Comparison of Random Forest Models, Support Vector Machine and Multivariate Linear Regression for Biodiversity Assessment in the Hyrcanian Forests

Biodiversity is an important structural feature of dynamic and complex forest ecosystems. One of the most challenging and important issues in assessing the structure of forest ecosystems is understanding the relationship between biodiversity and environmental factors. Hyrcanian Forests are considered a biodiversity hotspot in the world and have special and unique features that have led to an emphasis and importance of biodiversity conservation in these forests. The aim of this study was to investigate the effect of biotic and abiotic factors on the diversity and richness of tree species in Hyrcanian Forests from the west of Gilan province to the east of Golestan province. For this purpose, using 655 fixed sample plots (0.1 hectare), the diversity of trees in 3 provinces in the northern Iran from east to west of the Caspian Sea was analyzed. A combination of non-parametric models including random forest (RF) and support vector machine (SVM) and linear regression models were used to investigate the relationship between tree diversity and biotic and abiotic factors. Biotic and abiotic variables included the number of trees per hectare, diameter, respectively. Basal area (BA), Basal Area in Largest tree (BAL), slope, aspect and elevation. Evaluation statistics including the coefficient of determination, RMSE and percentage RMSE error showed that the random forest model was the best model to determine the relationship between biodiversity and environmental factors and has suitable accuracy for determining biodiversity changes in the northern forests of Iran.