Optimal Precipitation Estimation for Land Surface Modeling

Precipitation is the major meteorological forcing for land surface modeling, and therefore controls runoff, biogeochemical cycling, evaporation, transpiration, groundwater recharge, and soil moisture. However, precipitation estimates from rain gauge, ground-based radar, satellite, and numerical models are affected significant uncertainties, which can be amplified when exposed to highly non-linear land model physics.

This project uses a combination of precipitation data from different sources optimized to minimize hydrologic response error (soil moisture, runoff, evapotranspiration, etc.) in order to improve coupled model forecast skill. Ultimately, this solution would optimally merge a wide range of precipitation information from satellites, radars, gauges, and models trained to minimize land model evaporation, runoff, and soil moisture response resulting in coupled forecast improvements convection, clouds, precipitation, boundary layer processes, and atmospheric circulation.

We tested the hypothesis that by merging three precipitation products (satellite product, ground-based radar, and model estimates) trained to minimize newly emerging satellite soil moisture anomalies we can improve model forecast performance. We conducted an uncoupled demonstration over Oklahoma where ground observations are available for validation. This approach assimilates both remotely sensed precipitation and soil moisture in a hydrologic meaningful way to improve overall weather forecast model skill. Similar simulations could be also conducted for forcings with different spatial resolutions over different time periods.

Funded by NOAA Joint Center for Satellite Data Assimilation - 2015 Research in Satellite Data Assimilation for Numerical Environmental Prediction

Team: Dr. Maggioni (PI); Dr. Houser (co-I); Dr. Hazra (Postdoctoral Fellow)

Performance Period: August 2015 – July 2018