Comparison and verification of dynamical and statistical Lake Erie storm surge forecasts

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

COASTAL ZONE C
INFORMATION CENTER 4

NOAA Technical Memorandum NWS TDL-69

c't -
*

"'NTES O� F

COMPARISON AND VERIFICATION OF DYNAMICAL AND
~ STATISTICAL LAKE ERIE STORM SURGE FORECASTS

Techniques Development Laboratory
Silver Spring, Md.
J November 1979

QC
851
~. .

*:U6 NATIONAL OCEANIC AND National Weather
T3 ATMOSPHERIC ADMINISTRATION Service
no.69
Ai . gatCLL ;20

NOAA TECHNICAL MEMORANDUMS

National Weather Service, Techniques Development Laboratory Series

The primary purpose of the Techniques Development Laboratory of the Systems Development Office is
to translate increases of basic knowledge in meteorology and allied disciplines into improved operating
techniques and procedures. To achieve this goal, the laboratory conducts applied research and develop-
ment aimed at the improvement of diagnostic and prognostic methods for producing weather information.
The laboratory performs studies both for the general improvement of prediction methodology used in the
National Meteorological Service and for the more effective utilization of weather forecasts by the ulti-
mate user.

NOAA Technical Memorandums in the National Weather Service Techniques Development Laboratory series
facilitate rapid distribution of material that may be preliminary in nature and which may be published
formally elsewhere at a later date. Publications 1 through 5 are in the former series, Weather Bureau
Technical Notes (TN), Techniques Development Laboratory (TDL) Reports; publications 6 through 36 are in
the former series, ESSA Technical Memorandums, Weather Bureau Technical Memorandums (WBTM). Beginning
with TDL 37, publications are now part of the series NOAA Technical Memorandums National Weather
Service (NWS).

Publications listed below are available from the National Technical Information Service, U.S. Depart-
ment of Commerce, Sills Bldg., 5285 Port Royal Road, Springfield, VA 22161. Prices on request. Order
by accession number (given in parentheses). Information on memorandums not listed below can be obtained
from Environmental Science Information Center (D822), NOAA, 6009 Executive Boulevard, Rockville, MD
20852.

ESSA Technical Memorandums

WBTM TDL 17 Second Interim Report on Sea and Swell Forecasting. N. A. Pore and W. S. Richardson, Janu-
ary 1969, 7 pp. plus 10 figures. (PB-182-273)
WBTM TDL 18 Conditional Probabilities of Precipitation Amounts in the Conterminous United States. Don-
ald L. Jorgensen, William H. Klein, and Charles F. Roberts, March 1969, 89 pp. (PB-183-144)
WBTM TDL 19 An Operationally Oriented Small-Scale 500-Millibar Height Analysis Program. Harry R. Glahn
and George W. Hollenbaugh, March 1969, 17 pp. (PB-184-111)
WBTM TDL 20 A Comparison of Two Methods of Reducing Truncation Error. Robert J. Bermowitz, May 1969,
7 pp. (P8-184-741)
WBTM TDL 21 Automatic Decoding of Hourly Weather Reports. George W. Hollenbaugh, Harry R. Glahn, and
Dale A. Lowry, July 1969, 27 pp. (PB-185-806)
WBTM TDL 22 An Operationally Oriented Objective Analysis Program. Harry R. Glahn, George W. Hollen-
baugh, and Dale A. Lowry, July 1969, 20 pp. (PB-186-129)
WBTM TDL 23 An Operational Subsynoptic Advection Model. Harry R. Glahn, Dale A. Lowry, and George W.
Hollenbaugh, July 1969, 26 pp. (PB-186-389)
WBTM TDL 24 A Lake Erie Storm Surge Forecasting Technique. William S. Richardson and N. Arthur Pore,
August 1969, 23 pp. (PB-185-778)
WBTM TDL 25 Charts Giving Station Precipitation in the Plateau States From 850- and 500-Millibar Lows
During Winter. August F. Korte, Donald L. Jorgensen, and William H. Klein, September 1969,
9 pp. plus appendixes A and B. (PB-187-476)
WBTM TDL 26 Computer Forecasts of Maximum and Minimum Surface Temperatures. William H. Klein, Frank
Lewis, and George P. Casely, October 1969, 27 pp. plus appendix. (PB-189-105)
WBTM TDL 27 An Operational Method for Objectively Forecasting Probability of Precipitation. Harry R.
Glahn and Dale A. Lowry, October 1969, 24 pp. (PB-188-660)
WBTM TDL 28 Techniques for Forecasting Low Water Occurrences at Baltimore and Norfolk. James M. Mc-
Clelland, March 1970, 34 pp. (PB-191-744)
WBTM TDL 29 A Method for Predicting Surface Winds. Harry R. Glahn, March 1970, 18 pp. (PB-191-745)
WBTM TDL 30 Summary of Selected Reference Material on the Oceanographic Phenomena of Tides, Storm Sur-
ges, Waves, and Breakers. N. Arthur Pore, May 1970, 103 pp. (PB-193-449)
WBTM TDL 31 Persistence of Precipitation at 108 Cities in the Conterminous United States. Donald L.
Jorgensen and William H. Klein, May 1970, 84 pp. (PB-193-599)
WBTM TDL 32 Computer-Produced Worded Forecasts. Harry R. Glahn, June 1970, 8 pp. (PB-194-262)
WBTM TDL 33 Calculation of Precipitable Water. L. P. Harrison, June 1970, 61 pp. (PB-193-600)
WBTM TDL 34 An Objective Method for Forecasting Winds Over Lake Erie and Lake Ontario. Celso S. Barri-
entos, August 1970, 20 pp. (PB-194-586)
WBTM TDL 35 Probabilistic Prediction in Meteorology: a Bibliography. Allan H. Murphy and Roger A. 'Al-
len, June 1970, 60 pp. (PB-194-415)
WBTM TDL 36 Current High Altitude Observations--Investigation and Possible Improvement. M. A. Alaka
and R. C. Elvander, July 1970, 24 pp. (COM-71-00003)
NWS TDL 37 Prediction of Surface Dew Point Temperatures. R. C. Elvander, February 1971, 40 pp.
(COM-71-00253)
NWS TDL 38 Objectively Computed Surface Diagnostic Fields. Robert J. Bermowitz, February 1971, 23 pp.
(COM-71-00301)

(Continued on inside back cover)

NOAA Technical Memorandum NWS TDL-69

COMPARISON AND VERIFICATION OF DYNAMICAL AND
�. ~STATISTICAL LAKE ERIE STORM SURGE FORECASTS

William S. Richardson
Techniques Development Laboratory
Systems Development Office
and
: David J. Schwab
Great Lakes Environmental Research Laboratory
Ann Arbor, Mich.

j 4
Techniques Development Laboratory
Systems Development Office
Silver Spring, Md.
November 1979

U-S. DEPARTMENT OF COMMERCE NOAP
COASTAL SERVICES CENTER
2234 SOUTH HOBSON AVENUE
CHARLESTON, SC 29405-2413

Property of CSC Library

~~~ ~UNITED STATES NATIONAL OCEANIC AND National Weather
DEPARTMENT OF COMMERCE ATMOSPHERIC ADMINISTRATION S D itce
Richard A. Frank, Administrator Richard E. Haligren, Director4 1

CONTENTS
Page

Abstract . . ..............

1. Introduction........................ 1

2. Causes of storm surges. ..................2

3. Forecast methods. .....................2

4. Comparison and verification. ................3

5. Conclusions and recommendations. ..............6

Acknowledgments .. .......................7

References .. .........................7

Tables ............................. 8

Figures ............................. 9

COMPARISON AND VERIFICATION OF DYNAMICAL
AND STATISTICAL LAKE ERIE STORM SURGE FORECASTS*

William S. Richardson
Techniques Development Laboratory
National Weather Service, NOAA
Silver Spring, Md. 20910

and

David J. Schwab
Great Lakes Environmental Research Laboratory, NOAA
Ann Arbor, Mich. 48104

ABSTRACT. The Great Lakes Environmental Research Laboratory
and the Techniques Development Laboratory have compared Lake
Erie storm surge forecasts produced by a dynamical and a
statistical method for several months in 1977 and 1978. The
dynamical method yields much better forecasts at Buffalo and
slightly better forecasts at Toledo.

1. INTRODUCTION

Storm-generated lake level fluctuations (storm surges) can cause serious
problems on Lake Erie. Flooding and shoreline erosion occur when positive
surges are superimposed on high lake levels. During times of low lake
levels, negative surges can be hazardous to navigation and disturb hydro-
electric power generation. The National Weather Service (NWS) is responsible
for preparing and issuing storm surge forecasts for the lake. Since 1969
NWS has used a statistically derived storm surge forecast technique, developed
by Richardson and Pore (1969) at the Techniques Development Laboratory (TDL),
to generate automated storm surge forecast guidance for Buffalo, N.Y.
and Toledo, Ohio. The automated storm surge forecast guidance has proved
to be a useful forecast tool.

The development of an automated objective method to forecast winds over
the Great Lakes by Feit and Barrientos (1974) prompted an investigation
by the Great Lakes Environmental Research Laboratory (GLERI) to determine
whether a dynamical storm surge model developed by Schwab (1978) could
take advantage of these computer generated winds to improve the storm
surge forecasts for Lake Erie. The dynamical forecast model is a two-
dimensional refinement of one that was proposed by Platzman (1967) but
not implemented operationally. Preliminary tests with this linear-
finite difference model were so encouraging that GLERL and TDL decided
to make objective comparisons between storm surge forecasts made with the
GLERL dynamical model and forecasts produced by the TDL statistical method.

*GLERL Contribution No. 191

The comparison and verification of dynamical and statistical forecasts
with measured surge data will be the basis for making a recommendation
to NWS as to which method or combination of methods should be used to
produce Lake Erie storm surge forecast guidance. Before we present this
comparison and verification, the causes of Lake Erie storm surges and two
methods used to forecast these surges will be discussed briefly.

2. CAUSES OF STORM SURGES

Lake Erie storm surges are generally associated with the passage of
extratropical storms through the Great Lakes area. Winter storms that
approach the Great Lakes from the central part of the country can cause
strong winds over Lake Erie. When these strong winds blow along the main
axis of the lake, storm surges are generated. Strong southwest winds
blowing over the lake cause a tilted lake surface--the water level is
elevated in the eastern portion of the lake (Buffalo end) and lowered at
the western end (Toledo). Storm surge heights and surface winds associated with
an extreme example of this situation are shown in figure 1. Occasionally
a storm will pass south of Lake Erie causing northeast winds over the lake.
When this happens, the slope of the water surface is reversed with elevated
water levels at the Toledo end of the lake. An extreme event of this type is
shown in figure 2.

Storm surges on Lake Erie are caused primarily by wind stress on the
lake surface. The effect of atmospheric pressure, which causes higher
water levels in areas of low pressure, is less important. Storm surges
are especially pronounced on Lake Erie because of its shallow depth and
geographic orientation.

3. FORECAST METHODS

3.1 Dynamical

The dynamical method uses impulse response functions to calculate the storm
surge height. The surge at a given time is calculated as the weighted
sum of forcing terms during some period before the specified time, i.e.

m n +

i=l j=l
4.
Here the surge at time k is h k, g-,', is the*-water level response at time j
due to an impulse from forcing station i, Tik-j is the forcing function at
station i and time k-j, m is the number of forcing stations, and n is the
length of the response function. The forcing function is calculated as
_+ 4.
Ti cVij IVij , (2)

where V-~ is the wind vector at station i and time j, and c is a dimensionless
constant, 4xl06

The response functions g-were calculated by means of a linear finite
difference numerical model of Lake Erie as described by Schwab (1978). To
take into account hourly changes in the forcing function, the response functions
are recorded as hourly values. We determined that a 36-h response function
was sufficient for Lake Erie storm surges. Water level responses at Buffalo
and Toledo were computed for forcing at the two Lake Erie wind forecast points
shown in figure 3 (numbers 3 and 4) plus Buffalo and Toledo. Forcing functions
for storm surge forecasts are calculated according to (2) with hourly winds
interpolated from the 6-h automated wind forecasts over Lake Erie (Feit and
Barrientos 1974) and the 6-h automated surface wind forecasts at Buffalo and
Toledo (Carter 1975). It was also shown by Schwab (1978) that when hourly
winds observed at seven airports around the lake were used as forcing functions,
the dynamical model produced excellent results (see fig. 4).

3.2 Statistical

The currently operational statistical method was developed by Richardson
and Pore (1969). Water level deviations from the monthly mean at Buffalo
and Toledo (storm surges) were correlated with analysed 6-h sea level pressure
at grid points of the National Meteorological Center's (NMC's) Primitive
Equation model (Shuman and Hovermale 1968), shown in figure 3. A screening
correlation program (Miller 1958) was used to find the best predictors of
storm surges. The regression equations have the form

n
hk =AO + E A. P., (3)
j=l -I J

where hk is the storm surge at time k, AO is a constant, A. is a regression
coefficient, Pi is the sea level pressure at a NM~C grid point with lag times
of 0, 1, 2, 3, 4, or 5 hours, and n is the number of predictors. In 1973 the
regression equations were rederived with sea level pressure and storm surge
data from 1940 to 1971. Six equations were derived for Buffalo, one for each
hour. Three equations were derived for Toledo because water level data,
which were used in the development of the equations, were available for even
hours only. Observed pressures at NMC grid points, 2, 3, 7, 8, 12, and 13
in figure 3 were used as predictors with lags of 0, 1, 2, 3, 4, or 5 hours.
Operational storm surge forecasts were made with sea level pressure forecasts
of the NMC Limited-area Fine Mesh II model (National Weather Service 1977).
We found that the use of multiple equations caused spurious oscillations in
the statistical storm surge forecasts. We remedied this problem by applying
only one equation for each 6-h interval between pressure forecasts. The
equation used was the one which predicted the greatest water level fluctuation
during the 01- to 48-h forecast period.

4. COMPARISON AND VERIFICATION

The dynamical and statistical methods were compared and verified with
measured storm surge data from November and December 1977, and April, November,
and December 1978. Storm surge data from these months were chosen because an
earlier study (Pore et al. 1975) showed that the most significant surges
occurred during these months. Also, there is generally little lake ice to
interfere with storm surge generation during November, December, and April.

The comparison and verification consist of analyses of measured and forecast
storm surge events and overall statistical measurements for the entire 5-month
period. Storm surge events are times when significant storm surges (magnitude
of the surge is 3 ft or greater) occurred or were forecast to occur by either
method at Buffalo or Toledo1. The statistical measurements consist of cor-
relation coefficients and root-mean-square errors (rinse's).

While 0- to 36-h surge forecasts were verified and compared, only the ver-
ification and comparisons of the 13- to 24-h forecasts are presented in this
paper, as we determined this period to be the most significant for operational
forecasts. These midrange forecasts are representative of the total range of
forecasts. observed and forecast storm surge graphs, correlation coefficients,
and rmse's are presented first for Buffalo and then for Toledo. The observed
and forecast storm surge heights are plotted every hour for Buffalo and every
2 hours for Toledo. Figures containing pairs of surge graphs depict the ob-
served surge and dynamical forecast in the top panel of each pair. The lower
panel of the pair shows the observed surge and statistical forecast for the
same event. The dynamical method forecasts water level deviations from the
average lake level while the statistical method forecasts surges relative to
station mean levels. Therefore, observed water level deviations are shown
relative to the lakewide mean level for the dynamical method and relative to
the station mean for the statistical method. Lake and station means differ
by no more than 0.5 ft. No attempt is made to distinguish storm surges from
seiches. All departures of the lake level at Buffalo and Toledo from an
averaged lake level or a station mean are referred to as storm surges. Time
is GMT.

4.1 Buffalo

Figure 5 compares observed peak surges with forecast peak surges for Buffalo.
This comparison, which is shown for storm surge events when the magnitude of
an observed or forecast peak surge is equal to or greater than 3 ft, reveals
the following:

1. The statistical method forecasts large negative
surges when none are observed (15 cases) and
overforecasts by 4 ft the one observed large
negative surge.

2. The statistical method overforecasts minor positive
surges by about 2 ft (5 cases).

3. Both methods underforecast large positive surges
by 2 to 3 ft (6 cases).

The probable causes for these discrepancies are examined in a later section.
Here we will only describe and summarize the results of the comparison and
verification.

1 Storm surge height is given in feet, the unit used operationally by the
National Weather Service.

The observed large negative surge at Buffalo, and three cases for which
the statistical method overforecasts minor negative surges, are shown in
figure 6. For these four events, the dynamical forecasts are in excellent
agreement with measured surges. Minor positive surges (less than 3 ft)
are overforecast by the statistical method on November 27, 1977, December 17,
1978, and December 25, 1978 (fig. 7). The dynamical forecasts are again in
good agreement with measured surges.

Figure 8 shows examples of peak positive surges which were underforecast.
The dynamical method underforecasts the first surge of the "double peak"
event (December 2, 1977) by more than 3 ft. The statistical method under-
forecasts the same peak by about 2 ft. While both methods hint at a
second peak, this peak is also underforecast by both methods. This is an
interesting case, in that the two peak surges are associated with only one
storm system. The second peak appears to be related to the natural period
of Lake Erie which is about 14 hours. The storm surge on December 25, 1977
is also underforecast by about 2.5 ft by both forecast methods. Storm
surges that occurred on November 18, 1978 and December 5, 1978 are underforecast
by about 5 ft by both methods. The December 21, 1978 event is underforecast
by 3 ft by the dynamical method, while the statistical method underforecasts
this event by 2 ft. Both methods forecast the peak surges to occur i to 4
hours too late.

The comparison of the Buffalo surge forecasts generated by the dynamical
and statistical methods may be summarized as follows: In a number of cases,
significant surges are underforecast by both methods. Negative surges are
forecast very well by the dynamical method, while the statistical method
overforecasts these surges by significant amounts. The statistical method
sometimes overforecasts small positive surges. We therefore conclude that
the dynamical method yields much better storm surge forecasts at Buffalo than
the statistical method. The summary statistics, shown in table 1, reinforce
this conclusion. The correlations between the dynamical forecasts and the
observed water level fluctuations are about 5 to 15 percent higher than the
correlations associated with the statistical method. The rmse's associated
with the dynamical method are 0.3 to 1.0 ft lower (25% to 65% lower) than
the rmse's associated with the statistical method.

4.2 Toledo

A comparison of observed peak surges with forecast peak surges for
Toledo storm surge events when the magnitude of the observed or forecast
peak surge was equal to or greater than 3 ft is depicted in figure 9. This
figure points out the following:

1. The statistical method overforecasts minor positive
surges by about 2 ft (4 cases).

2. Both methods are in fair agreement with the commonly
observed negative surges (20 cases) and the single
observed large positive surge.

The statistical method overforecasts minor positive surges. Examples are
shown in figure 10. Minor positive surges are forecast very well by the
dynamical method. Figure 11 shows the one observed large positive surge and
three typical negative surges at Toledo. Both methods forecast the positive
surge on December 5, 1977. On December 9, 1977 the dynamical method overfore-

casts the large negative surge. Both methods slightly overforecast the smaller
negative surge on December 25, 1977. The statistical method overforecasts
the negative surge on December 21. 1978. Both methods tend to forecast peak
surges 3 to 6 hours too late.

For Toledo we conclude that the dynamical method is slightly better than the
statistical. However, the superiority of one forecast method over the other
is not nearly as clear as in the Buffalo comparison. The correlation co-
efficients and rinse's shown in table 2 also support the conclusion that the
dynamical method works better. The correlations for the dynamical fore-
casts are 10 to 15 percent higher than for statistical forecasts. The rinse's
for the dynamical forecasts are 0.3 to 0.5 ft lower (20 to 40 percent lower)
than those for the statistical forecasts.

5. CONCLUSIONS AND RECOMMENDATIONS

As pointed out by Richardson and Pore (1969), the statistical method
produces good forecasts for storm surge cases which are similar to cases
used in development. The negative surge cases observed at Toledo and
some positive surges at Buffalo were forecast well by the statistical method.
However, the statistical method produces poor forecasts at both Buffalo and
Toledo for small surges due to southwest winds, and for most surges generated
by northeast winds. This is because the cases used to develop the statistical
regression equations were usually large surges associated with southwest
winds. Very few cases of northeast winds and no cases of small (less than
3 ft) surges were included in the development data. This often causes
the statistical method to "cry wolf" by forecasting a large surge when none
is observed. The infrequent case of northeast winds causes high water at
the western end of Lake Erie which is highly susceptible to flooding; so an'
accurate forecast for these cases is important.

Both the statistical and dynamical methods underforecast peak positive
surges at Buffalo (see figure 5). Note that during this test period the
distribution of peak positive observed surges at Buffalo is bimodal (6 cases
about 3 ft and 6 cases about 6 ft). Both the statistical and dynamical
methods are in some sense "tuned" to average surge conditions (3 to 4 ft).
The statistical method is "tuned" with the developmental data while the
dynamical method is "tuned" by the choice of the drag coefficient. Because
peak positive surges at Buffalo were higher than average, the forecasts
tended to be too low.

The accuracy of the dynamical method is also directly affected by the
accuracy of the wind forecasts. Figure 4 is an example of the results that
can be obtained when hourly winds from seven stations around the lake are
used to drive the dynamical model. Increased temporal and spatial re-
solution in the wind forecasts would result in even better storm surge fore-
casts by the dynamical method.

On the basis of this comparison and verification of dynamical and statistical
storm surge forecasts at Buffalo and Toledo we recommend that the dynamical
forecast method be used in place of the currently operational statistical
method. We also recommend continued verification of the forecasts pro-
duced by the dynamical method once they become operational. Significant

storm surge events that are not forecast adequately, or events that
are forecast to be significant when they are not, should be investigated
event-by-event to determine whether the inadequate forecasts are the fault
of the method or the fault of the wind forecasts that serve as input to the
method. These studies may give us insight into how the forecast method can
be improved.

ACKNOWLEDGMENTS
The authors thank N. A. Pore and D. B. Rao for their helpful suggestions in
writing this paper. We also wish to thank the National Ocean Survey for lake
level data, Gary M. Carter and Anna E. Booth of TDL for furnishing Buffalo
and Toledo wind forecast data, and Franklin A. Rosenstein also of TDL for
tabulating and processing storm surge data.

REFERENCES

Carter, G. M., 1975: Automated prediction of surface wind from numerical
model output. Mon. Wea. Rev., 103, 866-873.

Feit, D. M. and C. S. Barrientos, 1974: Great Lakes wind forecasts based
on model output statistics. Proceedings of the 17th Conference of
Great Lakes Research, Hamilton, Ont., 725-732.

Miller, R. G., 1958: The screening procedure. A statistical procedure
for screening predictors in multiple regression, Part II. Studies
in Statistical Weather Prediction, Contract No. AF19 (604) - 1590,
Travelers Weather Research Center, Hartford, Conn., edited by B. Shorr,
86-95.

National Weather Service, 1977: High resolution LFM (LFM-II). NWS
Technical Procedures Bulletin No. 206, National Oceanic and Atmo-
spheric Administration, U.S. Dept. of Commerce, 6 pp.

Platzman, G. W., 1967: A procedure for operational prediction of wind
set-up on Lake Erie. Technical Report No. 11, The University of
Chicago, ESSA Contract E-91-67 (N).

Pore, N. A., H. P. Perrotti, and W. S. Richardson, 1975: Climatology of
Lake Erie storm surges at Buffalo and Toledo. NOAA Technical Memo-
randum NWS TDL-54, National Oceanic and Atmospheric Administration,
U.S. Dept. of Commerce, 27 pp.

Richardson, W. S., and N. A. Pore, 1969: A Lake Erie storm surge fore-
casting technique. ESSA Technical Memorandum WBTM TDL-24, Enviro-
mental Science Services Administration, U.S. Dept. of Commerce, 23 pp.

Schwab, D. J., 1978: Simulation and forecasting of Lake Erie storm surges.
Mon. Wea. Rev., 106, 1476-1487.

Shuman, F. G. and J. B. Hovermale, 1968: An operational Six-Layer Primitive
Equation model. J. Appl. Meteor., 7, 525-547.

Table l.--Summary of correlation coefficients and rmse's associated with
the dynamical and the statistical Buffalo storm surge forecasts for
November and December 1977,and April, November,and December 1978.

Dynamical method Statistical method

Correlation rmse Correlation rmse
coefficient (ft) coefficient (ft)

November 20 - 0.74 0.99 0.70 1.33
December 27, 1977

April 2-27, 1978 0.61 0.60 0.58 1.68

November 5 -
December 31, 1978 0.71 0.74 0.62 1.39

Table 2.--As in table 1, except for Toledo.

Dynamical method Statistical method

Correlation rmse Correlation rmse
coefficient (ft) coefficient (ft)

November 20 -
December 27, 1977 0.63 1.22 0.54 1.50

April 2 - 27, 1978 0.69 0.73 0.64 1.23

November 5 -
December 31, 1978 0.78 0.72 0.71 1.24

BUFFALO WINDS

8
8 7.79

6
%4-
-4

40 \~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

0
-2
o , I
i~ -4.
BUFFALO
-6~ ! .---,- TOLEDO
-6I
FEBI5
-~FEB 15 16 17

TOLEDO WINDS

Figure l.--Storm surge heights and wind observations for the highest storm surge (7.79 ft) at Buffalo
for the period 1940-72. The dates are placed at the 1200 EST position on the storm surge
graphs (Pore et al. 1975).

TOLEDO WINDS

6 t tas~~~~~~~.29

E2 ,..,2"d

0
�.- -4
BUFFALO
--.- TOLEDO
-6
APR26 27 28
-~~ 1966

BUFFALO WINDS
Figure 2.--Storm surge heights and wind observations for the highest storm surge (5.29 ft) at Toledo
for the period 1940-72. The dates are placed at the 1200 EST position on the storm surge
graphs (Pore et al. 1975).

5Q0

Q NMCI Grid Points

LZGreat Lakes Wind
Forecast Points

450

� ~~~~~~~~~~Toledo , I

Figure 3.--Location of National Meteorological Center (NMC) Primitive Equation Model grid Points which
surround the Great Lakes and Great Lakes wind forecast Points (Feit and Barrientos 1974).

6 - * Buffalo /
4 o Toledo U'

E2
0
.6-2
-4

E
S~ l I I I I I I.
0
u -6-4-2 0 2 4 6
Observed Storm Surge (ft

Figure 4.--Observed and computed storm surge peaks at Buffalo
and Toledo. Storm surge peaks were computed with a
dynamical storm surge model (Schwab 1978) which
used hourly observed winds as input data.

8
* Dynamical method
6 _AStatistical method

: a) A

> -2-

- 4 -A ~
- /
-6-

-8 ~ / I I
-8
-8 -6 -4 -2 0 2 4 6 8
Forecast Peaks (ft)

Figure 5.--Comparison of observed and forecast storm surge peaks
at Buffalo for cases when the magnitude of either the
observed or forecast peak surge equaled or exceeded
3 ft. Peak surge forecasts generated by the dynamical
method are denoted by dots. Forecasts generated with
the statistical method are depicted with triangles.

- 2- ... m.m.z.e*!2CO*@~b.IW.~~

-4 - - -4 --
g-6- ~~~- -6--

O- 2 '. - -..* ..0�_
~~~~~~~~~~~~~~~~~~~-4- eeee e, ee -2 seo, eeee

-6- e .em - -66-,

5 DEC 78 2 APR 78 3APR78

:.--.._...........-..~~~~~~~~~ ;... . . _.
- 2 - - ~~~~~~- 2 - -
-4-~~~~~~~~~~~~~~~- -l �

g-6---6

C,/)

- 2U. ..'"'""'..'"� ...- 2 - ' ......
-4 - - mme - 4 - *
ioii illolllme ~ ~ ~ ~ ~ ~ memm
(~~~~~~~~~~~~~~~)~~~~~~ leil * - m -
-6 - - -6 - %'e -

18 APR 78 19 APR 78 26NOV 78 27 NOV 78

Figure 6.,--Dynamical (upper panel) and statistical (lower panel) storm surge
forecasts (dots) and observed storm surges (solid) for Buffalo on
December 5, 1978, April 2-3, 1978, April 18-19, 1978, and
November 26-27, 1978. Time is GMT.

4 - - 4-

4: - *., - . -_4

26 NOV 7727 NOV 77 17 DEC 78 18 DEC 78
- -- 2 - _

anee 25-26
00 *-4 -
o0 0
-4- ,1*

26 NOV77 27 NOV577 17 DEC 78 218 DEC 78

- -15
25DC 826E7

Figure~4 7--As infgr xetfrNvme 6-7 97 eebr1-8 98

6- 6 -6-
� 4- 4 -
2= *'** ...*� 2-� * -.*.
E6 ~~.�2"~~~~ ~4- 4
2 -2-

0~~~~~~~~~~~~~~~~~~~2-
L#D 0 0 0
2 2 -..

~~~~~~~~~~~~~~~~~0'
6-:3~~ -~~-2

E 4 4 4

4 4
022 '\"�. 2- 2
2 - - -

� ,~~~~~~~~- -
1 DEC 77 2 DEC 77 25 DEC 77 18 NOV 78
r~~~~~~~~~~~~~~~~~~~~~~~
O",~~~~~~~~~~~~

.~_ 4
,, 4
2
2
~~~~~~0

6 4-
4 22
cO
2 o
o 2
O ~~~~~~~~~~-2- .......
-4
4 DEC 78 5 DEC 78 21 DEC 78

Figure 8.--As in figure 6 except for December 1-2, 1977, December 25, 1977, November 18, 1978, December 4-5,
1978, and December 21, 1978.

8
� Dynamical method
6 A Statistical method
6

-4
' 0

o A*~

-8 I ~ I I I I
-8 -6 -4 -2 0 2 4 6 8

Forecast Peaks (ft)

Figure 9.--Comparison of observed and forecast storm surge peaks at Toledo
for cases when the magnitude of either the observed or forecast
peak surge equaled or exceeded 3 ft. Peak surge forecasts gener-
ated by the dynamical method are denoted by dots. Forecasts
generated with the statistical method are depicted with triangles.
17

4- 2
@ -~~ 2 - -4-2

4- - 4 - ..... -
- 2-
'..-2. - 4 .,_.. .,_

(I)~~~~~~~~~~~~~~~~~~~~~~~~~~
.- 0

2- -- -

-4-a - A-4-
2 - 4- -20- -

20~~~~ NOV 7821NV726OV727OV8
0 - 2- ""* ..*

4-
~~~~~~~ll

E 2-
~~~-4-

9 DEC 78

Figure 10.--Dynamical (upper panel) and statistical (lower panel) storm
surge forecasts (dots) and observed storm surges (solid) for
Toledo on November 20-21, 1978, November 26-27, 1978, and
December 9, 1978. Time is GMT.
18

4 0

o 6 -2 6
0)-4 -2-

4- -

25 DEC 77 21 DEC 77
4- 2cm 2-ae,19 .

O_
*~~ ~ ~~~~~~~~~~/.[ -...
~~~~~~-2- -,"-4-
~2 -4
-4 -6'- -
5 DEC 77 9 DEC 77

2
-42

2-2

E o 2
' -4-
6

-6- ...6-
25 DEC 77 21 DEC 78

Figure 11.--As in figure 10 except for December 5, 1977, December 9, 1977,
December 25, 1977, and December 21, 1978.
19

(Continued from inside front cover)

NOAA Technical Memorandums

NWS TDL 39 Computer'\Prediction of Precipitation Probability for 108 Cities in the United States. Wil-
liam H. Klein, February 1971, 32 pp. (COM-71-00249)
NWS TDL 40 Wave Climatology for the Great Lakes. N. A. Pore, J. M. McClelland, C. S. Barrientos, and
W. E. Kennedy, February 1971, 61 pp. (COM-71-00368)
NWS TDL 41 Twice-Daily Mean Monthly Heights in the Troposphere Over North America and Vicinity. Au-
gust F. Korte, June 1971, 31 pp. (COM-71-00826)
NWS TDL 42 Some Experiments With a Fine-Mesh 500-Millibar Barotropic Model. Robert J. Bermowitz, Au-
gust 1971, 20 pp. (COM-71-00958)
NWS TDL 43 Air-Sea Energy Exchange in Lagrangian Temperature and Dew Point Forecasts. Ronald M. Reap,
October 1971, 23 pp. (COM-71-01112)
NWS TDL 44 Use of Surface Observations in Boundary-Layer Analysis. H. Michael Mogil and William D.
Bonner, March 1972, 16 pp. (COM-72-10641)
NWS TDL 45 The Use of Model Output Statistics (MOS) To Estimate Daily Maximum Temperatures. John R.
Annett, Harry R. Glahn, and Dale A. Lowry, March 1972, 14 pp. (COM-72-10753)
4NWS TDL 46 SPLASH (Special Program To List Amplitudes of Surges From Hurricanes): I. Landfall Storms.
Chester P. Jelesnianski, April 1972. 52 pp. (COM-72-10807)
NWS TDL 47 Mean Diurnal and Monthly Height Changes in the Troposphere Over North America and Vicinity.
August F. Korte and DeVer Colson, August 1972, 30 pp. (COM-72-11132)
NWS TDL 48 Synoptic Climatological Studies of Precipitation in the Plateau States From 850-, 700-, and
500-Millibar Lows During Spring. August F. Korte, Donald L. Jorgensen, and William H.
Klein, August 1972, 130 pp. (COM-73-10069)
NWS TDL 49 Synoptic Climatological Studies of Precipitation in the Plateau States From 850-Millibar
Lows During Fall. August F. Korte and DeVer Colson, August 1972, 56 pp. (COM-74-t0464)
- NWS TDL 50 Forecasting Extratropical Storm Surges For the Northeast Coast of the United States. N.
Arthur Pore, William S. Richardson, and Herman P. Perrotti, January 1974, 70 pp. (COM-74-
10719)
NWS TDL 51 Predicting the Conditional Probability of Frozen Precipitation. Harry R. Glahn and Joseph
R. Bocchieri, March 1974, 33 pp. (COM-74-10909)
->i NWS TDL 52 SPLASH (Special Program to List Amplitudes of Surges From Hurricanes): Part Two. General
Track and Variant Storm Conditions. Chester P. Jelesnianski, March 1974, 55 pp. (COM-74-
10925)
NWS TDL 53 A Comparison Between the Single Station and Generalized Operator Techniques for Automated
Prediction of Precipitation Probability. Joseph R. Bocchieri, September 1974, 20 pp. (COM-
74-11763)
,;NWS TDL 54 Climatology of Lake Erie Storm Surges at Buffalo and Toledo. N. Arthur Pore, Herman P.
Perrotti, and William S. Richardson, March 1975, 27 pp. (COM-75-10587)
NWS TDL 55 Dissipation, Dispersion and Difference Schemes. Paul E. Long, Jr., May 1975, 33 pp.(COM-
75-10972)
NWS TDL 56 Some Physical and Numerical Aspects of Boundary Layer Modeling. Paul E. Long, Jr. and
Wilson A. Shaffer, May 1975, 37 pp. (COM-75-10980)
NWS TDL 57 A Predictive Boundary Layer Model. Wilson A. Shaffer and Paul E. Long, Jr., May 1975,
44 pp. (PB-265-412)
_..NWS TDL 58 A Preliminary View of Storm Surges Before and After Storm Modifications for Alongshore-
Moving Storms. Chester P. Jelesnianski and Celso S. Barrientos, October 1975, 16 pp. (PB-
247-362)
NWS TDL 59 Assimilation of Surface, Upper Air, and Grid-Point Data in the Objective Analysis Procedure
for a Three-Dimensional Trajectory Model. Ronald M. Reap, February 1976, 17 pp. (PB-256-
082)
->NWS TDL 60 Verification of Severe Local Storm Warnings Based on Radar Echo Characteristics. Donald
S. Foster, June 1976, 9 pp. plus supplement. (PB-262-417)
-2>NWS TDL 61 A Sheared Coordinate System for Storm Surge Equations of Motion With a Mildly Curved Coast.
Chester P. Jelesnianski, July 1976, 52 pp. (PB-261-956)
-> NWS TDL 62 Automated Prediction of Thunderstorms and Severe Local Storms. Ronald M. Reap and Donald S.
Foster, April 1977, 20 pp. (PB-268-035)
NWS TDL 63 Automated Great Lakes Wave Forecasts. N. Arthur Pore, February 1977, 13 pp. (PB-265-854)
) NWS TDL 64 Operational System for Predicting Thunderstorms Two to Six Hours in Advance. Jerome P.
Charba, March 1977, 24 pp. (PB-266-969)
.-3NWS TDL 65 Operational System for Predicting Severe Local Storms Two to Six Hours in Advance. Jerome
P. Charba, May 1977, 36 pp. (PB-271-147)
NWS TDL 66 The State of the Techniques Development Laboratory's Boundary Layer Model: May 24, 1977.
P. E. Long, W. A. Shaffer, J. E. Kemper, and F. J. Hicks, April 1978, 58 pp. (PB-287-821)
NWS TDL 67 Computer Worded Public Weather Forecasts. Harry R. Glahn, November 1978, 25 pp. (PB-291-
517)
NWS TDL 68 A Simple Soil Heat Flux Calculation for Numerical Models. Wilson A. Shaffer, May 1979,
16 pp. (PB-297-350)

NOAA SCIENTIFIC AND TECHNICAL PUBLICATIONS

The National Oceanic and Atmospheric Administration was established as part of the Department of
Commerce on October 3, 1970. The mission responsibilities of NOAA are to asses% the socioeconomic impact
of natural and technological changes in the environment and to monitor and predict the state of the solid Earth,
the oceans and their living resources, the atmosphere, and the space environment of the Earth.

The major components of NOAA regularly produce various types of scientific and technical informa-
tion in the following kinds of publications:

PROFESSIONAL PAPERS - Important definitive TECHNICAL SERVICE PUBLICATIONS -Re
research results, major techniques, and special inves- ports containing data, observations, instructions, etc. 1
tigation s. A partial listing includes data serials; prediction and
outlook periodicals; technical manuals, training pa-
CONTRACT AND GRANT REPORTS - Reports pers, planning reports, and information serials; and
prepared by contractors or grantees under NOAA miscellaneous technical publications.
sponsorship. TECHNICAL REPORTS -Journal quality with

ATLAS - Presentation of analyzed data generally extensive details, mathematical developments, or data
in the form of maps showing distribution of rainfall,litns
chemical and physical conditions of oceans and at- TECHNICAL MEMORANDUMS - Reports of
mosphere, distribution of fishes and marine main- preliminary, partial, or negative research or technol-
mals, ionospheric conditions, etc. ogy results, interim instructions, and the like.

Information an availability of NOAA publications can he obtained from.

ENVIRONMENTAL SCIENCE INFORMATION CENTER (D822)
ENVIRONMENTAL DATA AND INFORMATION SERVICE
NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION
U.S. DEPARTMENT OF COMMERCE

6009 Executive Boulevard
Rockville, MD 20852

NOAA---S/T 79-306

3 6668 00002 9118