Spatial habitat suitability prediction of essential oil wild plants on Indonesia’s degraded lands

Background. Essential oils are natural products of aromatic plants with numerous uses. Essential oils have been traded worldwide and utilized in various industries. Indonesia is the sixth largest essential oil producing country, but land degradation is a risk to the continuing extraction and utilization of natural products. Production of essential oil plants on degraded lands is a potential strategy to mitigate this risk. This study aimed to identify degraded lands in Indonesia that could be suitable habitats for five wild native essential oil producing plants, namely Acronychia pedunculata (L.) Miq., Baeckea frutescens L., Cynometra cauliflora L., Magnolia montana (Blume) Figlar, and Magnolia sumatrana var. glauca (Blume) Figlar & Noot using various species distribution models. Methods. The habitat suitability of these species was predicted by comparing ten species distribution models, including Bioclim, classification and regression trees (CART), flexible discriminant analysis (FDA), Maxlike, boosted regression trees (BRT), multivariate adaptive regression splines (MARS), generalized linear models (GLM), Ranger, support vector machine (SVM), and Random Forests (RF). Bioclimatic, topographic and soil variables were used as the predictors of the model habitat suitability. The models were evaluated according to their AUC and TSS metrics. Model selection was based on ranking performance. The total suitable area for five native essential oil producing plants in Indonesia’s degraded lands was derived by overlaying the models with degraded land locations. Results. The habitat suitability model for these species was well predicted with an AUC value >0.8 and a TSS value >0.7. The most important predictor variables affecting the habitat suitability of these species are mean temperature of wettest quarter, precipitation seasonality, precipitation of warmest quarter, precipitation of coldest quarter, cation.