Landslide risk modeling using logistics regression model (Case study: Chaharmahal and Bakhtiari province)

Landslide is defined as the movement of a mass of rock, debris, or earth down a slope. Sedimentary mass movement can be various factors that play an important role in the destruction of communication roads and residential houses, the destruction of pastures and gardens, as well as erosion and sediment transport in the basins. Various studies have been conducted on landslide risk zoning using logistic regression model. Which factors such as distance from fault, slope, slope direction, land use, rainfall, lithology, elevation and distance from river and road were reported as the most important factors affecting landslides.

In this study, Chaharmahal and Bakhtiari province as an important mountainous region in the country was studied. Reviewing past research showed that among the various factors affecting landslides, 9 factors including distance from fault, slope, slope direction, land use, precipitation, lithology, elevation classes and distance from river and road as independent variables, are the most important factors. To analyze the data, first 200 landslide areas were randomly selected, and another 200 points in the whole area without landslides were also randomly selected. After preparing the layers, the logistic regression model was performed to investigate the role and weight of each of the 9 independent variables. The accuracy of the model results was checked using three statistics, R2 Naglerk, R2 Cox and Snell and factor-2LL.

The results of running the logistic regression model, showed that among the studied variables, the texture and slope of the land, respectively, can be considered as the most effective factors in creating landslides in the region. After that, the distance from roads and finally the amount of rainfall has the greatest impact on landslides in the studied region. The evaluation results of the model obtained from these five parameters, showed that the overall accuracy of the model is 90.9% and acceptable. After that, the distance from the roads is an effective factor in creating this phenomenon, and in the end, the amount of rainfall is effective on the landslides. The results of evaluating the correctness of the logistic regression model show that the factor -2LL in the last iteration of the model is equal to 117.893 and the coefficients of R2 Naglerk, R2 Cox and Snell are 0.65 and 0.42, respectively, which indicate the accuracy of the model. Based on the obtained results, a landslide risk zoning map was prepared for the study area.

Due to the high number of landslides recorded in Chaharmahal and Bakhtiari province and the risk of new landslides, this study was conducted on the area to determine the factors that have the greatest impact on this phenomenon and areas with a high risk of landslides. Based on the results, it can be stated that in addition to natural factors (land type, land slope and rainfall), some human factors, including unprincipled road construction, have an important role in landslides. These factors have caused 13.8% of Chaharmahal and Bakhtiari province is at high risk of landslides. To reduce the risks, ecosystem change and land use of the regions should be avoided as much as possible.