Abstract: Under the background of rapid urbanization, it is of great practical significance to reveal the spatio-temporal differentiation characteristics and influencing factors of non-agriculturalization of cultivated land in the Guangdong-Hong Kong-Macao Greater Bay Area for coordinating economic development and cultivated land protection in highly urbanized areas. Based on the multi-period land use data from 2000 to 2020, the spatial-temporal pattern and influencing factors of non-agriculturalization of cultivated land in the Guangdong-Hong Kong-Macao Greater Bay Area were systematically analyzed by integrating natural geography and socio-economic factors, and using the change index of cultivated land quantity, spatial autocorrelation analysis and interpretable machine learning method (XGBoost-SHAP-PDP). The results showed that the area of cultivated land in the Guangdong-Hong Kong-Macao Greater Bay Area continued to decrease by 270 000 hm² during the study period. The non-agriculturalization rate showed a stage evolution characteristic of rapid growth-fluctuating decline-slow recovery over time, and showed a significant “core-periphery” gradient differentiation pattern in space. The global Moran's I was 0.24-0.53, and the non-agriculturalization of cultivated land had significant spatial agglomeration. The high-value areas gradually expanded from the core region of Guangzhou, Foshan and Huizhou to the eastern coastal zones. The overall migration trend of the gravity center of cultivated land non-agricultural conversion followed the sequence of “southeastward-northwestward-northeastward”, and the migration distance increased from 4.24 km in 2005-2010 to 18.99 km in 2010-2015, and further expanded to 20.72 km in 2015-2020. The directional characteristic of its spatial distribution first weakened and then strengthened. In terms of influencing factors, population and economic conditions served as major drivers, while their impacts gradually declined over time. By contrast, agricultural mechanization and grain output exerted an increasingly prominent restraining effect. Complex non-linear synergistic relationships existed among various influencing factors.

Key words: cultivated land non-agriculturalization, spatio-temporal evolution, influencing factors, spatial autocorrelation, interpretable machine learning, Guangdong-Hong Kong-Macao Greater Bay Area