Prediction of habitat suitability of Morina persica L. species using artificial intelligence techniques

The Morina genus has 13 species in the world, out of which only M. persica L. is found to be growing wild in Iran. The aim of this research is to predict the spatial distribution and model the habitat suitability for M. persica species using four data mining models: maximum entropy (MaxEnt), support vector machine (SVM), generalized linear model (GLM), and boosted regression trees (BRT). A total of 404 M. persica locations were identified during extensive field surveys, and their geographical locations were recorded using a global positioning system (GPS) device. Furthermore, seventeen environmental predictors including topographical, geological, climatic, and edaphic factors were selected, and their thematic layers were mapped in ArcGIS. Lastly, habitat suitability was modeled using data mining techniques. The validity of the results was assessed using the area under the receiver operating characteristic curve (AUROC). Moreover, three cutoff-dependent metrics, Cohen's kappa, sensitivity, and specificity, were used for more scrutinized performance assessment. The results revealed that the highest effects on M. persica distribution were mostly associated with edaphic factors, followed by climatic, lithological, and topographical factors. The results showed that MaxEnt with an AUROC value of 95% showed an outstanding performance in terms of prediction power and generalization capacity, followed by SVM (94.1%), GLM (87.4%), and BRT (84.7%). Comparing the AUROC values, MaxEnt was selected as the premier model with the best performance for M. persica distribution modelling across the study area. Cutoff-dependent metrics were also in line with AUROC values; however, the latter made a more discernible distinction between the performance of SVM and MaxEnt. The GLM, BRT, SVM, and MaxEnt models classified 37.37%, 27.28%, 23.31%, and 6.51% of the study area as high and very high suitable habitats for M. persica, respectively. The inferences of this research would be of interest to authorities in the natural resources sector, the research community, local stakeholders, and biodiversity conservation agencies for use in conserving and reclaiming M. persica habitats in the study area.