Scenario analysis under climate extreme of carbon peaking and neutrality in China: A hybrid interpretable machine learning model prediction

Abstract

Climate extremes pose huge obstacles to societal development and unexpectedly elevate carbon emissions, emphasizing the significance of this study regarding China's carbon peaking and neutrality goals. To investigate these impacts, pertinent data from 1981 onwards in China were analyzed using grey relational analysis to identify key impact indicators. Based on the Impact of Population, Affluence, and Technology (IPAT) model, hierarchical clustering categorized these indicators, and multiple scenarios were constructed for evaluation. Utilizing a multi-factor approach, an optimal machine learning model was selected with nuanced accuracy to predict China's carbon peaking time, emission values, and the probability of achieving carbon neutrality by 2060. Through the combined application with Shapley Additive Explanations (SHAP), this study enhanced the model's interpretability and superiority. The findings yield distinctive insights: (1) Achieving the 2030 carbon peaking target is challenging due to extreme climatic influence on economic downturns, population decline, and technological barriers; (2) The 2060 carbon neutrality goal is attainable under certain conditions; and (3) Indicators within population, affluence, and technology segments show varying reduction effectiveness with different correlations and underlying factors. This study highlights China's capacity to achieve its carbon objectives, formulating targeted climate adaptation policies on promoting sustainable economic development and maintaining population dynamic structure.