Nonlinear Decline in Carbon Sequestration Potential of Dianchi Lake: Synergistic Impacts of Anthropogenic Pressures and Shrinkage Revealed by Multi-Source Remote Sensing and Explainable AI

City college, Kunming University of Science and Technology, 650500, Kunming, Yunnan, China

^* Corresponding author: luliwen@stu.kust.edu.cn

Abstract

Plateau lakes are critical yet vulnerable carbon sinks due to shallow depths and high anthropogenic disturbance under accelerating human pressures. This study quantifies the dual impacts of lake shrinkage and human activities on the carbon sequestration potential of Dianchi Lake, the largest freshwater lake on the Yunnan-Guizhou Plateau, China. Integrating multi-source remote sensing (Landsat 8/9, Sentinel-2, GF-5), hydrological modeling, and explainable artificial intelligence (XAI), we reveal a nonlinear 57% decline in carbon sink capacity (C_sink) from 2018 to 2023, driven by an annual lake area reduction of 0.68 km² and escalating anthropogenic stressors. Key findings indicate that anthropogenic factors, particularly industrial wastewater discharge and agricultural nitrogen runoff, significantly influence carbon flux variability, exceeding the impacts of climatic factors such as temperature and precipitation. While eutrophication- induced algal growth temporarily enhanced carbon fixation in the short term, long-term urbanization-driven wetland loss reduced organic carbon inputs. A novel Human Impact Index (H_index), incorporating comprehensive socio- environmental metrics, surged by 62%, highlighting intensified anthropogenic pressures. Scenario-based analyses demonstrate that stringent industrial controls could recover 60% of baseline C_sink by 2030, whereas integrated management stabilizes declines but fails to achieve full restoration. This study advances mechanistic insights into climate-human- carbon interactions in alkaline plateau lakes, advocating for policy prioritization of pollution mitigation and wetland restoration to safeguard fragile ecosystems.