Application of satellite data for mapping and monitoring wetlands fact finding report

[From the U.S. Government Printing Office, www.gpo.gov]

APPLICATION OF SATELLITE DATA
FOR MAPPING AND MONITORING WETLANDS
Fact Finding Report

Federal Geographic Data Committee
Wetlands Subcommittee

Technical Report 1
September 1992

Federal Geographic Data Committee
)artment of Agriculture - Department of Commerce - Department of Defense Department of Energy
)epartment of Housing and Urban Development Department of the Interior Department of State
TA595.5 Department of Transportation Environmental Protection Agency
.U55A66 I Federal Emergency Management Agency - Library of Congress
1992 National Aeronautics and Space Administration - National Archives and Records Administration
Tennessee Valley Authority

Federal Geographic Data Committee

Established by Office of Management and Budget Circular A-16, the Federal Geographic Data Committee
(FGDC) promotes the coordinated development, use, sharing, and dissemination of geographic data.

The FGDC is composed of representatives from the Departments of Agriculture, Commerce, Defense,
Energy, Housing and Urban Development, the Interior, State, and Transportation; the Environmental
Protection Agency, the Federal Emergency Management Agency-, the Library of Congress; the National
Aeronautics and Space Administration; the National Archives and Records Administration; and the
Tennessee Valley Authority. The Departments of Education, Health and Human Services, Justice, and
Labor; the General Services Administration; and the National Capital Planning Commission also participate
on FGDC subcommittees and working groups. The U.S. Geological Survey chairs the committee on the
behalf of the Secretary of the Interior.

FGDC subcommittees work on issues related to data categories coordinated under the circular.
Subcommittees establish and implement standards for data content, quality, and transfer; encourage the
exchange of information and the transfer of data; and organize the collection of geographic data to reduce
duplication of effort. Working groups have been established for standards, technology, and liaison with the
non-Federal community.

For more information about the committee, or to be added to the newsletter mailing fist, please contact:

Federal Geographic Data Committee Secretariat
U.S. Geological Survey
590 National Center
Reston, Virginia 22092

Facsimile: (703) 648-5755
Internet: [email protected]

For more information about the FGDC wetlands subcommittee contact:

FGDC Wetlands Subcommittee
U.S. Fish and Wildlife Service
1849 C Street, N.W., ARLSO 400
Washington, D.C. 20240

Facsimile: (703) 358-2232

The following is the recommended bibliographic citation for this publication:
Federal Geographic Data Committee. 1992. Application of Satellite Data for Mapping and Monitoring
Wetlands - Fact Finding Report; Technical Report 1. Wetlands Subconinuittee, FGDC. Washington, D.C.
32 pages plus Appendices.

Federal Geographic Data Committee
Department of Agriculture e Department of Commerce * Department of Defense 0 Department of Energy
Department of Housing and Urban Development 9 Department of the Interior * Department of State
Department of Transportation 0 Environmental Protection Agency
Federal Emergency Management Agency 9 library of Congress
National Aeronautics and Space Administration * National Archives and Records Administration
Tennessee Valley Authority

Contents.

Pne

EXECUTIVE SUMMARY ........................................... 1

INTRODUCTION ...................... ............ ..........

DISCUSSION .................................... .......... ........ 3

CONCLUSIONS ................................ ....................... 6

EXAMPLES OF COMBINING SATELLITE AND DIGITAL DATA DERIVED
FROM AERIAL PHOTOGRAPHY .................................. 8

FACT FINDING QUESTIONS AND ANSWERS .......................... 9
1. Is any data being currently collected over the conterminous United
States'. for which no, order has been previously placed? ........... 9
2. Is any Thematic Mapper data being processed- todjay? ........ 9
3. How much did the last satellite scene cost that your organization
purchased and how long did it take to receive it? ...... I.......... 11
4. Is each analysis performed as a unique operation, or can it be
performed on a routine basis on different scenes? ...........
5. Which of the Tbematic Mapper infrared bands, Number 4, 5, or 7,
are most important for wetland identification? ................... 12
6. How many wetland classes have you been able to successfully
discern using satellite data? Can you provide a Esting,of these
classes'? .......................... @! ...........*............ 13
7. Which wetland classes are the easiest to discern, and which classes
are the hardest? Please provide some statistical insight as to
relative accuracies . ...................................... 13
8. Can you provide the committee. with quantitative or qualitative
information on wetland identification which is accurately related to
wetlaAd size, water regimes, and wetland coverage classesand
types? ...................................... ...... 14
9. What do you believe is a reliable minimum mapping unit for
forested wetlands, scrub-shrub wetlands, open water ponds? ........ 16
10. What are the problems associated with identifying narrow bands of
wetlands such as those found along tidal creeks or riparian wetlands,
which are found along streams and are. often very in@portant? ....... 17
11. Dud to the scale of coverage, it has been reported that satellite data
consistently underestimate the* acreage of individual wetland basins.
If this is true, can a conversion factor be determined? Is the
underestimation a function of the size of the wetland basin or the
scale at which the satellite senses? ........................... 18
12. Have you had any success in identification of submerged aquatic
Gctn
vegetation and/or grass flats? .............................. 19
US Department of Commerce
NOAA Coastal Services Center Llbrar]Ki
2234 South Robson Avenue ftoperty Of c8c Libir"ry
Charleston, Sc 29405-2413

Page

13. Can the 10-meter panchromatic data from SPOT help compensate
for the lack of Band 5 (1.55 to 1.75 pm) midinfrared radiance? ..... 21
14. Have you had any success in using stereo data from SPOT to
increase your accuracy in wetland identification, classification, and
delineation? ........................................... 22
15. Can a satellite be reprogrammed in flight so as to capture data for
different purposes? ................... I ................... 23
16. Is georeferencing now sophisticated enough to allow for a pixel by
pixel analysis in order to perform change detection? ............... 24
17. How do you navigate on the ground to locate and identify 1 to 6 or
more pixels depicting change? ................................ 27
18. How is National Wetlands Inventory digital data processed so that
they can be interfaced with Thematic Mapper digital data used by
Ducks Unlimited, and is this process done on a unique basis or a
routine basis? ........................................... 28
19. Do you agree with the following statement made by Dr. Gregory T.
Koeln, Director of Ducks'Unhmited's Habitat Inventory Program? ... 29
20. At what point can we expect to have permanent storage capability
so as to maintain captured digital data on an indefinite basis? ...... 30
21. Data Quality Concerns .................................... 31

REFERENCE ................ I...................................... 32

APPENDIXES

A. Wetland classes used in waterfowl habitat and wetland inventory
.B. SPOT Image product fee schedule
C. EOSAT product fee schedule
D. Acronyms

Figure

1. An example of project rating error, where a small area is not registered
perfectly in all imagery .......................................... 25

Table

1. Undsat TM scenes being processed by'Ducks Unlimited on January 9, 1992 .... 10

iv
InOMV-1acisc SU
ywzdlol zo-2=o2 ir,7azo') AAON
omnevA vo-norl :iJL;02 t'E99

EXECUTIVE SUMMARY

The detail and reliability of information derived from satellite data have steadily
improved. These improvements include advancements in spatial and spectral resolution,
georeferencing, and digital image processing techniques, along with,growing experience
using satellite data. Significant strides have been made in integrating ancillary data, such
as soils and digital elevation models, into the classification of satellite data. This
integration is dependent upon the use of geographic information system (GIS)
technology. Stream gauging data and rainfall data are now being used to select the best
scenes for wetland identification. Even with these improvements, satellite data can not
match the accuracy of areal extent, classification detail, or reliability that can be
extracted from conventional aerial photography using manual photo-interpretation
techniques, such as those used by the U.S. Fish and Wildlife Service's (FWS) National
Wetlands Inventory (NWI) Project. However, for some regions, satellite remote sensing
may be the most cost-effective means for conducting reconnaissance wetland surveys.

The power of satellite imagery lies in its ability to be easily integrated with all other
sources of data in a GIS, contributing to the accuracy of the GIS. The U.S. Department
of Agriculture's Soil Conservation Service (SCS) believes that satellite technology can
help to classify certain administrative classes of wetlands legislated by the Farm Bills of
1985 and 1990. Many other resource managers have complained that, in practical
application, the promise of space-based remote sensing has not measured up to NWI
actual performance. The subcommittee believes satellite data, when used in conjunction
with NWI digital data produced through the use of aerial photography, can and do ,
provide a tool for monitoring water levels in wetlands and monitoring the cover change
of adjacent uplands. Synergistic effects created by combining both satellite data and
NWI digital data have greater value than using either data source alone. Such data sets
have the potential to be synoptic and accurate.

ENTRODUCTION

The President's Domestic Policy Council's Wetlands Task Force requested that the
Federal Geographic Data Committee's Wetland Subcommittee produce a report about
the application of satellite data for mapping and monitoring of wetlands. On January 14
and 15, 1992, the subcommittee held a meeting to discuss the current application of
satellite data for mapping and monitoring of wetlands. The subcommittee invited top
level technical experts from the following organizations to address a pres et list of
questions and describe their experiences:

Earth Observation Satellite Company (EOSAT)
SPOT Image Corporation (SPOT)
Ducks Unlimited (DU)
U.S. Environmental Protection Agency, Environmental Monitoring Systems
Laboratory (EPA)
National Oceanic and Atmospheric Administration (NOAA) Coast Watch

Change Analysis Program (C-CAP)
A. Oak Ridge National Laboratory
B. University of Delaware
Earth Satellite Corporation
U.S. Geological Survey (USGS), Earth Resources Observation System
(EROS) Data Center (EDC)
Maryland Department of Natural Resources, Water Resources
Administration (MD-DNR)

On January 15, 1992, the Wetlands Subcommittee met to discuss the information gained
from the previous days presentations, and to compare notes on the presentation of each
speaker. The subcommittee drafted a fact-finding report that analyzed and consolidated
the responses resulting from the January 14, 1992, meeting.

Upon review of the draft report, SCS reviewers believed that the report did not fully
reflect the latest findings in the use of satellites for mapping and monitoring wetlands.
NASA!s Space Remote Sensing Center (SRSC) is conducting research and mapping
wetlands for SCS in accordance with the Farm Bills of 1985 and 1990. SRSC's
experience in mapping wetlands with satellite data has been primarily related to
inventory efforts conducted by the SCS. Accuracy assessments conducted for the States
of Mississippi and Arkansas by the SCS for every county in the Yazoo basin showed
results that were superior to the original suggested SCS wetlands mapping techniques.
As a result, the wetland classification used by SRSC for the Yazoo basin in the States of
Mississippi and Arkansas is currently being used as baseline data for the SCS. SCS
personnel have commented that, for this region, satellite remote sensing proved to be the
only accurate and cost effective means of conducting a wetlands inventory for their
agency. SRSC's mapping approach has yet to be documented in detail or published in
the technical literature.

On May 4, 1992, the Wetlands Subcommittee invited SRSC to make a presentation on
their applications and to respond to the same questions answered by the speakers at the
January 14, 1992 meeting. Employees from the SCS's National Cartographic and GIS
Center (NCG) also made a presentation at the May 4th meeting. The SRSC and NCG
presenters were not questioned by peer technical experts as were the presenters at the
January 14, 1992, fact-finding meeting. This report includes the analysis and
consolidated responses from both the January 14, 1992, and May 4, 1992, meetings.

In many instances, DU answers and/or experiences are the only ones listed. This is
because DU's Dr. Gregory Koeln was the first speaker and provided his answers in
writing. Where subsequent speakers agreed with Dr. Koeln, his answer or experience
was used. If subsequent speakers had different answers or additional experiences, these
were included.

-DISCUSSION

The current limitations of satellite data that result in diminished utility for and accuracy
of wetland identification and habitat classification can be categorized into two areas:

a. Spectral Resolution

Spectral resolution problems result when the reflected radiance of different
land covers are very similar. Different land covers with similar radiance
have been placed in the same classification unit; for example, drier
forested wetlands being classified as upland forests; wetter scrub-shrub
wetlands being classified as either forested wetlands or emergent wetlands;
drier scrub-shrub wetlands being classified as upland forests; shadows from
forested areas falling onto agricultural lands being classified as emergent
wetlands; large, dense formations of tall emergent wetland plants, such as
Phragj]@ites spp., being classified as upland grassland or even upland
deciduous forests; and high coastal marsh, during low tide, being identified
as upland grassland. These problems in spectral resolution occur when the
satellite imagery is taken at a very wet time of the year or during a
drought. Many of the spectral resolution problems can be resolved or
ameliorated by using satellite imagery from several time periods that best
display wetland conditions. The process of merging multitemporal scenes
would result in profound cost increases for National programs, as well as
limit data useability for detecting change.

b. Spatial Resolution

Each thematic mapper (TM) pixel represents an area 3 0 by 30 meters.
Theoretically, spatial resolution is defined as two pixels -- that is, a box two
pixels on a side, or a total of four pixels (approximately 1 acre for TM
data). In practice, however, it generally takes a box with three identical
pixels on a side (nine pixels or approximately 1.5 acres) to consistently
identify an object. If the object being classified does not have a square
shape, or if the adjacent land cover has a similar reflected radiance, or if
one or more of the pixels has a slightly different radiance, then the number
of pixels needed to make a consistently confident determination increases.
Although wetlands smaller than the theoretical limit have been accurately
delineated, some scientists believe they need as many as 25 pixels.
(approximately 5 acres) to be confident about certain classification units.
Other scientists believe that 25 pixels is a worse case scenario. Through
the uses of ancillary data and visual interpretation, smaller wetlands can be
successfully identified in treeless areas when the basin is nonvegetated, full
of water, and surrounded by the bare soil of an agricultural field or much
drier prairie grass that has a significantly different spectral reflectance.

Spatial resolution has often been presented as the main shortcoming of wetlands
mapping by satellite. In reality, the greater concern is the spectral resolution that results
in the failure to distinguish wetlands from uplands. However, it is the spectral resolution,
that now allows us to see differences in biomass and may allow us in the future to
determine the health of wetlands.

Not all wetland mapping has the objective of assigning habitat classifications to mapped
wetlands, as does the NWI. For instance, the SCS's Swampbuster Program merely
attempts to identify wetlands and assign to each entity one of a few operational
classifications (prior'converted cropland, farmed wetland, wetland,@@ artificial wetland,
etc.). If the need for mapping U merely to identify- wetlands in farm fields and to update
wetland maps produced by other- means, then satellite imagery obtained when the
wetlands are flooded may be appropriate for this use,, especially when supplemented with
aerial photography; however,'this imagery may not be, appropriate for habitat
classification.

After deleting all point and linear wetland data DU compared wetland basins delineated
in the DU wetland classification using TM. data for basins identified on wetland maps
prepared by the NWI on the treeless prairies. With this method, approximately 20
percent of the wetlands smaller than- 2 acres (9 pixels) were detected; 70 percent of the
basins between,2 and 5 acres (9-25 pixels) were detected; 91 percent of the basins
between 5 and 10 acres (23-45 pixels) wore detected; and all basins greater than 25 acres
(112@pixels) were detected. Manual editing of the DU wetland data has allowed a
greater number of smaller basins to be detected. DU believes that under optimum
conditions, blocks of forested w6tland and scrub-shrub wetlands smaller than 10 acres
cannot be reliably detected using TM. SRSC, however, believes that deciduous forested
wetlands as small as one acre can be identified with TM data in the Yazoo basin of the
States of Mississippi and Arkansas. This assumes that multiple dates of imagery are
evaluated, at least one of which is acquired during a leaf off period when flooding has
occurred for the proper number of days during the growing season .or when the flood
event can be correlated to the appropriate river stage at a local stream gauging station.
ne criteria for the appropriate river stage is the minimum stage that produces 15 days
of consecutive flooding with 50 percent occurren ce, i.e. on the average once in two years.
Riparian specialists believe that the mapping width of at least 60-90 meters essentially
precludes the utility of satellite data for mapping most riparian areas.

Benefits in using satellite data for monitoring wetlands and upland land cover and land
use include:

L. the synoptic overview of redundant regions;
data'collected in a digital format, which facilitates the use of repeat
coverage to enhance results; -
3. ease of integration of the data,with other digital data themes;
4. repeat coverage, which facilitates monitoring seasonal or yearly
changes and allows data to be tied to rainfall events or river gauge
data;
5. lower cost per acre than aerial photography; faster results in mapping;

6. spectral sensing of mid-infrared bands of TM data, which provide
better wetland detection than data from the multispectral scanner
(MSS) or similar spectral range sensors;
7. , combination of the Landsat TM data with NWI data, which provided
better information as to the value of the wetland habitat for waterfowl
than the NWI data used alone.

Limitations in the use of satellite data for monitoring wetlands and upland habitats are:

1. reduced accuracy and detail-in identifying certain wetland habitat
types when compared to aerial photography; .
2. inability to classify more than a limited number of wetland classes;
3. inability to reliably and,routinely detect forested wetland and scrub-
shrub wetlands;
4. significant problems in trying to identify vegetated wetlands with drier
water regimes;
5. difficulty differentiating emergent vegetation growing in dry basins
from some upland vegetation types;
6. underestimations of the acreage of individual wetlands; the amount of
underestimation is not consistent;
7. difficulty detecting wetland basins less. than 2 acres (even in treeless
areas);
8. very limited ability. to detect by satellite observation alone small
wetland filling on the margins of wetlands ("nibbling");
.9. aggregation of complexes of wetlands with dendritic drainage patterns
and a variety of other wetland classes into one simple, class;
10. inability to map most riparian areas because of spatial resolution;
11. weather conditions preclude data acquisition on desired dates (this
also applies to aerial photography but not to the same extent);
12. clouds, cloud shadows, and, in areas of high relief, terrain shadows
cause spectral signatures that are confused with land cover types;
13. uncertainty as to whether data from Landsat 6 (to be launched in
early 1993) will be collected if there is no order for the data. Such a
practice will reduce the number of scenes that will be collected and
available for use in multi-temporal analysis.

These limitations apply to wetland classification schemes that are focused on mapping
wetland habitats. They do not apply to all classification systems. For example,
limitations 1, 2, 3, 4, 5,.9, and 10 do not apply to mapping conducted as part of the SCS
Swampbuster Program, which is based primarily on recertification and compliance
checking of nonvegetated farmed wetlands for the 1985 and 1990 Farm Bills.

CONCLUSIONS

Satellite data can be used to monitor water levels of some wetlands and changes in a
number of habitat types.' and to inventory and determine changes in upland cover.
Satellite technology cannot be used for inventories of wetlands for the purpose of
wetland identification and classification without also using other data sources. If there
are no other data sources available, it can be used for reconnaissance wetland surveys.
Digital satellite data facilitates integration of multiple data sets in GIS applications.

The current satellite technology is'most valuable when used in conjunction with digital
data derived from aerial photography and other sources; this technology cannot be used
to classify wetland types as defined by the FWS wetland classification system entitled
"Classification@ of wetlands and deepwater habitats of the United States" (Cowardin and
others 1979). Most scientists believe that to adequately delineate and classify wetlands,
one needs 5-meter resolution (or better) color-infrared data viewed stereoscopically. In
certain instances, satellite information is not only extremely helpful in inventorying
wetlands, it may be the best or only tool available at any given time. A case in point is
farmed wetlands located in the Yazoo basin of Mississippi and Arkansas. These
wetlands have had their natural vegetation removed and hydrology altered. Therefore,
soils and vegetation offer little in terms of identifying wetland characteristics. The only
practical way to identify these wetlands is by documenting wetland hydrology from offsite
information (e.g. stream gauge data, aerial photographs, satellite information, etc.).
Because aerial photographs are not generally available in a sequence that would depict
frequency and duration of inundation, their utility is limited in this basin. However,
satellite data repeat coverage correlated to steam gauge data can provide information of
frequency and duration of inundation, which can be directly tied to wetland hydrology
criteria, thus identifying certain types of wetlands. Acquiring cloud-free data during the
optimum time period (e.g. flooding) for delineating wetlands is at best a hit or miss
proposition. Even if satellite collection coincides with flooding, lingering clouds often
reduce data usefulness'.

The following points must be considered prior to an effort to expand the satellite
capabilities of the industry.

1. The best technique for initial wetland habitat mapping and inventory is the
technique currently used by the FWS's NWI project, which uses aerial
photography..

2. Digital NWI data and satellite data are. easily- merged to provide products
with greater Wetland evaluation and monitoring capabilities than either
type of data used alone. Digital data for all NWI maps @are required. The
subcommittee believes it makes little sense to expand the satellite
capabilities for wetland monitoring in the U.S. if the NWI process is not
accelerated and all NWI final maps are not digitized.

3. An important obstacle in using current satellite technology for wetland
monitoring is the lack of cloud-free data acquired during the optimum time
period (when all wetlands are inundated or fully saturated to their upland
margin and not beyond). During years of normal precipitation, this may
only be for a two to three week period each year. Currently, the
opportunity to acquire TM data only every 16 days (8 days if you include
Landsat 4) is not sufficient., With adjustable viewing. angles, SPOT- has a
very high rate of acquisition opportunities, up to,, 11 times- every 26 @ days. -
Oblique views, however, will increase the difficulty of ground registration
particularly in areas of high relief displacement. 'SPOT also offers radar
imagery. Limited resolution of radar data, however, limits its effectiveness
for monitoring surface hydrology in wetland areas. There are various
means of overcoming this obstacle. One way to increase the odds of
obtaining, cloud-free data is. to use multiple satellites. The costs of
maintaining multiple satellites will be high, but in the early years of the
Landsat program, it was generally accepted that multiple satellites would
be needed if acquisition of cloud-free data for operational programs was
required. Another means of partially overcoming the problem of cloud
cover is to develop satellites that can readily modify the look angle (as
SPOT can) to provide more frequent coverage for areas of interest.

DU has worked with radar data. Many technical obstacles remain before
radar data can be used operationally to monitor wetland inundation and
saturation, but DU is encouraged that radar may provide an additional tool
for wetland monitoring. Development of satellite systems with radar
capabilities could be one solution foracquiring cloud-free data at the
optimum time for wetland monitoring.

4. In efforts, to promote the expansion of . satellite systems for wetland
monitoring, we should de-emphasize the use of today's.satellite systems as
the primary source.for wetland inventories and emphasize the use of
satellite systems for monitoring wetlands and evaluating wet.land functions
in conjunction with other data sources. Wetland mappers .must make their
specifications and data requirements known to the people designing
Landsat 7, Which is scheduled for launch in 1991. -Information and data
needs should be forwarded through appropriate organizational channels.

Current satellite data and the capabilities of satellite systems in the near
future cannot be used to accurately classify the wetland type nor, to
accurately delineate the wetland margins for all existing wetlands. Using
satellite data in addition to NWI digital data provides a tool for monitoring
water conditions of wetlands, examining impacts to wetlands, and
ascertaining the point-in-time value of wetlands.

In many countries, either the value of wetland has not been recognized or
funds for the type of mapping provided by the NWI are not available. DU
has worked very successfully in both Canada and Mexico to provide

reconnaissance wetland maps using Landsat TM technology. ne products
are not as accurate or detailed as those produced in the United States by
the NWI, but in many cases, they represent the best and only source of
wetland data over these regions.

EXAMPLES OF COMEBEqNG SATELLITE AND DIGITAL DATA
DERIVED FROM AERIAL PHOTOGRAPHY

The following are examples of how FWS digital NWI data could. be combined with
satellite data to produce a product having greater value than using either data source
alone.

1. Maps have been produced for Maryland using SPOT 10-meter data as an
image base and using NWI vectors a's an overlay. This combination
provides a valuable and cost effective tool for informing and educating the
public on wetlands.

2. Many environmental impacts to wetlands result from misuse of the
surrounding uplands. Satellite data can be used to analyze these impacts at
a lower per acre cost and results can be produced faster than with aerial
photogramm.etry.

3. Satellite data can be used to identify regions where rapid changes in
wetlands are taking place. As a result, these areas may require frequent
updates. Updates should use aerial-photography-based NWI techniques.

4. Current satellite capabilities limit the use of satellite data for updating
NWI maps. Some of the satellite systems being pla -nned may have the
capabilities to update NWI maps. However, even with the improved
satellite capabilities, it is not certain whether these systems can adequately
implement the full FWS wetland classification system, or whether they can
implement a simplified version.

5. Satellite technology is a tool to provide current status of water levels in
wetlands. Satellite data can readily be used to ascertain' the extent of
inundation and saturation of nonvegetated wetlands at a point in time.

FACT FINDING QUESTIONS AND ANSWERS

The following questions and answers provide specific insights into the status of mapping
wetlands using satellite imagery.

1. Is any data being currently collected over the conterminous United States (U.S.)
for which no order has been previously placed.)

Answer: Yes

EOSAT

EOSAT's data acquisition, policy is to acquire daytime data on all passes over the
conterminous U.S. whether or not an order has been placed for it.

SPOT

The SPOT satellite is constantly acquiring imagery over the U.S., including Alaska and
Hawaii. Client acquisition requests are priority. Regardless of client acquisition
requests, it is SPOT's goal to acquire total area coverage of the U.S. Intelligent archive
building is used to build a historical database.with the first priority. being near-vertical
panchromatic 10-meter and multispectral 20-meter data. The second priority is oblique
panchromatic and multispectral. imagery for change detection and stereoscopic terrain
modeling.

2. Is any TM da ta being processed today?

Answer: Yes

EOSAT

EOSAT acquires Landsat TM data every day. Customer orders are processed on an
image generation computer -system designed to process data from Undsats 4, 5, and 6.
This system became operational on October 1, 1991.

The Landsat 4 and 5 archive contains 28225 TM scenes in the conterminous United
States with less than 10 percent cloud cover for the period'July 1982 to January 1991.
Of these scenes, 1598 were acquired of the East Cost between the months of March. and
October, inclusive. EOSAT can provide more detailed information about specific
regions at any time.

SPO

SPOT also acquires multispectral data everyday. SPOT imagery is being processed today
from the SPOT 2 satellite. SPOT 1 has been reactivated to increase acquisition of
imagery during the growing season, specifically for current clients and for archive
building in relation to vegetation and wetland studies. The panchromatic 10-meter-
resolution imagery is being used in forestry, oil and gas research, urban planning,
communications, and defense applications. The multispectral 20-meter-resolution
imagery is being used in land cover classifications and vegetative analysis. Multispectral
imagery is being used by several Federal agencies due to high resolution, rapid- revisit
capability (capability to acquire 11 images every 26 days), and commercial delivery
capability.

SPOT is devoted@ to total quality management in regards to- every step in the delivery of
its data and service to its clients. There is a 10 line drop minimum data specification in
its quality control process. Cloud cover rating is confirmed prior to final production of
all deliveries. Full technical support is provided after delivery of product.

Ducks Unlimited

On January 9,.1992, the Landsat TM scenes listed in table 1 were being processed.
Some of the scenes were being classified, but the majority of the scenes were being
grouped and edited to wetland and, in. some -cases, upland classes.

Table 1. Landsat TM scenes being processed by Ducks Unlimited on January 9, 1992

Path Row Geographic Area Ducks Unlimited Office

22 26 Central Ontario Winnipeg, Manitoba
26 42 Laguna Madre, Mexico Long Grove, Illinois
26 43 Laguna Madre, Mexico Long Grove, Illinois
26 42 Laguna Madre, Mexico Monterrey, Mexico
29 28 North Dakota, South Dakota Long Grove, Illinois
Minnesota
32 43 Pabellon, Mexico- Long Grove, Illinois
33 25 Southwest Manitoba Winnipeg, Manitoba
39. 23 West Central Saskatchewan Regina; Saskatchewan
40 23 East Central Alberta Edmontonj @ Alberta
66 14 Yukon Flats, Alaska Long Grove, Illinois

3. How much did the last satellite scene cost that your organization purchased and
how long did it take to receive it?

Answer:
EOSAT sells a fall, system-corrected scene for $ 4,400. SPOT data costs $2,450
per computer-compatible scene, either panchromatic or multispectral. Both
EOSAT and SPOT currently deliver products in less than quoted delivery time.
Landsat MSS data more than.2 years old is available for $200 from EDC. Special
acquisition fees are extra. Scenes take 4-8 weeks for delivery. Standard archived
scenes are delivered in less than one week.

See Appendixes B and C for SPOT and EOSAT fee schedules and detailed cost
information for a variety of formats. When comparing the price lists for SPOT
and EOSAT data it is important to realize that a full SPOT scene covers
approximately one-eighth the area of a full Landsat TM scene.

Ducks Unlimited

In its most recent order, DU purchased three full Landsat TM scenes of Mexico. The
order was faxed to EOSAT on December 18, 1991. DU received the order 19 days later
on January 6, 1992. Cost per satellite scene was $2200 (one-half the normal price). DU
was able. to buy, the data at one-half the normal price because of a sale EOSAT offered
1991 buyers of satellite data. The.,scenes DU purchased in 1991 will be used to examine
wetland changes. The scene previously ordered on September 9, 1991; was received. on
December 7, 1991, more than, 12 weeks later.

EPA

It took.the Environmental Protection Agency (EPA) over a year to acquire all 16 scenes
necessary for the Chesapeake Bay watershed characterization project. The scenes varied
over a four year time span. This was due to cloud cover problems, even though one base
year would have been desirable. This experience showed EPA how helpful multiple
satellites and constant acquisition would have been for obtaining fall study area coverage
with better date,,consistency.

4. Is each analysis perform ed as a unique operation, or can it be performed on a
routine basis on different scenes?

Answer:
Ea ch data user performs a standard set of operations on raw satellite data as a
first step, but each.user performs a different initial set of operations. After the
standard operations are complete each user has a set of additional operations or
procedures that may or may not be used to draw out the information required
from the scene.

Ducks Unlimited

DU uses a standard image processing procedure for every one 1of its scenes. For each
scene DU has a file with 230 spectral classes. These spectral classes are grouped into
informational classes such as wet meadow, shallow marsh, deep marsh, and open water.
The grouping is unique for each scene and variable based on the experience of the
digital image interpreter, time of year that the data were acquired, precipitation patterns,
available ground truth, and many other factors. Because of spectral class confusion,
DU's digital image interpreters use their-visual interpretation skills to assure that the
information group assigned for the pixel agrees with their visual interpretation of the
data. For example, shadows, whether cast onto the image from clouds, terrain, or forest,
typically have a spectral signature identical to one of the spectral signatures assigned to
shallow marsh. Usually the interpreter can visually distinguish a shallow marsh spectral
class resulting from shadows and will edit these to a non-wetland class. In DU's work in
Alaska and northern regions of Canada, black spruce may frequently be assigned a
spectral class that has been assigned to shallow marsh, yet the interpreter can usually
visually distinguish black spruce from shallow marsh and edit the black spruce pixels
classified as shallow marsh to a forested class.

SRSC

A few preparatory operations can be applied to most satellite scenes regardless of
location of the scene's coverage. These include such things as image-to-image
registration, post classification georeferencing, and creating vegetation versus
nonvegetation masks. The list of preparatory operations increases when the analysis is
restricted to a geomorphic region with a consistent land-surface form, vegetative cover,
or regional climate.

5. Which of the TM infrared bands, Number 4, 5, or 7, are most important for
,wetland identification?

Answer:
Band 5 has traditionally been considered the single most important band for its
capability to discriminate levels of vegetation and soil moisture. However, a
combination of bands is needed for wetland detection.

Ducks Unlimited

The capability of Band 5 to detect moisture makes it the most important band for
wetland identification. However, a combination of TM Bands 3, 4, and 5 usually is the
best combination of bands for@wetland detection. If water quality is of interest, Band 3
should be replaced with Band 1. It is possible that Band 1 may also help in separating
open water from open water with submerged aquatic vegetation (SAV).

Earth Satellite Coj:poratio

Band 5 (1.55-1.75 pm) has traditionally been considered the single most important for its
capability to discriminate levels of vegetation and soil moisture. In Earth Satellite's
experience, Band 4 (0.76-0.90 jAm) is also important in the classification of vegetative
communities and vegetation moisture, from which wetlands may be inferred. Earth
Satellite's approach to land cover classification utilizes all six TM reflective bands in
order to identify homogeneous classes. Additionally, Earth Satellite utilizes a proprietary
program, GEOVUE, to simultaneously analyze multiple bands of data, either in a
multispectral or multitemporal fashion, or in combination.

SRSC

In general Band 5 is the most important for wetland identification due to its sensitivity to
soil and vegetation moisture. But this is not always the case; it depends on what type of
wetland is being classified. For example, Band 5 was most important for identifying
standing water and saturated soils associated with farmed wetlands in the Mississippi
Delta. Band 4 was most important in identifying patches of healthy natural vegetation
located within recently plowed agricultural fields in the Prairie Pothole region of North
Dakota. The patches of vegetation are surrogate indicators of temporary wetlands where
the potholes were filled with water when the farmer began plowing the field. Because
these potholes were filled with water, the farmer simply plowed around the potholes. At
the time the satellite scene was acquired the open water in the pothole had evaporated
and the natural vegetation now growing in the pothole contrasted both spectrally and
spatially with the surrounding bare soil.

NCG

NCG has found that TM Bands 5 and 7 to be extremely useful in differentiating
vegetation.

6. How many wetland classes have you bee n* able to successfully discern using
satellite data? Can you provide a listing of these classes?

Wetland classes used in waterfowl habitat and wetland inventory.
(See Appendix A for categories identified by DU and SRSQ

7. Which wetland classes are the easiest to discern, and which classes are the
hardest? Please provide some statistical insight as to relative accuracies.

Answer:
Although quantitative statistical data are not available, the consensus is that the
easiest classes to discern are permanently flooded or intermittently exposed open
water ponds (pahistrine unconsolidated bottoms). Tle difficulty increases, moving
through the wettest to driest marshes, deciduous forested wetlands, evergreen

forested wetlands, and any type of scrub-shrub wetlands. Mangroves are an
exception; due to their unique spectral reflectance, they are the easiest vegetated
wetland type to identify. For coastal wetlands the ease and ability depends on
tide conditions. With appropriate tide conditions, mangroves are the easiest class
to discern, then moving to salt marshes to forested wetlands with scrub-shrubs
(other than mangroves) being the hardest class to discern.

Ducks Unlimited

DU's work in the Prairie Pothole region found that shallow marsh, deep marsh, and
open water classes are the easiest to discern without any editing. The open water classes
are always easy to discern; spectral class confusion has never been a problem with these
classes. In general, the deep marsh class is as easy to discern as the open water class;
seldom has spectral class confusion been a problem with the deep marsh class. Wetter
shallow marsh classes are nearly as easy to discern as the deep marsh classes.

Dry shallow marsh classes frequently have a great deal of spectral class confusion. For
these spectral classes, which typically describe the temporary basins or the margins of
seasonal or semipermanent wetlands, spectral class confusion is often a problem.
Spectral classes which discern temporary wetlands and wetland margins of seasonal and
semipermanent wetlands often will cause commission errors, identifying moist areas in
fields as wetlands. DU removes these commission errors using its image editing
procedures.

For example, E10W (estuarine subtidal open water), LlOW (lacustrine limnetic open
water), and POW (palustrine open water) are spectrally identical and must be edited to
these classes based upon location and size of wetland. In a similar manner, emergent
vegetation of estuarine, lacustrine, and palustrine systems are spectrally identical and
must be separated by an editing or post-processing step.

8. Can you provide the committee with quantitative or qualitative information on
wetland identification which is accurately related to wetland size, water regimes,
and wetland coverage classes and types?

a. Wetland size

Answer:
ne accuracy of satellite data for identifying wetlands is heavily dependent on
collecting satellite data when the wetlands are inundated. A multiacre temporary
basin that is dry and cultivated when the satellite data are collected usually cannot
be identified using satellite data. However, stock ponds (or dugouts) in Montana
range land are readily identified using Landsat TM data even though these ponds
represent very small areas (10 ft by 20 ft) of open water.

Ducks Unlimited

Using Landsat TM data collected in late May 1986, and employing DU's 1987 image
processing procedures, which did not include an interactive editing step, only 22 percent
of the wetland basins (identified from NWI data) less than 2 acres in size were classified
as wetlands as shown below:

Wetland basin Percent classified by
size (acres) thematic mapper data
0-1.9 22
2-4.9 70
5-9.9 91
10-24.9 96
> 25 100

The drought of the 1980's severely impacted the available habitat for waterfowl. By the
end of May 1986 most of the temporary basins were dry and many of the seasonal and
some of the semipermanent basins were dry. Also during the 1980's, many temporary
and seasonal basins were cultivated, making wetland detection very difficult, even using
conventional aerial photography techniques.

b. Water regimes

Answer:
In the Prairie Pothole region, satellite data collected in late April or very
early May in years of normal precipitation are ideal for delineating
wetlands. However, it is difficult to ascertain from spectral information if a
wetland has a temporary, seasonal, semipermanent, or permanent water
regime. Satellite data collected late in the year provides information on
water regimes, but can make wetland detection very difficult.

C. Wetland coverage classes and types (i.e. forested, scrub-shrub, emergent,
etc.)

Answer:
DU's biologists, using their habitat inventory products derived from
satellite data, have been pleased with the separation of shallow marsh,
deep marsh, and open water when using data collected from mid-May to
June. When using late April or early May data, deep marsh typically has a
spectral signature that is difficult to separate from open water. DU
frequently debates the trade-off between using late April or early May data
to improve detection of temporary and seasonal wetlands or mid-May to
June data to improve separation of wetland types.

SRSC

An important factor in the size question is how much spectral contrast there is between
pixels of certain wetland classes and surrounding pixels of other land cover classes. In
the Prairie Pothole region of North Dakota SRSC had success in classifying open water,
deep marsh and shallow marsh as small as one-half acre in size using SPOT multispectral
data. It should be pointed out that SRSC believes wetlands were detected at this size
because of optimum conditions: open water was generally surrounded by the bare soil of
an agricultural field or much drier prairie grass vegetation which bad a significantly
different spectral reflectance. The same could be said for the deep marsh and shallow
marsh. Since SRSC has only conducted one wetlands study in the Prairie Pothole region,
they can not state with certainty that satellite remote sensing is capable of repetitively
identifying wetlands at this size. But they are encouraged by these early results and
intend to do further wetlands research in this region. The SRSC mapping procedures
have not been documented in detail or published in the technical literature where they
would be subjected to peer review of the process and map products. Quantitative testing
against NWI digital wetland data will be performed -on other maps in the Prairie Pothole
region.

9. What do you believe is a reliable minimum mapping unit for forested wetlands,
scrub-shrub wetlands, open water ponds?

Answer: No definite answers were provided.

NOAA-C-CAP

Jerry Dobson, of Oak Ridge National Laboratory, reported that on the Salisbury,
Maryland 1:100,000 scale USGS quad, his initial analysis only identified 30 percent of the
forested wetlands. This accuracy was doubled to 60 percent after some. initial field work
using a fall satellite scene. Accuracy was further improved to 67 percent through
additional field work and the use of a spring scene.

EPA

Ross Lunetta, of EPA, has used a wetland/upland mask developed through the use of
satellite information gathered during the leaf-off wet season of a normal year to help
identify deciduous forested wetlands in subsequent data gatherings. He has not tried this
technique on evergreen forested wetlands. The wet season scene is selected using
rainfall data and/or river stage data.

SRSC

SRSC has successfully classified hardwood forested wetlands in the Yazoo basin as small
as one acre using TM data. This assumes the use of multiple dates of imagery and
acquiring scenes when the hardwoods were inundated with flood water. This does not

imply that they can map all forested wetlands down to one acre. Quantitative testing
against other wetland mapping efforts have yet to be performed.

Ducks Unlimited

DU has had only minimum experience in delineating scrub-shrub and forested wetlands.
From their limited experience, they believe that these features may be two of the most
difficult habitat types to accurately delineate using satellite data. These habitats are best
identified in leaf-off periods when the soil is fully inundated or saturated. Even under
these optimum conditions, blocks of forested and scrub-shrub wetlands smaller than 10
acres cannot be reliably detected.

Earth Satellite Corporatio

Earth Satellite Corporation believes that a minimum mapping unit of 5 to 10 acres is
possible for detection of forested and shrub-scrub wetlands by digital methods using
satellite imagery. Thissize may be reduced by using additional data sources, GIS data,
and/or other visual (monoscopic and/or stereoscopic) data; the presence of spectrally
unique wetland forest species in an area, e.g., mangrove, would also help reduce this size.

Emergent wetlands and deep water habitats are detectable to as little as 3-5 acres,
depending upon water regime at the time of imagery. Smaller than 3 acres, with current
spatial resolutions, the accuracy of classification decreases rapidly.

10. What are the problems associated with identifying narrow bands of wetlands such
as those found along tidal creeks or riparian wetlands, which are found along
streams and are often very important?

Answer:
The drier the adjacent upland the easier it is to identify riparian wetlands. In
general the riparian strips need to be at least 60-90 meters wide to be identified.
Most riparian specialists believe that a minimum mapping width of 60-90 meter
essentially precludes the use of satellite data for mapping many riparian areas.

Ducks Unlimited

DU's work in Alaska found that most narrow drainages. have varying widths of sedge,
small willows scattered amongst the sedges, or sedges primarily dominated by small
.willows or small coniferous shrubs. If the linear feature is narrow (90 m wide or less),. it
can be very difficult to properly classify. Also, many of the drainages have small slow-
flowing creeks that are often not delineated by NWI. Drainage without such creeks offer
little value for waterfowl except as nestirig habitat. However, drainage with slow-flowing
creeks have value for waterfowl. Even though these creeks are often not displayed by
NWI on their 1:63,360 scale maps, DU's Landsat TM data have enabled the display of
these narrow ribbons of water.

One additional problem with delineating narrow linear features using satellite data is the
mixed pixel problem. A narrow band of water in a creek surrounded by upland grasses
may be erroneously classified as a shallow marsh. In this situation, water may cover one-
third of a pixel while upland grasses cover the remaining two-thirds of the pixel. The
spectral response for this pixel will be a mixture of water and vegetation, which is
identical to the spectral response of shallow marsh.

SPOT

Mapping of a 60-meter wide swath would be possible with SPOT multispectral (3 pixels
wide) or with merged panchromatic and multispectral. (6 pixels wide), provided these
wetlands are not obstructed by upland overstory. If overstory is present and is composed
of deciduous trees, imagery can be acquired during the spring wet season, before leaf-out
occurs.

SRSC

Using ancillary data such as digital line graph (DLG) hydrography data and digital
elevation models (DEM's), a mask can be developed that identifies areas that are a
predetermined elevation above the nearest water source. Using this mask the feature
space over which spectral analyses are performed can be reduced. A classification can
be done on the reduced area, which facilitates delineation of narrow bands of wetlands.
Applying this technique along with other techniques should permit the identification of
riparian areas less than 90 meters wide using satellite imagery in certain geomorphic
regions.

11. Due to the scale of coverage, it has been reported that satellite data consistently
underestimate the acreage of individual wetland basins. If this is true, can a
conversion factor be determined? Is the underestimation a function of the size of
the wetland basin or the scale at which the satellite senses?

Answer:
Satellite data consistently underestimate the acreage of individual wetland basins.
There is ongoing debate as to whether an expansion factor or factors, can be
developed to convert these underestimates.

Ducks Unlimited

Satellite data consistently underestimate the acreage of individual wetland basins because
the drier margins of these wetland basins are inaccurately classified as upland habitat. In
classification of satellite data the analyst is constantly making trade-offs between
omission and commission errors. Commission errors (i.e., identifying a wet area in an
upland field as a wetland class) are increased as omission errors (i.e., identifying the dry
margin of a wetland as an upland class) are decreased. Typically when grouping spectral
data for wetlands,-those spectral classes representing dry wetland margins are not

assigned as a wetland class. If they were, considerable commission error would occur
and nonwetlands (moist areas in fields) would be assigned as wetlands.

Labor intensive techniques can be used ' such as DU's image editing procedures to help
reduce this problem. In addition, all spectral classes that represent not only wetland
margins, but also nonwetland areas, can be identified and wetlands can be allowed to
11grow" into these margins.

A conversion factor developed for a Prairie Pothole scene would likely be meaningless in
other geographic regions. Conversion factors could possibly be developed for other
wetland types in different regions but would be scene-specific due to water conditions at
the time of data capture.

Correction factors could possibly be used to correlate the actual acreage of wetlands
mapped using aerial photography to satellite-derived wetlands acreage. For a
representative sample of wetland basins in the study area, various dependent variables
from the satellite data are needed such as area of basin, length of basin perimeter, shape
index, and square and cubic transformations of these variables. Using the actual
acreage of the wetland basins as the independent variable and using regression analysis
techniques, it is possible to ascertain an appropriate regression model to correct the
acreage of wetland basins as derived from satellite data.

NCG

NCG believes there are scientifically valid sampling and estimation procedures whereby
adjustments can be made to figures generated by 100 percent mapping, but it is not
likely that several simple adjustment factors will provide desired reliability.

EPA

Although a conversion factor may aid applications that require only acreage totals over
broad areas, it is unlikely that a conversion factor could improve mapping of individual
wetland boundaries.

12. Have you had any success in identification of SAV and/or grass flats?

Answer:
No. The satellite can "see" some percentage of the SAV, but no verification
studies have been performed. The USGS estimates that using TM data, they have
been able to identify approximately 70 percent of the submerged aquatic beds in
the tidal Potomac River that can be detected using 1:24,000 scale photographs.
Best results are gained when both the satellite imagery and aerial photography are
captured at low tide. As with aerial photography, the ability of satellite imagery
to identify SAV is heavily influenced by wind conditions, depth, and turbidity.
Even small ripples on the water can prevent identification of SAV. Turbidity in
the water conceals much such vegetation.

Ducks Unlimited

In some DU study areas in Alaska, spectral classes were identified that had a degree of
correlation with areas identified as aquatic beds on the available NWI maps. Also, one
of the scenes DU used in Mexico provided data for parts of southern Texas that NWI
maps cover; some spectral classes appeared to correlate with L2AB (lacustrine littoral
aquatic bed) on the NWI map. DU researchers do not believe that satellite data actually
detects the SAV. They believe they are detecting shallow water that could be providing
adequate conditions @water depth and light penetration) for the growth of SAV.

Maryland Department of Natural Resources

The Maryland Department of Natural Resources (MD-DNR) delineated and classified
SAV using SPOT satellite multispectral. digital data (acquisition date of SPOT data,
September 27, 1987). The results of the effort are published in a report entitled
"Delineation and classification of submerged aquatic vegetation (SAV) using SPOT
satellite multispectral digital data." The following is from a report that was written by
Dr. K. Peter Lade of Salisbury State University under contract to MD-DNR:

"Aerial photography has been, to date, the principal data source other than field
verification. There is little question that good quality aerial photography, taken
under the right circumstances, at the right scale, and supplemented by good field
testing, can better serve the purpose of documenting the distribution of SAV than
any other data collection method.

Satellite data are, perhaps, the next most obvious data source. Synoptic in its
coverage, digital in format, satellite data have the potential of not only automating
much of the tedious process of matching "photography" to standard orthographic
projections, but has the further advantage of being comparatively inexpensive.
The disadvantage is mostly one of resolution.

SAV bedsWere identified on the SPOT satellite data of September 27, 1987, by
displaying segments of the satellite tract on screen for computer aided analysis.
In each case, it was assumed that a bed density of less than 25 percent would
probably not be visible at the 20 meter resolution of the multispectral satellite
data. A screen cursor was used to point at SAV colonies and the computer was
permitted to suggest additional locations of submerged aquatic vegetation within
the experimental frame.

The effect of this method was to reduce the total estimate of SAV in two ways.
First there is the -loss of statistics for those beds with low density. Since there is
no reliable method for estimating the total contribution of low density beds as a
fraction of total acreage for higher density beds, there will inevitably be an error
that may account for as much as 10 percent of the potential acreage of SAV in
the Chesapeake Bay and tributaries. Second, area calculations were not based on
polygons circumscribing an area of identified SAV, but rather on the point
distribution of targeted SAV coverage. This resulted in a more nearly accurate

mapping of aerial "tent of SAV, as opposed to the representational mapping of
previous studies where polygons were circumscribed around observed SAV beds.
In this case, the lower acreage figures to be expected through feature mapping of
the satellite data would, in fact, be a better estimate of the actual distribution of
SAV than the photo-mapping of previous years."

SPOT

The National Park Service has used SPOT multispectral imagery to locate and track the
movement of large blooms of SAV, specifically hydrilla, which are causing navigational
problems on the Potomac River. SPOT was able to acquire imagery at very specific
times, which coincided with low tides and maximum SAV exposure.

SPOT multispectral imagery was used as a rapid and accurate means of making long-
term measurements of biomass for monitoring purposes. The imagery allowed for the
detection of small changes in vegetation and environmental characteristics. The ability
to acquire the imagery at specific times was critical to the study. Spartina alterniflor
within the Great Marsh near Lewes, Delaware, was studied to correlate changes in
wetlands biomass distribution with precipitation. Three SPOT scenes, approximately one
year apart, were digitally compared. A supervised classification was performed to
identify those wetland areas dominated by Spartina alterniflora. A vegetation index,
based on the relative reflectance values of the Spartina, was then developed and used to
automatically map Spartin according to relative density, and thus biomass. This proved
to be a highly accurate, nondestructive, and rapid means of assessing and monitoring
biomass distribution.

13. Can the 10-meter panchromatic data from SPOT help compensate for the lack of
Band 5 (1.55 to 1.75 gm) midinfrared radiance?

Answer:
The 10-meter panchromatic data from SPOT cannot identify wetlands as well as
the 30-meter TM data with Band 5. But this is not a fair question because 10-
meter panchromatic SPOT data are not a replacement for Landsat Band 5 data.
Neither is Landsat Band 5 data a replacement for 10-meter SPOT data. Both
make valuable and unique contributions in extracting information about ground
cover. They do not replace each other, but are viewed by most users as
complementing each other.

TM's midinfrared bands. (especially Band 5) have unique capabilities for detecting
wetlands. Wetlands that are completely vegetated and not inundated, and dry
mud Rats are very difficult to classify as wetlands using panchromatic data alone.
A combination of visual and near-infrared data, or panchromatic data in
combination with visual and near-infrared data are needed. Mapping of a swath
at least 60-meters wide would be possible with SPOT multispectral (3 pixels wide)
or with merged panchromatic and multispectral (6 pixels wide), provided these
wetlands are not obstructed by upland overstory. H overstory is present and is

composed of deciduous trees, imagery can be acquired during the spring wet
season, before leaf-out occurs. If mid-infrared data were available at 10-meter
resolution, SPOT would have a very powerful tool for wetland detection.

Ducks Unlimited

DU has successfully combined 10-meter panchromatic data from SPOT with TM data
(including Band 5). They believe that spectral classifications using combined data
provided little improvement over using TM data alone. The visual enhancement of data
allowed DU's digital image interpreters to slightly improve their editing decision. For
this technique, DU used TM Bands 3, 4, and 5, and a hue, saturation, and intensity
transformation. The result of this transformation was a three band file. One band
depicted values for hue, the second band depicted values for saturation, and the third
band depicted values for intensity. The intensity band was replaced by the values from
the panchromatic band. Using a reverse transformation, the hue, saturation and intensity
files were transformed to a 3-band file representing red, green, and blue. This 3-band
file was used for the classification procedure.

SRSC

SRSC has developed a proprietary method of merging SPOT panchromatic data with
SPOT multispectral data. Unlike some of the traditional merging techniques (i.e.
intensity hue saturation transformation) SRSC's method of data integration does not
significantly change the original digital number (DN) values of the SPOT multispectral
data. This method reportedly gives their analysis the best of both worlds by producing a
data product that has the improved spatial resolution of the panchromatic data and
retaining the spectral resolution of the SPOT multispectral data. SRSC is in the process
of classifying wetlands in the Prairie Pothole region of North Dakota. Early results have
indicated that the merged product has identified open water ponds, deep marshes, and
shallow marshes as small as between one-quarter and one-half acre. Many of these
small wetlands were not classified when analyzing SPOT multispectral data (20-meter
resolution) over these same areas. SRSC believes that for this unique geomorphic
region, the SPOT panchromatic data merged with the SPOT multispectral data using
SRSC method may compensate for the lack of a middle infrared band because of its
improved spatial resolution. SRSC plans to conduct more research with the merged data
in this region and will determine if classification of potholes is improved significantly
over classification with SPOT multispectral data. Quantitative testing against NWI
digital wetland data will be performed.

14. Have you had any success in using stereo data from SPOT to increase your
accuracy in wetland identification, classification, and delineation?

Answer: No experience at all.
Theoretically, if stereo air photos were not available for a desired time period and
suitable SPOT stereo pairs were available, they could be used in film or print

form, or in digital form in a digital stereoplotter workstation to provide stereo
data.

15. Can a satellite be reprogrammed in flight so as to capture data for different
purposes?

Answer:
The SPOT satellite viewing angles can be reprogrammed to acquire imagery. The
viewing angles can be adjusted as much as 27' on either side of nadir (54' angle
of adjustment) to acquire imagery anywhere within a 1000 krn swath.

SPOT

The uniqueness and strength of the SPOT satellite system is based partially on the
capability to program the satellite upon request. A conterminous U.S. acquisition
request must be received one week prior to the start date. A programming request must
be received three weeks prior to the start date for Alaska and Hawaii.

Specific project requests can be accommodated with only one day notice to program the
satellite (e.g. to monitor flood waters and capture a scene at highest flood level to
update the one-hundred-year flood plain).

The SPOT satellite has two high-resolution visible sensors (HRV's) enabling
programming of a twin pass. A twin pass can provide coverage of 117 kilometers (two
scenes overlap). The HRV's can be programmed to acquire a scene within a + 27*
swath. The ability to program the satellite enables SPOT to capture a particular point
on the earth three times a week.

EOSAT

Landsats 4 and 5 have fixed data strearns; the 85-megabit-per-second seven-band TM
data and the 15-megabit-per-second four-band multispectral. scanner data. EOSAT
believes there is no ability to reprogram for different purposes.

EOSAT transmits two days of commands to each satellite every day. The commands for
the second day are made as insurance against loss of transmission capability. EOSAT
experiences extremely few such losses. The command loads include the start and stop
transmission times for transmitting all seven bands to all ground stations.

Inclination adjustments necessary to maintain the satellite over the Landsat world
reference system paths need to be performed less than once per year. Orbit adjustments
for altitude correction are performed during satellite nighttime orbit without interrupting
service to the ground stations. EOSAT does not acquire data during periods of orbit
adjustment or when they are outgassing the cold focal plane (about once every four
months).

16. Is georeferencing now sophisticated enough to allow for a pixel by pixel analysis in
order to perform change detection?

Answer:
Yes, but with some qualifications. Relative positional accuracy of a geocoded
satellite image is a function of spatial resolution (pixel size), the quality and
accuracy of the paper maps used for ground control points, the global positioning
system (GPS) setup if used (differential GPS is the most accurate), the skill of the
image analyst, and the quality and appropriateness of the computer algorithms
used to process the data. It is easier to get high accuracy for flatter, well surveyed
areas, although a properly orthocorrected image can be highly accurate. In
general, root mean square (RMS) errors can be as low as 0.5-1 pixel (5-10 meters
for SPOT panchromatic, 10-20 meters for multispectral) for most U.S. areas.
When SPOT reports error(s) on a product, it usually is referring to the
specification, which indicates the greatest error allowable. In the case of
SPOTView 7.5-minute quadrangles, which are panchromatic data orthocorrected
using 7.5-minute quadrangles digital elevation models, the error specification is 15
meters, or 1.5 pixels.

EOSAT

EOSAT provides map-oriented precision-corrected imagery which is accurate to within
.:L 15 meters (one standard deviation) with respect to a map, exclusive to terrain effects.
Within a scene, the pixel-to-pixel distances are accurate to less than one meter, also
exclusive of terrain effects. Many users perform change detection using two such scenes.

If elevation effects are important, EOSAT can provide map-oriented terrain-corrected
data, which is corrected using control points from maps and digital elevation maps. The
algorithms used in correcting Landsat data are designed to the specification of plus-or-
minus one-third pixel for scene-to-scene registration. These specifications are for the
entire scene; they are not random errors at each pixel within a scene.

Ducks Unlimited

DU typically registers full scenes of TM data to the Universal Transverse Mercator
(UTM) grid system using 20 to 30 control points and a linear transformation. For
regions where accurate UTM coordinates for the control points are available from
published maps or using a GPS, DU can obtain RMS errors of less than 30 meters.
(Obtaining UTM coordinates for control points in Mexico using the 1:50,000-scale map
sheets results in a RMS error of 40 to 45 meters.) For example, using these techniques
in the U.S. and Canada, DU would register -two different scenes to the UTM grid system,
and the overlay of these two images would not be a perfect fit.

Pixel-to-pixel registration is not required to identify wetland change. Registration of
pixel-to-pixel or its near neighbor is readily achieved in the simple georeference
procedures used by DU. In its wetland change detection procedures, by using connective

components techniques, DU identifies all pixels from either scene that are connected to
the wetland basin. For example, a small temporary basin may have 4 pixels of shallow
marsh in one scene and four pixels of shallow marsh in the other scene, but because of
project rating error, only two of these pixels are identified as shallow marsh in both
scenes (see figure 1).

X10 0 1
X 0.0

T_ I

Figure 1. An example of project rating error, where a small area is not registered
perfectly in all imagery. (Each box represents one pixel; X, small area in one scene;
0, same small area in comparison scene.)

Another procedure is similar to the procedure DU is using, but replaces the linear
transformation with a high-order polynomial transformation. The number of control
points required is dependent upon the order of the polynomial transformation used. DU
is not using the polynomial transformation but believes that when 100 or more of its
control points are collected and a third or fourth order polynomial is used, pixel-to-pixel
registration can be obtained.

Using connective components techniques, the basin in figure 1 would be identified as a
six-pixel basin. In the Ent year, four pixels are wet and two pixels are dry. In the
comparison year, the same four pixels are wet and two are dry. Consequently, no change
has occurred in the wetland. Even if small change is reported because of registration
errors or classification errors, only those basins exceeding a threshold of change would
be classified as changed basins.

Earth Satellite Co=ratio

For imagery acquired over the U.S., using the most recent USGS 1:24,000-scale
topographic quadrangles, image georeferencing RMS errors can be often less than one-
half pixel (5 to 15-meters, depending upon image source). Earth Satellite Corporation
routinely georeferences imagery to subpixel accuracies. Earth Satellite's standard
procedure in change detection activities is to georeference the first image to the best
available map source, then to register the second image to the first via image-to-i
LI I H

correction. This minimizes displacement to as little as one-half pixel between the image
dates. Theoretically, a pixel-by-pixel analysis could be conducted; in reality, one would
expect to encounter edge differences of up to one-half pixel on unchanged parcels
between image dates.

To avoid numerical problems with the polynomial approach and remove the dorhiinant
error source, ephemeris error, along with smaller errors such as clock error, attitude bias,
and spacecraft-instrument misalignment bias, the Earth Satellite Corporation uses the
following procedure:

0 Perform a systematic geometric correction on a single band (usually
Band 4).

0 Use USGS (or customer-supplied, non-U.S.) maps to identify control points
within the scene.

0 Given the error in location of the control points from their correct
position, determine the errors in the ephemeris of the spacecraft.

0 Correct the ephemeris error and regenerate the geometric correction
model coefficients.

SRSC

As the first step in the georeferencing process, SRSC generally performs image-to-image
registration. One image is chosen as the base and the remaining images are registered
to that base. This is different from georeferencing each scene to a base map, where a
half-pixel error in one scene and a half-pixel error in another scene could potentially
result in a composite error of one pixel, which can be significant when doing change
detection. By performing image-to-image registration, any error that occurs during the
georeferencing process is consistent for all the images since they all now overlay each
other and the same georeferencing equation is used for all scenes.

Secondly, SRSC georeferences data using a nearest neighbor resampling routine after the
raw data has been classified. Georeferencing after the classification ensures that pixel
values have not been altered to such a degree that small areas and borders cannot be
correctly classified. But even when using a nearest neighbor resampling routine, SRSC's
analysts check the georeferenced classified image against the nongeoreferenced classified
image to check for potential alterations produced by the georeference transformation.

NCQ

NCG has rectified SPOT and TM data to an RMS error of 0.7 pixels or less using three
or four ground control points per quad, classified respective scenes, performed change
detection on as many as four scenes over the same area, and experienced no known
registration related problems. Film-based orthophotoquads are preferred for picking
ground control points. Ground registration is the most time consuming step in

processing satellite data because of the amount of time required to select ground control
points.

17. How do you navigate on the ground to locate and identify 1 to 6 or more pixels
depicting change?

Answer:
Identifying the precise location of one to six pixels on the ground can be very
difficult. Pixels adjacent to readily identifiable fixed features (such as the
intersection of section lines) can be readily identified. However, trying to use the
shape of a wetland on the ground to identify a particular pixel is difficult and can
lead to errors. The wet border of wetland can rapidly change from week to week
and certainly changes seasonally and annually. Such borders are not usable to
identify a particular pixel on the ground.

Ducks Unlimited

Use of the GPS offers a potential for pixel identification on the ground. One of DU's
employees using the GPS in Alaska felt that he could indeed identify the location of
individual pixels. Other employees used a GPS in Mexico and felt that they could not
rely on the GPS to identify the location of individual pixels. In Mexico, DU employees
identified various road intersections and obtained the UTM coordinates for the road
intersections using the GPS. They also obtained the UTM coordinates for the same road
intersections from various published map sheets at a scale of 1:50,000. Differences
between the UTM coordinates obtained from a Mexican map and from the GPS ranged
from 26 meters to over 1200 meters. Using the GPS in Alaska, it only took a few
minutes to obtain the UTM coordinates. In Mexico, employees frequently waited 15
minutes or longer before data from three GPS satellites were obtained. The longest wait
before obtaining data from three GPS satellites occurred early in the mornings and late
in the afternoons. The signals from the GPS satellites may have been degraded by the
U.S. Department of Defense when DU was in Mexico. By using two GPS receivers, one
located on a benchmark, it should be possible to determine the location of pixels on the
ground.

Earth Satellite Colporation

Earth Satellite Corporation has successfully identified change polygons using
georeferenced imagery. Pixel column and row locations of a wetland polygon centroid
are translated into their UTM or latitude and longitude coordinates. Using GPS
receivers and maps and imagery for corollary reference, it is then possible to locate the
pixels on the ground within the accuracy of the GPS receiver. In the case of curvilinear
or oddly shaped wetlands, where the centroid falls outside the polygon, a GIS editing
step would be used to move the coordinate inside the polygon. This GIS edit step could
also select any particular points or clusters that are of specific interest to the analyst
from visual examination.

18. How is NWI digital data processed so that they can be interfaced with TM digital
data used by DU, and is this process done on a unique basis or a routine basis?

Answer:
DU has interfaced NWI digital data with satellite data in two ways. For primarily
visual comparisons, DU overlays the NWI vector data directly onto its raster data.
These types of overlays of vector data onto raster data are less than ideal when
using a monitor on which one raster cell is presented as one pixel on the monitor.
Using these techniques, the NWI vectors totally obscure the raster data of small
wetlands. Using a plotter, however, each raster cell of the satellite data can be
represented by a 2 by 2, 3 by 3, 4 by 4, or greater block of plotter pixels. The
NWI vector lines only require a single plotter pixel. Therefore, fewer of the cells
from the raster satellite data will be obscured by the NWI vectors.

For analytical work, DU converts NWI data to raster format at the same
resolution as its raster satellite data. Using NWI data as fact, DU determines the
maximum extent of each wetland basin. For each wetland basin, DU measures
the acreage of various NWI wetland types and the acreage of wetland types
derived from satellite data. These data are maintained in a wetland basin data
base. By using both the NWI data and the satellite-derived wetland information,
DU can better evaluate the wetland as waterfowl habitat. As an example, two
hypothetical basins of the same size might be reported by NWI as 100 percent
PEMC (palustrine, emergent, seasonally flooded). For basin A, DU's satellite
data recorded 60 percent of the basin as shallow marsh and 40 percent of the
basin as deep marsh. For basin B, DU's satellite data reported 60 percent of the
basin as shallow marsh and 40 percent as nonwetland. Basin A, at the time that
the satellite data were collected, was providing better waterfowl habitat than basin
B.

SPOT's smaller pixel size for 20-meter multispectral data would help to alleviate
the problem that DU reported when overlaying vector data over raster data. A
one-pixel wetland area interpreted from Landsat TM 30-meter data would most
likely occupy from two to four pixels with SPOT 20-meter data. A rasterized
polygon boundary (from vector data) of one-pixel SPOT 20-meter data would no
longer obscure an entire one-pixel TM 30-meter wetland.

A merged and properly georeferenced data set incorporating SPOT panchromatic
(10-meter) with Landsat TM (30-meter) data would allow for easier and more
accurate registration of satellite data with the NWI vector data base, due to the
enhanced geometric fidelity of the satellite data.

Mar3LIand Department of Natural Resources

In April 1989, the State of Maryland enacted the Nontidal Wetlands Protection Act.
This law established a program to prevent net losses of nontidal wetlands. One of the
program's first requirements was creation of new maps showing nontidal wetlands and
relating them to identifiable ground features. The maps also needed to show areas that

were designated as Wetlands of Special State Concern that contain rare, threatened, or
endangered species, or that have unique habitat value. In December 1989, the MD-
DNR completed a project using SPOT panchromatic imagery as a georeferenced base
and overlaid it with NWI digital wetlands data.

It was not possible to create a series of orthophoto base maps for the State in this period
of time. It would have been inappropriate to use uncorrected aerial photography. SPOT
10-meter panchromatic data provided imagery in time to meet project deadlines and at a
lower cost than aerial photography. SPOT imagery can be printed at 1:24,000 scale and
provides ground resolutions that allow the average person to orient themselves on the
ground and identify features with which they are familiar.

The full scenes provided by SPOT were reformatted to match the USGS 7.5-minute
topographic quadrangle series for Maryland. This process involved extraction of the
individual quad images, rotation to true north, and registration to existing map vector
data sets. Where the overlap between scenes was not great enough to allow extraction
of a full 7.5-minute quad from one scene, portions of two or three scenes were extracted
and merged to create one 7.5-minute base map. The NWI data files, as well as files of
the transportation network, hydrology, and place names were procured in 7.5-minute
quadrangle format. These data layers were plotted into the image. raster and then
converted to a print file that could be used for on-demand printing of paper maps using,
an electrostatic printer.

This work was, a one-time effort that provided the basis for a tum-key GIS that is
currently in use for project review and planning purposes.

19. Do you agree with the following statement made by Dr. Gregory T. Koeln,
Director of DU's Habitat Inventory Program?

"Satellite data are best used as a monitoring tool for wetlands and are a poor tool
for establishing baseline data on wetlands. Combining NWI digital data with
satellite data provides much greater information than either product used alone."

Answer:
Most people would agree in general with Dr. Koeln's statement.

Ducks Unlimited

Dr. Koeln replied "Yes, but I wish I had better stated my point. I am not certain of the
source of the above statement, but I have expressed the sentiment of this statement
many times. Perhaps. I can better describe my views thus:

'Current satellite capabilities are best used for monitoring wetlands and evaluating
selective functions of wetlands (i.e., evaluating waterfowl habitat). As a
monitoring and evaluation tool of wetlands functions, satellite data are best used
in conjunction with NWI digital data. Combining NWI digital data with satellite

data for evaluating wetland functions provides much greater information than
either product used alone. Current satellite technology cannot delineate and
describe wetland types as accurately as the procedures used by NWI. Where NWI
data does not exist, the currently available satellite technology should not be used
if the objective is wetland delineation. However, if the objective is to evaluate
wetland functions at a point in time (i.e., available waterfowl habitat at the time
the data were collected), current satellite technology offers an economically
feasible solution for rapidly appraising selected wetland functions."'

SRSC-

SRSC's biggest objection to Dr. Koeln's statement deals with his comment that satellite
data "is a poor tool for establishing baseline data on wetlands." SRSC kindly disagrees
with this statement, because the SCS is now using SRSC's wetland classifications for the
States of Mississippi and Arkansas as their wetland base for administering the 1.990 Farm
Bill. Also, their preliminary'work in North Dakota demonstrated the possibility that
satellite data could be used to establish a wetland base, especially in the agricultural
prairie areas where wetlands have not been inventoried. SRSC cannot yet provide any
statistical insight into the comparison of their mapping efforts with those of NWI, but
they believe their wetland mapping compares favorably with the NWI effort in the SRSC
study area in North Dakota. They also believe that satellite data are a useful tool for
monitoring change in wetlands because they have done this work for the SCS to identify
converted wetlands. The idea of integrating satellite data with NWI data is appealing
because the NWI Would serve as an excellent mask to locate wetlands prior to processing
the satellite data.

20. At what point can we expect to have permanent storage capability so as to
maintain captured digital data on an indefinite basis?

Answer:
CD-ROM technology offers the greatest potential for maintaining near-permanent
storage of digital data.

Until there is a storage media capable of lasting a human lifetime or longer, there
will always be a need to perform archive maintenance. EDC has a project to
copy the Landsat 4 and 5 archive to a helical scan magnetic tape. This activity
will begin in October 1992. The computer system is now being built to perform
this activity.

EOSAT will be working with EDC personnel in this project and expect to initiate
a similar project for EOSATs own data in the near future. Landsat 4 and 5 data
will be available from 1982 to present.

There have been recent advances in recording media. There is now an optical
-tape that has a very long predicted shelf life. It is in use at Landsat's Canadian

receiving station and is about to be installed in the European Space Agency
-facilities.

The tradeoffs usually considered in archive media are long term costs and
retrieval speed. At the moment, tradeoff studies are making EOSAT lean toward
helical scan tape recording.

EDC

EDC believes the issue is not the existence of near-permanent storage media, but rather
media with a moderate lifetime (e.g. 10-20 years) from which data can be rapidly and
cheaply transcribed to the next generation of storage media, and for which hardware
necessary for the transcription can be readily available. EDC expects that no matter
what media are chosen today for archive storage, a new, more desirable media. will be
available in 10-20 years, and it will want to then convert to that media. EDC's biggest
problem with the existing Landsat archive has not been media performance, but rather,
the fact that very specialized hardware was required to transcribe the data and, in the
case of the older Wide-Band Video tapes, the hardware system is no longer functional.

EDC hopes to have converted all @Landsat 1 through 5 data to helical scan magnetic tape
within the next five years. Once that is accomplished, subsequent transcriptions will be
easier, as the new media will greatly reduce the volume of the data and increase the
speed with which the data can be transcribed.

SPOT

SPOT is now implementing an archiving program that uses CD-ROM. A study is
currently underway at SPOT that will likely result in an in-house capability to deliver
products on CD-ROM later this year. Already several projects on CD-ROM have been
delivered to clients, using outside vendors to convert the data to CD.

21. Data Quality Concerns

Ducks Unlimited

DU returned the EOSAT TM tape they had received on December 7, 1991. The scene
was, returned because of numerous line drops and extensive salt-and-peppering. The
replacement tape had not been received by DU as of January 14, 1992.

DU typically receives a product from EOSAT 6 to 8 weeks after the order is placed.
DU is greatly concerned over scenes received with line drops or salt-and-pepper data
gaps. Rather than return nearly every scene recently purchased, they have found it more
efficient to correct these problems in-house. When an unacceptable TM tape is
immediately returned to EOSAT for replacement, the replacement typically takes longer
to receive than did the original order. DU has waited nearly a year for some of its
replacement scenes, and often the replacement scene had numerous line drops and salt-

and-pepper data gaps. Fortunately, the data gaps in the first delivery did not occur at
the same location as the data gaps in the replacement, and DU was able to patch an
acceptable scene together.

DU returned many of the scenes that it had purchased from 1984 through 1986 because
of data gaps. EOSAT always provided a replacement and after one or more
replacements, DU was usually able to create a complete image. From 1987 through
1989, DU returned few if any scenes, thinking this problem had been resolved. In 1990,
however, these data gaps started to occur again.

EPA

Ross Lunetta, at EPA, reported he had similar problems and told the group he did not
plan to purchase any TM data until it was available from Landsat 6. He thought the
data problem was with the satellite itself. Dr. David Fischel, Chief Scientist for EOSAT,
responded that the problem was not with the satellite, but with the processing unit.

EOSAT

The EOSAT computer system in use until October 1, 1991, was the original system built
in 1982. Many of the problems associated with the bad data were caused by this aging
equipment. The salt-and-pepper effect was principally caused by a deteriorating array
processor. EOSAT has experienced a dramatic decrease in rejects since October 1, 19.91.
In the new system, EOSAT is able to perform considerably more quality assurance
checks to catch bad data before it goes to the customer, but EOSAT is still shaking out
some insidious bugs. The following problems remain:

Dropped scan lines (about 16 output image lines), which can occur for
scans that are acquired with the antenna looking close to the horizon.
Some salt-and-pepper also occurs with low antenna angle.

Dropped "minor frames" (16 lines high and 16 to 32 pixels long), which are
usually caused by tape aging.

The TM tape DU received from EOSAT on December 7, 1991, was produced during the
turnover to the new system. Just before EOSAT turned the new system on, the old
system failed. The data set that took so long to process was on a particularly
obstreperous tape. EOSAT occasionally comes across a tape that is very difficult to read
correctly.

REFERENCE

1. Cowardin, Lewis M., and others, 1979, Classification of wetlands and deepwater
habitats of the United States: Washington, D.C., U.S. Government Printing Office.

Appendix A

Wetland Classes Used in Waterfowl Habitat and Wetland Inventory

Ducks Unlimited

Listed below are the wetland classes that DU uses in its waterfowl habitat inventory
work in the Prairie Pothole region of Canada and the United States. These classes
cannot be separated strictly by their spectral characteristics. DU uses ancillary data and
visual interpretation of the data using their software system, DISP/TRAIN/EDIT. For
example, mud flats may be confused spectrally with various other types of bare soil;
however, using visual interpretation skills and the DISP/TRAIN/EDIT software, skilled
image interpreters can delineate bare soil categories adjacent to wetlands or occurring on
known wetland areas (from ancillary data) as mud flats.

WETLAND CODE DESCRIPTION

Wetland

100 Undetermined Wetland
110 Open Fresh Water
ill. Open Saline Water
112 Open Turbid Water
120 Deep Marsh
130 Shallow Marsh
140 Wet Meadow
150 Mud Flat
160 Dry Wetland
170 Forested Wetland
180 Riverine Water

Bogs

200 Undetermined Bog
210 Open Bog
220 Treed Bog

In the last six months, DU began a waterfowl habitat inventory program in Alaska. They
currently try to use the following wetland cover classes, a simplification from Cowardin
and others (1979):

WETLAND CODE DESCRIPTION

ElAB Estuarine Subtidal Aquatic Bed
E10W Estuarine Subtidal Open Water
E2EM Estuarine Intertidal Emergent

A-1

DU wetland cover classes for waterfowl habitat inventory program in Alaska (continued)

WETLAND CODE DESCRIPTION

E217L Estuarine Intertidal Flats
L10W Lacustrine Linmetic Open Water
L2AB, Lacustrine. Littoral Aquatic Bed
L2FL Lacustrine Littoral Flats
PAB Palustrine Aquatic Bed
PEMFL Palustrine Emergent-Flat
POW Palustrine Open Water
PUBX Palustrine Unconsolidated Bottom
Excavated
R Riverine
RX Riverine Excavated
UPAGR Upland Agriculture
UPBAR Upland Barren
UPDEV Upland Developed
UPRAN Upland Range
UPSS Upland Scrub-Shrub

SRSC

Listed below are the wetland classes used by SRSC.

Delta Region of Mississippi and Arkansas

Farmed Wetlands
Natural Wetlands
Converted Wetlands
Prior Converted Wetlands

Prairie Pothole region of North Dakota

Open Water
Deep Marsh
Shallow Marsh
Shallow Marsh (located in buffer zone from open water)
Vegetated Potholes (natural vegetation patches surrounded by bare soil of.
an agricultural field)

Coastal Plain of South.Carolina

Palustrine Emergent
Palustrine,Woody

A-2,

Appendix B

SPOT Image Product Fee Schedule
When comparing the price lists for SPOT and EOSAT data it is important to realize that
a fufl SPOT scene covers approximately one-eighth the area of a M Landsat scene.
SPOT Product Fee Schedule Effective March, W2

1. Standard SPOT Products
Panchromatic/Muitfspectral

Computer Compatible Level IA. IB (Full Scene) $2.450
Tapes (CC71 6250 or 1600 bpi

Film Level 1A, 16
(30% discount w/corre- 1:400.000 (Full Scene) $1,800
1200,000 (Full Scene) $1.Soo
sponding CCT) 1200,000 (1/4 Scene) $1,800
1:100,000 (1/4 Scene) $1,800

When ordered alone $950
Photographic Prints
When ordered with corresponding $300
Level 1 CCT or film

2. SPOTView@ GIS related products
Dlgftal Photographic Print

7.5 - (corresponds to USGS 7.5 minute $950 $950
map series) (1:24,000 7.5 XS
P - Panchromatic XS - Munispectral not available)

15 - Four 7.5 minute SP0TVlews $2.000 $2,000
creating 15 x 15 minute area (1:50,000 6-r
P - Panchromatic XS - MuRLSWral 1:63,360)

FS - Full SPOT scene $3.000 $3,000
(37 x 37 miles) (1:100,000)
P - Panchromatic XS - Multispectraj

I Other SPOT Products
Digital P ho tographt FlimlPrint

SPOT Digital Terrain Model (OTM)
(604/6-100% of full scene size/ $15.000
minimum of 820 sq. mi.)

SPOT Quarter scene DTM $10.000

SPOT BasinViewrm - complete
coverage of any geologic basin in the world $1.80/sq. mi.
B/W only minimum 2,500 sq. mi. W

4. Almaz Radar Image Products

Standard Product - Approx. 1:100,000 scale photographic $1,600 each
print - mosaic of 6 image strips covering 40 x 40 km area

Specialty Products -
Level A CCT 1600 bpi density $1,600 each
Level 8 CCT 6250 bpi density $2.400
Level C CCT 1 $2.800
9" x 9' negative of level B AJmaz image $2,400
9" x 9' negative of level C Almaz image $2,800

5. Promotional Products

Promotional Data Sets Digital Photographic

SPOT Education and Evaluation Data Set (S.E.E.D.S.) $900 $ 90 (slides)
(33 subscene set)

spor Art- $300 (unmounted)
S400 (mounted)

SPOT Posters $25 (in mailing tube)

SPOT User's
Handbook (3 volumes) - $150

6. Satellite Acquisition Programming
Standard High Priority

$2,000+
Programming Fee $6001scene $3001attempt
(does not include final product price) (up to 10 attempts)

Fees DO NOT include Shipping.
AD licensed SPOT Data YAN be delivered 'FOB SPOT image Corporation.* (Reston, VA).
Scene shiMng for all standard products included at no charge.
Call for information on duplicate copy prices, rush services. processing options, photographic scales
and non-standard products.

Additional information can be found in the document SPOT Prodticts and Services.
Please note: All prices subject to change without notice.
SPOT Image CoMoration
1897 Preston White Drive Reston, Virginia 22091-4368
(703) 620-2200 - Fax (703) 648-1813

B-2

Appendix C

EOSAT Product Fee Schedule

When comparing the price lists for SPOT and EOSAT data it is important to realize that
a fall SPOT scene covers approximately one-eighth"the area of a full Landsat scene.
PRODUCTS PRICE SHEET
(Effective octotw 1. 1991)

Observation Satellite ComPany - 4M Forbes Boulevard Lanham, MO 2OMS-9954

DIGITAL PRODUCTS MAP ORIENTED

Product Code Price Copy

FuH Scene - Terrain Corrected

6250 Bpi CCr TWMr6F $5950 $90
gmm Cartridge TM"NF 5950 90

Full Scene - Precidan Corrected

6250 Bpi CC`r TMFT6F $55M S90
8aun Cartridge TNMNF 5500 90

Full Scene - System Corrected

6250 Bpi CC`r TMST6F $4400 $90
8nun Cartridge TMS8NF 4400 90
-------------
S-nb-sm-neTic;@ - - - - - - - - - - - - - - -
X 100. km) - Terrain Corrected

6250 Bpi CCr TMIT6Q $4650 $90
1600 Bpi Ccr TmTnQ 5115 90
8rrun Cartridge TMTSNQ 4650 90

Subscene (100 km X 100 km) - Precision Corrected

6250 Bpi CC`r TM?T6Q $4200 S90
1600 Bpi CC`r TmprriQ 4620 90
8mm Owaidge TMPSNQ 4200 00

Subscene (100 km X 100 km) - System Corrected

6250 Bpi CCT TMST6Q S3100 $90
1600 Bpi CC`r TMSTIQ 3410 90
Sam Cartridge TMS8NQ 3100 90
----------------------------------
Map Sheet - Terraix Corrected

6250 Bpi CC`r TNw6M $4= $90
1600 Bpi CCT TMTMM 4455 90
grain Cartridge TMnNM 4050 90

Map Shed - Precidon Corrected
6250 Bpi C& TN[FT6?A $3600
1600 Bpi CCT TMPTIM 3960 90
Sam Catidge TMPSN'M 3600 90

Map Sheet - System Corrected

6250 Bpi CC`r TMS`T6M S2500 S90
1600 Bpi CC`r ThfSTIM 2750 90
gazin Cartridge TMS8NM 2500 90

C-1

DIGITAL PRODUCTS PATH ORIENTED

Product Code Price Copyt
Full Scene - System Corrected
6250 Bpi CCT TPST6F S4400 $90
1600 Bpi CCT TPSTIF S4840 90
8mm Cartridge TPS8NF 4400 90

Subscene (100 km X 100 km) - System Corrected

6250 Bpi CCT TPST6Q $3100 $90
1600 Bpi CCT 17STIQ 3410 90
8mm Cartridge TPS8NQ 3100 90

PHOTO PRODUCTS MAP ORIENTED

Color Film Subscene (100 km X 100 km) - System Corrected

1:500,000 Positive and NegativeTransparencies TMSCTQ $2700
1:500,000 Print TMSCIQ 2W*
1:250,000 Print TMSC2Q 200*
1:100,000 Print TMSC5Q 2000

Must Mullen Trmgpam= first
PHOTO PRODUCTS PATH ORIENTED

Color Film Full Scene - System Corrected

1: 1,000,000 Positive and Negative Transparencies TPSCTF $2700
1: 1,000.000 Print TPSCIF 200@
1:500,000 Print TPSC2F 2W*
1:250,000 Print TPSC4F 2W*

Must Mashase Tymungarency fint

MULTI SPECTRAL SCANNER (MSS) PRODUCTS
ORBIT-PATH ORIENTED, SYSTEM CORRECTED ONLY. FULL SCENE ONLY

DIGITAL

625D Bpi Onease specify BSQ or BL) NfPST6F slim
16M Bpi Mleew specify BSQ or BIL) hdPSTIF 1.000 90

PHOTOGRAPHIC PRODUCTS

1: 1.000.OW Color Positive Transpitlency NMC77 $600 $120
1:250,000 Color Print NMSC4F LOW 2w
1: 1,000.000 B/W Positive Transparency hVSBTF 155 31
1: 1,000.000 B/W Negative Transpuency h1PSBNF 175 35
1: 1.000,000 B/W Print N4PSBlF 95 19

NOTE' MW prim listed above for MSS a*y to data aoqukW within 24 mondis of do date am oider is receive&
For Price quoudow derived ftm older MSS dwa. caum SWAT Cuawmer Services.

t Copy Prke appose whem Go d W" "111mal lw@w@
All priew are ps led Is U.S. Deasm

C-2

EOSAT'S PRODUCT LINE DEFINITIONS

With the introduction of the new image processing system on October 1, EOSAT changed its product codes and its product
descriptions to better reflect the system's capability. The product catalog uses the new names; the box highlights the major
changes. The terms are explained further below.

OLD NEW

Standard Path Oriented, System Corrected
Geocoded Map Oriented, Precision Corrected
Movable Scene Floating Scene
Quarter Scene Or Quadrant Scene Subscene (100 X 100km)
(100 X 85km)

Names have been chosen to encourage customers to tailor their orders to their individual needs. Every order is a custom
order. The phrase "non-standard product" has been banished; if EOSAT can produce the requested item, the order will be
filled. The more popular products are listed on the current price sheets available from Customer Services.

Scenes, Subscenes - Digital TM products are available as full scenes, subscenes and map sheets. TM photo products are
available as full scenes and subscenes, and MSS products are sold as full scenes.

Floating scenes - Scenes crossing row boundaries are floating scenes, identified by row numbers with decimals. Path 34/
row 34.2 begins 20% below the top of path 34/row 34. Except for MSS photographic products, a customer may
order data from any area within a path.

Imagery orientation - Landsat products are correlated to the Earth's surface in one of two ways:
Path-oriented products have the spacecraft's orbital orientation, with north off to one side of the image. the
category includes what used to be called the standard digital product.

Map-oriented products are usually north up. This adjustment facilitates co-registration of Landsat imagery with
other digital data, as a geographic information system. The customer can select pixel size, map projection and
Earth ellipsoid model.

Data corrections - Three lecels of Landsat data correction are available.
System-corrected - All products, digital and photographic, are radiometrically and geometrically corrected. Radio-
metric corrections are made within a band so that a given detector value always represents the same radiance level
for the scene. Geometric corrections reorient the image data to compensate for the Earth's rotation and variations in
spacecraft position and altitude. Landsat data has always been system corrected.

TM Map Oriented Digital Products can be geometric adjusted to two additional levels of positional accuracy.

Precision-corrected data incorporates ground control points such as road intersections to relate the spacectaft's
predicted position to its actual geodetic position. This kind of data used to be called geocoded data. EOSAT has
ground control points for areas within the United States. Customers must furnish topographic maps of 1:50,000 scale
or better for all other areas. Maps will be returned.

Terrain-corrected data eliminates the distortion that results from recording a three-dimensional view in two
dimensions by including relief adjustments from a digital elevation model; this kind of correction is especially
helpful if the height difference across a scene is more than 500 feet. EOSAT has models for areas within the United
States. For other areas of the world, customers must supply digital terrain model.

January 1993

C-3

Appendix D

Acronyms

C-CAP Coast Watch Change Analysis Program, NOAA
DEM digital elevation model
DLG digital line graph
DN digital number
DU Ducks Unlimited
EDC EROS Data Center, USGS
EOSAT Earth Observation Satellite Company
EPA U.S. Environmental Protection Agency
EROS 'Earth Resources Observation Systems
FGDC Federal Geographic Data Committee
FWS U.S. Fish and Wildlife Service, U.S. Department
of the Interior
GIS geographic information system
GPS global positioning system
HRV high resolution visible (refers to sensors in SPOT
satellite)
In meter
MD-DNR, DNR State of Maryland, Department of Natural
Resources
MSS multispectral. scanner
NASA National Aeronautics and Space Administration
NCG SCS National Cartographic and GIS Center
NWI National Wetlands Inventory
NOAA National Oceanic and Atmospheric
Administration, U.S. Department of Commerce
PRE project rating error
RMS root mean square
SAV submerged aquatic vegetation
SCS Soil Conservation Service, U.S. Department of
Agriculture
SPOT Satellite Pour I'Observation de la Terre
SRSC Space Remote Sensing Center, NASA
TM Thematic Mapper
U.S. United States of America
USGS U.S. Geological Survey, U.S. Department of the
Interior
UTM Universal Transverse Mercator (a grid system
based on the transverse mercator map
projection)
/4m micrometer or micron (one-millionth of a meter)

D-1

338184