Enhancing Subseasonal and Seasonal (S2S) Forecasts

Computer forecasts of weather and climate variability created by the U.S. and other nations on the subseasonal scale (weeks) and on the seasonal scale (months) are increasingly used by private and public activities to mitigate risk and seize opportunity. These forecasts are produced as ensembles of tens of individual computer forecasts and are most useful when calibrated by comparing histories of both forecasts and observations and then presented as probabilities of future events. This project will explore several techniques for creating new forecast systems in two phases: first by using superensembles assembled from forecasts by U.S. agencies and then by using broader collections that include forecasts from some of the most prominent national and international forecast centers. The goal is: Improve the skill and statistical reliability of subseasonal to seasonal (S2S) forecasts in ways that lead to more effective decisions and actions by the users. First, the project will explore how the performance of superensembles varies with the number and skill of contributed forecast systems. The second effort will attempt to create hybrid forecasts that combine the conventional computer forecasts with statistical forecasts. The third effort will examine improvements in forecast performance achieved with innovative new forecast calibration strategies.