Filling in the gaps: Inferring spatially distributed precipitation from gauge observations over complex terrain

Abstract

AbstractIn recent decades, computational hydrology has trended toward higher‐resolution distributed models of the land surface. The accuracy of these models is limited, by uncertainty in distributed precipitation forcings. In this research, different precipitation distribution schemes were compared through inter‐station transfer experiments, as well as within a distributed hydrologic model applied at ≤150 m resolution over four study catchments in complex terrain. Distributed precipitation estimates were derived using multiplicative spatial scaling (MSS) and map‐based precipitation (MBP) techniques including both climatological and time‐varying spatial information from a range of native spatial resolutions (500 m–4 km). The primary interest was to evaluate a novel application of satellite‐based snow water equivalent (SWE) reconstruction (RSWE) as a surrogate for cold season precipitation against a common source of spatial precipitation information: the Parameter‐elevation Regressions on Independent Slopes Model (PRISM). An elevation‐based orographic precipitation gradient and simple inverse‐distance interpolation were also included as a baseline. For the case of RSWE, MSS was very sensitive to differences between observed SWE and reconstructed SWE, producing positive biases in the catchment water balance. Over 12 year simulations, daily streamflow correlations from the uncalibrated model were highest for RSWE when adjusted for accumulation‐season sublimation, and for monthly PRISM, both achieving R=0.8, where the former performed better during anomalous years for both MSS and MBP. Annual water balance ratios were much more sensitive to the choice of distribution scheme, with large overestimates, > 30%, for RSWE products using the MSS techniques versus MBP (< 15%). Overall, PRISM performed best using MSS, while RSWE performed best using MBP.