Effect of Rainfall Measurement Errors on Nonpoint-Source Pollution Model Uncertainty

Abstract

Rainfall data are generally considered the most important input in watershed models and a major source of total uncertainty. This paper investigated the effects of rainfall measurement errors on hydrologic and nonpoint source pollution (H/NPS) modeling in the Daning River watershed in the Three Gorges Reservoir Region (TGRR) of China. The daily rainfall values were randomly permutated by Monte Carlo (MC) sampling, and 150 combinations of rainfall inputs were estimated using the Soil and Water Assessment Tool (SWAT). Based on the results, the rainfall measurement error is transformed into hydrologic modeling uncertainty and further propagates into even larger NPS modeling uncertainty. It was expected from the SWAT applications that the rainfall measurement error would introduce considerable prediction uncertainty especially during high-flow periods. Additionally, the model outputs become more accurate at the expense of a wider 90% confidence interval (90CI) when more possible error values were included. In this case, this paper combined the stochastic modeling and establishing a multi-event uncertainty analysis.

Keywords: rainfall, uncertainty, nonpoint-source pollution, SWAT, Monte Carlo method, Three Gorges Reservoir Region