Sensitivity Analysis of Fully Distributed Parameterization Reveals Insights Into Heterogeneous Catchment Responses for Water Quality Modeling

Spatially distributed parameterization is preferable in capturing the heterogeneity of catchment properties and in allowing a better model representation of catchment responses. In hydrological modeling, sensitivity analysis is recommended to address the high-dimensional parametric problems. However, less has been focused on water quality modeling, presumably due to the lack of suitable fully distributed models. Based on the newly developed mesoscale hydrological-nitrate model, we investigated for the first time the spatially distributed sensitivity of nitrate model parameters and correlated the sensitivity indices with multiple catchment characteristics. The study was conducted in the highly heterogeneous Selke catchment, central Germany. Three nested catchments were defined based on the heterogeneous catchment responses (gauged by three nested stations). Results showed that parameters of soil denitrification, in-stream denitrification, and in-stream assimilatory uptake were the most sensitive parameters throughout the catchment. They all showed high spatial variability, which also varied when different gauging stations were considered. Spearman rank correlation indicated that the sensitivity of soil denitrification was controlled mainly by the relative limitations between the terrestrial hydrological transport capacity and the soil nitrate availability; the sensitivity of in-stream processes was predominated by the spatial variability within the river network (e.g., the proximity to the gauging station), rather than the local biogeochemical factors. Based on the insights gained from the spatial sensitivity and correlation analyses, we suggested that an appropriate monitoring scheme is important in reflecting actual catchment responses, and a cautious statistical correlation is informative in benefiting future parameter regionalization of water quality models.