Ecological niche model transferability of the white star apple (Chrysophyllum albidum G. Don) in the context of climate and global changes

Chrysophyllum albidum is a forest food tree species of the Sapotaceae family bearing large berries of nutrition, sanitary, and commercial value in many African countries. Because of its socioeconomic importance, C. albidum is threatened at least by human pressure. However, we do not know to what extent climate change can impact its distribution or whether it is possible to introduce the species in other tropical regions. To resolve our concerns, we decided to model the spatial distribution of the species. We then used the SDM package for data modeling in R to compare the predictive performances of algorithms among the most commonly used: three machine learning algorithms (MaxEnt, boosted regression trees, and random forests) and three regression algorithms (generalized linear model, generalized additive models, and multivariate adaptive regression spline). We performed model transfers in tropical Asia and Latin America. At the scale of Africa, predictions with respect to Maxent under Africlim (scenarios RCP 4.5 and RCP 8.5, horizon 2055) and MIROCES2L (scenarios SSP245 and SSP585, horizon 2060) showed that the suitable areas of C. albidum, within threshold values of the most contributing variables to the models, will extend mostly in West, East, Central, and Southern Africa as well as in East Madagascar. As opposed to Maxent, in Africa, the predictions for the future of BRT and RF were unrealistic with respect to the known ecology of C. albidum. All the algorithms except Maxent (for tropical Asia only), were consistent in predicting a successful introduction of C. albidum in Latin America and tropical Asia, both at present and in the future. We therefore recommend the introduction and cultivation of Chrysophyllum albidum in the predicted suitable areas of Latin America and tropical Asia, along with vegetation inventories in order to discover likely, sister or vicarious species of Chrysophyllum albidum that can be new to Science. Africlim is more successful than MIROCES2L in predicting realistic suitable areas of Chrysophyllum albidum in Africa. We therefore recommend to the authors of Africlim an update of Africlim models to comply with the sixth Assessment Report (AR6) of IPCC.