Peri-urban farmland zoning based on morphology and machine learning: a case study of Changzhou City, China

Farmland zoning is a critical initiative for differentiated protection of farmland and promoting sustainable regional development. Landscape pattern index was predominantly utilized to reflect farmland landscape patterns, while the morphological characteristics did not receive enough attention. Additionally, the zoning methods were difficult to balance the evaluation results of multiple factors. To address these issues, this study conducted Morphological Spatial Pattern Analysis (MSPA) on peri-urban farmland in the main urban area of Changzhou City and evaluated it from the perspectives of production capacity, location condition, landscape pattern, and policy. A Particle Swarm Optimization-based K-Means clustering algorithm was then applied to farmland zoning. The results of the morphological analysis indicate that peri-urban farmland located in the direction of urban expansion and closer to the city is more susceptible to occupation and less spatially continuous. The multi-perspective evaluation results with obvious spatial differentiation characteristics were divided into four regions using the clustering algorithm, with villages as the minimum cluster unit in the study area. Finally, targeted management suggestions were proposed based on the clustering indicator characteristics. Overall, MSPA and the clustering algorithms leveraged in this study offer valuable methods for analyzing and zoning farmland. The findings are helpful in promoting the diversified protection and sustainable utilization of farmland resources.