Seasonal Land-surface Hydrology Prediction and Predictability over CONUS

Seasonal forecasting in operational centers has mostly focused on prediction of temperature (T2m) and precipitation (P). Recent research from our team has shown snow water equivalent (SWE) can be even more skillful from late winter through early summer for snow dominated regions. Here, in addition to T2m, P, and SWE, we explore the predictability, assessed using a perfect model approach, and prediction skill of other components of land-surface hydrology: soil moisture (SM), total water storage (TWS), evapotranspiration (ET) and Runoff (Ro). Using seasonal forecasts from NCEP-CFSv2 and ECMWF-SEAS5 forecasting systems, we ask the following questions: What components of the water cycle have the most predictability and actual prediction skill? Is there a relationship between the prediction skill and the predictability? How do these results change across different hydrologically defined regions of the conterminous U.S.?

We compare our results in two hydrologic regions of the U.S. that have different annual water cycles: Missouri River region which has more consistent precipitation throughout the year versus California which receives most its precipitation from October through April. Preliminary results show variables with high memory such as SM, TWS and SWE have both high predictability and prediction skill followed by ET and Ro which are highly dependent on water storage. Results also show prediction skill is positively related to predictability for all variables both by region and time of year. When comparing regions, the Missouri River region showed relatively consistent predictability and prediction skill throughout the year, however, in California, predictability and especially prediction skill was high during the spring and early summer and dropped off considerably for forecasts valid late summer through fall.