Landslide Susceptibility Zonation applying frequency ratio models and statistical index in in Oghan watershed

This research was conducted to prepare landslide susceptibility zonation (LSZ) map for the Oghan basin using Frequency Ratio(FR) and Statistical Index (SI) models. For this purpose, the most important factors affecting land sliding including slope, aspect, plane curvature, profile curvature, elevation, TWI, precipitation, land use, lithology, the distance from fault, the distance from drainage, the distance to road were determined. Then, the landslide inventory map was prepared by using field digital checks, GPS and Google earth. ROC curve and Area under the curve (AUC) were used to validation. research findings show that statistical index model for training data and validation are attributed as 0.925 and 0.916, respectively and they are more effective than frequency ratio model to prepare landslide susceptibility map, according to which 34.91, 28.51, and 36.59 % of basin area and 5.42, 20.46, and 35.72 % of population of the province encounter very low, low, moderate, high and very high risk of landslide. Also, geological and precipitation factors were introduced as the most important factors in the occurrence of landslides in the study area. 

Landslides are natural geological processes that change and reform Earth's surface relief. Landslides occur directly due to gravity and it can cause rapid and considerable movement of scree, soil, and rocks down the slopes. Knowing the landslide mechanism and zoning the landslide susceptible areas is necessary for land use planning. Obviously, preparing landslide susceptibility map (LSM) can give insight to land use managers and decision makers. Oghan watershed is located in Northern Iran, with geographical position between of 37ᵒ 9ʹ to 37ᵒ 15ʹ northern latitude and 55ᵒ 5ʹ to 43ʹ 55ᵒ Eastern longitude and it covers an area of 40352 Hectares. An important feature of this area is its height difference which varies between 191m to 2500m from sea level. Because of heavy rain and landslide prone formations, the area is potentially prone to mass wasting. As in points with high percentage slope and areas influenced by such human factors as road and land use changes, mass wasting is aggravated.

To identify landslides in the area under study, field visits, aerial photography, and Google Earth program were used, as a result of which 143 landslides were identified and their positions were spotted in Google earth and the final spot layer of landslides was produced in the form of shapefile. Finally, to conduct the research and execute models, 70 % of landslides (100 locations) were used as model training and the other 30% (43 landslide locations) for the validation. 12 factors affecting slope instability in the region under study were considered which include: elevation, slope, Aspect, profile curvature and plan curvature, lithology, distance from fault, land use, percipitation, distance from drainage, topographic wetness index, and distance from road. To draw susceptibility map for Oghan region, two models of frequency ratio and statistical index were considered and ROC curve was used for their comparison and validation. Frequency ratio (FR) refers to the ratio of landslide surface area in each class to the surface area of the related class. In fact, frequency ratio indicates little relationship between landslide occurrence and different variables affecting it. Statistical index model is a two-variable statistical analysis where the weight for each class factor affecting landslide occurrence is obtained by natural logarithm of landslide density of that class factors affecting landslide density of the whole region.

Most landslides occurred in heights of 168 to 1600m. Many villages, with approximate population of 9656 are in Oghan area, most of which are located in heights of 1600m. Further, most landslides are found in road sides and high traffic communication networks or near canal networks. This proves that shear stresses contribute to the landslide occurrence and, therefore, are used as effective factor in susceptibility assessment. Most landslides in the study area occur in slopes over 15%, in the East, South East, South, South West, in heights of 168 to 1200m, within surface curvature and Convex profile curvature. Moreover, TWI values smaller than 1.5 and larger than 4 are significant. Lar and Ilikah formations indicate the highest susceptibility. Distances between 1000-3000m for faults, smaller than 250m for drainage, and smaller than 500m or 3500-4000m for roads should be considered.

Comparison of maps produced by the given methods indicates that in frequency ratio method, 69% of Oghan watershed area (most of the area) is located in very low and low risk zone which houses 21.6 of the whole population, and the rest of the area is located in moderate to very high risk zone. In statistical index model, 34.91% of the watershed area is located in very low and low risk zone and most part of the area is located in moderate to very high risk zone and 5.42% of the population of the watershed live in very low and low risk zone, while most of the population reside in moderate, high and very high risk zones. Validation results show that the area under ROC curve for training data and validation in statistical index model is a little bit larger, excellent; and this proves that statistical index method has higher confidence coefficient and has relatively better performance than frequency ratio method.