Spatiotemporal dynamics of land use transformation and its impact on ecosystem services in the Poyang lake urban agglomeration
摘要
Rational utilization of land resources is crucial for the sustainable development of regional ecosystems. Using the Poyang Lake Urban Agglomeration (PLUA) in Jiangxi Province, China, as a case study, this study investigates land use transformation (LUT) characteristics via the land use transition matrix, land use structure (LUS), land use intensity (LUI), and land use dynamics (LUD). Ecosystem services were evaluated using the InVEST model and remote sensing data. The nonlinear effects, thresholds, and spatial relationships of LUT on ecosystem services were analyzed using random forest models, constraint lines, and bivariate spatial autocorrelation methods. The findings indicate that (1) Between 2000 and 2020, woodland, grassland, and arable land decreased significantly by 671.02 km2, 247.92 km2, and 953.92 km2. In contrast, urban, rural construction, and industrial lands expanded by 573.42 km2, 185.79 km2, and 1187.39 km2. Significant land type conversions occurred, accompanied by notable differences in LUS, LUI, and LUD changes. (2) Average values of food supply (FS) and carbon sequestration (CS) increased by 33.4 t/km2 and 43.11 t/km2, while habitat quality (HQ) and water conservation (WC) decreased by 0.05 and 16.52 mm. (3) LUT exhibited nonlinear relationships with ecosystem services. LUI was identified as the predominant driver of FS and HQ, with %IncMSE values of 85.28 and 132.60. LUD exerted the strongest influence on WC (%IncMSE = 30.49), while LUS had the greatest impact on CS (%IncMSE = 36.60). A single threshold (0.26) was identified in the relationship between LUS and FS. For LUI, thresholds were identified at 2.10 for WC, 2.24 for HQ, and 1.54 for CS. No clear threshold was observed between LUD and any of the evaluated ecosystem services. (4) Spatially, an overall opposing distribution pattern was observed. FS, HQ, and CS primarily exhibited low-high and high-low clustering areas in relation to LUT, while WC showed a high-low clustering area.