A novel ensemble approach of bivariate statistical-based logistic model tree classifier for landslide susceptibility assessment

Abstract

This study addresses landslide susceptibility mapping (LSM) using a novel ensemble approach of using a bivariate statistical method (weights of evidence [WoE] and evidential belief function [EBF])-based logistic model tree (LMT) classifier. The performance and prediction capability of the ensemble models were assessed using the area under the ROC curve (AUROC), standard error, 95% confidence intervals and significance level P. Model performance analyses indicated that the AUROC values of the WoE–LMT ensemble model using the training and validation data-sets were 86.02 and 85.9%, respectively, whereas those of the EBF–LMT ensemble model were 88.2 and 87.8%, respectively. On the other hand, the AUC curves for the four landslide susceptibility maps indicated that the AUC values of the ensemble models of WoE–LMT (85.11 and 83.98%) and EBF–LMT (86.21 and 85.23%) could improve the performance and prediction accuracy of single WoE (84.23 and 82.46%) and EBF (85.39 and 81.33%) models for the training and validation data-sets.     

Prediction of the soil erosion in a forest and sediment yield from road network through GIS and SEDMODL

This study was conducted to classify water erosion risk for a deciduous forest and to predict the amount of sediment yield from forest road network. GIS in combination with AHP was used for determining the soil erosion risk degrees of forest. Beside, sediment yield from forest roads with gravel and asphalted surfacing was estimated using SEDMODL. Rainfall simulator was used in calibration and validation process of model. Results showed that 47.9% of forest soil is classified from moderate to very high vulnerability. 32.3% of roads were located in soil erosion risk class of very high. This class is generally found in the eastern region of forest, while areas with very low risk are found in the south western part. 14.6% of the forest roads were ranked as having very low sediment yield. Only 61 segments out of 339 segments of forest road network delivered sediment to the ravine network. The estimated annual sediment yield for all of road sections by SEDMODL and rainfall simulator were 10,935.45 and 10,509.29 g m-2, respectively. Results of the calibration and validation process showed that the variation accounted for in the predicted values by SEDMODL with the observed values under rainfall simulation was 3.90%. Best management practices （BMP） must be considered for the areas with high degrees of erosion risk.     

Deforestation effects on soil properties,runoff and erosion in northern Iran

Improper cultivation practices are seriously degrading native forest ecosystems in northern Iran. Hence, the objectives of this study are to compare selected soil properties, runoff amount, erosion and also introducing equations to predict the runoff and soil erosion in three types of land use (forest, garden and cultivated). A simple portable rainfall simulator has been set in 90 random points to create experimental rainfall. Result showed that changes in natural forest led to a significant clay, organic carbon of soil, total N and antecedent soil moisture decrease and sand, pH and bulk density increasing. The rainfall runoff experiments indicate that runoff content of the natural forest soils was 35 % and respectively 38.45 % higher than the garden and cultivated land soils .This result could be related to the higher antecedent soil moisture in natural forest compared with the other land uses. According to the obtained results, garden soil erosion and cultivated land was 1.351 and respectively 1.587 times higher than the forest. The correlation matrix revealed that runoff content was positively correlated with antecedent soil moisture, bulk density and silt, and negatively with soil organic carbon, total N and sand. Also, soil organic carbon, total N, clay and sand showed negative correlation with soil erosion, while there is a positive correlation between erosion and silt, bulk density, pH and antecedent soil moisture. The results of multiple linear regression showed that runoff in forest, garden and cultivated land can be predicted with correlation coefficient of 0.637, 0.547 and 0.624, respectively. The correlation coefficients of 0.798, 0.628 and 0.560 in equations indicate their moderate potential in simulating soil erosion.