Remote sensing driven multilayer perceptron markov modelling of land use change and ecosystem service decline in the Eastern Mediterranean

Abstract

Landscape transformation in biodiversity-sensitive Mediterranean regions presents a major threat to ecosystem integrity and long-term sustainability. This study applies an integrated modeling framework combining classified land use and land cover (LULC) maps (1985, 2024) and simulated projections (2050, 2070) using a Multilayer Perceptron (MLP) neural network and Markov Chain simulations to evaluate land dynamics in the Amanos Mountains of Turkiye, a critical biodiversity hotspot in the Eastern Mediterranean. Using high-resolution remote sensing, spatial drivers, and global ecosystem valuation coefficients, we examine both LULC transitions and their impact on ecosystem service values (ESVs). Projections indicate a 47% increase in agricultural land and a 54% expansion of artificial surfaces by 2070, primarily at the expense of forests and mixed vegetation. ESVs are expected to decline from similar to$1 billion/year in 1985 to below $620 million/year by 2070. Transition potential maps identify high-probability zones (P > 0.90) of LULC conversion, especially near road networks, urban peripheries, and forest-agriculture interfaces. These areas are at elevated risk of ecosystem service decline and fragmentation. The findings underscore the urgency for spatially targeted land governance and ecological corridor protection strategies in rapidly transforming Mediterranean systems.