Combining snow, streamflow, and precipitation gauge observations to infer basin-mean precipitation

Abstract

Precipitation data in mountain basins are typically sparse and subject to uncertainty due to difficulties in measurement and capturing spatial variability. Streamflow provides indirect information about basin-mean precipitation, but inferring precipitation from streamflow requires assumptions about hydrologic model structure that influence precipitation amounts. In this study, we test the extent to which using both snow and streamflow observations reduces differences in inferred annual total precipitation, compared to inference from streamflow alone. The case study area is the upper Tuolumne River basin in the Sierra Nevada of California, where distributed and basin-mean snow water equivalent (SWE) estimates have been made using LiDAR as part of the NASA Airborne Snow Observatory (ASO). To reconstruct basin-mean SWE for years prior to the ASO campaign, we test for a robust relationship between SWE estimates from ASO and from snow courses and pillows, which have a longer record. Relative to ASO’s distributed SWE observations, point SWE measurements in this part of the Sierra Nevada tend to overestimate SWE at a given elevation, but undersample high-elevation areas. We then infer precipitation from snow and streamflow, obtained from multiple hydrologic model structures. When included in precipitation inference, snow data reduce by up to one third the standard deviations of the water year total precipitation between model structures and improve the consistency between structures in terms of the yearly variability in precipitation. We reiterate previous findings that multiple types of hydrologic data improve the consistency of modeled physical processes and help identify the most appropriate model structures.