Development of an Improved Week 3-4 Temperature Consolidation First Guess

Each Friday, forecasters from the Climate Prediction Center (CPC) produce a Week 3-4 Temperature Outlook for the U.S. that indicates the probability of two-week averaged temperatures being above or below the climatological mean for the forecast valid period. The forecasters have a variety of tools at their disposal including dynamical model guidance from the European Centre for Medium-range Weather Forecasting (ECMWF), the Global Ensemble Forecasting System Version 12 (GEFSv12), the Climate Forecast System Version 2 (CFSv2), the Environment and Climate Change Canada (ECCC), and the Japan Meteorological Agency (JMA). In addition, the forecasters make use of a statistical Multiple Linear Regression (MLR) model that harnesses tropical conditions and the long-term trend as predictors. With this broad range of tools, it is often challenging to synthesize the various model output into a single outlook under the time constraints imposed by a real-time environment. This project sets out to identify a skillful consolidation of the models with the intent that it may be used as a first-guess outlook for the forecasters. We rigorously evaluate the retrospective skill of 1) the individual models, 2) equal-weighted blends of the models, 3) real-time skill-based consolidations of the models, and 4) reforecast skill-based consolidations of the models. Initial results indicate that amongst the individual models, the ECMWF provides the most skillful forecast, whereas the CFSv2 provides the least skillful forecast. Equal-weighted blends of the models tend to underperform the ECMWF alone. However, blends that provide additional weight to the ECMWF, while still including other model information from, for example, the GEFSv12, JMA or MLR, add some skill beyond the ECMWF alone. While work is ongoing, it is anticipated that both real-time skill-based and reforecast skill-based consolidations of the models will improve further upon the blends mentioned above. The most successful consolidations will be monitored in real-time over the course of the coming year, with the most skillful consolidation being provided to the forecaster as a first-guess outlook.