Under the sponsorship of the National Ground Intelligence Center (NGIC), the National Center for Atmospheric Research (NCAR) has developed the Global Climatology Analysis Tool (GCAT) to generate “on demand” large databases of atmospheric parameters at high resolution, tailored to the specific transport and Dispersion needs of NGIC. The WRF model is applied to dynamically downscale NNRP global analyses and to generate long records, up to 30 years, of hourly fine scale (~3km) gridded data. The goal is to refine large scale seasonal predictions from the Climate Forecast System (CFS) at a regional scale.

Artificial intelligence techniques such as the Self Organizing Maps (SOMs) are locally applied to a pre-defined area to automatically classify large volumes of data and extract weather regimes representative of the region at different scales: global (NNRP, CFS), and local (WRF). The first step consists in identifying local meteorological effects related to the global patterns selected over a training period covering the past 30 years. An inter-comparison is then made to evaluate the skills of the different data sets. In a second step, SOMs are applied to CFS seasonal forecasts to quantify the likelihood of the identified local events to occur.