Impact of uncertainty in precipitation forcing datasets on the hydrologic budget of an integrated hydrologic model in mountainous terrain

Precipitation is a key input variable in distributed surface
water-groundwater models, and its spatial variability is expected to
impact watershed hydrologic response via changes in subsurface flow
dynamics. Gridded precipitation datasets based on gauge observations,
however, are plagued by uncertainty, especially in mountainous terrain
where gauge networks are sparse. To examine the mechanisms via which
uncertainty in precipitation data propagates through a watershed, we
perform a series of numerical experiments using an integrated surface
water-groundwater hydrologic model, ParFlow.CLM. The Kaweah River
watershed in California, USA is used as our virtual catchment laboratory
to characterize watershed response to variable precipitation forcing from
headwaters to groundwaters. By applying the three cornered hat method, we
quantify the spatially distributed uncertainty in four publically
available precipitation forcing datasets and their simulated hydrology.
Simulations demonstrate that uncertainty in the simulated groundwater
storage is primarily a result of topographic redistribution of uncertainty
in precipitation forcing. Soil water redistribution is the primary pathway
that redistributes uncertainty downslope. We also find that topography
exerts a larger impact than variable subsurface parameters on propagating
uncertainty in simulated fluxes. Finally, we find that improvement in
model performance metrics is higher for a single simulation forced with
the mean precipitation from the available datasets than the averaged
simulated results of separate simulations forced with each dataset.
Results from this study highlight the importance of topography-moderated
flow through the critical zone in shaping the groundwater response to
climate variability.