Combining Indirect Observations and Models to Resolve Spatiotemporal Patterns of Precipitation in Complex Terrain
MetadataShow full item record
Mountain precipitation in the Western United States is critical for the water resources of the region, but resolving spatial and temporal patterns of precipitation in complex terrain is challenging due to lack of observations, measurement uncertainty and high spatial variability. We examine several gridded precipitation datasets over the Sierra Nevada mountain range of California, and find that these widely-used products exhibit substantial variation in water-year total precipitation over different areas of the range. In addition, trends in precipitation and snow computed from different datasets vary substantially. Both findings suggest that further work is needed to better resolve spatial and temporal patterns of precipitation in complex terrain. Streamflow observations are widely made and reflect the basin’s hydrologic response to precipitation input. We develop a methodology for inferring basin-mean precipitation using lumped hydrologic models and Bayesian model calibration, which infers water-year total precipitation given daily streamflow observations. We apply this approach to several basins around Yosemite National Park in the Sierra Nevada in order to assess the sensitivity and robustness of inferred precipitation. We find that patterns of precipitation can be inferred from streamflow, both in terms of spatial and year-to-year variability. However, by using a small ensemble of hydrologic model structures to test the sensitivity of inferred precipitation, we also show that the absolute amounts of inferred precipitation are subject to significant uncertainty. Higher-elevation basins of the Sierra Nevada are hydrologically snow-dominated, and we hypothesize that the uncertainty in inferred precipitation can be reduced by calibrating the hydrologic model to both snow and streamflow observations. We leverage the recent availability of a high-resolution distributed snow dataset from the Airborne Snow Observatory (ASO) to determine basin-mean snow water equivalent (SWE) over the upper Tuolumne River basin. We also compare point and distributed SWE measurements over the basin, to assess the reliability of using point measurements to estimate basin-mean SWE. In this case, point measurements show bias in estimating basin-mean ASO SWE, largely due to non-representative sampling with respect to elevation. When basin-mean SWE is included with streamflow in model calibration, uncertainty in inferred precipitation is reduced by up to half, and model ensemble consistency is improved. To resolve patterns of precipitation over the Sierra Nevada, we infer precipitation from streamflow using 56 stream gauges that measure runoff from relatively unimpaired basins, over 1950-2010. We compare inferred precipitation to gauge-based gridded precipitation data, finding that significant differences exist between the mean spatial patterns of precipitation over the range. In particular, inferred precipitation suggests that gridded products underestimate precipitation for higher-elevation basins whose aspect faces prevailing winds. Better agreement is found in lower-elevation and leeward basins. Collectively, the findings suggest that the development of spatially distributed precipitation datasets should not consider precipitation gauge data in isolation, but should also consider other related hydrologic observations in order to better resolve patterns of precipitation in complex terrain.
- Civil engineering