Toward Improvement in Seasonal Forecasting in the Southwest United States Using Regional Climate Product at Convective-Permitting Scale

Accurate regional and local scale information about seasonal climate variability and its impact on water availability is important in many practical applications like agriculture, water resource planning, long term decision making etc. Presently, the primary source for real time seasonal climate forecast comes from the Climate Prediction Center (CPC) within the NOAA National Center for climate Prediction (NCEP) which uses its model forecast component (CFSv2) of North American Multi-Model Ensemble (NMME). However, it has been observed that in comparison to the cool season, the level of skill in warm season seasonal forecasts of precipitation produced by the NMME is much lower (Kirtman et al. 2014) due to the poor climatological representation of warm season convective precipitation. To fully realize the potential in improving warm season seasonal forecasts using a dynamical modeling approach requires dynamical downscaling of NMME models to better improve their representation of convective precipitation. Specifically, a convective-permitting (3km) scale is required to explicitly represent thunderstorms in a regional model. Also, for basin scale study, coarse resolution models must be downscaled to create high spatial resolution information for efficient hydrologic forecasting. Driven by these motivations, this study addresses towards a method useful to improve the seasonal forecasting and to get reliable precipitation and streamflow projection for use in practical purposes. Reanalyses of Climate Forecast System (CFSR) is used for purpose of creating initial conditions for creating initial conditions for CFSv2 retrospective forecasts (Saha et al. 2014). CFSR is dynamically downscaled to analyze the credibility of Regional Climate Model products when seasonality and interannual variability of mean and extreme precipitation is concerned. A decade long (2000-2010) dynamically downscaled RCM simulation is generated using Weather Research and Forecasting model (WRF) with a 12-km spatial resolution covering the Colorado River basin by dynamically downscaling CFSR data. An additional convective-permitting nested domain (3 km resolution) is included for the WRF simulation for specific sub basins of Southwest U.S region. In this study, we have shown the improvement of convective permitting model product in representing mean climatology as well as climatology of extreme precipitation events in Upper and Lower Colorado basin. It is evident that use of regional model adds value to the reanalyses in terms to better spatial and temporal representation which is also consistent with previous studies (Prien et al. 2015, Liu et al. 2016). Hence it appears to be an important initial step towards seasonal to subseasonal (S2S) forecasting using downscaled product from global CFS forecast models.