AUTHORS: Mehdi Ghasemizade, Gabriele Baroni, Karim Abbaspour, and Mario Schirmer
Physically based models for simulating environmental processes are usually criticized due to having many parameters. This issue leads to over-parameterization and can finally reduce the uncertainty (reliability) of the simulated outputs. Sensitivity and identifiability analyses are common diagnostic tools to address over-parametrization in complex environmental models. In this study, we performed a temporal global sensitivity and identifiability analyses of HydroGeoSphere (HGS) model parameters. HGS was used to simulate daily evapotranspiration, water content, and recharge based on high quality data of a weighing lysimeter. Figure below shows the schematic of the lysimeter as well as the conceptual model for simulating the lysimeter. The model has four soil layers in addition to a preferential flow component. We found that identifiability of a parameter does not necessarily reduce output uncertainty. It was also found that the sensitivity of the model parameters is required to allow uncertainty reduction in the model output.