470
The discrepancy between observed and DPM simulated hydrographs can be due to
a number of reasons. Firstly, DPM is highly sensitive to the temporal distribution
of precipitation. Using 15 minute rainfall instead of the actual temporal
distribution is a possible error source. Secondly, DPM is highly sensitive to the
channel dimensions. The cross-sectional data available on channels in the study
area were very limited. Using approximate channel dimensions is another possible
error source. Thirdly, DPM is highly sensitive to land use parameters that affect
infiltration. The land use information obtained from the satellite data was not in
sufficient detail to distinguish between different agricultural crops. No attempt
was made to better discriminate between the different agricultural crops. Using
average parameter values to represent agricultural land is another possible error
source.
CONCLUSIONS
The two models tested in this study adequately simulated the peak flow rates for a
selected storm using a parameter set partially derived from Landsat data.
Identification of different land uses within a Landsat scene was found to be a
straightforward process. The determination of the areal percentage of each land
use within the watershed using the image processing software simplifies the
watershed discretization process.
The accuracy of the modeling results obtained using Landsat data is comparable to
results obtained using the conventional land use maps. Using the Landsat imagery
however, has one distinct advantage over land use maps: Once the spectral
signatures of various land uses within a given area are determined, the variation in
land use with time can easily be traced and incorporated into a model's parameter
set. In studies where the impact of land use changes on watershed behavior is
being assessed, an accurate record of the land use variation can be extremely
helpful.
For long term flow simulation, where evapotranspiration rates play an important
role in soil moisture variations, Landsat data should be supplemented with more
detailed ground truthing to identify specific crops. The monthly flow simulation
results reported herein could possibly have been improved with more accurate
crop parameters.
With the increasing concern in water quality and quantity in both urban and rural
areas, hydrologic modeling is rapidly becoming a popular planning tool. Use of
remotely sensed land use data and related image processing hardware and
software can be very helpful in modeling efforts that require timely, accurate
representation of the spatial distribution of land uses within a relatively large
watershed.
The utility of using remotely sensed data in defining the inputs for the two models
discussed herein should continue to be investigated.
ACKNOWLEDGEMENTS
The study reported herein is partially supported by funds from the Civil
Engineering Department and the Remote Sensing System Laboratory (RSSL) of the
Civil Engineering Department of the University of Maryland. The authors are
grateful to Ms. Olivia Carter for her assistance in the preparation of the
manuscript and to Ms. Janet Woodruff for typing the final paper.
REFERENCES
Century Engineering, Inc. (1981) Anacostia Watershed Erosion Sedimentation
Control Study. Prepared for the Washington Suburban Sanitary Commission, WSSC
Contract No. 81 PS-06-34A, Towson, Maryland.