This event will be held remotely over Zoom.
Title: Regime shifts in seasonal California rainfall: Any long-term predictability?
Abstract: Most of California precipitation falls during the water year October to April, and the amount of rainfall the state receives is critical for water resource management. Although some progress has been made on subseasonal forecasting of atmospheric rivers, that bring a large portion of California rainfall when they hit the coast, extending forecasts beyond 2-3 weeks is still out of reach. This is problematic for the Department of Water Resources (DWR) of California, since decisions for water management must be made months ahead, at seasonal rather than subseasonal time scales. Of particular interest for decision makers is anticipating persistence in anomalous hydrological conditions (for example, whether a dry year will stay dry), and be confident that no seasonal regime shifts may occur that would make water management efforts irrelevant. Historically, seasonal forecasting of California rainfall has relied on the teleconnection with El Nino Southern Oscillation (ENSO), but the relationship has weakened in recent years. As the current skill of California rainfall seasonal forecasting is low, there is a need to identify and evaluate other sources of predictability.
In this project, our focus is to identify potential term sources of predictability that may help anticipating regime shifts in California water year. Using seasonal hindcast/forecasts from the North American Multi-Model Ensemble (NMME), we explore whether some climate drivers can explain the ensemble spread and what large-scale dynamic features differentiate good from poor forecasts. We also perform AMIP-type experiments with the Whole Atmosphere Community Climate Model (WACCM) in which tropical and/or high-latitude variability is imposed through nudging of the temperature and wind fields. This allows us to estimate whether/how? seasonal predictability can be gained from a better representation of tropical and high-latitude variability in the models. The project also includes the development of machine learning-oriented statistical models to explore potential skill in seasonal California rainfall and regime shifts from empirical methods.