Landslide hazard mapping in Bukan - Sardasht road using the weight of evidence and evidential belief function models

Introduction
Communication ways can be one of the most basic elements in the progress of civilizations and provide the grounds for economic growth and development of different regions. Natural disasters, including landslides, floods, earthquakes, storms, soil erosion, and tsunamis, cause great damage to property and human life, among which landslides are recognized as one of the most important natural disasters worldwide. Landslides are one of the most important environmental hazards in mountainous areas, they have adverse consequences such as the destruction of human lives, economic problems, death, and harm to the lives of humans and other living beings, disrupting infrastructure and Vital arteries causing the destruction of the tourism industry and historical and cultural monuments.
Landslide occurrences in Iran are one of the most important natural disasters, and they suddenly disturb the morphology of the earth's surface in mountainous areas and lead to a lot of financial and human losses. According to preliminary estimates, 500 billion rials of financial damage is caused to the country by landslides in Iran every year. Natural factors such as rainfall, geology, soil type, river flooding, and earthquakes due to the activation of faults along with other human factors such as road construction and road trenches for the purpose of road development and widening cause roadside landslides in mountainous areas every year. Therefore, for the purpose of zoning the landslides occurring on the roads, it is very necessary to use GIS spatial prediction models. According to the above-mentioned materials, the current research was conducted in order to zone the risk of landslides in the Bukan-Sardasht mountain road using the Evidential Belief Function and Weight of Evidence models.
Materials and Methods
After the investigations carried out in the studied area, twelve factors including lithological layers, distance from the fault, degree of slope, direction of slope, vegetation differentiation index, distance from the waterway, normal curve shape of the range, vertical curvature of the range, horizontal curvature of the range, Rainfall, distance from the road and height above sea level were found to be effective factors in the occurrence of landslides in the study area. During field investigations and control of satellite images, 109 cases of landslides were observed on the surface of the Bukan-Sardasht Road, and the reason for these instabilities was excavation and loss of the foot of the slope as a result of road construction activities, as well as erosion of waterways and erosion of river banks. By identifying the landslide areas, the first step in preparing the landslide susceptibility map was done.
Result and discussion
The communication axis of Bukan-Sardasht is a mountain road and has many landslides. According to the visits made in the study area, 109 small and large landslide points were recorded in the studied basin. From a total of 109 observed slip points. There are factors related to topography, geology, climate, permeability, slope, waterways network, and human factors such as road network and land use changes that cause landslides. Therefore, twelve influential factors including geological factors, slope degree, slope direction, normal slope curvature, transverse curvature, longitudinal curvature, height above sea level, NDVI, distance from faults, distance from road, distance from waterway and rainfall using two the weighted evidence model (WOE) and the definite evidence function model (EBF) were used to determine the landslide risk potential.
They had made comparisons between statistical and probabilistic methods and concluded that WOE and EBF methods had a higher area under the ROC curve and showed that these methods have high accuracy in preparing landslide susceptibility maps, which the present study also expresses their opinion. Yalchin et al. (2011) [1] also compared the methods of frequency ratio, weighted evidence, and logistic regression, and they obtained the area value under the ROC curve for each of the mentioned methods as 0.890, 0.903, and 0.840 respectively. brought that it indicated a very good estimate of the weighted evidence method compared to other methods. According to the area under the curve obtained from the definite evidence function method in the present study, the results of this research are consistent with the studies of Shahabi et al., (2023) [2] and Tawakolifar et al., (2023) [3]. The above researchers had proposed various statistical methods such as weighting of evidence, frequency ratio logistic regression, and artificial neural network as suitable tools in preparing the landslide susceptibility map, and the results obtained from this research also express the opinion. it is them.
Conclusion
In this study, evaluation and comparison between two data mining methods of weighted evidence and conclusive evidence function were done. The validation of the results with the Rock Index (ROC) indicated that the methods used in this study had a very good ability to predict the areas with landslide susceptibility in the area around Bukan-Sardasht road, and in the meantime, the method is subject to definitive evidence with the level Under the ROC curve equal to 0.910 with a standard error of 0.044, compared to the weighted evidence method with the area under the ROC curve equal to 0.893, had better predictions. There are many influencing factors on the occurrence of landslides and the selection The most important factors in the occurrence of landslides are of great importance in this study twelve factors were used. It seems that other factors such as the intensity of rainfall, soil texture, and underground water (springs) are also factoring that trigger and cause landslides in the study area and it is suggested that they be considered in future studies.