Dynamical Downscaling in Seasonal Climate Forecasts: Comparison between RegCM- and WRF-Based Approaches

The purpose of the present study is to assess the large-scale signal modulation produced by two dynamically downscaled Seasonal Forecasting Systems (SFSs) and investigate if additional predictive skill can be achieved, compared to the driving global-scale Climate Forecast System (CFS). The two downscaled SFSs are evaluated and compared in terms of physical values and anomaly interannual variability. Downscaled SFSs consist of two two-step dynamical downscaled ensembles of NCEP-CFSv2 re-forecasts. In the first step, the CFS field is downscaled from 100 km to 60 km over Southern Europe (D01). The second downscaling, driven by the corresponding D01, is performed at 12 km over Central Italy (D02). Downscaling is performed using two different Regional Climate Models (RCMs): RegCM v.4 and WRF 3.9.1.1. SFS skills are assessed over a period of 21 winter seasons (1982–2002), by means of deterministic and probabilistic approach and with a metric specifically designed to isolate downscaling signal over different percentiles of distribution. Considering the temperature fields and both deterministic and probabilistic metrics, regional-scale SFSs consistently improve the original CFS Seasonal Anomaly Signal (SAS). For the precipitation, the added value of downscaled SFSs is mainly limited to the topography driven refinement of precipitation field, whereas the SAS is mainly “inherited” by the driving CFS. The regional-scale SFSs do not seem to benefit from the second downscaling (D01 to D02) in terms of SAS improvement. Finally, WRF and RegCM show substantial differences in both SAS and climatologically averaged fields, highlighting a different impact of the common SST driving field.