Model Parameter Sensitivity Across Time Scales

Abstract

Sensitivity analysis (SA) has been a topic of increasing application on environmental modeling. Several toolboxes available for different programming languages have been released, including a great variety of approaches and methodologies for investigation and interpretation of SA results. Such techniques have been applied to improve the performance of hydrologic models by focusing on the most relevant parameters. Despite studies indicating the importance of parameters varying over time, little research has explored how sensitivity varies across different simulation time scales. In this study, we evaluate how the time scale of the data used in rainfall-runoff model affects the sensitivity of model parameters for a collection of catchments located in Brazil and the USA. We compare the sensitivity arising from a hydrological model calibrated to monthly time resolution streamflow data with the sensitivity arising from the same model calibrated to daily time resolution data with outputs aggregated to a monthly resolution. Additionally, we also evaluate how the importance of the model parameters change depending on the choice of the objective function. The results demonstrate that time resolution influences the sensitivity of the model parameters and that addressing the temporal dynamics of parameter sensitivity might represent important gains in model simulations depending on the time scale of interest.