Spatio-temporal sensitivity analysis of the wetland modules of a semi-distributed hydrological model

Abstract

Hydrological models offer the opportunity to evaluate wetland conservation networks as nature-based solutions to mitigate hydrological extremes. To optimize the replication of a hydrological model response (i.e., stream flows), the models rely on input data (land cover, hydrographic network, soil types, topography, meteorological data) and parameters. However, the effect of parametric uncertainty on the simulated variables and the spatio-temporal variability of this sensitivity are insufficiently documented. This study presents an application of the novel global sensitivity analysis method, the Variogram Analysis of Response Surface, to the isolated and riparian wetland modules of HYDROTEL, a semi-distributed, process-based, deterministic model, on two watersheds located in the province of Quebec, Canada. The analysis accounted for: (i) the spatial variability by assessing the sensitivity at various locations in the watersheds and by compiling sensitivity indices for wetland networks discretized following the Strahler order classification which quantifies the structure of hydrographic networks and (ii) the temporal variability of the sensitivity using the Generalized Global Sensitivity Matrix. The results indicate that the parameters defining the geometry of wetlands (area, depth) are the most sensitive for both isolated and riparian wetlands, and for all Strahler orders. The temporal assessment of sensitivity highlights the seasonal controlling processes, including a peaking sensitivity of the maximum water depth in wetlands during spring snowmelt, whereas the sensitivity of the evapotranspiration parameter increases during summer, but is null during winter. Results also indicate that the parameters of the isolated modules are more sensitive for wetlands located on lower stream order (upstream/headwaters) while those of the riparian modules display a greater sensitivity when wetlands are located on higher Strahler orders (downstream/main channel). These findings deepen our understanding of the impact of wetland features on stream flows and provide guidelines to plan future data acquisition campaigns in wetlandscapes.