Multidimensional effects of history, neighborhood, and proximity on urban land growth: A dynamic spatiotemporal rolling prediction model (STRM)

Accurate prediction of future urban land demand is essential for effective urban management and planning. However, existing studies often focus on predicting total demand within an administrative region, neglecting the spatiotemporal heterogeneities and interrelationships within its subregions, such as grids. This study introduces a dynamic spatiotemporal rolling prediction model (STRM) that integrates historical trends, neighborhood status, and spatial proximity for spatially explicit prediction of urban land demand at a grid level within an administrative region. STRM leverages historical urban land demand and proximity information from neighborhood grids to predict future demand of the foci grid. By integrating history and neighborhood information into a deep forest model, STRM provides an approach for rolling predictions of grid‐level urban land demand. Parameter sensitivity and structural sensitivity analyses of STRM reveal the impact of historical lags, neighborhood size, and spatial proximity on urban land demand predictions. Application of STRM in Wuhan demonstrated the performance of STRM over a 17‐year period (2000–2017), with an average adjusted R2 of 0.89, outperforming other urban land demand prediction models. By predicting demand on a year‐by‐year basis, STRM effectively captures spatiotemporal heterogeneity and enhances the resolution of urban land demand prediction. STRM represents a shift from static macroscopic to dynamic microscopic prediction of urban land demand, offering valuable insights for future urban development and planning decisions.