Distributed hydrologic modeling for flow forecasting using high-resolution data

Abstract

The recent availability of NEXRAD rainfall data, along with high-resolution Digital Elevation Maps (DEM), soil, land-use, and land-cover data, has motivated the use of such data to improve distributed hydrologic modeling. The National Weather Service Hydrologic Lab (NWSHL) is promoting the Distributed Modeling Intercomparison Project (DMIP) to encourage the use of spatially distributed data to improve flow modeling and prediction along the entire river system The main goal of DMIP is to promote the development of models and modeling systems that best utilize NEXRAD and other spatial data sets to improve River Forecast Center (RFC)-scale river simulations. The goals of this study are to forecast streamflow along the river and apply a modeling approach that maps the 3-dimensional rainfall, DEM, and soil data to the streamflow profile along the river and fmally to zero-dimensional streamflow forecast at the outlet. The study had four main directions: ( 1) comparison of uncalibrated and calibrated model performance, (2) ·comparison of lumped and distributed model performance, (3) study of the model performance at interior evaluation points, and ( 4) comparative evaluation of the SAC-SMA (Sacramento Soil Moisture Accounting) and the HYMOD (Hydrologic Model). This study explores the use of NEXRAD rainfall data in the context of hydrologic modeling of the Illinois River basin at Watts, OK, using lumped and distributed versions of the HYMOD and SAC-SMA models. HYMOD is a recently developed hydrologic model, containing five parameters and much less complexity as compared to the 16 parameter SAC-SMA model. The distributed approach developed in this study uses the DEM data to subdivide the basin and to determine the length of each river reach Each river reach is divided into a number of smaller segments for which the slope is estimated from the DEM data. Each rainfall-runoff model was used to compute the precipitation excess for each sub-basin. Runoff was assigned to each river segment based on the contributing area of that segment and then routed down river to the basin outlet using the kinematic wave routing scheme. Both models have been calibrated by running the SCE-UA model (Shuffled Complex Evolution- University of Arizona, Duan et al., 1992). The comparison of the two model systems based on their accuracy and efficiency is investigated.