Modeling Farmland Conversion Dynamics with a Discrete-Time Hazard Approach and Spatial Autocorrelation Adjustment

The conversion of farmland for non-agricultural purposes has become a critical issue worldwide, raising concerns about food security, sustainable land management, and rural development. While many governments have implemented regulations to control industrial activities on agricultural land, the challenge of residential housing development in farmland areas has often been overlooked.

This study adopts a 5-meter grid as the spatial unit and applies a discrete-time hazard model to investigate the dynamics of farmhouse development. Unlike conventional applications, the approach is enhanced by explicitly addressing the issue of spatial autocorrelation. First, the global Moran’s I statistic is employed to examine the spatial clustering of farmhouse distribution, which informs the inclusion of farmhouse density as an explanatory factor when appropriate. For other variables, Moran’s I is again used to evaluate spatial dependence. Variables with low spatial autocorrelation are directly incorporated into the model. For those with high spatial autocorrelation, ordinary least squares (OLS) regression and variance inflation factor (VIF) diagnostics are conducted to determine whether they serve as core explanatory factors. Core factors are transformed into neighborhood lag variables to capture local spatial effects while mitigating multicollinearity, whereas non-core factors are either excluded or combined with related variables to reduce redundancy. The refined set of variables is subsequently integrated into the discrete-time hazard model to estimate the drivers of farmhouse development.

The findings underscore the importance of integrating spatial statistics with hazard modeling to more effectively capture the complex processes underlying farmland residential development. This framework not only enhances methodological rigor but also provides a transferable approach for examining farmland conversion in diverse geographic contexts, offering valuable insights for land-use planning and agricultural land protection policies worldwide.