عنوان مقاله [English]
Landslides are one of the most important geomorphological processes that affect the evolution of landscapes in mountainous areas, which in some cases also lead to catastrophic events (Hattanji and Moriwaki, 2009). In recent decades, due to the increasing trend of damages caused by natural disasters, especially landslides, the category of forecasting and preventing damages has been seriously discussed (Izadi and Entezari, 2013). Landslide risk zoning maps can provide an effective and efficient tool for planners and decision-makers to identify suitable zones for future development (Ghobadi Alamdari, 2019). The use of GIS and multi-criteria decision-making methods, with an integrated approach, can accelerate the planning process in the diagnosis of critical and emergency cases and lead to the issuance of appropriate results.
Givi Chay watershed with an area of 1814 square kilometers is one of the sub-basins of Sefidroud, which is in the coordinates of 48 degrees and 4 minutes and 20 seconds to 48 degrees and 40 minutes and 12 seconds east longitude and 37 degrees and 26 minutes and 3 seconds. Up to 37 degrees and 55 minutes and 55 seconds north latitude, it is located in the south of Ardabil province and within the city of Khalkhal and Givi. This watershed is from the north to Qarasu watershed and the heights of Turka, Pileh, Chaleh Marz and, Gondab, from the west to Qarnaqo river catchment and from the east to Agh-e-Uler, Navroud and, Talesh mountain ranges and from the south to the basin The catchment area of the Ghezel Ozan River is limited. The minimum and maximum height of the Givi Chay watershed, are respectively; 873 and 3025 meters. Geologically, the study area is located in the West Alborz-Azerbaijan tectonic zone.
The present research is of applied type and its research method is analytical based on the combination of data analysis, GIS, and the use of multi-criteria analysis techniques. ENVI, Arc GIS, Idrisi, and Excel software have been used for image processing and data analysis. In this study, first the effective factors in Landslide (including slope, aspect, Elevation classes, lithology, land use, precipitation, distance from the communication road, distance from the waterway and distance from the fault), according to the natural and human conditions of the region were identified. In the next step, information layers related to each factor were prepared in the GIS environment. The information layers of curved lines, communication roads, and waterway networks were obtained by digitization from the topographic map of Givi and Khalkhal cities with a scale of 1: 50,000, and the slope and aspect layers were prepared using a digital elevation model. Information layers related to lithology (rock resistance) and faults, by digitization from the geological map of Givi, Khalkhal-Rezvanshahr, and Masouleh; Prepared at a scale of 1: 100,000. To extract the land use of the study area, first geometric and atmospheric corrections were made on the images using the Flaash method in Envi software. Then the classification was done by object-oriented method and nearest neighbor algorithm in Ecognition software, and the results obtained from the classification of uses in the present study, both in terms of individual uses and in terms of total accuracy and kappa statistics, are acceptable (greater than 85 Percent), in relation to the information produce. The precipitation map of the basin was drawn using meteorological and rainfall station data and the method of precipitation gradient equation (P: 0.11104H + 193.24). To prepare a landslide risk map, WLC, OWA, VIKOR, and MABAC multi-criteria decision algorithms, fuzzy standardization and cortical weighting method have been used. Landslide zoning maps have been evaluated using the relative performance detection curve (ROC).
Results and discussion
According to the output obtained by using the WLC method, 427.352 square kilometers of basin area is in the high-risk class and 599.237 square kilometers is in the high-risk class. According to the landslide hazard zoning map obtained from the OWA method, respectively; 284.262 and 670.628 square kilometers of the basin area are very high-risk and high-risk classes. According to the hazard map obtained from the VIKOR method, high-risk and high-risk classes, respectively, occupied 745.457 and 394.471 square kilometers of the basin area. Also, the results obtained using the MABAC method show coverage of 572.900 square kilometers of high-risk floor and 551.030 square kilometers of the high-risk floor of the basin area.
Results of output overlap of the studied models, with a scattering of landslide points; Showed that according to WLC, OWA, VIKOR, and MABAC multi-criteria decision algorithms, respectively, 37.84, 46.73, 59.46, and 48.65% of the slip points in the high-risk category and 37.84, 51.35, 24.33 and 35.14% of slip points are in a high-risk category. The matching of slippery surfaces and hazardous zones shows that at the output of all the studied algorithms, the areas in the high-risk, high-risk category have the largest area of landslide surfaces. In addition, in the low-risk classes introduced by the multi-criteria algorithms, a limited number of landslides are observed, and in the low-risk classes, no distribution of landslides occurs in the basin. Therefore, it can be concluded that due to the distribution of landslides in each of the hazard classes, all the studied algorithms and especially the Vikor method by covering 22 landslides in a very high-risk class, of high relative accuracy in They have a landslide risk assessment.
According to the results, respectively; Slope factors with a weight of 0.16, lithology with a weight of 0.15, and land use with a weight of 0.13 had the most role in the occurrence of landslides in the basin, according to the output of the studied algorithms, area Low and very low-risk areas have the lowest area in the basin. On the other hand, medium, high, and very high-risk zones have the largest area of the basin. It can be said that the results of this study indicate the high power of the Givi-chay basin in terms of the occurrence of landslides. Due to the distribution of landslides in each of the hazard classes, all studied algorithms have a high relative accuracy in landslide risk assessment. According to the validation results, the area under the ROC curve for OWA, VIKOR, and MABAC methods was calculated to be 0.72, 0.73, 0.85 and 0.76, respectively. Therefore, the accuracy of OWA, WLC and MABAC methods is very good and the accuracy of the VIKOR method is excellent.