Multidimensional factors and mitigation scenarios for carbon emissions (2014–2022) in urban agglomeration: a machine learning analysis of the Nansi Lake Basin, China

Urban agglomerations are key areas for carbon emissions, and lake ecosystems are statistically associated with the regulation of their carbon cycles. However, existing research on carbon emissions in urban agglomerations largely focuses on macro-scale studies. It lacks a systematic analysis of the multidimensional factors associated with carbon emissions at the fine community scale in lake-type urban agglomerations, particularly neglecting watershed-specific ecological processes that are associated with carbon regulation such as aquatic vegetation. This study uses the Nansi Lake Basin urban agglomeration as its research site, aiming to quantify the multidimensional factors associated with community-scale carbon emissions and simulate spatially gradient-based carbon reduction optimization strategies. Based on multi-source data from 2014 to 2022, XGBoost was used for regression prediction and feature importance analysis, combined with scenario design and optimal strategy simulation. The results show that multicollinearity and weak linear correlation among independent variables, but the ensemble model has high prediction accuracy (average R² = 0.866). Feature importance analysis indicates that carbon emission factors exhibit a dual characteristic of “socio-economic dominance” and “prominent watershed characteristics.” This study designed three-tiered intervention scenarios: key intervention, moderate intervention, and comprehensive optimization. Based on the model’s predictions, adjusting policy-adjustable indicators such as Temperature and Population is associated with carbon emission reductions of approximately 7,000 tC. DEM is treated only as a static topographic control for model sensitivity diagnosis and is not considered a policy-adjustable mitigation option. This research provides a quantifiable and spatially explicit framework for exploring carbon reduction options for the Nansi Lake Basin and similar lake-type urban clusters and has significant decision-making reference value for formulating differentiated regional low-carbon development plans.