Comparative evaluation of the use of statistical models of frequency ratio and Shannon entropy in order to zone landslide-sensitive areas (case study: Middle Tajen watershed)

The complete or partial movement of rock or soil caused by gravity is called a landslide. Landslides are considered among natural disasters and account for at least 17% of casualties caused by natural disasters, which sometimes cause the loss of life, property and infrastructure. Therefore, the first step to understand and recognize landslides and then reduce the damages caused by them is to analyze landslides using statistical methods. The frequency ratio method is a statistical analysis model and one of the most popular approaches for the analysis of landslides, which is still widely used by many researchers. Another two-variable statistical model in landslide susceptibility studies is the Shannon entropy model. Entropy is a management approach that estimates dispersion and disorder in natural phenomena and is widely used in environmental sciences. Middle Tajen watershed is one of the sub-basins of Tajen basin in Mazandaran province and due to special environmental conditions, it is prone to mass movements and landslides have occurred in different parts of it during previous periods. Therefore, it seems necessary to prepare a landslide susceptibility map with acceptable accuracy for this area. Based on this, the aim of the current research is to evaluate the comparison of two statistical models of frequency ratio and Shannon entropy to prepare a landslide susceptibility map in the middle Tajen watershed.

In order to prepare a landslide susceptibility map with Shannon's frequency ratio and entropy statistical models, first 68 landslide points were identified for the region. Also, according to the field observations of the conditions of landslides recorded in the region, as well as a review of the sources of ten factors affecting the occurrence of landslides, including slope, aspect, distance to stream, elevation, land use, lithology, Plan Curvature, Profile Curvature, rain and Topographic Wetness Index were identified and modeling was done using these factors. The frequency ratio statistical model is a simple and understandable probabilistic model that shows the correlation between landslides in the region and the factors that cause them in the region. In the statistical model, the frequency ratio of the values of one, less than one and more than one respectively indicates the average, low and high correlation between landslides and their causing factors in the region. Entropy measures abnormal behavior, variability, instability, degree of disorder and uncertainty of a system compared to its possible initial state. In general, the concept of entropy describes the abnormality between causes and consequences or the degree of abnormality; Therefore, the Shannon entropy model is used to show the thermodynamic conditions of a system. In the prepared landslide maps, the most important question that is raised is the discussion of the accuracy and correctness of the prepared model, so the process of validating the built models is important. Therefore, In the present study, Receiver Operating Characteristics (ROC) Area Under the Curve (AUC) were used to train and validate the built model.

The results of the effect of factors affecting the occurrence of landslides in Shannon's entropy model showed that the parameters of topographic humidity, land use and longitudinal curvature with the weight values of Wj equal to 0.73, 0.66 and 0.65 respectively are the most have an effect on the occurrence of landslides in the region. The results of evaluating the success rate and predicting the model according to the ROC curve and the area under the curve (AUC) for the training and validity of the two models, frequency ratio and Shannon's entropy are respectively equal to 0.76, 0.72 and 0.62, 0.59 to It was found that it indicates the relatively good performance of the abundance ratio model and the weak to poor performance of the Shannon entropy model for evaluating and preparing a landslide susceptibility map in the studied basin. Also, the results of landslide susceptibility zoning in the middle Tajen basin using frequency ratio model showed that 30.156, 17.066 and 5.206 percent of the basin area are in the range of medium, high and very high sensitivity, respectively.

Landslides are one of the natural hazards that cause many human and financial losses. The zoning of landslide-prone areas, while identifying sensitive and high-risk areas, can provide the basis for the implementation of proper management plans in the slopes in order to reduce damages. In the present study, frequency ratio and Shannon's entropy models were used to evaluate the susceptibility of landslides in the middle Tajen watershed located in Mazandaran province. The results of the landslide susceptibility zoning map in the basin showed that more than half of the basin's area (52.428%) is in the moderate to very high landslide susceptibility range. According to the results obtained from the ROC curve, the frequency ratio model according to the values of the area under the curve (AUC) with success rate (0.76) and prediction rate (0.72) respectively compared to the Shannon entropy model with success rate and prediction 0.62 and 0.59, respectively, have a relatively good performance for the zoning of the landslide susceptibility map in the study area of the middle Tajen basin. Since in the last few decades, due to the change of use and destruction of natural resources, landslides have become a serious and big danger in many parts of the world, Therefore, the preparation of the landslide susceptibility zoning map plays an important role in the management of high-risk slopes and land preparation.