Content area
Abstract
Sensitivity analysis of a model can identify key variables affecting the performance of the model. Uncertainty analysis is an essential indicator of the precision of the model. In this study, the sensitivity and uncertainty of the Long-Term Hydrologic Impact Assessment-Low Impact Development 2.1 (L-THIA-LID 2.1) model in estimating runoff and water quality were analyzed in an urbanized watershed in central Indiana, USA, using Sobol''s global sensitivity analysis method and the bootstrap method, respectively. When estimating runoff volume and pollutant loads for the case in which no best management practices (BMPs) and no low impact development (LID) practices were implemented, CN (Curve Number) was the most sensitive variable and the most important variable when calibrating the model before implementing practices. When predicting water quantity and quality with varying levels of BMPs and LID practices implemented, Ratio_r (Practice outflow runoff volume/inflow runoff volume) was the most sensitive variable and therefore the most important variable to calibrate the model with practices implemented. The output uncertainty bounds before implementing BMPs and LID practices were relatively large, while the uncertainty ranges of model outputs with practices implemented were relatively small. The limited observed data in the same study area and results from other urban watersheds in scientific literature were either well within or close to the uncertainty ranges determined in this study, indicating the L-THIA-LID 2.1 model has good precision.