PREDICTING EDELWEISS (Anaphalis javanica) HABITAT SUITABILITY UNDER CLIMATE SCENARIOS: MACHINE LEARNING APPROACHES

Abstract

Climate change threatens species distribution by altering environmental conditions, posing severe risks to high-elevation endemic plants like Anaphalis javanica (Javan Edelweiss). Research on the performance of machine learning models in predicting the habitat suitability of Anaphalis javanica under various climate scenarios is crucial for generating accurate scientific information to support conservation efforts, given that climate change has the potential to affect the future distribution and mortality of this species. This study evaluated current and future habitat suitability for A. javanica across Java, Bali, and Nusa Tenggara using Decision Tree (DT), Random Forest (RF), and Maximum Entropy (MaxEnt) models. Species occurrence records were compiled from GBIF and published literature, while topographic, edaphic, climatic, and bioclimatic predictors. Model evaluation relied on Accuracy, Precision, Recall, F1-score, Cohen’s Kappa, and AUC. All models demonstrated outstanding predictive performance (accuracy: 0.953–0.974; AUC: 0.976–0.999), with Random Forest achieving the highest performance (accuracy: 0.974; Kappa: 0.947; AUC: 0.999). Soil type, elevation, temperature, and precipitation emerged as primary determinants. Currently, highly suitable habitats are concentrated in montane regions. However, projections for 2080 under RCP 2.6 and RCP 8.5 scenarios indicate a progressive decline in habitat suitability, most severely under RCP 8.5. These findings highlight that climate change will shrink suitable habitats and confine A. javanica to high-elevation climate refugia, underscoring the urgency of targeted conservation strategies.