Estimating and Simulating Precipitation Uncertainty Data for Large-Scale Hydrologic Applications

Abstract

The usefulness of satellite multi-sensor precipitation and other large-scale precipitation products in hydrologic applications can be hindered by substantial uncertainty. In parts of the world with few ground observations of precipitation, such uncertainty is difficult to quantify. At the same time, how to cope with the characterize and model the spatiotemporal structure of this uncertainty has been called a grand challenge within the precipitation community. We present progress on two fronts which, when combined, addresses this grand challenge. Rather than relying on ground reference data to quantify uncertainty in NASA’s IMERG precipitation dataset, we instead use the dual-frequency precipitation radar aboard the NASA/JAXA GPM platform. This uncertainty information is then fed into the Space-Time Rainfall Error and Autocorrelation Model (STREAM), which uses an uncalibrated anisotropic and nonstationary spatiotemporal correlation modeling approach to stochastically generate ensemble precipitation fields that depict the uncertainty inherent in IMERG. We then use these ensemble fields to examine the effects of precipitation uncertainty in several hydrologic applications, including flood monitoring and prediction of water and energy fluxes. Ensemble-based hydrologic simulations outperform those based on IMERG and help reveal the spatiotemporal scales and hydrologic variables for which precipitation uncertainty is critical. The approach is compatible with other continental-to-global scale precipitation estimates such as those from numerical weather models, and can also be used in precipitation downscaling contexts. We are developing a set of open-source tools to facilitate its usage.