Land subdivision in the law's shadow: Unraveling the drivers and spatial patterns of land subdivision with geospatial analysis and machine learning techniques in complex landscapes

Abstract

Land subdivisions, especially in rural areas, pose a significant threat to sustainable development in many regions of the world. This issue is particularly challenging to understand in complex landscapes, where many biophysical and anthropic drivers interact without the necessary land regulatory guidance. We combined kernel density analysis and machine learning modeling to unravel the spatial patterns of land subdivisions and the complex relationships between their drivers. We used the Los Lagos region in southern Chile as a study case because it is a global biodiversity hotspot where land subdivisions are constantly increasing. We identify a significant increasing trend of subdivisions. Our modeling approach showed robust performance with an R² of 0.727, RMSE of 5.109, and a bias of −0.009. The proximity to urban areas, to the coast, distance to electric mains, demographic structure, and proximity to protected areas were significant predictors of land subdivision. Fertile lands, particularly those near urban centers, have become prime targets for subdivisions, exacerbating the conflict between urban development and agricultural sustainability. We highlight the increasing number of subdivisions on threatened ecosystems and highly productive soils. We discuss the interrelationship between the drivers and conclude that subdivision is primarily associated with conventional urban sprawl, although other urbanization phenomena could also be observed in some areas. These findings provide challenges and opportunities for global spatial planning and harmony with biodiversity conservation.