Hierarchical Sensitivity Analysis For Large-scale Process-based Hydrological Modeling With Application In An Amazonian Watershed

Sensitivity analysis methods have lately received great attention for identifying important uncertainty sources and improving model calibration and predictions for hydrologic models.However,the high computational cost for large-scale process-based hydrological models has made it challenging to apply the quantitative and comprehensive global sensitivity analysis method to such sophisticated models.In an effort to make up the paucity of previous researches,we improved the hierarchical sensitivity analysis method by replacing the conventional Monte Carlo sampling method with Latin hypercube sampling method and defining a new set of subdivided parametric sensitivity indices.We implemented the improved hierarchical sensitivity analysis method to quantify three types of uncertainties in an ~9,000 km2 Amazon catchment associated with a three-dimensional,large-scale and process-based hydrologic model(PAWS).The importance of different uncertainties was elucidate by calculating sensitivity indices for two hydrologic outputs:evapotranspiration(ET)and groundwater contribution to streamflow(Qg).In general,parameters,especially the vadose zone parameters,are the most important uncertainty contributors for all sensitivity indices.In addition,the influence of climate scenarios on ET predictions is also nonignorable.Furthermore,the thickness of the aquifers along the river grid cells is important for both ET and Qg.These results provide a deeper understanding about how these models function and about how the natural systems function.This research unveils the absolute necessity to consider various uncertainties when performing sensitivity analysis on PBHMs and prove the scientific feasibility of improved hierarchical sensitivity analysis method.This paper provides insights into implementing a comprehensive global sensitivity analysis to large-scale PBHMs.