Comparing six different species distribution models with several subsets of environmental variables: Predicting the potential current distribution of Guettarda speciosa in Indonesia

Abstract

Abstract. Yudaputra A, Pujiastuti I, Cropper Jr. WP. 2019. Comparing six different species distribution models with several subsets of environmental variables: predicting the potential current distribution of zebra Guettarda speciosa in Indonesia. Biodiversitas 20: 2321-2328. There are many algorithms of species distribution modeling that widely used to predict the potential distribution pattern of diverse organisms. Finding the best model in terms of predicting the potential distribution of many species remains a challenge. The objective of this study is to compare six different algorithms for predicting the potential current distribution pattern of Guettarda speciosa (zebra wood). The occurrence records of G. speciosa are derived from herbarium database, Bogor Botanic Gardens’s plant inventory database and direct field surveys through NKRI expedition. Seven climatic variables and elevation data are extracted from global data. R open-source software is used to run those algorithms and QGIS is used to prepare the spatial data. The result shows that MAXENT outperforms other predictive models with the highest AUC score 0.89, followed by SVM (0.87), RF (0.86), and GLM (0.82), DOMAIN (0.73), and BIOCLIM (0.62). Based on the AUC score, the four predictive models (MAXENT, SVM, RF, GLM) are categorized into good predictive models, indicating those are quite better to predict the potential current distribution pattern of G. speciosa. Whereas, DOMAIN is fair predictive model and BIOCLIM is poor predictive model. The predictive map derived from four models (MAXENT, SVM, RF, and GLM) shows almost similar appearance in predicting of potential current distribution of G. speciosa. The predictive map of current distribution would be useful to provide information regarding the potential habitat of G. speciosa across the landscape of Indonesia.