Enhanced Identification of hydrologic models using streamflow and satellite water storage data: a multi-objective calibration approach

Abstract:

The conventional procedure for parameter identification of hydrological processes through conditioning only to streamflow data is challenging in physically based distributed hydrologic modelling. The challenge increases for modeling the landscapes where vertical processes dominate horizontal processes, leading to high uncertainties in modelled state variables, vertical fluxes and hence parameter estimates. Such behavior is common in modeling the prairie region of the Saskatchewan River Basin (SaskRB, our case study), Canada, where hydrologic connectivity and vertical fluxes are mainly controlled by surface and sub-surface water storage.

To address this challenge, we developed a novel multi-criteria framework that utilizes total column water storage derived from the GRACE satellite, in addition to streamflows. We used a multi-objective optimization algorithm (Borg) and a recently-developed global sensitivity analysis approach (VARS) to effectively identify the model parameters and characterize their significance in model performance. We applied this framework in the calibration of a Land Surface Scheme-Hydrology model, MESH (Modélisation Environmentale Communautaire – Surface and Hydrology) to a sub-watershed of SaskRB.

Results showed that the developed framework is superior to the conventional approach of calibration to streamflows. The new framework allowed us to find optimal solutions that effectively constrain the posterior parameter space and are representative of storage and streamflow performance criteria, yielding more credible prediction with reduced uncertainty of modeled storage and evaporation.