Climate-driven vegetation greening in Southwest China's Karst region: A multi-scale kNDVI analysis

Vegetation in Karst regions is highly sensitive to climate change, yet vegetation-climate relationships remain poorly quantified across spatial scales in these complex landscapes. Using 2000-2022 MODIS data, we apply kernel Normalized Difference Vegetation Index (kNDVI)-which reduces soil/rock background effects-to examine climate-vegetation dynamics at pixel, vegetation-type, and regional scales in Southwest China Karst Typical Region (SWCKTR). Multi-method correlation analyses (Pearson, detrended, and moving-window partial correlations) reveal scale-dependent patterns. Regional analysis shows persistent greening (0.0048 yr^-1) despite warming (0.028 °C/year) and drying (-5.19 mm/year), with temperature as the dominant driver (R = 0.3289, p < 0.01). At the pixel scale, 61.32% of areas show positive temperature correlations, while 92.65% show negative precipitation correlations, reflecting karst hydrological constraints. Vegetation-type analysis reveals divergent sensitivities: Northern Tropical Humid Semi-Evergreen Seasonal Rainforest exhibits the fastest spring greening (0.0086 yr^-1, p < 0.01) and positive autumn precipitation correlation (0.3705, p < 0.01), while Subtropical Mountain Coniferous Forest shows negative precipitation responses across all seasons (summer: 0.5808, p < 0.01) and autumn degradation (-0.0014 yr^-1). Seasonally, spring shows the fastest regional greening (0.0069 yr^-1, p < 0.01) with positive climate correlations, while summer exhibits the slowest growth (0.0039 yr^-1) despite strongest warming. Residual analysis indicates climate factors dominate (>80% of pixels), though human activities contribute significantly near urban centers (>10% positive residuals). Multi-scale integration reveals hierarchical climate-vegetation coupling, with temperature effects consistent across scales while precipitation effects vary by scale, season, and vegetation type.