Land capability evaluation of afforestation using Random Forest algorithm (Kan Watershed, Tehran)

It is necessary to evaluate sustainable spatial allocation of afforestation. For this purpose, this study was conducted in the Kan watershed of Tehran province to assess the suitability of land for afforestation. First, suitable tree species were chosen based on land characteristics of study area and purpose of restoration. Then, the ecological demands of tree species were investigated and effective Indicators that affect the evaluation process were identified. After processing, classification and integration of spatial layers in GIS using the system analysis method, a random forest algorithm was trained and suitability map of afforestation was produced. Results show that Random Forest method has high accuracy in predicting suitable areas for afforestation. Also, 2116 ha of the study area is moderately suitable for afforestation. Based on Boruta algorithm Soil depth, growing season precipitation, elevation, soil texture, slope, and aspect are considered as the most important to the least important features, respectively and it is not necessary to carry out weighting methods for evaluation of afforestation capability. Generally, the Random Forest method can be used as a capable way to prepare ecological capability maps.