Sensitivity to spatial and temporal resolution in the performance of a parsimonious hydrological model for quantifying water balance partitioning in dryland regions.

Dryland regions cover around one-third of the global land surface and are naturally prone to water scarcity. Drylands typically experience highly variable precipitation, both spatially and temporally, with potential evapotranspiration (PET) greatly exceeding annual rates of precipitation. However, our ability to accurately quantify the key components of the water partitioning in these regions is hampered by the scarcity of data and the highly dynamic nature of the hydrological processes. In this context, assessing the ability of models to represent the key hydrological processes at different spatial and temporal scales is of key importance to enhance our understanding and quantification of the water balance in dryland regions. Here, we have assessed the impact of the model grid size and the temporal scale of climatic forcing in combination with variations in model structure in the description and representation of key hydrological processes that influence the water partitioning in dryland regions. The analysis was performed in the Walnut Gulch Experimental Watershed where a dense network of hydrological measurements is readily available for model evaluation. We show that our parsimonious model, DRYP, can describe well the water partitioning across a range of different temporal and spatial scales. However, we find that sub-daily time steps of precipitation and PET, combined with a fine spatial resolution of less than or equal to 1km grid size, are needed for robust quantification of the water partitioning, which is also very sensitive to the choice of infiltration model. The results highlight the important role of channel losses through the streambed of ephemeral streams (~7% of the precipitation), and the impact of the underlying alluvial riparian area in the partitioning of water fluxes between riparian vegetation evapotranspiration (~60 % of transmission losses) and the production of focused groundwater recharge (~3 % of the precipitation). These results have important implications for the potential for improving the performance of large scale models in dryland regions by indicating the appropriate temporal and spatial scales required for the proper representation of dryland hydrological processes.