A statistical and machine learning methodology to model rural depopulation risk and explore its attenuation through agricultural land use management
Daniel Jato‐Espino, Fernando Mayor-Vitoria
Abstract
The abandonment of rural areas has become a major demographical challenge in recent years, especially in Spain and, more specifically, in the Valencian Community. A classification released by the government of this region revealed that almost a third of its municipalities are at depopulation risk. This classification is based on demographic variables, which are valid for identifying the phenomenon but insufficient to provide insight into how to counteract it. Instead, this study developed a methodology to model rural depopulation risk from land use and socioeconomic variables. Correlation analysis and principal component analysis enabled identifying which variables were meaningful for rural depopulation. Then, support vector classification was used to fit the demography-based depopulation classification used by the regional government. The mean accuracy reached was above 80%, which validated the proposed model and variables. Since crop areas was found to be one of the most influential variables in such model, the potential of agroeconomic measures to counter depopulation was examined. The results achieved suggested that depopulation might be reduced by 25% if exploiting 25% of the areas suitable for agriculture. In view of these outputs, public administrations may promote the implementation of land use spatial strategies based on sustainable agriculture.