Integrated ecological suitability assessment for sustainable ecotourism development: A case study of khodabandeh county, Iran

Sustainable ecotourism development requires an integrated evaluation of ecological capacity, environmental constraints, and spatial accessibility. This study assesses the ecological suitability of Khodabandeh County, northwestern Iran, for sustainable ecotourism development by integrating fuzzy logic, the Analytic Network Process (ANP), Random Forest (RF), and Geographic Information Systems (GIS). Six key ecological and infrastructural criteria—water availability, climate, landform, land use, conservation status, and accessibility—were identified through expert consultation and authoritative sources. Criterion weights were determined using a Delphi-based ANP framework implemented in Super Decisions software. To address uncertainty and gradual transitions in environmental variables, all criteria were standardized using fuzzy membership functions and integrated through weighted linear combination to generate a fuzzy ecological suitability map. In parallel, a five-class Random Forest model was trained using ۴۱۱ spatial reference points labeled by a panel of ۲۰ experts. Established natural, cultural, and historical tourist attractions were assigned to the highest suitability class, whereas the remaining points were classified through joint evaluation of the relevant thematic layers. Because the labels relied on variables also used in the ANP–fuzzy analysis, the RF model was treated as a complementary spatial-consistency assessment rather than as an independent validation framework. On the test subset, the RF model achieved an overall accuracy of ۰.۸۲۲۶ and a macro-averaged F۱-score of ۰.۷۹۳۵, with all misclassifications occurring between adjacent suitability classes. The final ANP–fuzzy–GIS suitability map classified ۸% of Khodabandeh County as completely suitable, ۲۶% as suitable, ۳۱% as relatively suitable, ۲۲% as unsuitable, and ۱۳% as completely unsuitable, with the first three classes collectively accounting for ۶۵% of the study area.. The overlay comparison between the ANP–fuzzy–GIS and RF maps yielded an exact spatial agreement of ۳۹.۸۳% and a Cohen’s kappa coefficient of ۰.۲۴۷, whereas ۹۰.۵۷% of pixels were classified within one adjacent suitability class. This contrast indicates that most disagreements involved shifts between neighboring ordinal categories rather than opposing suitability levels, supporting broad spatial consistency but not pixel-level equivalence. These findings can support differentiated spatial planning aimed at balancing ecotourism development with ecological protection in Khodabandeh County. Subject to local recalibration of the criteria, suitability thresholds, expert weights, and spatial datasets, the framework may also be applicable to other ecologically sensitive regions.

کلیدواژه ها:

ANP&ndash ، Fuzzy&ndash ، Random Forest framework ، Ecotourism suitability ، GIS-based land suitability analysis ، Khodabandeh County ، Sustainable tourism planning