Statistically Forecasting Early Warm Season Precipitation in the CONUS Using a Canonical Correlation Approach Applied to CFSv2 Reforecast Data

Forecasts at the subseasonal to seasonal timescale have significant value during the warm season. Though global S2S models are somewhat skillful in forecasting warm season atmospheric teleconnection patterns, they are comparatively less skillful in forecasting precipitation within the CONUS. In this study, observed 500-hPa atmospheric circulation anomalies in the Northern Hemisphere are related to the dominant patterns of observed two-month standardized precipitation index (SPI) in the CONUS during the early warm season (June-July) from 1982-2010. A canonical correlation analysis is then applied to determine the dominant coupled modes between observed teleconnection patterns and SPI. The early warm season is a potential period of 'forecast opportunity. ' Atmospheric teleconnections related to coupled ocean-atmosphere variability project relatively strongly onto the continental-scale patterns of SPI variability at this particular time of year. The same procedure is performed on CFSv2 reforecast data (1982-2010), considering forecasts initialized in May, to determine dominant coupled modes between modeled geopotential height anomalies and model SPI. A new statistical downscaling forecast technique combines information from both of these CCA analyses to produce a new SPI forecast product. The skill of the new SPI forecast product is evaluated against the original CFSv2 forecast SPI to its assess value added. Possible next steps toward creating a new operational SPI forecast product at S2S timescales will also be discussed.