Modeling and Spatial Prediction of Landslide Risk Using Advanced Data Mining Algorithms (Case Study: Kalat County)

Landslides are one of the most important land hazards in the world that occur from falling or moving in an integrated and often rapid volume of sedimentary material along the slopes (Sharafi et al., 1399: 128). Instability of natural slopes is one of the geomorphological and geological phenomena that has an effective role in changing the shape of the earth's surface (Saraskanrood et al., 1399: 2). Landslide studies indicate that landslides are a hazard that often occurs frequently, and is present on all continents and is a global threat to humans, infrastructure and the environment (Brooke et al., 2018: 125).

In this study, first the effective factors in landslide occurrence include altitude, slope direction, slope, distance from fault, waterway density, distance from river, soil, land use, distance from road, NDVI vegetation index, geology, slope curvature, Land type, slope transverse profile, slope longitudinal profile, canal power index (SPI) and topographic moisture index (TWI) were identified. In the next step, information layers related to each factor were prepared in the GIS environment. Information layers of curve lines, communication paths, waterway network and drainage density were obtained by digitization from a topographic map with a scale of 1: 50000 and slope and slope direction layers were prepared using DEM digital elevation model with 10 m spatial resolution. Took. Geological and fault layers were prepared by digitization from a 1: 100000 geological map. A 1: 50000 land use map was used to prepare the land use information layer. To prepare a soil map, a soil map of 1: 50,000 has been used.Resul In the weighted evidence model and the definitive evidence function, the western direction (1.03) had the most impact and the northern directions (-1.22) and flat (0.03) had the least impact on the occurrence of landslides. In both models, a slope of 20 degrees (2.17) had the most impact and less than 10 degrees (0.0) had the least impact on the occurrence of landslides. The highest number of landslides in both models was observed at an altitude of 2000 meters (1.44) and the lowest amount of landslides at an altitude of less than 1000 meters (0.0). The soil type of alpha sol has the most effect (1.28) and Inseptol has the least effect (0.0). In the weighted evidence model and the definitive evidence function, medium drainage density (1.6) and (1.12) have the most effect and low drainage density (0.2) has the least effect, respectively. As can be seen in the map, the distance from the fault in both models is 0-200 meters with the highest impact (2.19) and the lowest impact at a distance of more than 1000 meters (-1.16). Regarding the parameter of distance from the road in the weighted evidence model and the definitive evidence function, distance of 0-500 meters with (1.18) and (1.33) have the most impact and distance of 1500 meters have the least impact (1-6), respectively. is. Examination of the geological map showed that in both models, Sarcheshmeh Formation had the greatest impact (2.28) and Neogene sediments had the least impact (0.16) on landslides. In the parameter of distance from waterway in both models, distance is 0-300 m with the highest (2.22) and lowest (0.11), land use has the highest impact of rangeland use with the highest (1.33) and the lowest forest lands (0.33). ), Vegetation index has the most impact (1.81), in the slope curvature of flat surfaces with the most impact (1.77) and the lowest concave surfaces (0.2), waterway strength index with the most impact (1.88) and the lowest ( 0.27), topographic moisture index with the most impact (1.66) and the least impact (0.11), mountainous land type with the most impact (1.99) and the least impact (0.15) and longitudinal profile and land profile Domain has the highest impact (1.85), (1.63) and the lowest impact (-1.1) and (-1.3), respectively.

The combination of landslides in the distance layer from the waterway also indicates that the highest distribution of landslides is in the floor with a distance of 200-400 meters. Regarding the vegetation index, it can be concluded that the lower the vegetation density, the more landslides occur. Because a large part of the study area is mountainous with high walls and deep valleys. Therefore, the reason that the sensitivity of landslides increases in the study area with increasing slope can be attributed to the high amount of rock falls and overturning in the area. In general, regarding the shape of the slope in various forms, it should be said that flat slopes have less strength than convex and concave slopes. Waterway power index, topographic moisture index, longitudinal profile index and slope transverse profile index also have a great impact on landslide occurrence. The results show higher accuracy of the definitive evidence function (EBF) model than the support vector machine and weighted evidence function models. The reason for the high accuracy of the EBF method compared to the other two methods is that this method is a combination of other methods. Also, the higher accuracy of statistical methods than decision-making methods is another result of this research. Finally, by preparing a landslide risk zoning map, it is possible to help planners and managers in order to reduce potential damages and find safer places to develop construction and road construction.