Contrasting causal pathways of vegetation greening between economically strong and weak towns in China’s greater bay area

Abstract

Vegetation dynamics in rapidly urbanizing regions reflect complex interactions among climate drivers and human pressures, yet current studies have not fully captured fine-scale spatiotemporal heterogeneity at the township level. Here, we presented the first systematic town-scale analysis of fractional vegetation coverage (FVC) across the Guangdong-Hong Kong-Macao Greater Bay Area (GBA) from 2000 to 2020. Leveraging a high-precision MultiVI-derived FVC dataset alongside five drivers: human footprint, precipitation, air temperature, soil moisture, and vapor pressure deficit, we decomposed FVC trajectories into three pathway metrics: mean cover, long-term trend, and interannual variability. We then applied Local Moran’s I to detect spatial clustering, Space-Time Cube and Emerging Spatiotemporal Hot Spot Analysis to map cold-hot spot evolution, Geographical Convergent Cross Mapping to infer causality, Geographically Weighted Regression to reveal spatial non-stationarity of each explanatory power, Geographically Optimal Zones-based Heterogeneity and Locally Explained Stratified Heterogeneity models to quantify driver interactions, and Structural Equation Modelling to disentangle direct and climate-meditated anthropogenic effects. Our results showed that precipitation primarily dominated the causal strength of all three FVC pathways across the GBA, while human footprint and atmospheric drought exerted significant secondary influences. For the FVC mean pathway, economically strong urban cores experienced temperature-mediated suppression offset partially by direct anthropogenic greening efforts, while economically weak peripheral towns relied on soil moisture. By contrast, human activities predominated directly FVC pathways of trend and variability in both strong and weak towns, which specifically suppressed the growth trend but promoted the stability from year to year. These insights suggest that cores should prioritize land use controls and climate-adaptive greening, while peripheries invest in water-management infrastructure. Our framework of systematic spatiotemporal statistics and causal inference offers a transferable approach for precision ecological governance in global megaregions.