Advancing annual global mean surface temperature prediction to 2 months lead using physics based strategy

Abstract

Abstract Interannual global mean surface temperature (GMST) forecast provides critical insights into the economic and societal implications of climate variability. The pronounced GMST elevation in 2023–2024 indicates that the Earth may have accumulated enough heat to cause widespread disasters, underscoring the necessity for establishing accurate short-term GMST predictions to offer timely and sustainable public service. However, capturing high-frequency annual variability (ANV) component of GMST poses challenges due to its susceptibility to intraseasonal-to-interannual (ISI) noises, particularly across the Northern Hemisphere’s mid-to-high latitudes. Averaging these ISI variations in November and December effectively enhances signal clarity, especially over oceans, and masks unpredictable noises on land. By forecasting the average GMST for November and December to extract ANV predictability, a strategy for annual GMST prediction was established. This approach successfully advanced precise GMST hindcasts by up to 2-months during 1980–2022, exceeding performance of existing climate models and boosting early warning for interannual GMST shifts.