Detecting Regional Climate Reversibility and Stabilization After Temperature Overshoot

Due to insufficient climate action to date, the world is on track to exceed 1.5°C of global warming in the coming decade. Stringent climate action towards net zero, followed by continued net negative carbon emissions, may allow temperatures to be brought back below that level after a prolonged period of climate overshoot. Even if global mean temperatures are reversed, how such overshoot shapes regional climate patterns in the long term remains poorly understood. Here, we investigate the long-term effects of climate overshoot using explainable machine learning models to identify persistent and reversible changes in regional temperature patterns for ensembles of two different overshoot scenarios until 2300. Our approach allows for robust detection of statistically significant differences on the regional level. We address three questions: (1) which regional temperature distributions return to their pre-overshoot state, (2) which stabilize at altered conditions, and (3) how distinguishable high overshoot and low overshoot pathways remain up to 2300. To complement the machine learning analysis, we apply principal component analysis to compare pre- and post-overshoot climate states and assess their degree of convergence. Our analysis provides a methodological framework to detect climate reversibility and stabilisation on the regional level, highlighting where long-term changes persist despite global temperature decline.