Pendekatan Machine Learning dalam Memetakan Kesesuaian Habitat Mal

Abstract

Climate change and increasing anthropogenic pressure pose serious threats to endemic species with restricted distributions, such as the Maleo (Macrocephalon maleo) of Sulawesi. This study aims to model habitat suitability and potential distribution of the Maleo using an integrated Geographic Information System and Maximum Entropy approach. Presence-only occurrence data were combined with bio-physical and anthropogenic environmental variables to generate spatial predictions of habitat suitability across coastal and lowland landscapes. The model demonstrated strong predictive performance, indicating that the selected variables effectively captured the ecological requirements of the species. Habitat suitability patterns revealed that sandy soil characteristics, proximity to natural heat sources, and river systems were the most influential factors enhancing habitat suitability, reflecting the species’ unique reproductive ecology. In contrast, proximity to roads and settlements consistently reduced suitability, highlighting the negative impact of human disturbance. The continuous suitability output was further classified into core habitat and buffer zones to support conservation-oriented spatial planning. The resulting zoning framework identifies priority areas for protection and management, particularly outside formal protected areas where development pressure is high. Overall, this study provides robust spatial evidence for understanding Maleo habitat requirements and offers a transferable methodological framework for modeling other endemic species. The findings underscore the importance of integrating ecological and human dimensions in habitat modeling to support effective, evidence-based conservation strategies