Blending satellite-based snow depth products with in situ observations for streamflow predictions in the Upper Colorado River Basin

Abstract

In snowmelt-driven river systems, it is critical to enable reliable predictions of the spatiotemporal variability in seasonal snowpack to support local and regional water management. Previous studies have shown that assimilating satellite-station blended snow depth data sets can lead to improved snow predictions, which however do not always translate into improved streamflow predictions, especially in complex mountain regions. In this study, we explore how an existing optimal interpolation-based blending strategy can be enhanced to reduce biases in satellite snow depth products for improving streamflow predictions. Two major new considerations are explored, including: (1) incorporating terrain aspect and (2) incorporating areal snow coverage information. The methodology is applied to the bias reduction of the Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E) snow depth estimates, which are then assimilated into the Noah land surface model via the ensemble Kalman Filtering (EnKF) for streamflow predictions in the Upper Colorado River Basin. Our results indicate that using only observations from low-elevation stations such as the Global Historical Climatology Network (GHCN) in the bias correction can lead to underestimation in streamflow, while using observations from high-elevation stations (e.g., the Snow Telemetry (SNOTEL) network) along with terrain aspect is critically important for achieving reliable streamflow predictions. Additionally incorporating areal snow coverage information from the Moderate Resolution Imaging Spectroradiometer (MODIS) can slightly improve the streamflow results further. Key Points: Blending satellite and in situ snow data improves streamflow prediction Incorporating terrain aspect in the blending can improve the results Additionally incorporating MODIS snow cover can further improve the results