Landslides susceptibility zonation with bivariate statistical and fuzzy logic models (Case Study: Balekhloo catchment)

Introduction Landslide susceptibility (LS) is the likelihood of a landslide occurring in an area on the basis of local terrain conditions (Brabb, 1984). In recent decades number of different types of hazards is greatly increased due to human activities and its encroachment on the natural environment. Compared with other natural hazards such as volcanic eruptions and floods, landslides cause considerable damage to human beings and massive economic losses (Guzzetti, 2005). According to preliminary estimates, about 500 billion riyals annual are caused economic damage in Iran by landslide occurrence (Hosseinzadeh et al., 1388:27). Today Because of the rapid development of computing power and Geographical Information System (GIS) technology, a vast number of quantitative or statistical Methods have been applied to assess landslide susceptibility (Wang, 2013:81). Among these models can be pointed to such as logistic regression model, Analytical Hierarchy Process (AHP), Analytic Network process (ANP), artificial neural network (ANN), the bivariate statistical models, fuzzy logic model. LNRF model and etc. Selecting the most appropriate approach and model is done based on the data type, scale and scale of of the analysis. The work done at the country and abroad can be pointed following cases: Gharahi et al (1390) landslide susceptibility Alborz Dam zonation by using bivariate statistical and AHP model, Results showed that the compared to the weighting factor method statistical indicators zonation provides more realistic distribution of landslide susceptibility. Bidar (1391) has attempted to zonation mass movements in road Meshkinshahr - Movil the advantage of Analytic Hierarchy Process (AHP). He combined with 9 parameters catchment to zonation into four sections with a high risk, high, medium and low risk. Foreign researchers such as Yaklyn (2008) using the Analytic Hierarchy Process (AHP) based on GIS to landslide hazard zonation is investigated according to the characteristics geological formation in the Turkey, And has concluded that in the study area (Ardesen) 98% of the landslides occurred in units with geologic formations susceptible to weathering, steep and bare land. Sabuya et al (2006) the fuzzy logic model used for the assessment of slope instability in Rio de Janeiro of Brazil And found that, in this model the expert can be weighting from zero to one Classes of factors, so the results are better than other models. In this study balekhloo catchment (Ardabil) has Zonation for landslide susceptibility By using fuzzy logic and bivariate statistical Methods And landslide predisposing factors such as (Lithology, distance from fault, distance from river, drainage density, land slope, aspect and land use, distance from roads, vegetation density (NDVI) and elevation). 2- Methodology Fuzzy logic model fuzzy logic model is Generalization of the classical set theory in mathematical science and is a new approach to the expression of uncertainty and confusion routine. Fuzzy sets are defined by membership functions. For each fuzzy set is a number between zero and one Where zero indicates the lack full membership and one indicates full membership (Hosseini et al, 1390). Fuzzy model is done using several functions. The most important operators of the fuzzy logic can be pointed to Fuzzy Product, Fuzzy Sum gamma operator and etc. Gamma operator is defined based on Fuzzy Product and Fuzzy Sum (Equation 1). Equation (1 μ_(combination=((Fuzzy Algebraic Sum)(Fuzzy Algebraic Product))^(1-γ)) Where μ_combination layers obtained from the fuzzy gamma and the γ parameter is defined in the zero and one. Considering studies conducted and their results and compare different values of gamma, in this study, the Gamma 8/0 was used and landslide susceptibility map was produced. Bivariate statistical models Bivariate statistical models have based on overlapping parameters and density of landslides occurred. In these models the relative importance of Classes of each parameter calculated using the Landslide density in its and using the formula. In this method Landslide and parameters are handled as a dependent variable and independent variables respectively. There are several statistical methods to calculate the weighted values that here used the information value and Area density models. 1- Information value method Overall, the combination of quantitative variables (such as slope) and qualitative variables (such as land use) does cause problems in statistical analysis. This problem can be resolved in considering that each class variable individually can to have one of two modes present (1) or absent (0). This method can determine whether the class variable is present or absent (niyazei et al 1389: 12). Mentioned methods runs using the equation 2 (Naderi and Karimi, 1390: 98). Equation (2) I_i=LN((S_i/N_i)/(S/N)) 〖 S〗_i Landslide area occurred in class variable, 〖 N〗_i class variables area, 〖 S〗_. The total area of landslide, 〖 N〗_. The total area of the study area and LN is Logarithm. 2- Area density method In the methods of Area density weight of each class variables is calculated by using the equation (3) (Afjeh Nasrabadi and et al, 1387). Equation (3) Wa=1000(A/B)-1000(C/D) In above equation A is the area landslide happened in each class variables, B area of each class, C the total area of the landslide And D is the total area of the study area. With obtained weight of each class variables, weights of classes applied in software and landslide susceptibility maps were produced for both statistical methods. 3– Discussion According to the method mentioned above layers intended after preparation Were used for zonation And landslide susceptibility zonation map was obtained using mentioned three methods. Maps obtained using the method of natural breaks was classified In 5 risk class (very low susceptibility, low susceptibility, moderate susceptibility, high susceptibility and high susceptibility). For the compare models, were used from Landslide Index (Equation 4) and using it were evaluated entire risk Category. Equation (2) Li=((Si/Ai))/((∑n(Si/Ai)))*100 In above equation Si is area landslide occurred In each zone risk, Ai area of each risk zone, N number of risk Category and Li index landslide occurrence in each risk zone is to Percentage. 4– Conclusion Models used in the study Acceptable results are provided considering was to evaluate using landslide index. Information value models with values of 10 and 87percent in the high and very high risk Classes Have a better Evaluation than two other models. Area density model with values of 9% and 87% and fuzzy logic models with values of 24% and 60% in the high and very high risk Classes Are placed respectively the next ranks. Evaluation results indicate that in Information value models about 98 percent of landslide occurred in high and very high risk classes. This value has been estimated 97 percent for Area density model and 84 percent for fuzzy logic model. The results from each of the three models indicate a high potential for landslides occurred in western and southwestern parts of the basin. This areas are formed more of sedimentary rocks, old terraces and various combinations of clay, marl, limestone and Lahar. This is indicative of great influence of lithology and Type of formation to create landslide.