Towards quick parameter estimation of hydrological models with large number of computational units

Abstract

Distributed hydrologic models represent the spatial heterogeneity within the watersheds, but are computationally intensive. The problem of computational effort is further exacerbated in the process of the model calibration, where iterative simulations are performed to estimate optimal parameter values. This study evaluates a parameter transfer approach to address this issue and is demonstrated on two watersheds. The results indicate that the parameters from a coarse resolution model that can reasonably capture the spatial heterogeneity of the watershed, can be used to simulate a fine resolution model with satisfactory performance (NSE > 0.60 in both watersheds). The models with transferred parameters did preserve sufficient watershed heterogeneity to still provide acceptable prediction, with performance resulting in an NSE > 0.6 and NSE > 0.90 for both stream flow and water quality variables respectively at a location upstream. The parameter transfer approach can be used with lesser computational effort along with good simulation accuracy, if the spatial heterogeneity of the watershed is preserved at the coarse resolution.