To examine climate predictability on seasonal time scales (3-6 month forecasts) using regional models, in this study we continue our advancement and testing of a high resolution (32-km), Eta model-based Regional Climate Model (Eta RCM). The model is an adaptation of the NCEP operational Eta NWP model as of 24 July, 2001 (which is the Eta model version used in the NCEP 25-year Regional Reanalysis), with changes made to make the configuration of the model execution consistent with the longer time scales of seasonal forecasts, including daily updates to the fields of sea surface temperature (observed or predicted), sea ice cover (observed or predicted), green vegetation cover (climatology), and albedo (climatology).

To test the skill of the Eta RCM in predicting warm-season anomalies of precipitation, two summertime cases were chosen: 1988 and 1993, where 1988 is known for severe drought in the most part of the U.S., whereas 1993 manifested heavy flood-producing rainfall in the central United States. In contrast to many previous RCM studies driven by analysis lateral boundary conditions and initialized from one single date (i.e., one member realization), we used ensemble approach and both analyzed and predicted lateral boundary conditions. For each year, the model was run from three different initial dates starting from late May. For each run, two convection schemes (Betts-Miller-Janjic and Kain-Fritsch) were used with either analyzed lateral boundary conditions from the NCEP Global Reanalysis II or predicted lateral boundary conditions from the NCEP global Seasonal Forecast Model (SFM). Our focus in this study is the total precipitation and interannual variability.

We examine the resulting ensemble mean and individual members to demonstrate a) whether the Eta RCM successfully captures both wet and dry anomalies in total precipitation over the U.S. in the two years, and b) the predictability of such extreme events relative to the choice of both convection schemes and lateral boundary conditions. The results show that the Eta RCM can capture the dry/wet bias in the precipitation during both years and has notable sensitivity to the choice of convection schemes and lateral boundary conditions, with substantial member-to-member variability. The latter variability indicates that previous RCM studies that employed only "one member" initialized from one single date may be misleading by overlooking inherent internal variability.