Effect of spatial variability on river basin stream flow modeling

Sudhakar Mamillapalli, Purdue University

Abstract

Distributed parameter hydrologic water quality models integrated with GIS facilitate simulation of large watersheds with numerous subbasins. Application of such distributed parameter models to large watersheds has given rise to many problems--three problems were examined in this work. The effect of watershed representation detail on runoff simulation results was studied. Simulations were conducted on 7 watersheds with different levels of representation detail. Results indicate that improved accuracy could be obtained either by finer discretization of the watershed into sub-basins or by choosing smaller land use and soil thresholds so that more combinations of soil and land use are simulated. A threshold configuration seems to exist with little or no increase in accuracy beyond that level of discretization. A simple method is proposed to determine the threshold configuration. The second problem examined is selection of weather data for each subbasin within the watershed. Four methods for selection of weather data have been studied, the first being the Thiessen polygon average of all gages within the basin. The second is the use of the Thiessen polygon average of each subbasin. The third method is the use of the gage closest to centroid of the subbasin for each subbasin, and fourth the use of the gage closest to the centroid of the whole basin as the gage to use for all subbasins. These methods were applied to 7 watersheds and the results indicate that the Thiessen polygon average over the whole basin was the best method for all basins analyzed. Finally, the calibration of large watersheds with numerous subbasins has been examined. Calibration is complicated due to interaction between parameters in different subbasins because of flow routing. A new calibration procedure was designed and implemented which applies the Golden Section Search Procedure for finding the calibration parameters in an optimal fashion, starting upstream and moving downstream to avoid parameter interaction problems. The procedure was demonstrated on two watersheds using the basin scale model SWAT, and results are presented. Calibration using this approach provided runoff simulation results that improved considerably over uncalibrated simulations.

Degree

Ph.D.

Advisors

Engel, Purdue University.

Subject Area

Agricultural engineering|Hydrology|Geography

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS