Modeling tourism-driven coastal land-use change and habitat quality under alternative development pathways: Implications for environmental management from explainable AI

Abstract

The study area is a tourism-driven coastal city, which is one of the most sensitive areas regarding the combination of development pressure and ecological sensitivity in the Mediterranean region. In this study, we integrated Random Forest-based transition potential modeling, Markov-based baseline demand, cellular automatabased spatial allocation, SHAP-based interpretation of the driving factors, and the InVEST Habitat Quality model to project future land-use changes and assess their implications for habitat quality. Our findings suggest that there has been a 3.5-fold increase in artificial surfaces from 1990 to 2025 with a major focus on coastal and valley areas, resulting in the fragmentation of semi-natural vegetation over time. This is due to increasing pressure from tourism development. The selected CA configuration achieved an Overall Accuracy of 0.715, indicating moderate agreement between the observed and simulated 2025 land-use patterns. By 2075, artificial surfaces reach 62.6 km2 under SSP1, 66.8 km2 under BAU, and 76.6 km2 under SSP5. Mean habitat quality declines from 0.566 in 2025 to 0.548, 0.530, and 0.508, respectively. These contrasting trajectories demonstrate the importance of compact development and stronger control of tourism-driven coastal expansion for sustainable coastal land management.