Application of SWAT model in predicting water quantity and quality for U.S. and Thailand watersheds

摘要

Evaluating processes which influence water quantity and quality can be achieved through either long-term on-site monitoring or with the use of simulation models. On-site monitoring can be time-consuming, labor-intensive, and expensive. Therefore, the use of simulation models has become viable and cost-effective. However, to ensure that the model is capable and reliable to describe hydrologic processes in various hydrologic conditions and to use it as an assessment tool, there is a strong need in testing the model against extensive field measured data for different scales, land use, topography, climate, and soil conditions prior to its application for solving natural resource problems within watersheds. For this reason, the objective of this research was to make an effort to calibrate and validate the surface and subsurface components of the Soil and Water Assessment Tool (SWAT) model for various hydrologic conditions in predicting surface and subsurface drain flows and their water quality.;Calibration and validation of the SWAT for the Upper Maquoketa River Watershed were performed by comparing measured and predicted stream flows and NO3-N losses at the watershed outlet during the period of 1999-2001. The coefficient of determination (r2) statistics found for the monthly stream flows and NO3-N losses were equal to 0.73 and 0.72, respectively indicating that model performed reasonably well.;Evaluation of the SWAT tile flow components were performed by comparing the measured tile flows with the predicted tile flows at the Iowa State University\u27s Northeastern Research Center for five years (1993-1997) using both STATSGO and SSURGO soil databases. SWAT simulation results indicated that the model reasonably predicted the cumulative annual tile flow volumes and reasonably tracked the observed trends for the calibration year. The SWAT model did not accurately predict cumulative annual tile flows and monthly tile flows for the validation years. Therefore, the simulation results showed that the model predicted similar results regardless of soil data set.;Application of the SWAT was conducted for the Chi River Subbasin II located in the northeastern Thailand by comparing predicted stream flows and NO 3-N losses with corresponding in-stream for five years (2000-2003, 2005). Statistical comparisons between the simulated results and the observed data for the calibration year gave a reasonable agreement for both monthly r 2 and Nash-SutCliffe model efficiency (E) within ranges of 0.77-0.88 and 0.55-0.79, respectively, where as validation results showed lower values of r2 and E values ranging from 0.23 to 0.77 and -7.98 to 0.66.