Inverse Mapping for Rainfall-Runoff Models using History Matching Approach
In this paper, we consider two rainfall-runoff computer models. The first model is Matlab-Simulink model which simulates runoff from windrow compost pad (located at the Bioconversion Center in Athens, GA) over a period of time based on rainfall events. The second model is Soil Water Assessment Tool (SWAT) which estimates surface runoff in the Middle Oconee River in Athens, GA. The input parameter spaces of both models are sensitive and high dimensional, the model output for every input combination is a time-series of runoff, and these two computer models generate a wide spectrum of outputs including some that are far from reality. In order to improve the prediction accuracy, in this paper we propose to apply a history matching approach for calibrating these hydrological models, which also gives better insights for improved management of these systems.
READ FULL TEXT