Integrating MaxEnt and CA–Markov–MLP for Multi-Temporal Landslide Susceptibility Modelling

Landslide susceptibility is often treated as a static assessment of present-day conditions, overlooking the temporal evolution of geomorphological and anthropogenic drivers. This limitation is particularly relevant in rapidly urbanising areas, where land use change continuously alters environmental conditions influencing slope stability. This study examines the temporal evolution of landslide susceptibility in the Grigorescu neighbourhood of Cluj-Napoca, Romania, using environmental datasets representing conditions in 1971, 2009, and 2025, along with a projected land use scenario for 2047. The proposed framework integrates multi-temporal landslide inventories and conditioning factors with Maximum Entropy (MaxEnt) modelling and CA–Markov–MLP land use simulation (MOLUSCE). Results indicate a progressive shift towards higher susceptibility classes over time, accompanied by urban expansion onto increasingly steep terrain. However, slope gradient remained the dominant conditioning factor throughout all analysed periods, while land use change influenced the temporal evolution and spatial redistribution of susceptibility through progressive urban expansion into terrain already predisposed to instability. The 2047 scenario suggests that continued urban expansion may increase the exposure of built-up areas to zones of elevated susceptibility. Model performance was robust (AUC > 0.8; Kappa > 0.9). Beyond site-specific findings, the framework provides a transferable methodology for integrating urban growth dynamics into landslide susceptibility assessment, supporting sustainable spatial planning and risk-informed urban development in rapidly urbanising hilly environments.