“This proposal responds to the 2016 solicitation for CPO’s Modeling, Analysis, Prediction and Projection (MAPP) program Competition 2: “Research to Advance Prediction of Subseasonal to Seasonal Phenomena.” The proposed project focuses on some of MAPP’s primary objectives, namely, “improving methodologies for global to regional-scale analysis, predictions, and projections” and “developing integrated assessment and prediction capabilities relevant to decision makers based on climate analyses, predictions, and projections.”
A large number of forecasts from a suite of models are routinely provided by the Subseasonal to Seasonal (S2S) Prediction Project and the North American Multi-Model Ensemble (NMME) Project. To develop a reliable and timely climate product from these datasets, we propose a new methodology to assess an individual model’s forecast skill, generate statistical weights based on the skill of member model forecasts of slowly-varying surface states, and use aforementioned weights to produce an optimized single forecast. We will compare this methodology to traditional multi-model combination techniques. The new methodology has unique advantages: a) It provides an ideal framework for regional analyses and prediction; b) It allows the combined atmospheric forecast to rely more on models with superior forecast skill of surface anomalies, which are the main drivers of the S2S forecast skill; c) Calculations of forecast skill and weights for each model are highly flexible, and the methodology has many potential applications. The weight of each model member can be calculated from the latest evaluation of the model’s forecast performance and may evolve over time.
Preliminary results show that the new methodology outperforms individual models and can increase the one-month lead forecast skill of surface air temperature by 50% over the simple multi-model average across much of the area of focus. Even though the forecast skill improvement of precipitation (P) and surface air temperature (T2m) over North America is our primary target, the effects are expected to reach all forecast variables over the globe. We propose to identify regions where there is significant forecast skill of North American P and T2m and diagnose the dominant factors influencing such skill. We seek to understand how these factors contribute to the forecast skill of P and T2m, especially the role of land surface processes in achieving S2S forecast skill, through crafted numerical experiments with the Climate Forecast System (CFS). The project will also explore the potential to improve S2S forecast skill by improving the quality of land surface initial states in CFS and examine impacts of land initialization on S2S forecast skill. The overall goal of the proposal is to enhance the Nation’s capability to predict variability on S2S time scales. By performing our analysis with the NMME and S2S forecast datasets and adapting it to operational settings, this proposal directly contributes to the NOAA Next-Generation Strategic Plan objectives of “an improved scientific understanding of the changing climate system and its impacts” and “mitigation and adaptation choices supported by sustained, reliable, and timely climate services.””