Investigation of Effective Factors and Landslide Hazard Zoning Using Density Area, Analytical Hierarchy Process and Logistic Regression Methods in the Ashvand Watershed

EXTENDED ABSTRACT      Recognition of effective factors on landslide occurrence in the area leads to an important set of solutions to control and take suitable actions upon this phenomenon. This paper aims at recognizing such factors and also landslide occurrence and zoning landslide hazard, using density area, Analytical Hierarchy Process and logistic regression, and also investigating the accuracy of the proposed models through the Ashvand Watershed at Nahavand city. To this end, the author conducted a field study and also reviewed the related research that resulted to the identification of 10 factors as effective on landslides including: slope, aspect, elevation, distance from river, distance from road, distance from fault, distance from village, geology, land use, and precipitation. In the next step, the data layers for these factors and landslide distribution have been generated. The factors were prioritized using AHP method and weight maps were generated for each identified factor. Landslide hazard zoning maps were also prepared for all the three conducted methods. Finally, the output maps from the previous step were categorized into four groups including low risk, medium risk, high risk, and very high risk. Through the assessment of the zoning maps by means of the ROC curve, the authors concluded that the logistic regression with AUC 0/891 and AHP with AUC 0/844 are the optimal models for the aim of this study.   1- INTRODUCTION      If only one factor was causing a phenomenon in nature, it would be very simple to decide on it and to predict its occurrence, but in general, the phenomena in nature have many qualitative and quantitative factors. One of the most dangerous type of erosion phenomena can be mass and land slide movements that have a lot of financial dangers. According to the definition of the International Geological Engineering Association in 1990, the movement of land constituents, from slope downwards is called landslide or sloping instability. In Iran, landslide is one of the most important natural disasters that has a significant role in destroying communication roads, pastures, Gardens and residential areas, as well as erosion and transfer of high volumes of sediment in our country's catchment areas. Therefore, in order to manage and mitigate the risk, it is necessary to predict the occurrence and preparation of a landslide hazard zonation map, and this has led to the development of numerous empirical and statistical models emphasizing the use of the Geographic Information System (GIS). In the Ashwand Basin in Nahavand, due to the sensitivity of the area, the greatest amount of erosion is due to landslides. Therefore, the assessment and management of the region is necessary in order to compensate for any damage and injuries incurred. 2- Methodology      The watershed of Aswand is located in Hamadan province and northeast of Nahavand city with an area of ​​about 96/47 square kilometers and an environment of about 24/24 km. After reviewing the field and reviewing similar studies, 10 factors including slope, tilt direction, elevation, distance from the waterway, distance from the road, distance from the fault, distance from the village, geology, land use and precipitation were extracted as factors affecting the occurrence of the known landslide. Then, the ten layers of the data and the distribution of the landfall event were prepared. Using AHP, 10 factors investigated were prioritized. After drawing up weight maps for each factor, zoning maps were prepared in all three models, in which the maps were classified into four categories including low risk, medium risk, high risk and very high risk.   3- Results      In this study, we tried to determine the factors affecting landslide according to past studies as well as their existence. Selection of these factors plays an important role in the accuracy of the landslide zonation map in the region. The first step in zoning landslide is the recognition and selection of effective factors in its occurrence, which indicates the accuracy of the zoning map. The first step in the zoning of land scaling is the identification and selection of effective factors in its occurrence, which indicates the accuracy of the zoning map. According to the review of available resources and available data, 10 geological factors including land use, elevation, slope, slope direction, distance from the road, distance from the waterway, distance from the village, distance from fault and precipitation for zoning the land scarcity risk in the region were identified and classified. The purpose of this study was to determine the landslide hazard zonation using the three methods of AHP, surface density and logistic regression using operational maps. At first, the maps of the agent were prepared and, according to each method, a landslide hazard zonation map was prepared. Then, using three levels of surface density, AHP and logistic regression, a zoning map, and in the next step, other zoning maps were prepared. The results were categorized into four classes of low risk, moderate, hazardous and very high risk. The results of the evaluation of the models based on the ROC are presented in Figures 17-19. AUC below represents the predictive value of the system by describing its ability to accurately estimate events occurring (landslide occurrence) and its failure to occur. AUC values vary from 0.5 to 1. In these curves, as stated above, the most ideal model has the highest level below the curve, which is a logistic with a curve surface of 0.891 and AHP with a surface below the curve of 0.484 with a slight difference in the surface area under the curve. Ideal models are as much as 766/0 compared to the level-level congestion model. 4- Discussion & Conclusions In general, identification of the most important factors affecting the occurrence of landslide with the use of the AHP model is presented in Figure 3. Accordingly, the important factors in landslide occurrence are respectively: distance from fault, distance from the road, slope, distance from the village, geology, land use, distance from the waterway, altitude, precipitation and direction of slope. The density of landslides in each class of factors was investigated and it was determined that in each factor which class has the most slip. In considering the distance from the road, the maximum slip is less than 1000 m from the fault, because the fault is one of the most important tectonic factors that can potentially sensitize the slopes. In the study of the effect of the distance factor on the road landslide, it was concluded that in the distance of 1000-400 meters from the road, the highest slip is observed and the reason why the slipping distance is less than 400 meters is that the protective structures in this distance disturb the natural state of the area and the slope of the region, and create vertical cuts. In the study of slope factor, the maximum slip occurred in slope categories of 45-30 degrees and then 15-15 degrees. In the downward slopes, due to the rocky extent of the slopes and the small thickness of the detached materials, the occurrence of low-lying slopes is usually less and sloping slopes, due to lower shear stress, are usually less susceptible to ground occurrence. The distance from the village showed that the maximum slip occurred up to 1000 meters. The presence of the village in the area causes the disturbance of the gradient equilibrium, the creation of vertical cuts, the change of use around the village, and the compaction of the soil and its degradation. In the geological factor, the greatest slip occurred in the AN class (anidrite and gabbro), due to its high sensitivity to erosion and fragmentation of the existing minerals and formation of surface-shaped structures, which are sensitive to slip, water penetration and mass movements. The created land use in the area shows that a high percentage of landslides has occurred in the rangelands with the number of landslides in semi-dense, low-density rangeland and dense pasture, respectively. Investigating the distance factor from the waterway network showed that the slides occurred in the studied area at a distance of 200 meters from this complication. The reason for this is that the drainage is due to the river's dredging and erosion, which carries the material to the slopes and maintains the retaining factor from the slopes, and the rubbing of the wall along the river causes a collapse of the slope and, consequently, the instability of the slope overlooking the river. From the study of elevation classes, it was also found that the classes with an average height had the highest sensitivity to the slippage, due to the low rainfall in the lower floors and the phenomenon of glaciation and slowness of the clay process in the upper classes. In the results of precipitation analysis in the region, it was concluded that the greatest effect on slippage is, respectively, on the average rainfall levels (455-445 mm). The increase is expected to increase by more than 465 mm for reasons other than excavation, but the result shows that there is less slip in the classes with a rainfall of more than 465 mm, which can be due to other effective factors such as Gradient and geological factor or elevation of the area. In the southern slopes that are considered as slopes, they have the highest slip in comparison with other directions in the region, which can be due to weathering. In these slopes, the high humidity of the area, along with heat from the sunlight, provides suitable conditions for chemical weathering. The intensity of weathering is inversely related to the amount of adhesion of materials, which provides a ground for landslide occurrence. Using three models of surface density and AHP and logistic regression, zoning maps were prepared and the area was classified into 4 categories with low risk, medium risk, high risk and very high risk (Figure 4). Also, by evaluating research done using ROC and accuracy, it is shown in Figure 5 that the logistic regression model and then the hierarchical analysis process model have been used as important and efficient models in the zoning of ground-sensitive sensitivity. The study area is due to a combination of natural and human factors that have a great potential for earthquake occurrence. These landslides annually cause many damage to the road, residential areas, agricultural lands and other resources in the region. Fault factors, geology and network of inland waterways are irrevocable and the only way to prevent their damage is to not disturb these areas and avoid them. Road, village and land use planning and management factors can be prevented by stimulating and increasing movements in these areas by constructing a road based on environmental conditions and preventing the construction of unpredicted roads and proper use in these areas. . It is suggested that other models be used in this area in order to map the zoning and compare different models in order to determine the best model for zoning of landslide hazard and also to consider more effective factors in the region and economic losses caused by landslides and the costs of slip control are evaluated and calculated.