Using machine learning to unravel the driving mechanisms of carbon balance in resource-based cities: A “Pattern-Process-Service-Sustainability” framework

Resource-based cities (RBCs) have borne carbon emissions far exceeding their own development needs during rapid urbanization, making them critical areas for achieving carbon neutrality. However, existing studies have largely focused on either carbon emissions or carbon sequestration as separate processes, thereby limiting refined spatial governance. Using multi-source remote sensing data for 125 resource-based cities, this study proposes an integrated analytical framework that incorporates the Pattern-Process-Service-Sustainability (PPSS) paradigm and combines interpretable machine learning with partial least squares structural equation modeling (PLS-SEM) to identify and explain the driving mechanisms of carbon balance. The results show that, from 2002 to 2023, China’s RBCs generally remained in a state of carbon surplus, but their carbon balance declined by 69.73%, while the proportion of carbon-deficit cities increased from 28.00% to 48.80%. Spatially, carbon deficits were mainly concentrated in coal- and oil & gas-based cities in North China, whereas carbon-surplus cities were primarily distributed in forestry-based cities and cities with favorable ecological backgrounds in Northeast and Southwest China. The SHAP contribution results indicate that soil retention (SDR), green space (GS), and habitat quality (HQ) are closely associated with carbon balance and exhibit pronounced threshold effects. The PLS-SEM results further reveal that ecosystem services serve as an important link between landscape patterns and carbon balance. Rather than exerting a simple linear influence, landscape patterns mainly affect carbon balance indirectly through the “pattern-process-service” pathway. These findings provide a theoretical framework and methodological reference for optimizing source-sink spatial configurations and formulating differentiated carbon-neutrality strategies in resource-based cities.