Identification of the most suitable afforestation sites by Juniperus excels specie using machine learning models: Firuzkuh semi-arid region, Iran

Abstract

Choosing Selecting suitable sites for afforestation is a complex process that is influenced by various factors that require the use of new models and methods in order to create better results. The main purpose of this study is to investigate the use of a machine learning framework to map the best sites for afforestation with J. excelsa, an important species for soil and water conservation in Firuzkuh County, Tehran Province, Iran. Existing stands of J. excelsa were located. Measures of 14 environmental variables were compiled at each site. Three machine learning algorithms–Fuzzy ARTMAP (FAM), Multi-layers perceptron (MLP), and Classification tree analysis (CTA) – were used to model ideal locations for growing the tree. They were compared in terms of success rate. The best performance was achieved by CTA (area under curve (AUC) = 0.899). MLP (AUC = 0.892) was second best, and FAM (AUC = 0.835) had the lowest success. All three models achieved very good to excellent results; however, the CTA model was the most effective. Locations of high and very high favorability for J. excelsa comprise between 8% and 18% of the study area. The factors that are most important for the locations of replanting are those with bedrock of the Cl geological group and where rainfall ranges from 350 mm/year and 450 mm/year. This study offers support to decision makers for improving (lower cost and less time) selection of planting sites that are more likely to support tree survival to achieve natural restoration.