Estimating Seasonal to Multi-year Predictability of Statistics of Climate Extremes using the CESM2-based Climate Prediction System

Climate extremes, such as heat waves, heavy precipitation, intense storms, droughts, and wildfires, have become more frequent and severe in recent years as a consequence of human-induced climate change. Estimating the predictability and improving prediction of the frequency, duration, and intensity of these extremes on seasonal to multi-year timescales are crucial for proactive planning and adaptation. However, climate prediction at regional scales remains challenging due to the complexity of the climate system and limitations in model accuracy. Here we use a large ensemble of simulations, assimilations, and reforecasts using Community Earth System Model version 2 (CESM2) to assess the predictability of statistics of climate extremes with lead times of up to 5 years. We show that the frequency and duration of heat waves during local summer in specific regions are predictable up to several months to years. Sources of long-term predictability include not only external forcings but also modes of climate variability across time scales such as El Niño and Southern Oscillation, Pacific Decadal Variability, and Atlantic Multidecadal Variability. This study implies opportunities to deepen our scientific understanding of sources for long-term prediction of statistics of climate extremes and the potential for the associated disaster management.