Ecotourism suitability assessments increasingly rely on ecological indicators that capture spatial heterogeneity, ecosystem sensitivity, and biodiversity value. This study develops a hybrid indicator-based framework that combines fuzzy expert knowledge with ensemble machine learning to quantify ecotourism suitability in Malatya Province, Türkiye. Fifteen ecological and socio-environmental predictors, including elevation, slope, climate variables, river proximity, biodiversity richness, endemic species distributions, and land-cover patterns, were incorporated into a GIS-based analytical environment. Expert-derived fuzzy weights were computed using the Fuzzy Logarithmic Methodology of Additive Weights (F-LMAW) to generate an Ecological Suitability Indicator (ESI). Complementarily, K-Means clustering was used to derive data-driven suitability classes, which were modeled using Random Forest and XGBoost with spatial-block cross-validation. XGBoost demonstrated superior classification performance (accuracy = 66.8%; κ = 0.585). Across all models, biodiversity richness, endemic species presence, slope gradients, and river corridors consistently emerged as key ecological determinants. While the ESI produced conservative suitability zones, ensemble learning identified broader high-quality ecological landscapes, including the Levent Valley corridor, Nemrut foothills, and river-based habitats. The convergence between expert- and data-driven indicators demonstrates the reliability of hybrid ecological assessment. The proposed framework offers a transparent and transferable approach for constructing ecological suitability indicators in data-scarce, biodiversity-rich regions.
Karadeniz et al. (2026) studied this question.