Spatiotemporal heterogeneity and response quantification of agricultural carbon dynamics and non-point source pollution in Guangdong Province, China

In the context of global climate governance, agricultural non-point source pollution (ANPSP) and carbon dynamics (CD) are major challenges to agricultural sustainability, particularly in rapidly urbanizing coastal regions worldwide. This study used geographic information system, structural equation model, and SHapley Additive exPlanations to investigate the spatiotemporal heterogeneity and response quantification of ANPSP and CD in Guangdong Province from 1994 to 2023. The results indicated that (1) total nitrogen (TN), total phosphorus (TP), and chemical oxygen demand (COD) declined stepwise overall, and the relationship between carbon emissions (CE) and carbon uptake (CU) was complex. The hot spots of ANPSP became increasingly concentrated in southwestern Guangdong. (2) Livestock production (LP) showed the strongest associations with TN, TP, and COD. The FAR, CP, and LP had potential indirect pathways. CU was significantly negatively associated with CE. (3) LP features accounted for the largest contribution across ANPSP, with goat husbandry accounting for 17.75%, 38.73%, and 40.61% in TN, TP, and COD, respectively. FAR accounted for the largest contribution to CE (62.68%), while rice (44.77%) and sugarcane (31.11%) accounted for the largest contributions to CU. These results could provide scientific support for quantitative characterization of ANPSP and CD in the Guangdong-Hong Kong-Macao Greater Bay Area.