Spatiotemporal evolution of land use in Ili, Xinjiang, China: a comparative evaluation of PLUS and Logistic-CA-Markov models

This study employed the PLUS and Logistic-CA-Markov models to simulate the spatiotemporal evolution and expansion pattern of land use within the Ili River Valley, Xinjiang, China. Integrating 1-km spatial resolution land use data from 2005 to 2015 with nine driving factors, we conducted a comparative assessment. Contrasting the model predictions against observed 2020 land use data confirmed the feasibility of our methodology. Key findings indicate: (1) significant land use transitions occurred between 2005 and 2010, characterized by expansions in cropland, grassland, and construction areas, alongside declines in forestland and water bodies; (2) GDP, precipitation, and elevation emerged as significant drivers influencing land use expansion; (3) The PLUS model outperformed Logistic-CA-Markov, achieving a higher Kappa coefficient (0.68 vs. 0.62) and overall accuracy (0.82 vs. 0.77), although both models exhibit relatively low FoM (Figure of Merit) indices (0.01 vs. 0.05). Specifically, the PLUS model attained superior accuracy for grassland (88.34% vs. 81.04%) and unused land (76.37% vs. 71.84%), while Logistic-CA-Markov model demonstrated marginally better accuracy for forestland (52.39% vs. 48.66%). These results demonstrate that the PLUS model possesses stronger reliability and practicality for simulating land use dynamics in this region. While both models effectively captured the overarching trends of land use expansion, their predictive ability for localized land use changes remains constrained.