Subseasonal-to-Seasonal (S2S) Forecast Skill Attribution across the United States

Subseasonal-to-seasonal (S2S) precipitation forecast skill is critical for sectors that depend on medium-range forecasts, such as energy grid management, irrigated agriculture, drought and flood mitigation, and long-term water supply planning. Reliable S2S forecasting models are essential for adapting water resource management practices to shifts in hydroclimatology. To improve these forecasts, it is important to understand the factors that influence S2S precipitation. While short and long-range forecasts are relatively accurate, the skill of S2S forecasts—ranging from 15 to 90 days—is often less reliable. Understanding the current skill of S2S precipitation forecasts and identifying the factors that contribute to this skill is key to operationalizing these models. Known influences on S2S precipitation across the Conterminous United States (CONUS) include large-scale atmospheric circulation patterns and climate oscillations, such as the El Niño-Southern Oscillation (ENSO), Madden-Julian Oscillation (MJO), Pacific Decadal Oscillation (PDO), and Atlantic Multidecadal Oscillation (AMO). This study investigates the contribution of several different hydroclimatic indices on S2S precipitation forecast skill of the European Centre for Medium-Range Weather Forecast (ECMWF) Model.  The attribution analysis considers the correlation of the absolute error between forecast and observed precipitation values with each of the indices including Niño-3.4, MJO, PDO, AMO, Pacific North American Pattern (PNA), and North Atlantic Oscillation (NAO) for lead times of 15-90 days and for four forecast-initialized seasons: a) JFM, b) AMJ, c) JAS, and d) OND. Additionally, feature importance is evaluated using lasso regression for feature selection and principal component analysis. As climate change exacerbates hydroclimatic extremes, developing accurate forecasting models is essential for preparing for future uncertainties.