12 - Prediction of Soil Disturbance Susceptibility Maps of Forest Harvesting Using R and GIS-Based Data-Mining Techniques

Abstract

Soil disturbance susceptibility mapping is essential for forest harvesting management in the mountainous forests of northern Iran. The main objective of this study was to model and predict soil disturbance patterns by three data-mining techniques including logistic regression (LR), the classification and regression tree (CART), and the General additive model (GAM). For this purpose, soil disturbance locations were determined in the study area by field surveys. The soil disturbance conditioning attributes such as slope degree, slope aspect, altitude, slope length (LS), topographic position index, topographic wetness index (TWI), soil texture, forest type, forest density, and distance from roads and skid trails were extracted from the spatial database. The validation results (Akaike Information Criterion value and kappa coefficient) showed that GAM has a desirable ability to produce a soil disturbance susceptibility map compared to LR and CART. Therefore, according to the results of the best model (GAM), slope degree was the most important factor followed by soil texture, LS, TWI, and distance from roads and skid trails. The findings of the present study could be useful for forestry management to mitigate soil disturbances caused by forest harvesting in the Shourab Forest, Iran.