Spatial assessment of land suitability for agro-marine tourism on Wawonii Island, Indonesia

For the sustainable development of integrated coastal-based economic activities, spatially informed planning is needed to address environmental protection and livelihood improvement, particularly in small island environments that are increasingly impacted by land use pressures and environmental risks. In this study, a GIS-based machine learning method was used to evaluate and classify land suitability for maritime agrotourism development on Wawonii Island, Southeast Sulawesi, Indonesia. Twelve physical, environmental, and accessibility factors were combined within a geospatial framework and evaluated using Support Vector Machine (SVM) and Random Forest (RF) models, trained and validated on 52 existing maritime agrotourism sites. Model performance was evaluated using Receiver Operating Characteristic (ROC) curves, with the RF model having a higher predictive accuracy of approximately 80% (AUC ≈ 0.80) compared to the SVM model (AUC ≈ 0.60). Using the selected models, land suitability was categorized into four classes. The findings indicate that low suitability areas dominate the island, covering approximately 52.01% of the total area, while medium, high, and very high suitability classes cover 16.22%, 17.76%, and 14.01%, respectively. High and very high suitability areas are mostly found in coastal and lowland areas with gentle slopes, good accessibility, and a combination of agricultural and marine resources. Subdistrict-level analysis indicates that Central and South Wawonii are priority areas for marine agrotourism development, while other areas require more careful land use management and a conservation-oriented approach. These results clearly demonstrate the efficiency of GIS-based machine learning as a spatial decision support system for small island planning and provide quantitative and spatial evidence for the formulation of sustainable tourism policies.