. Unusual conditions, such as a heavy rainfall, occa- coniferous and broad-leaved evergreens are included in this sionally may result in a short-lived, luxuriant plant cover. category. Coniferous evergreens predominate except in tropi- Wet, nonvegetated exposed lands are included in the wet- cal regions where broad-leaved evergreens are indigenous. land categories. Coniferous evergreens, often called softwoods, include such Categories of Bare Land include Dry Salt Flats; Beaches; eastern species as the longleaf pine, Pinus palustris, slash Sandy Areas other than Beaches; Bare Exposed Rock; Strip pine, P. ellioti, shortleaf pine, P. echinata, loblolly pine, P. Mines, Quarries; Gravel Pits; Transitional Areas; and Mixed taeda, and other southern yellow pines; various spruces, Picea, Barren Land: and balsam fir, Abies balsamea; white pine, P. strobus, red pine, P. resinosa, and jack pine, P. banksiana; and hemlock, Tsuga Dry Salt Flats are level bottoms of interior desert basins that canadensis-, and such western species as Douglas fir, Pseudotsuga capture infrequent rainfall and do not qualify as Wetland. menziesii, redwood, Sequoia sempervirens, ponderosa pine, Salt concentrations result in highly reflective surfaces. P. monticola, Sitka spruce, P. sitchensis, Engelmann spruce, P. engelmanni, western redcedar, ThyJa plicata, and western Beaches are the smooth sloping accumulations of sand and hemlock, Tsuga heterophylla. Evergreen species commonly associ- gravel along shorelines. The inland face is usually stable, but ated with Wedand, such as tamarack, Larix laficina, or black the shoreward face is subject to erosion by wind and water spruce, P. mariana, are not included in this category. and subject to deposition in protected areas. 1.422-Scrub/Shrub-Evergreen Scrub/Shrub includes ar- eas in which @!67% of the shrubs remain green throughout Sandy Areas other than Beaches are composed primarily of the year. Anderson et al. (1976) states dunes-accumulations of sand transported by the wind. Sand accumulations most commonlyare found in deserts although "Both coniferous and broad-leaved evergreens are in- they also occur on coastal plains, river flood plains, and cluded in this category. The typical Shrub Lands are deltas and in periglacial environments. When such sand ac- found in those and and semiarid regions characterized cumulations are encountered in tundra areas, they are not by such xerophytic vegetative types with woody stems as included here but are placed in the Bare Ground Tundra big sagebrush, A?Iemisia tridentata, shadscale, Atiiplex category. confertifolia, greasewood, Sarcobatus vermiculatus, and creosotebush, Larrea divaricata. When bottom lands and Bare Exposed Rock includes areas of bedrock, desert pave- moist flats are characterized by dense stands of typical ment, scarps, talus, slides, volcanic material, glacial debris, wetland species ... they are considered Wetland. Where and other accumulations of rock without vegetative cover, highly alkaline soils are present, halophytes such as desert with the exception of such rock exposures in tundra regions. saltbush, Atyiplex may occur. The type, density, and asso- ciation of these various species are useful as indicators of Strip Mines, Quarries, and Gravel Pits are areas of extractive the local hydrologic and pedologic environments. Also mining activities with significant surface expression. Vegeta- included in this category is chaparral, a dense mixture tive cover and overburden are removed to expose such de- of broadleaf evergreen sclerophyll shrubs, and the oc- posits as coal, iron ore, limestone, and copper. Quarrying of currences of mountain mahogany, Cercocarpus ledifolius building and decorative stone and recovery of sand and gravel and scrub oaks, Quercus." deposits also result in large open-surface pits. Active, inac- 72 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program tive, and unreclaimed strip mines, quarries, borrow pits, and rennial Snow/Ice cover is of paramount interest. A combina- gravel pits are included in this category until other cover has tion of environmental factors may cause snow and ice to survive been established, after which the land is classified in accor- the summer melting season. Areas of Perennial Snow/Ice cover dance with the resulting cover. Unused pits or quarries that are defined as those where snow, firn (coarse, compacted granu- have been flooded, however, are placed in the appropriate lar snow), or ice accumulation exceeds ablation. Ablation is the Water category. combined loss of snow or ice mass by evaporation and melt- water run-off (Anderson et al., 1976). The class Snow/Ice con- Transitional Areas are dynamically changing from one land tains two subclasses: 1.71-Perennial Snow/Ice and 1.72-Glacier. cover to another, often because of land use activities. This transitional phase occurs when, for example, forest lands are 1.71-Perennial Snow/Ice-This class contains areas covered cleared for agriculture and wetlands are drained for develop- year-round with snow and ice but which have not accumu- ment. Often land becomes temporarily bare as construction lated sufficient ice to be considered Glaciers. Snowfields can initiates the transition from Woody Land or Grassland to a be extensive and thus representative of a regional climate, or future cover associated with residential, commercial, or other can be isolated and localized, where they are known by vari- intensive land use. Lands, such as spoil banks and sanitary ous terms, such as snowbanks. The regional snowline is con- landfills, that are temporarily altered by grading and filling trolled by general climatic conditions and closely parallels are considered transitional. the regional 32* F (0' Q isotherm for the average tempera- ture of the warmest summer month. The use of the term "line" is somewhat misleading because the "snowline" repre- 1.6-Tundra sents an irregular transitional boundary, which is determined at any single location by the combination of snow accumula- Tundra is the term applied to the treeless cover beyond the tion, snow melt, and ablation, variables that can change rap- latitudinal limit of the boreal forest in poleward regions and idly within short distances because of changes in local topo- above the elevation range of the boreal forest in high moun- graphy and slope orientation. Snowfields normally can be tains. In the United States, tundra is found primarily in distinguished from the following Glacier subclass by their Alaska, several areas of the western high mountain ranges, relative lack of flow features (Anderson et al., 1976). and isolated enclaves in the high mountains of New England and northern New York. 1.72-Glaciers-Glacial ice originates from the compaction of The vegetative cover of the tundra is low and dwarfed, snow into firn and finally into ice under the weight of several often forming a continuous mat. Plant characteristics are an successive annual accumulations. Refrozen melt water usu- adaptation to an extreme physical environment in which ally contributes to the increasing density of the glacial ice temperatures may average above freezing only one or two mass. With sufficient thickness, weight, and bulk, flow begins; months of each year, strong desiccating winds may occur, all glaciers exhibit evidence of present or past motion in the great variation exists in solar energy, and permafrost is ubiq- form of moraines, crevasses, and other glacial geomorphic uitous beneath the surface. features. The number of species in the flora is relatively small com- Where the snowline of adjacent ice-free areas extends across pared with typical middle- and low-latitude flora, and the the glacier, it is known as the firn limit, which represents the number decreases as the environment becomes more severe dividing line between the glacier's two major zones, the zone with increasing latitude and elevation. The tundra vegetation of accumulation and the zone of ablation. While glaciers is most luxuriant near the boreal forest. Conversely, plant normally are recognized easily, certain glacial boundaries density and species diversity are lowest near the boundaries may be subject to misinterpretation, even by the experienced of permanent ice and snow areas, where only isolated patches interpreter. Flow features up-glacier from the firn limit typi- of vegetation occur on generally bare surfaces. cally are obscured by fresh snow, forcing the image inter- Tundra may be further subdivided into Shrub and Brush preter to depend on secondary information, such as valley Tundra, Herbaceous Tundra, Bare Ground Tundra, Wet Tun- shapes, or to seek a more discriminating sensor. Similarly, gla- dra, and Mixed Tundra (Anderson et al., 1976). cial drift materials (rock and soil) may stripe the surface of a glacier, and moraine material may cover the ter-minus (or snout) because of ablation, making boundary determination in that 1.7-Snow/Ice vicinity difficult. This later problem occasionally is compounded by the presence of considerable vegetation rooted in the insulat- The temporal dimension is crucial in determining snow and ing blanket of ablation moraine (Anderson et al., 1976). ice cover. Any snowfall, for example, deep enough to conceal another land cover, no matter how briefly, comprises the visible surface at that time and technically constitutes the 2.0-Wetland land cover for the period of its duration. As a practical mat- ter, of course, analysts usually need to characterize the land Cowardin et al. (1979) define wetlands as lands where satura- cover persisting for a greater portion of the year. At higher tion with water is the dominant factor determining soil devel- latitudes and elevations, snow and ice persist for greater opment and the types of plant and animal communities portions of the year, and seasonal coverage becomes a more living in the soil and on its surface. The single feature that all important concern. At extreme latitudes and elevations, Pe- wetlands share is soil or substrate that is at least periodically Appendix 3: C-CAP Land-Cover Classification Definitions 73 saturated with or covered by water. The upland limit for wetland or deepwater habitat under the Cowardin et al. (1979) vegetated wetlands with soil is 1) the boundary between land classification system. The Level. 11 C-CAP Water and Sub- with predominantly hydrophytic cover and land with pre- merged Land classes are modified from Cowardin et al. (1979) dominantly mesophytic or xerophytic cover; 2) for nonvege- (Appendix 2), and include Water, Reef, and Aquatic Beds. tated wetlands with soil, the boundary between soil that is Marine and Estuarine Reefs and Marine and Estuarine Aquatic predominantly hydric and soil that is predominantly Beds are combined into two classes, 3.2-Marine/ Estuarine nonhydric; or 3) in the case of wetlands without vegetation or Reef and 3.3-Marine/EstuarineAquatic Bed. Aquatic bed in soil, the boundary between land that is flooded or saturated rivers, lakes, and streams are assigned to 3.4-Riverine Aquatic sometime during the growing season each year and land that Bed, 3.5-Lacustrine Aquatic Bed, and 3.6-Palustrine Aquatic is not. Most wetlands are vegetated and are found on soil. Bed classes. This last class also, includes Cowardin et al.'s In the C-CAP Coastal Land-Cover Classification System (1979) Rock Bottom and Unconsolidated Bottoms. (Table 2), "Wetland" includes all areas considered wetland Most C-CAP products will designate water as a single class by Cowardin et al. (1979) except for Wetland Bottoms, Aquatic (3.1) regardless of system type. fir is recognized, however, that Beds, and Nonpersistent Emergent Wetlands. Subdivision of the ma or systems (Marine/Estuarine, Riverine, Lacustrine, the Wetlands class closely resembles the Cowardin et al. sys- Palustrine) are ecologically quite different from one another. tem (Appendix 2). At Level II, C-CAP uses certain Cowardin Hence, the four systems at Level III are shown as subclasses: et al. classes (e.g. Rocky Shore, Unconsolidated Shore, Emer- 3.1 I-Marine/Estuarine, 3.12-Riverine, 3.13-Lacustrine, and gent Weiland) or grouped Cowardin et al. classes (e.g. Woody 3.14-Palustrine. Even though C-CAP does not commit itself to Wetland = Scrub-Shrub + Forested Wetland) in combination provide the subclass data, this option is encouraged for re- with Cowardin et al. systems (i.e. Marine, Estuarine, Riverine, gional participants. Incorporating water system information Lacustrine, Palustrine). Thus, C-CAP Level 11 wetland classes makes the C-CAP scheme more compatible with the Cowardin became 2.1-Marine/Estuarine Rocky Shore, 2.2-Marine/Es- et al. system. The subclass 3.1l.-Marine/Estuarine includes tuarine Unconsolidated Shore, 2.3-Marine/ Estuarine Emer- bottoms and undetected reefs and aquatic beds. The sub- gent Wetland, 2.4-Estuarine Woody Wetland, 2.5-Riverine classes 3.12-Riverinc, 3.13-Lacustrine, and 3.14-Palustrine in- Unconsolidated Shore, 2.6-Lacustrine Unconsolidated Shore, clude bottoms and undetected aquatic beds or non-persis- 2.7-Palustrine Unconsolidated Shore, 2.8-Palustrine Emergent tent emergent wetlands. Wetland (persistent), and 2.9-Palustrine Woody Wetland. 3.3-Marine/Estuarine Aquatic Beds includes the subclass Salinity displays a horizontal gradient in marshes typical of Rooted Vascular, which is broken into High Salinity (>5 ppt) coastal plain estuaries. This is evident not only through the and Low Salinity (<5 ppt). The break was made at 5 ppt direct measurement of salinity but also in the horizontal salinity because it separates true! seagrasses that require high distribution of marsh plants in marshes with positive correla- salinity from low salinity species that are tolerant of or re- tions between vertical rise and landward location (Daiber, quire fresh water. Both low and high salinity types of SRV are 1986). Therefore Marine Estuarine Emergent Wetland was important to the C-CAP project. High Salinity includes partitioned into Haline (Salt) and Mixohaline (Brackish) Cowardin et al.'s mesohaline, polyhaline, euhaline, and Marshes. For both subclasses, the definitions used in Cowardin hyperhaline salinity categories. Low Salinity includes Cowardin et al. (1979) were used, i.e. the salinities for Mixohaline et al.'s oligohaline and fresh categories. range from 0.5 to 30 ppt, and Haline include salinities >30 ppt. Within a marsh, plant zonation is usually quite evident. Along the Atlantic coast of North America the pioneer plant Systems and Classes of Cowardin et al. is saltmarsh cordgrass, Spa?tina alterniflora, which often ap- pears in pure stands. Higher up the slope saltmeadow hay, Most of the C-CAP wetland and water definitions are taken Spa7fina patens, becomes dominant, while the upland edges directly from Cowardin et al. (1979). This classification is are bordered by marsh elder, Ivafrutescens, and groundsell hierarchical, progressing from systems and subsystems, at the tree, Bacchaiis halimifolia. Thus, salt marshes could be subdi- most general level, to classes, subclasses, and dominance vided further into High Marsh and Low Marsh. types. Appendix Table 2 illustrates the hierarchical structure C-CAP does not attempt to identify Nonpersistent Emer- to the class level. Modifiers for water regime, water chemistry, gent Wetlands, because they are seasonal. These wetlands are and soils are applied to classes, subclasses, and dominance classified as "Riverine Water" and "Lacustrine Water." Ma- types. Special modifiers describe wetlands and deepwater rine and Estuarine Rocky Shores were combined into a single habitats that have been either created or highly modified by class, Marine/Estuarine Rocky Shore. The same logic was human or beaver activity. applied to create Marine and Estuarine Unconsolidated Shores, Aquatic Beds, and Water. Systems 3.0-Water and Submerged Land The term system refers to a complex of wetlands and deepwater habitats that share the influence of similar hydro- All areas of open water with <30% cover of trees, shrubs, logic, geomorphologic, chemical, or biological factors. Sys- persistent emergent plants, emergent mosses, lichens, or other tems are subdivided into subsystems. land cover are grouped under the heading, Water and Sub- The characteristics of the five major systems-Marine, Es- merged Land, regardless of whether the area is considered tuarine, Riverine, Lacustrine, and Palustrine-have been dis- 74 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program cussed at length in the scientific literature and the concepts wetlands dominated by trees, shrubs, persistent emergents, are well recognized. However, there is disagreement as to emergent mosses, or lichens and 2) habitats with water con- which attributes should be used to bound the systems in taining ocean-derived salts >0.5%. A channel is "an open space. For example, both the limit of tidal influence and the conduit either naturally or artificially created which periodi- limit of ocean-derived salinity have been proposed,as defini- cally or continuously contains moving water, or which forms tions of the upstream limit of Estuarine Systems (Caspers, a connecting link between two bodies of standing water." 1967). As Bormann and Likens (1969) affirm, boundaries of Limits. The Riverine System is bounded on the landward ecosystems are defined to meet practical needs. side by upland, by the channel bank (including natural and man-made levees), or by wetland dominated by trees, shrubs, persistent emergents, emergent mosses, or lichens. In braided Marine System streams, the system is bounded by the banks forming the outer limits of the depression within which the braiding Definition. The Marine system consists of the open ocean occurs. overlying the continental shelf and its associated high-energy The Riverine System terminates downstream where the coastline. Marine habitats are exposed to the waves and cur- concentration of ocean-derived salts in the water exceeds rents of the open ocean, and the water regimes are deter- 0.5% during the period of annual average low flow, or where mined primarily by the ebb and flow of ocean tides. Salinities the channel enters a lake. It terminates upstream where exceed 30%, with little or no dilution except near the mouths tributary streams originate or where the channel originates of estuaries. Shallow coastal indentations or bays without from a lake. Springs discharging into a channel are consid- appreciable freshwater inflow, and coasts with exposed rocky ered part of the Riverine System. islands that provide the mainland with little or no shelter from wind and waves, are also considered part of the Marine System because they generally support typical marine biota. Lacustrine System Limits. The Marine System extends from the outer edge of the continental shelf shoreward to one of three lines: 1) the Definition. The Lacustrine System includes wetlands and landward limit of tidal inundation (extreme high water of deepwater habitats with all of the following characteristics: 1) spring tides), including the splash zone from breaking waves; situated in a topographic depression or a dammed river 2) the seaward limit of wetland emergents, trees, or shrubs; channel; 2) lacking trees, shrubs, persistent emergents, emer- or 3) the seaward limit of the Estuarine System, where this gent mosses or lichens with >30% areal coverage; and 3) total limit is determined by factors other than vegetation. Deepwater area >8 ha (20 acres). Similar wetland and deepwater habitats habitats lying beyond the seaward limit of the Marine System are totaling less than 8 ha are also included in the Lacustrine outside the scope of this classification system. system if an active wave-formed or bedrock shoreline feature makes up all or part of the boundary, or if the water depth in the deepest part of the basin is >2 in (6.6 feet) at low water. Estuarine Svstem Lacustrine waters may be tidal or nontidal, but ocean-derived J salinity is always <0.5%. Definition. The Estuarine system consists of deepwater tidal Limits. The Lacustrine System is bounded by upland or by habitats and adjacent tidal wetlands that are usually semi- wetland dominated by trees, shrubs, persistent emergents, enclosed by land but have open, partly obstructed, or spo- emergent mosses, or lichens. Lacustrine systems formed by radic access to the open ocean, and in which ocean water is at damming a river channel are bounded by a contour approxi- least occasionally diluted by freshwater runoff from the land. mating the normal spillway elevation or normal pool eleva- The salinity may be periodically increased above that of the tion, except where Palustrine wetlands extend lakeward of open ocean by evaporation. Along some low-energy coast- that boundary. Where a river enters a take, the extension of the lines there is appreciable dilution of sea water. Offshore Lacustrine shoreline forms the Riverine-Lacustrine boundary. areas with typical estuarine plants and animals, such as red mangroves, Rhizophora mangle, and eastern oysters, Crassostrea virginica, are also included in the Estuarine system. Palustrine System Limits. The Estuarine system extends 1) upstream and landward to where ocean-derived salts measure <0.5% during the period of Definition. The Palustrine System includes all nontidal wet- -average annual low flow; 2) to an imaginary line closing the lands dominated by trees, shrubs, persistent emergents, emer- mouth of a river, bay, or sound; and 3) to the seaward limit of gent mosses or lichens, and all such wetlands that occur in wetland emergents, shrubs, or trees where they extend beyond tidal areas where salinity due to ocean-derived salts is below the river mouth defined by (2). The Estuarine System also 0.5%. It also includes wetlands lacking such vegetation, but includes offshore areas of continuously diluted sea water. with all of the following four characteristics: 1) the area is <8 ha (20 acres); 2) active wave-formed or bedrock shoreline features are lacking; 3) water depth in the deepest part of Riverine System basin is <2 in at low water; and 4) salinity due to ocean- derived salts is <0.5%. Definition. The Riverine system includes all wetlands and Limits. The Palustrine System is bounded by upland or by deepwater habitats contained within a channel, except 1) any of the other four Systems. Appendix 3: C-CAP Land-Cover Classification Definitions 75 Description. The Palustrine System was developed to group In the Marine and Estuarine Systems, Bottoms are Subtidal, the vegetated wetlands traditionally called by such names as whereas Streambeds and Shores are Intertidal. Bottoms, marsh, swamp, bog, fen, and prairie pothole, which are found Shores, and Streambeds are further divided at the class level throughout the United States. It also includes small, shallow, on the basis of the important characteristic of rock versus permanent, or intermittent water bodies often called ponds unconsolidated substrate. Subclasses are based on finer dis- (except in New England and New York where the term pond tinctions in substrate material unless, as with Streambeds and often refers to substantial lakes). Palustrine wetlands may be Shores, the substrate is covered by, or shaded by, an areal situated shoreward of lakes, river channels, or estuaries on coverage of pioneering vascular plants (often nonhydro- river floodplains, in isolated catchments, or on slopes. They phytes) @!30%. The subclass is then simply "vegetated." Fur- may also occur as islands in lakes or rivers. The erosive forces ther detail as to the type of vegetation must be obtained at of wind and water are of minor importance except during the level of dominance type. Reefs are a unique class in which severe floods. the substrate itself is composed primarily of living and dead animals. Subclasses of Reefs are designated on the basis of the type of organism that formed the reef. Classes and Subclasses As shown in Appendix Table 2, the classes defined in Cowardin et al. (1979) include The class is the highest taxonomic unit below the Subsystem level. It describes the general appearance of the habitat in Rock Bottom (not used in the C-CAP system) terms of either the dominant life form of the vegetation or Unconsolidated Bottom (not used in the C-CAP system) the physiography and composition of the substratc-features Aquatic Bed that can be recognized without the aid of detailed environ- Reef mental measurements. Vegetation is used at two different Streambed (not used in the C-CAP system) levels in the classification. Five life-forms-trees, shrubs, Rocky Shore emergents, emergent mosses, and lichens-are used to de- Unconsolidated Shore fine classes because they are relatively easy to distinguish, do Emergent Wetland not change distribution rapidly, and have traditionally been Scrub-Shrub Wetland used as criteria for classification of wetlands. Other forms of Forested Wetland vegetation, such as submersed or floating-leaved rooted vas- cular plants, free-floating vascular plants, submergent mosses, and algae, though frequently more difficult to detect, are Aquatic Bed used to define the class Aquatic Bed. Pioneer species that briefly invade wetlands when conditions are favorable are Definition. The Aquatic Bed class includes wetlands and treated at the subclass level because they are transient and deepwater habitats dominated by plants that grow principally often are not true wetland species (Cowardin et al., 1979). on or below the surface of the water for most of the growing Using life-forms at the class level has two ma or advantages: season in mostyears. Water regimes include subtidal, irregularly 1) extensive biological knowledge is not required to distin- exposed, regularly flooded, permanently flooded, intermittently guish between various life-forms and 2) various life-forms are exposed, semi-permanently flooded, and seasonally flooded. easily recognizable on a great variety of remote sensing prod- Description. Aquatic beds represent a diverse group of ucts (Anderson et al., 1976) If vegetation (except pioneer plant communities that require surface water for optimum species) covers @!30% of the substrate, classes are distinguished growth and reproduction. They are best developed in rela- on the basis of the life form of the plants that constitute the tively permanent water or under conditions of repeated flood- uppermost layer of vegetation and that occupy an areal cover- ing. The plants are either attached to the substrate or float age @!50% of vegetative cover. Finer differences in life-forms freely in the water above the bottom or on the surface. The are recognized at the subclass level. For example, in the subclasses are Algal, Aquatic Moss (not used by C-CAP and C-CAP system Estuarine Woody Wetland is divided into the not defined here), and Rooted Vascular. subclasses Scrub-Shrub and Forest categories, each of which may be further characterized as Deciduous, Evergreen, and Algal-Algal beds are widespread and diverse in the Marine Mixed on the basis of the predominant life-form. This differs and Estuarine Systems, where they occupy substrates charac- somewhat from the Cowardin et al. system which distinguishes terized by a wide range of sediment depths and textures. trees from shrubs at the class level. They occur in both the Subtidal and Intertidal subsystems If vegetation covers <30% of the substrate, the physiogra- and may grow to depths of 30 in (98 ft). Coastal algal beds are phy and composition of the substrate are the principal char- most luxuriant along the rocky shores of the Northeast and acteristics used to distinguish classes. The nature of the sub- West. Kelp (Macrocystis) beds arc especially well developed on strate reflects regional and local variations in geology and the the rocky substrates of the Pacific coast. Dominance types influence of wind, waves, and currents on erosion and depo- such as the rockweeds Fucus and Ascophyllum and the kelp sition of substrate materials. The classes Bottoms, Shores, Lamina7ia are common along both coasts. In tropical regions, and Streambeds are separated on the basis of duration of green algae, including forms containing calcareous particles, inundation. In the Riverine, Lacustrine, and Palustrine Sys- are more characteristic; Halimeda and Peniellus are common tems, Bottoms are submerged all or most of the time, whereas examples. The red alga Laurencia and the green algae CauLe@pa, Streambeds and Shores are exposed all or most of the time. Enteromorpha, and Ulva are also common Estuarine and Ma- 76 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program rine dominance types; Enteromorpha and [Rva are tolerant of singly or in combination have an areal cover :@75% and an fresh water and flourish near the upper end of some estuar- areal coverage by vegetation of <30%. Water regimes are ies. The stonewort Chara is also found in estuaries. restricted to irregularly exposed, regularly flooded, irregu- Inland, the stoneworts Chara, Nitella, and Tolypella are ex- larly flooded, seasonally flooded, temporarily flooded, and amples of algae that look much like vascular plants and may intermittently flooded. grow in similar situations. However, meadows of Chara may Description. In Marine and Estuarine Systems, Rocky Shores be found in Lacustrine water as deep as 40 m (131 ft) where are generally high-energy habitats that lie exposed as a result hydrostatic pressure limits the survival of vascular submergents of continuous erosion by wind-driven waves or strong cur- (phanaerogams). Other algae bearing less resemblance to rents. The substrate is stable enough to permit the attach- vascular plants are also common. Mats of filamentous algae ment and growth of sessile or sedentary invertebrates as well may cover the bottom in dense blankets, may rise to the as attached algae or lichens. Rocky shores usually display a surface under certain conditions, or may become stranded vertical zonation that is a function of tidal range, wave action, on unconsolidated or rocky shores. and degree of exposure to the sun. In the Lacustrine and Riverine Systems, Rocky shores support sparse plant and Rooted Vascular-Rooted Vascular beds include a large ar- animal communities. The subclasses are Bedrock and Rubble. ray of vascular species in the Marine and Estuarine Systems. More detailed definitions are provided in Cowardin et al. They have been referred to as temperate grass flats (Phillips, (1979). 1974), tropical marine meadows, as well as eelgrass beds, turtlegrass beds, and seagrass beds. The greatest number of species occur in shallow, clear tropical, or subtropical waters Unconsolidated Shore (Tidal Flats) of moderate current strength in the Caribbean and along the Florida and Gulf Coasts. Principal dominance types in these Definition. The Unconsolidated Shore class includes all wet- areas include turtle grass, Thalassia testudinum, shoalgrass, land habitats having three characteristics: 1) unconsolidated Halodule unightfi, manatee grass, Cymodoceafiliformis, widgeon substrates with < 75% areal cover of stones, boulders, or grass, Ruppia ma7itima, sea grasses, Halophila spp., and wild bedrock; 2) <30% areal cover of vegetation other than pio- celery, Vallisneria anzeyicana. neering plants; and 3) any of the following water regimes: irregularly exposed, regularly flooded, irregularly flooded, Reef seasonally flooded, temporarily flooded, intermittently flooded, saturated, or artificially flooded. Intermittent or Definition. The Reef class includes ridgelike or moundlike intertidal channels of the Riverine System and intertidal chan- structures formed by the colonization and growth of seden- nels of the Estuarine System are classified as Streambed. tary invertebrates. Water regimes are restricted to subtidal, Description. Unconsolidated Shores are characterized by irregularly exposed, regularly flooded, and irregularly flooded. substrates lacking vegetation except for pioneering plants Description. Reefs are characterized by being elevated above that become established during brief periods when growing the surrounding substrate and interfering with normal wave conditions are favorable. Erosion and deposition by waves flow; they are primarily subtidal, but parts of some reefs may and currents produce a number of landforms such as beaches, be intertidal as well. Although corals, oysters, and tube worms bars, and flats, all of which are included in this class. Uncon- are the most visible organisms and are mainly responsible for solidated Shores may be found adjacent to Unconsolidated reef formation, other mollusks, foraminifera, coralline algae, Bottoms in all systems. As in the class Unconsolidated Bot- and other forms of life also contribute substantially to reef toms, the particle size of the substrate and the water regime growth. Frequently, reefs contain far more dead skeletal ma- are the important factors determining the types of plant and terial and shell fragments than living matter. The subclasses animal communities present. Different substrates usually sup- are Coral, Mollusk, and Worm. Only the first subclass is port characteristic invertebrate fauna. The subclasses are emphasized by C-CAP; the other two definitions are omitted. Cobble-gravel, Sand, and Mud. More detailed definitions are provided in Cowardin et al. (1979). Coral-Coral reefs are widely distributed in shallow waters of warm seas, in Hawaii, Puerto Rico, the Virgin Islands, and southern Florida. Odum (1971) characterized them as stable, Emergent Wedand well-adapted, highly diverse, and highly productive ecosys- tems with a great degree of internal symbiosis. Coral reefs lie Definition. The Emergent Wetland class is characterized by almost entirely within the Subtidal subsystem of the Marine erect, rooted, herbaceous hydrophytes (excluding mosses System, although the upper part of certain Reefs may be and lichens) which are present for most of the growing exposed. Examples of dominance types are the corals Porites, season in most years. These wetlands are usually dominated Acropora, and Montipora. The distribution of these types re- by perennial plants. All water regimes are included except flects primarily elevation, wave exposure, and reef age. subtidal and irregularly exposed. Description. In areas with relatively stable climatic condi- Rocky Shore tions, emergent wetlands maintain the same appearance year after year. In other areas, such as the prairies of the central Definition. The Rocky Shore class includes wetland environ- United States, violent climatic fluctuations cause them to ments characterized by bedrock, stones, or boulders which revert to open water in some years. Emergent wetlands are Appendix 3: C-CAP Land-Cover Classification Definitions 77 found throughout the United States and occur in all systems species include true shrubs, young trees, and trees or shrubs except Marine. Emergent wetlands are known by many names, that are small or stunted because of environmental condi- including marsh, meadow, fen, prairie pothole, and slough. tions. All water regimes except subtidal are included. Areas dominated by pioneer plants that become established Description. Scrub-shrub wetlands may represent a succes- during periods of low water are not emergent wetlands and sional stage leading to forested wetlands, or they may be should be classified as Vegetated Unconsolidated Shores or relatively stable communities. They occur only in the Estua- Vegetated Streambeds. The subclasses in the Cowardin et al. rine and Palustrine Systems but are one of the most wide- system are Persistent and Nonpersistent. spread classes in the United States. Scrub-shrub wetlands are known by many names, such as shrub swamp, shrub, bog, and pocosin. The C-CAP category includes forests composed of young trees <6 m tall regardless of potential height at maturity. Woody Wedand Broad-leaved. Deciduous--In Estuarine System wetlands the predominant deciduous and broad-leaved trees or shrubs The Woody Wetland class includes any species with an aerial are plants such as sea-myrtle, Bacchafis halimifolia, and marsh stem that persists for more than one season. The Woody elder, Iva frutescens. In the Palustrine System typical domi- Wetland class is divided into three subclasses: Deciduous, nance types are alders, Alnus spp., willows, Salix spp., button- Evergreen, and Mixed. bush, Cephalanthus cocidentalis, red osier dogwood, Cornus stolonifera, honeycup, Zenobia pulverulenta, spirea, Spiraea Deciduous-The Deciduous WoodyWeaand subclass includes douglasii, bog birch, Betula pumila, and young trees of species all wetland forest and shrub areas having a predominance of such as red maple, Acer rubrum, or black spruce, Picea mariana. trees and shrubs that lose their leaves or needles at the end of Needle-leaved Deciduous-This group, consisting of wet- the frost-free season or at the beginning of a dry season. This lands where trees or shrubs are predominantly deciduous category contains greater than two-thirds deciduous trees and needle-leaved, is represented by young or stunted trees and shrubs. The Deciduous Woody Wetland category can be such as tamarack or bald cypress, Taxodium dislichum. divided into three categories: Forest, Scrub-Shrub, and Dead. Dead-Definition. Dominated by dead deciduous woody Forest-Definition. Forested wetland is characterized by vegetation, these wetlands are usually produced by a pro- woody vegetation 2:6 m in height. All water regimes are in- longed rise in the water table resulting from impoundment cluded except subtidal. of water by landslides, human activity, or beaver activity. Such Description. Forested wetlands are most common in the wetlands may also result from various other factors such as eastern United States and in those sections of the West where fire, salt spray, insect infestation, air pollution, and moisture is relatively abundant, particularly along rivers and herbicides. in mountains. They occur only in the Palustrine and Estua- rine Systems and normally contain an overstory of trees, an Evergreen-The Evergreen Woody Wetland subclass contains understory of young trees or shrubs, and a herbaceous layer. wedand forests and shrubs that do not lose their leaves or Forested wetlands in the Estuarine System, including the needles at the end of a frost-free season or at the beginning of a mangrove forests of Florida, Puerto Rico, and the Virgin dry season. The Evergreen Woody Wetland category is subdi- Islands, are known by such names as swamps, hammocks, vided into two additional categories: Forest and Scrub-Shrub. heads, and bottoms. These names often occur in combina- Forest-Definition. Forested wetland is characterized by tion with species names or plant associations, such as cedar woody vegetation @!6 m in height. All water regimes are in- swamp or bottom land hardwoods. cluded except subtidal. Broad-leaved Deciduous-Dominant trees typical of broad- Broad-leaved Evergreen-In the Southeast, broad-leaved leaved deciduous wetlands, which are represented through- evergreen wetlands reach their greatest development. Red out the United States, are most common in the South and bay, Persea borbonia, loblolly bay, Gordonia lasianthus, and sweet East. Common dominants are species such as red maple, bay, Magnolia virginiana, are prevalent, especially on organic American elm, Ulmus americana, the ashes Fraxinus soils. This group also includes red mangrove, black man- pennsylvanica and F. nigTa, black gum, Nyssa sylvatica, tupelo grove, Avicennia germinans, and white mangrove, Langunculayia gum, N. aquatica, swamp white oak, Quercus bicolor, overcup racemosa, which are adapted to varying levels of salinity. oak, Q lyrata, and basket oak, Q michauxii. These wetlands Needle-leaved Evergreen-Black spruce, growing on or- generally occur on mineral soils or highly decomposed or- ganic soils, represents a major dominant of the Needle-leaved ganic soils. Evergreen subclass in the north. Though black spruce is Needle-leaved Deciduous--The southern representative of common on nutrient-poor soils, northern white cedar, Thuja the Needle-leaved Deciduous Wetland subclass is bald cy- occidentalis, dominates northern wetlands on more nutrient- press, Taxodium distichum, which is noted for its ability to rich sites. Along the Atlantic Coast, Atlantic white cedar, tolerate long periods of surface inundation. Tamarack is Chamaecypaiis thyoides, is one of the most common dominants characteristic of the boreal forest region, where it occurs as a on organic soils. Pond pine, Pinus serotina, is a common dominant on organic soils. Relatively few other species are necdle-leaved evergreen found in the Southeast in associa- included in this subclass. tion with dense stands of broad-leaved evergreen and decidu- ousshrubs. Scrub-Shrub-Definition. Scrub-Shrub wetland includes ar- Scrub-Shrub-Definition. Scrub-shrub wetland includes eas dominated by woody vegetation <6 m (20 ft) tall. The areas dominated by woody vegetation <6 m (20 ft) tall. The 78 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program species include true shrubs, young trees, and trees or shrubs an area, the land is classified as Mixed Woody. The Mixed that are small or stunted because of environmental condi- Woody category is subdivided into two additional categories: tions. All water regimes except subtidal are included. Forest and Scrub/Shrub. Broad-leaved Evergreen-In the Estuarine System, vast wet- Forest-This category includes all forested areas where land areas are dominated by mangroves (Rhizophora mangle, both evergreen and deciduous trees are growing and neither Languncularia racemosa, Conocarpus erectus, and Avicennia predominate. genninans) that are I m to <6 m tall. In the Palustrine System, Scrub/Shrub-This category includes all shrub areas where the broad-leaved evergreen species are typically found on both evergreen and deciduous shrubs are growing and nei- organic soils. Northern representatives are labrador tea, Ledum ther predominate. groenlandicum, bog rosemary, Andromeda glaucophylla, bog lau- Dead-Wetland areas dominated by dead mixed woody rel, Kalmia polifolia, and the semi-evergreen leatherleaf, vegetation are, like dead deciduous woody wetlands, most Chamaedaphne calyculata. In the south, fetterbush, Lyonia lu- common in, or around the edges of, impoundments and cida, coastal sweetbells, Leucothoe axillaris, inkberry, Ilex glabra, beaver ponds. The same factors that produce dead deciduous and the semi-evergreen black ti-ti, Cyyilla racemiflora, are char- woody wetlands produce dead mixed woody wetlands. acteristic broad-leaved evergreen species. Needle-leaved Evergreen-The dominant species in needle- leaved evergreen wetlands are young or stunted trees such as Water black spruce or pond pine. Cowardin et al. (1979) define deepwater habitats as perma- Dead-Definition. These wetland areas are dominated by nently flooded lands lying below the deepwater boundary of dead evergreen woody vegetation. Like dead deciduous woody wetlands. Deepwater habitats include environments where wetlands, they are most common in, or around the edges of, surface water is permanent and often deep, so that water, rather impoundments and beaver ponds. The same factors that than air, is the principal medium within which the dominant produce dead deciduous woody wetlands produce dead ever- organisms live, whether or not they are attached to the substrate. green woody wetlands. As in wetlands, the dominant plants are hydrophytes. However, Mixed-The Mixed Woody Wetland subclass includes all the substrates are considered nonsoil because the water is too forest and shrub wetland areas where both evergreen and deep to support emergent vegetation. The class Water includes deciduous trees and shrubs grow and neither predominates. Marine/Estuarine, Lacustrine, Palustrine, and Riverine When evergreen and deciduous species each occupy2!33% of Deepwater subclasses as defined by Cowardin et al. (1979). Appendix 4 GCAP Wo?*shops This guidance document results from the participation of mend protocols for accuracy assessment of C-CAP TM data. more than 200 scientists, technical specialists, managers, and The accuracy assessment procedures outlined in this report regional experts in nine protocol-development workshops. Four are a direct result of those workshops. The classification regional workshops were held in the Northeast, Southeast, West workshop was a multi-agency group organized to develop a Coast, and Great Lakes regions to address a broad range of classification system that would suit C-CA-P needs. Through issues including those specific to each major coastal region. the workshop and a long iterative process thereafter, the Also, a major national workshop was held that focused specifi- classification presented in this report was developed. The cally on mapping submerged aquatic vegetation. Participation Maryland Field Reconnaissance workshop was unique in that was encour-aged across all Federal and State agencies involved in the organizers were not from C-CAP but from the Maryland coastal research, management, and policy and many other agen- Department of Natural Resources and Salisbury State Univer- cies concernedwith remote sensing and land-cover analysis. The sity. The principal objective of the workshop was to provide seagrass workshop followed the same for-mat, focusing specifi- recommendations concerning C-CAP products. The recon- cally on submersed habitats. Each workshop was presented with naissance consisted of field visits to sites in the vicinity of a draft C-CAP protocol, based initially on the Chesapeake Bay Salisbury, Maryland, identified in a preliminary version of the prototype, and participants were encouraged to refine the draft C-CAP Chesapeake Bay Land-Cover Change Database. and resolve remaining issues. Issues not resolved in the five In addition, the preliminary C-CAP data were compared major workshops were addressed through dedicated topical with other types of ancillary data supplied by workshop workshops. Finally, the issues that could not be resolved through participants. workshops were explored through research funding proposals., Findings and recommendations from these workshops and Two accuracy assessment workshops involved leading spe- from other meetings of specialists (not listed here) were cialists in spatial error estimation who were asked to recom- crucial in the development of this document. Southeast Regional Workshop Classification Scheme Workshop Location: University of South Carolina Location: Silver Spring, MD Columbia, SC Date: 12 February 1991 Dates: 29-31 May 1990 Co-Chairs: James Johnston Host: University of South Carolina Vic Klemas Co-Chairs: Jerome Dobson Kenneth Haddad West Coast Regional Workshop Seagrass Mapping Workshop Location: Embassy Suites Hotel Location: Embassy Suites, Tampa Airport Hotel Seattle, WA Tampa, FL Dates: 29 April-1 May 1991 Dates: 23-25 July 1990 Co-Chairs: Jerome Dobson Co-Chairs: Randolph Ferguson Kenneth Haddad Robert Orth Accuracy Assessment Workshop Maryland Field Reconnaissance Workshop Location: National Marine Fisheries Service Location: Salisbury State University Beaufort Laboratory Salisbury, MD Beaufort, NC Dates: 16-18july 1991 Date: 25 September 1990 Host: Salisbury State University Host: Beaufort Laboratory Workshop William Burgess, Maryland Department of Chair: Jerome Dobson Design: Natural Resources; Edward W. Christoffers, NOAA/NMFS Northeast Regional Workshop Jerome Dobson, Oak Ridge National Laboratory Location: Whispering Pines Conference Center Randolph Ferguson, NOAA/NMFS W. Altonjones Campus Adam Fisch, Virginia Council on the University of Rhode Island Environment Kingston, RI K. Peter Lade, Salisbury State University Dates: 8-10january 1991 James Thomas, NOAA/NMFS Host: University of Rhode Island Bill Wilen, USFWS, National Wetlands Inventory. Co-Chairs: Jerome Dobson Sponsors:, Salisbury State University, Maryland Department Kenneth Haddad of Natural Resources, NOAA, and USFWS 79 80 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program Accuracy Assessment Workshop Great Lakes Regional Workshop Location: Oak Ridge National Laboratory Location: Best Western Ann Arbor Regent Oak Ridge, TN Ann Arbor, MI Date: 1-2 August 1991 Dates: 19-27 August 1991 Host: Oak Ridge National Laboratory Host: NOAA Great Lakes Environmental Research Lab Chair: Jerome Dobson Co-Chairs: Jerome Dobson Kenneth Haddad Appendix 5 C-C4P Notocol Development Research C-CAP funded research to refine various aspects of the proto- University of South Carolina col based on workshop recommendations and on findings Determine the impact of tides on coastal change detection from the upland and wetland prototype (Chesapeake Bay), the water and submerged land prototype (North Carolina North Carolina State University Coast), and the Salisbury field experience. These research Develop methodologies for accuracy assessment for change projects are intended to increase the geographical coverage detection databases of the C-CAP change detection database. The following is a list of the institutions that performed the research and the University of Virginia topics addressed in fiscal year 1991: Examine influence of tides on TM data with the aid of digital elevation models University of South Carolina � Test change detection methodologies University of Maine � Identify optimum pattern recognition algorithms Develop improved methodologies for detecting forested wetlands North Carolina State University Develop a seamless database from two independently de- University of New Hampshire veloped land-cover databases derived from TM data Develop methodologies for accuracy assessment of change Universities of Rhode Island and Connecticut detection databases � Test change detection methodology � Test the use of available digital wetlands data as an aid for Two other studies were also funded by C-CAP in fiscal year classifying TM imagery 1992: Beaufort, NMFS Universities of Rhode Island and Connecticut o Develop change detection methodologies for SAV 9 Funded for six months to finish work started in 1991 An announcement of availability of funds for protocol devel- Beaufort, NMFS opment research was distributed in March 1992. The follow- Develop change detection methodologies for SAV using ing five studies were funded: CPS technology. 82 NOAA Technical Report NMFS 123: Dobson et aL: Coastal Change Analysis Program Appendix 6 Workshop Pat-ficipants Mr. Bob Ackerman Thomas E Bigford MD, DNR, FPWS NOAA/National Marine Fisheries Service Chesapeake Bay Programs (Forestry) Habitat Conservation Branch Tawes State Office Building One Blackburn Drive Annapolis, MD 21401 Gloucester, MA 01930-2298 Steve Ackleson Elaine Blok Bigelow Laboratory for Ocean Science Geonex Martel, Inc. McKown Point 8950 9th Street North West Boothbay Harbor, ME 04575 St. Petersburg, FL 33703 Dean Albro Nate Boyer Division of Freshwater Wetlands EOSAT Rhode Island Dept. of Environmental Management 4300 Forbes Blvd. 291 Promenade St. Lanham, MD 20706 Providence, RI 02908 Earl Bradley Warren Alward Coastal Resources Division Dept. of Natural Resources Tawes State Office Bldg. B-3 Land and Water Management Division Annapolis, MD 21401 P.O. Box 30028 Lansing, MI 48909 James Brewer USDA,SCS Roy A. Armstrong 339 Revell Highway NASA, Ames Research Center Annapolis, MD 21401 37250 Sequoia Terrace #1032 Freemont, CA 94536 Douglas A. Bulthuis Padilla Bay National Estuarine Peter V. August Research Reserve Environmental Data Center 1043 Bayview-Edison Road Dept. of Natural Resources Science Mount Vernon, WA 98273 University of Rhode Island Kingston, RI 02881 Bill Burgess Maryland Water Resources Admin. John Banta Tawes State Office Building, D-2 Director of Planning Annapolis, MD 21401 Adirondack Park Agency Ray Brook, NY 12977 Alice Chalmers University of Georgia Franklin S. Baxter Marine Institute U.S. Geological Survey Sapelo Island, GA 31327 Mail Stop 514 National Center Michael Chambers Reston, VA 22092 USGS, National Mapping Division 590 National Center A] Beeton Reston, VA 22092 Director Great Lakes Enviromental Heather Cheshire Research Lab/NOAA Computer Graphics Center 2205 Commonwealth Box 7106, North Carolina State University Ann Arbor, MI 48105 Raleigh, NC 27695 83 84 NOAA Technical Report NMYS 123: Dobson et al.: Coastal Change Analysis Program Alexander J. Chester Joseph E. Costa NOAA/National Marine Fisheries Service Massachusetts Coastal Zone Management Beaufort Laboratory Buzzards Bay Project Beaufort, NC 28516-9722 2 Spring Street Marion, MA 02738 Nicholas Chrisman Dept. of Geography Paul Crawford University of Washington Olympic National Park Smith Hall/DP-10 National Park Service Seattle, WA 98195 600 East Park Avenue Port Angeles, WA 98362 Eric Christensen Science Technology Laboratory Ford A Cross, Director Lockheed 1210 NOAA/National Marine Fisheries Service Stennis Space Center, MS 39529 Southeast Fisheries Science Center Beaufort Laboratory Edward W. Christoffers; 101 Pivers Island Road NOAA/CBP Science Coordinator Beaufort, NC 28516-9722 Tawes State Office Building, C-4 Annapolis, MD 21401 Pat Cummens Division of Science and Research Barbara Cintron GIS Laboratory Puerto Rico Dept. of Natural Resources 401 East State Street P.O. Box 5887 Trenton, NJ 08625 Puerta de Pierra, PR 00906 Thomas E. Dahl Daniel Civco National Wetlands Inventory Dept. of Natural Resources U.S. Fish and Wildlife Service Management and Engineering, Box U-87 9720 Executive Center Drive 1376 Storrs Road Suite 101, Monroe Building University of Connecticut St. Petersburg, FL 33702 Storrs, CT 06269-4087 Barbara D'Angelo Elaine Collins 3WMOO U.S. EPA Region III NOAA/NESDIS 841 Chestnut Building 1825 Connecticut Ave., N.W., Room 406 Philadelphia, PA 19107 Washington, DC 20235 Rick Dawson Commander U.S. Department of the Interior U.S. Army Corps of Engineers National Park Service Waterways Experiment Station Southeast Regional Office Attn: CEWES-ER-W/Buddy Clarain 75 Spring Street SW Vicksburg, MS 39180-6199 Atlanta, GA 30303 Russell G. Congalton Michael DeMers Dept. of Natural Resources New Mexico State University 215 James Hall Dept. of Geography University of New Hampshire Box MAP Durham, NH 03824 Las Cruces, NM 88003 Robert Costanza Larry Deysher Chesapeake Biological Laboratory Coastal Resources Assoc. Solomons Island, MD 20688-0038 2270 Camino Vina Roble Suite L Dave Cowen Carlsbad, CA 92009 Statistical and Behavioral Sciences Laboratory University of South Carolina Columbia, SC 29208 Appendix 6: Workshop Participants 85 Jerome E. Dobson Don Field Oak Ridge National Laboratory NOAA/National Marine Fisheries Service Geographic Information Systems Southeast Fisheries Science Center and Computer Modelling Beaufort Laboratory P.O. Box 2008, 4500 N, MS 6237 101 Pivers Island Road Oak Ridge, TN 37831 Beaufort, NC 28516-9722 Bill Dunn J. Michael Flagg USDA SCS Virginia Dept. of Conservation and Recreation 339 Revell Highway Division of Soil and Water Annapolis, MD 21401 203 Governor St., Suite 206 Richmond, VA 23219 Sandy Wyllie Echeverria Institute of Marine Science Ms. Bellory Fong University of Alaska California Dept of Water Resources Fairbanks, AK 99775 3251 "S" Street (RM B-5) Sacramento, CA 95816 Robert Emmett NOAA/National Marine Fisheries Service Andrew Frank Pt. Adams Biological Field Station Dept. of Civil Engineering P. 0. Box 155 119 Boardman Hall Hammond, OR 97121 University of Maine Orono, ME 04469 William Enslin Center for Remote Sensing Adam Frisch Michigan State University Virginia Council on the Environment 302 Berkey Hall 202 N. 9th Street, Suite 900 East Lansing, MI 48824-1111 Richmond, VA 23219 Ron Erickson Ellen Fritts U.S. Fish and Wildlife Service Alaska Dept. of Fish and Game 4101 E. 80th Street Habitat Division Bloomington, MN 55425-1600 P.O. Box 3-2000 1255 West 8th Street Maggie Ernst Juneau, AK 99802-2000 NOAA/Coastal Ocean Program Office 1315 East-West Highway Gregory Fromm Silver Spring, MD 20910 NOAA/National Ocean Service Photogrammetry Branch Tamra Faris 1315 East-West Highway NOAA/National Marine Fisheries Service Silver Spring, MD 20910 P.O. Box 21668 Juneau, AK 99802 Carolyn Gates California Coastal Commission Robin Fegeas 45 Frcemont St. U.S. Geological Survey Suite 2000 Mail Stop 521 San Francisco, CA 94105 National Center Reston, VA 22092 Len Gaydos Ames Research Center Randolph L. Ferguson P.O. Box 1000 NOAA/National Marine Fisheries Service Moffett Field, CA 94035-1000 Southeast Fisheries Science Center Beaufort Laboratory Bess Gillelan 101 Pivers Island Road Chief Beaufort, NC 28516-9722 NOAA/Chesapeake Bay Office 410 Severn Me, Suite 107A Annapolis, MD 21403 86 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program Frank Golet Michael E. Hodgson Dept. of Natural Resources Science Oak Ridge National Laboratory University of Rhode Island Geographic Information Systems and Computer Modelling Kingston, RI 02881 P.O. Box 2008, 4500 N. MS6274 Oak Ridge, TN 37831 Al Green Texas Parks and Wildlife Dept. Sherman Hollander 4200 Smith School Road Michigan Resource Information Program Austin, TX 78745 P.O. Box 30028 Lansing, MI 48909 Christie Guy Assateaque Island National Seashore Bob Holman Rt 611-7206 WRRI National Seashore Lane Box 7912 Berlin, MD 21811 North Carolina State University Raleigh, NC 27695-7912 Kenneth D. Haddad Marine Research Institute Frank Horvath FL Dept. of Environmental Protection Chief Scientist 100 8th. Avenue SE Great Lakes Information System St. Petersburg, FL 33701-5059 Michigan Dept. of Natural Resources Land and Water Management Division Anne Marie Hale Miglarese Lansing, MI 48909 South Carolina Water Resources Commission 1201 Main Street, Suite 1100 Karl Huber Columbia, SC 29201 Virginia Dept. of Conservation and Recreation Division of Soil & Water Larry Handley 203 Governor Street, Suite 206 U.S. Fish and Wildlife Service Richmond, VA 23219 National Wetlands Research Center 700 Cajun Dome Blvd. Kent Hughes Lafayette, LA 70506 Dept. of Commerce, NOAA/NESDIS Federal Office Building 41069 Bud Harr-is Washington, DC 20233 University of Wisconsin at Green Bay 2420 Nicolet Drive Paul G. Huray Green Bay, WI 54311 106 Osborne Administration Building University of South Carolina Greg Hellyer Columbia, SC 29208 U.S. Environmental Protection Agency Region 1, WWP 424 Merton Ingham JFK Federal Building NOAA/National Marine Fisheries Service Boston, MA 02203 Northeast Fisheries Science Center Narragansett Laboratory Gary Hendrix 28 Tarzwell Drive U.S. Department of the Interior Narragansett, RI 02882-1199 National Park Service Room 1092, Sciences and Natural Resources Harry Iredale 75 Spring Street SW NOAA/NESDIS Atlanta, GA 30303 Universal Building 1825 Connecticut Avenue, NW Rex C. Herron Washington, DC 20235 NOAA/National Marine Fisheries Service Building 1103 EugeneJaworski Stennis Space Center, MS 39529 SER-GEM Center 3075 Washtenaw Avenue Carl Hershner Ypsilanti, MI 48197 Virginia Institute of Marine Science College of William and Mary Gloucester Point, VA 23062 1 Appendix 6: Workshop Participants 87 John Jensen Vic Klemas Dept. of Geography Director University of South Carolina Center for Remote Sensing Columbia, SC 29208 College of Marine Studies University of Delaware Lisajipping Newark, DL 19716 U.S. Army Corps of Engineers Great Lakes Hydraulics and Hydrology Branch Frederick Kopfler P.O. Box 1027 U.S. Environmental Protection Agency Detroit, MI 48231 Gulf of Mexico Program Bldg. 1103 JimmyJohnston Stennis Space Center, MS 39529 National Biological Survey National Wetlands Research Center K Koski 700 Cajun Dome Blvd. NOAA/National Marine Fisheries Service Lafayette, LA 70506 Auke Bay Lab 11305 Glacier Highway James R. Karr Auke Bay, AK 99821 Institute for Environmental Studies Engineering Annex, FM-12 Rose Kress University of Washington U.S. Army Corps of Engineers Seattle, WA 98195 Waterways Experiment Station 3909 Halls Ferry Road Dick Kelly Vicksburg, MS 39180 Maine State Planning Office 184 State Street Tom Kunneke Augusta, ME 04333 Geonex Martel, Inc. 8950 9th Street North Dick Kempka St. Petersburg, FL 33703 Ducks Unlimited, Inc. 9823 Old Winery Place, #16 Kathy Kunz Sacramento, CA 95827 U.S. Army Corps of Engineers Seattle District Siamak Khorram P. 0. C-3755 Computer Graphics Center Seattle, WA 98124-2255 5112jordan Hall, Box 7106 North Carolina State University Charles LaBash Raleigh, NC 27695-7106 Environmental Data Center Dept. of Natural Resources Science Donley Kisner University of Rhode Island Bionetics Corp. Kingston, RI 02881 P.O. Box 1575 VHFS K.-Peter Lade Warrenton, VA 22186 Salisbury State University Image Processing Center Bjorn Kjerfve I 10 Power Street Baruch Institute Salisbury, MD 21801 University of South Carolina Columbia, SC 29208 Rene Langi Dept. of Biology Richard Kleckner San Diego State University U.S. Geological Survey San Diego, CA 92182 Mail Stop 590 National Center Lewis A. Lapine Reston, VA 22092 NOAA, National Ocean Service Photogrammetry Branch, N/CG23 1315 East-West Highway Silver Spring, MD 20910 88 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program M. Larson Carl Markon No Address U.S. Geological Survey 4230 University Drive George Leshkevich Anchorage, AK 99508-4664 Great Lakes Environmental Res. Lab. 2205 Commonwealth Blvd Garry Mayer Ann Arbor, MI 48105-1593 NMFS Restoration Center 1315 East-West Highway Roy R. Lewis, III Silver Spring, MD 20910 Lewis Environmental Services, Inc. P.O. Box 20005 L. Nelson Mayjr. Tampa, FL 33622-0005 Coastal Fisheries Institute Louisiana State University Nancy Liebowitz Baton Rouge, LA 70803 U.S. Environmental Protection Agency 1600 S.W. Western Blvd. Bruce McCarnmon Corvallis, OR 97333 USDA/USFW Ecology Range and Watershed Katherine Lins P.O. Box 3623 U.S. Geological Survey Portland, OR 97208 Mail Stop 512 National Center Keith McLaughlin Reston, VA 22092 1720 Peachtree Road, N.W. Atlanta, CA 30367 Robyn Loudermilk Hawaii Office of State Planning Peter McRoy State Capitol - Room 406 Institute of Marine Science Honolulu, HI 96813 University of Alaska Fairbanks, AK Ross Lunetta or U.S. Environmental Protection Agency Dept of Botany Environmental Monitoring System University of Hawaii Laboratory 3190 Mailae Way P.O. Box 93478 Honolulu, HI 96822 Las Vegas, NV 89193-3478 Todd Mecklenberg David Lusch Geonex Martel, Inc. Center For Remote Sensing 8950 9th Street North 302 Berkey Hall, Michigan State University St. Petersburg, FL 33703 East Lansing, MI 48824-1111 Norman Melvin Anne Lynn Plant Ecologist State Biologist USDA\SCS USDA SCS 339 Revell Highway 339 Revell Highway Annapolis, MD 21401 Annapolis, MD 21401 Robert E. Menzer Andreas Magerjr. Director NOAA/National Marine Fisheries Service U.S. Environmental Protection Agency Southeast Region Environmental Research Laboratory 9450 Koger Blvd. Sabine Island St. Petersburg, FL 33702 Gulf Breeze, FL 32561-5299 Leslie Manfull Carolyn Merry National Park Service Dept of Civil Engineering ADNR 470 470 Hitchcock Hall P.O. Box 37127 Ohio State University Washington, DC 20013-7127 2070 Neil Avenue Columbus, OH 43210-1275 Appendix 6: Workshop Participants 89 William Michner Robert Orth Baruch Institute Virginia Institute of Marine Science University of South Carolina The College of William and Mary Columbia, SC 29208 Gloucester Point, VA 23062 Chris Mobley Jacquelyn Ott NOAA/National Marine Fisheries Service ERIM-Room C148 777 Sonoma Avenue - Room 325 Box 8618 Santa Rosa, CA 95404 Ann Arbor, MI 48107 John Mooney Keith Patterson NOAA/National Ocean Service Geonex Martel, Inc. Photogrammetry Branch, N/CG2 8950 9th Street, North 1315 East-West Highway St. Petersburg, FL 33703 Silver Spring, MD 20910 Brian Pawlak Tom Mumford NOAA/National Marine Fisheries Service Division of Aquatic Lands Southeast Fisheries Science Center Washington Dept. of Natural Beaufort Laboratory Resources, EX12 101 Pivers Island Rd. Olympia, WA 98504 Beaufort, NC 28516-9722 Ron Myszka Ed Pendleton U.S. Environmental Protection Agency U.S. Fish and Wildlife Service Great Lakes National Program Office (SGB) 180 Admiral Cochran Drive, Suite 535 230 South Dearborne Street Annapolis, MD 21401 Chicago, IL 60604 Robert Peplies Scott Nixon Dept. of Geography Rhode Island Sea Grant Program East Tennessee State University Graduate School of Oceanography P.O. Box 22870A Narragansett, RI 02882-1197 Johnson City, TN 37614 Douglas Norton Dennis Peters U.S. Environmental Protection Agency U.S. Fish and Wildlife Service WH553, Room E-7,431) NWI Regional Wetland Coordinator Washington, DC 20460 Eastside Federal Complex 911 NE Ilth Ave. Richard P. Novitzki Portland, OR 97232-4181 U.S. Environmental Protection Agency 1600 S.W. Western Blvd. Ron Phillips Corvallis, OR 97333 Beak Consultants, Inc. 12931 NE 126th Place Charles E. Olsonjr. Kirkland, WA 98034-7716 School of Natural Resources University of Michigan Steve Phillips Samuel Trask Dana Building Division of Forestry 430 East University Street Alaska Dept. of Natural Resources Ann Arbor, MI 48109-1115 P.O. Box 107005 Anchorage, AK 99510-7005 Chris Onuf U.S. Fish and Wildlife Service Larry Pomatto National Wetlands Research Center Wetlands and Aquatic Protection Campus Box 339, CCSU DNREC 6300 Ocean Drive P.O. Box 1401 Corpus Christi, TX 78412 Dover, DE 19903 90 NOAA Technical Report NMFS 123:. Dobson et al.: Coastal Change Analysis Program John Posenau Tony Reyer Virginia Institute of Marine Science NOAA/National Ocean Service College of William and Mary Strategic Environmental Assessments Div. Gloucester Point, VA 23062 1305 East West Highway SSMC-4, 9th Floor Sandy Prisloe Silver Spring, MD 20910 Natural Resources Center Dept. of Environmental Protection Steve Robb Room 545 Salisbury State University State Office Building Image Processing Center Hartford, CT 06106 110 Power Street Salisbury, MD 21801 Jane Provancha NASA Robert Rolley Biomedical and Environmental Laboratories Division of Fish and Wildlife Mail Code BIO-2 300 West Ist Street Kennedy Space Center, FL 32899 Bloomington, IN 47403 Warren Pulich Jr. Charles Roman Texas Parks and Wildlife Dept. U.S. Department of the Interior Resource Protection Division National Park Service 4200 Smith School Road Cooperative Research Unit Austin, TX 78745 Graduate School of Oceanography University of Rhode Island Charles Racine Narragansett, RI 02882-1197 U.S. Army Corps of Engineers Cold Regions Research and Lynn Sampson Engineering Laboratory USDA/Soil Conservation Service 72 Lyme Road Manley Miles Building Hanover, NH 03755 1405 South Harrison East Lansing, MI 48823 David Rackley NOAA/National Marine Fisheries Service Don Scavia, Director Southeast Region NOAA/Coastal Ocean Program Office 9450 Koger Blvd. 1315 East-West Highway St. Petersburg, FL 33702 Silver Spring, MD 20910 Ann Rasberry Charles Simenstad Programmer/Analyst Wetland Ecosystem Team Forest Park and Wildlife Service Fisheries Research Institute Tawes State Office Bldg. B-2 University of Washington Annapolis, MD 21401 Seattle, WA 98195 Jill Reichert Tim Schlagenhaft Maryland Water Resources Admin. Minnesota Dept. of Natural Resources Tawes State Office Bldg. D-2 Section of Fisheries Annapolis, MD 21401 Box 12 500 Lafayette Road Kenneth R. Reisinger St. Paul, MN 55155 Div. of Rivers and Wetlands Conservation Bureau of Water Resources Management Peter F. Sheridan P.O. Box 8761 NOAA/National Marine Fisheries Service Harrisburg, PA 17105-8761 4700 Avenue U Galveston, TX 77551-5997 Becky Ritter Ducks Unlimited, Inc. One Waterfowl Way Long Grove, IL 60047 Appendix 6: Workshop Participants 91 Fred Short Kevin Summers Jackson Estuarine Laboratory U.S. Environmental Protection Agency University of New Hampshire EMA-P-NC Gulf Province Manager RFD 2 Gulf Breeze Environmental Research Manager Adams Point Gulf Breeze, FL 32561-5299 Durham, NH 03824 John Sutherland Karen Siderelis NOAA/National Sea Grant College Program North Carolina Center for Geographic R/ORI Information and Analysis 1335 East-West Highway North Carolinia Dept. of Environment, Health Silver Spring, MD 20910 and Natural Resources Raleigh, NC 27611-7687 Billy Teels, National Biologist USDA Soil Conservation Service Gene Silberhorn South Ag Bldg. Virginia Institute of Marine Science P.O. Box 2890 The College of William and Mary Room 6144 Gloucester Point, VA 23062 Washington, DC 20013 Jerry Spratt Michael Thabault California Dept. of Fish and Game NOAA/National Marine Fisheries Service 2201 Garden Road 777 Sonoma Avenue, Rm. 325 Monterey, CA 93940 Santa Rosa, CA 95404 Louisa Squires James P. Thomas U.S. Environmental Protection Agency NOAA/National Marine Fisheries Service 1600 SW Western Blvd. Restoration Center Corvallis, OR 97333 Office of Protected Resources, F/PR5 1315 East-West Highway Kristen Stout Silver Spring, MD 20910 Bionetics Corp. P.O. Box 1575 (VHFS) Greg Tilley Warrenton, VA 22186 Maryland Water Resources Admin. Tawes State Office Bldg. D-2 Judy Stout Annapolis, MD 21401 University of South Alabama Dauphine Island Sea Laboratory Ralph Tiner Marine Environmental Sciences Consortium U.S. Fish and Wildlife Service P.O. Box 369 NWI Regional Wetland Coodinator Dauphine Island, AL 36528 300 Westgate Center Drive Route 2 & 116 North Larry Strong Hadley, MA 01035 U.S. Fish and Wildlife Service Northern Prairie Wildlife Research Center Tom Tiner Route 1, Box 96C Salisbury State University Jamestown, ND 58401 Image Processing Center 110 Power Street Richard P. Stumph Salisbury, MD 21801 USGS Center for Coastal Geology 600 4th Street South Bill Tippets St. Petersburg, FL 33701 California Dept. of Parks & Recreation 8885 Rio San Diego Drive Michael Sullivan Suite 270 Mississippi State University San Diego, CA 92108 Dept. of Biological Sciences P.O. Iirawer GY Robert Virnstein Mississippi State, MS 39762-5759 St. Johns River Water Management District P.O. Box 1429 Palatka, FL 32178 92 NOAA Technical Report NMFS 123: Dobson et al.: Coastal Change Analysis Program Joel Wagner Donald Williams National Park Service - Air c/o Army Corps of Engineers 12795 W. Alameda Planning Division Lakewood, CO 80225 P.O. Box 1027 Detroit, MI 48231 Carl Weaver MDE, Chesapeake Bay and Ken Winterberger Special Programs United States Forest Service 2500 Broening Highway Forestry Science Laboratory Baltimore, MD 21224 201 E. 9th Avenue Suite 303 David Weaver Anchorage, AK 99501 Massachusetts Office of Environ. Affairs GIS Division Lisa Wood 20 Somerset St., 3rd Floor NOAA/National Marine Fisheries Service Boston, MA 02108 Southeast Fisheries Science Center Beaufort Laboratory Jeff Weber 101 Pivers Island Road Oregon Coastal Program Beaufort, NC 28516-9722 Dept. of Land Conservation & Development 320 SW Stark, Rm 530 Jay Zieman Portland, OR 97204 University of Virginia Dept. of Environmental Sciences Fred Weinmann Clark Hall U.S. Environmental Protection Agency Charlottesville, VA 22903 (WD-138) 1200 6th Avenue Kevin Zytkovicz Seattle, WA 98101 Minnesota Dept. of Natural Resources Ecological Services, Box 25 jennette Whipple 500 Lafayette Road NOAA/National Marine Fisheries Service St. Paul, MN 55155 Tiburon Laboratory 3150 Paradise Dr. Tiburon, CA 94920 Bill Wilen National Wetlands Inventory U.S. Fish and Wildlife Service 400 ARL - SQ l8th and C Streets, NW Washington, DC 20240 NOAA TECHNICAL REPORTS NMFS The major responsibilities of the National Marine Fisheries Service (NMFS) are to monitor and assess the abundance and geographic distribution of fishery resources, to understand and predict fluctuations in the quantity and distribution of these resources, and to establish levels for their optimum use. NMFS is also charged with the development and implemen- tation of policies for managing national fishing grounds, with the development and enforcement of domestic fisheries regulations, with the surveillance of foreign fishing off U.S. coastal waters, and with the development and enforcement of international fishery agreements and policies. NMFS also assists the fishing industry through marketing services and economic analysis programs and through mortgage insurance and vessel construction subsidies. It collects, analyzes, and publishes statistics on various phases of the industry. Recently Published NOAA Technical Reports NMFS 111. Control of disease in aquaculture: proceed- California species of rockfishes (Scorpaenidae: ings of the nineteenth U.S.-Japan meeting on aquacul- Sebastes) from rearing studies, by Guillermo Moreno. ture; Ise, Mie Prefecture, Japan, 29-30 October 1990, November 1993, 18 p. edited by Ralph S. Svrjcek. October 1992, 143 p. 117. Distribution, abundance, and biological char- 112. Variability of temperature and salinity in the acteristics of groundfish off the coast of Washington, Middle Atlantic Bight and Gulf of Maine, by Robert L. Oregon, and California, 1977-1986, by Thomas A. Dark Benway, Jack W. Jossi, Kevin P. Thomas, and Julien R. and Mark E. Wilkins. May 1994, 73 p. Goulet. April 1993, 108 p. 118. Pictorial guide to the groupers (Teleostei: Ser- 113. Maturation of nineteen species of finfish off the ranidae) of the western North Atlantic, by Mark Grace, northeast coast of the United States, 1985-1990, by Kevin R. Rademacher, and Mike Russell. May 1994, 46 p. Loretta O'Brienjay Burnett, and Ralph K. Mayo. June 1993, 66 p. 119. Stocks of dolphins (Stenella spp. and Del- 114. Structure and historical changes in the phinus delphis) in the eastern tropical Pacific: a groundfish complex of the eastern Bering Sea, by phylogeographic classification, by Andrew E. Dizon, Wil- Richard G. Bakkala. July 1993, 91 p. liam F. Perrin, and Priscilla A. Akin. June 1994, 20 p. 115. Conservation biology of elasmobranchs, edited 120. Abundance and distribution of ich- by Steven Branstetter. September 1993, 99 p. thyoplankton along an inshore-offshore transect in Ouslow Bay, North Carolina, by Allyn B. Powell and 116. Description of early larvae of four northern Roger E. Robbins. June 1994, 28 p. Copyright Law Although the contents of there reports have not been copyrighted and may be reprinted entirely, reference to source is appreciated. The National Marine Fisheries Service (NMFS) does not approve, recommend, or endorse any proprietary product or proprietary material mentioned in this publication. No reference shall be made to NMFS, or to this publication furnished by NMFS, in any ad- vertising or sales promotion which would indicate or imply that NMFS approves, recommends, or endor- ses any proprietary product or proprietary material mentioned herein, or which has as its purpose an in- tent to cause directly or indirectly the advertised product to be used or purchased because of this NMFS publication. NOAA Technical Reports NMFS guarantee a good final printed copy. Technical Reports of the Fishery Bulletin Figures are to be labeled with author's name and number of figure. Use Times Guide for Contributors Roman font (upper and lowercase letters) to label within figures. Avoid vertical let- tering except for y-axis labels. Zeros should precede all decimal points for Text footnotes should be numbered values less than one. Figures may be sub- Preparation mitted as computer software files (along with Arabic numerals and typed on a with laser-printed copies), as photo- separate sheet of paper. Footnote all per- mechanical transfers (PMTs), or as high Title page should include authors' full sonal communications, unpublished data, quality photographic prints. Send only names and mailing addresses and the and unpublished manuscripts with full xerox copies of figures to the Scientific senior author's telephone and FAX address of the communicator or author, Editor; original figures will be requested number. or, as in the case of unpublished data, later when the manuscript has been ac- where the data are on file. Authors are cepted for publication. Figure legends Abstract should not exceed one double- advised to avoid references to nonstan- should explain all symbols and abbrevia- spaced typed page. It should state the dard (gray) fiterature, such as internal, tions and should be double-spaced on a main scope of the research but em- project, processed, or administrative re- separate page at the end of the phasize its conclusions and relevant find- ports, wherever possible. Where these manuscript. ings. Because abstracts are circulated by references are used, please include abstracting agencies, it is important that whether they are available from NTIS Copyright law does not cover govern- they represent the research clearly and (National Technical Information Service) ment publications, they fall within the concisely. or from some other public depository. public domain. If an author reproduces lAterature cited comprises published any part of a government publication in Text must be typed double-spaced works and those accepted for publication his work, reference to source is con- throughout. A brief introduction should in peer-reviewed literature (in press). Fol- sidered correct form. portray the broad significance of the paper; the remainder of the paper should low the name and year system for cita- be divided into the following sections: tion format. In the text, cite Smith and Submission Materials and methods, Results, Jones (1977) or (Smith andjones, 1977). Discussion (or Conclusions), and Ac- If there is a sequence of citations, list Send printed copies (original and two knowledgments. Headings within each chronologically: Smith, 1932; Green, copies) to die Scientific Editor: section must be short, reflect a logical se- I.947; Smith andjones, 1985. Abbrevia- quence, and follow the rules of multiple tions of serials should conform to ab- Dr. Ronald Hardy, Scientific Editor subdivision (i.e. there can be no subdi- breviations given in Serial Sources for the Northwest Fisheries Science Center, vision without at least two items). The BIOSIS Previews Database. Authors are F/NWC3 entire text should be intelligible to inter- responsible for the accuracy and com- National Marine Fisheries Service, NOAA disciplinary readers; therefore, all acro- pleteness of all citations. 2725 Montlake Boulevard East nyms, abbreviations, and technical terms Tables should not be excessive in size Seattle, WA 98112-2097 should be spelled out the first time they and must be cited in numerical order in Once the manuscript has been accepted are mentioned. The scientific names of the text. Headings should be short but for publication, you will be asked to sub- species must be written out the first time ample enough to allow the table to be mit a software copy of your manuscript they are mentioned; subsequent mention intelligible on its own. All unusual sym- to die Managing Editor. The software of scientific names may be abbreviated. bols must be explained in the table head- copy should be submitted in WordPeEfecl Follow the U.S. Government Printing Offi'e ing.,Otlier incidental comments may be text format (or in standard ASCII text Syle Manual (1984 ed.) and the CBE Syle footnoted with italic numerals. Use format if WordPerfect is unavailable) and Manual (5th ed.) for editorial style, and asterisks for probability in statistical should be placed on a 5.25-inch or 3.5- the most current issue of the American data. Because tables are typeset, they inch disk that is double-sided, double or Fisheries Sociey's Common and Scientific need only be submitted typed and for- high density, and that is compatible with Names of Fishes from the United States and matted, with double-spaced legends. Zeros either DOS or Apple Macintosh systems. Canada for fish nomenclature. Dates should precede all decimal points for should be written as follows: I I Novem- values less than one. ber 1991. Measurements should be ex- Copies of published articles and pressed in metric units, e.g., metric tons Figures include line illustrations and notes are available free of charge to the as (t); if other units of measurement are photographs (or slides) and must be cited senior author (50 copies) and to his or used, please make this fact explicit to the in numerical order in the text. Unless her laboratory (50 copies). Additional reader. The numeral one (1) should be photographs are submitted on glossy copies may be purchased in lots of 100 typed as a one, not as a loWer-case el (1). paper with good contrast, we cannot when die author receives page proofs. UNITED STATES DEPARTMENT OF COMMERCE BULK RATE NATIONAL OCEANIC AND ATMOSPHERIC I ADMINISTRATION POSTAGE & FEES PAID NATIONAL MARINE FISHERIES SERVICE SCIENTIFIC PUBLICATIONS OFFICE U.S. Department of Commerce BIN C15700 Permit No. G-19 SEATTLE, WA 98115 OFFICIAL BUSINESS Penalty for Private Use, $300 NOAA SCIENTIFIC AND TECHNICAL PUBLICATIONS The National Oceanic and Atmospheric Administration was established as part of the Department of Commerce on October 13, 1970. The mission responsibilities of NOAA are to assess the socioeconomic impact of natural and technological changes in the environment and to monitor and predict the state of the solid Earth, the oceans and their living resources, the atmosphere, and the space environment of the Earth. The major components of NOAA regularly produce various types of scientific and technical information in the following kinds of publications: PROFESSIONAL PAPERS-Important definitive TECHNICAL SERVICE PUBLICATIONS-Reports research results, major techniques, and special containing data, observations, instructions, etc. A par- investigations. tial listing includes data serials; predictions and outlook periodicals; technical manuals, training papers, plann- CONTRACT AND GRANT REPORTS-Reports ing reports, and information serials; and miscellaneous prepared by contractors or grantees under NOAA technical publications. sponsorship. TECHNICAL REPORTS-Journal quality with exten- ATLAS-Presentation ofanalyzed data generally in the sive details, mathematical developments, or data listings. form of maps showing distribution of rainfall, chemical and physical conditions of oceans and atmosphere, TECHNICAL MEMORANDUMS-Reports of distribution of fishes and marine mammals, ionospheric preliminary, partial, or negative research or technology conditions, etc. results, interim instructions, and the like. Information on availability of NOAA publications can be obiainedfrom: U.S. Department of Commerce National Technical Information Service 5285 Port Royal Road Springfield, VA 22161 U.S. GOVERNMENT PRINTING OFFICE: 1995 - 691-393 3 6668 -00003 7 1- t_