پهنه‌بندی حساسیت وقوع زمین‌لغزش با استفاده از مدل هیبریدی قضیه بیز –ANP (مطالعه موردی: کرانه جنوبی حوضه آبریز اهر چای از روستای نصیرآباد تا سد ستارخان)

نویسندگان

1 دانشگاه محقق اردبیلی

2 دانشگاه تبریز

چکیده

  نقشه‌های حساسیت وقوع زمین‌لغزش یکی از مهم‌ترین ابزارهای لازم برای برنامه­ریزان و تصمیم گیران محیطی به‌ویژه در مناطق کوهستانی است. در این پژوهش، پهنه­بندی پتانسیل وقوع زمین‌لغزش در کرانه جنوبی حوضه آبریز اهرچای از روستای نصیرآباد تا سد ستارخان با استفاده از روش تحلیل شبکه‌ای بر پایه زمین‌لغزش‌های رخ‌داده درگذشته که به‌وسیله روش شی گرا استخراج و شناسایی گردیده است، و تئوری احتمالات شرطی (قضیه بیز)، موردمطالعه قرارگرفته است. تئوری بیز ، به‌عنوان مدلی سودمند و آزمایش‌شده در زمینه­های مختلف، همچون مطالعه حرکات توده­ای، تحقیقات معدنی و نقشه­کشی چشمه­های­ آب زیرزمینی می‌باشد.  جهت وزن دهی به لایه‌ها در این روش‌ها،  از10 پارامتر شیب زمین (به درجه)، جهت شیب، کاربری زمین، سنگ‌شناسی، بارش، شاخص تراکم پوشش گیاهی(NDVI)، شاخص طول شیب(LS)، شاخص رطوبت توپوگرافیک (TWI)، شاخص قدرت آبراهه (SPI)، فاصله از گسل و فاصله از آبراهه استفاده شده است. تعداد 70 درصد زمین­لغزش­ ها (25 عدد) برای اجرای مدل و 30 درصد (10 عدد) دیگر برای اعتبارسنجی به‌کاررفته است. نتایج حاصله از روش مذکور و اعتبار سنجی میدانی مؤید دقت شناسایی 71.11٪ و دقت کلاس‌بندی 91.4٪ می­باشد. با توجه به نتایج به‌دست‌آمده 02/34 درصد از اراضی محدوده موردمطالعه از پتانسیل بسیار بالایی برای وقوع زمین‌لغزش برخوردار می­باشند. همچنین، درصد قابل‌توجهی از زمین‌لغزش­ها در طبقه با حساسیت زیاد (05/57) قرار دارند. لذا می­توان گفت دقت مدل­های بکار رفته در پهنه­بندی حساسیت وقوع زمین­لغزش قابل‌قبول و خوب است.

کلیدواژه‌ها


عنوان مقاله [English]

Landslides susceptibility mapping using hybrid model of Bayes' theorem & ANP, Case Study: Ahar drainage basin South boundary (From Nasirabad to Sattar Khan dam)

چکیده [English]

Introduction
Mass movements and landslides are considered the most destructive natural hazards. The study predicts landslide location using conditional probability theory (Bayesian theorem), on the southern edge of the Ahar drainage basin (From Nasirabad to Sattar Khan dam),, on the landslide occurred in the past by the object-oriented approach extracting and identifying been conducted. Using Bayesian probability theory correlation between parameters and landslide areas (two-thirds of landslide areas) to determine the weight of all categories of parameters were obtained. According to the map obtained by any of the parameters weight class, class of high silt old alluvial terraces are in layers, average pasture land in between classes, directions north and northwest, steep grade 5-25 the distance of 270-125 meters from the river greatest impact on weight and landslides in the area. The accuracy of landslide susceptibility maps using a third (12 points slip) landslide areas were evaluated. The result showed that the model with the predictability and kappa coefficient 0/93   and 0/945 percent of high and very high risk of landslides in class acceptable accuracy in evaluating and landslide susceptibility mapping.
Materials & Methods
In this study, object-oriented method to extract landslides, the IRS P 6 bands of satellite images were proper. Multiresolution segmentation algorithm was then applied for the segment. Among the different scales, scale 35 was chosen as the appropriate scale and form factor and compression factor each respectively 0.2 and 0/5  were selected and specific conditions of each class at the end of the definition, classification was carried out in the last stage to win to assess the landslide susceptibility in the region of evidence weight method is used. This model is useful as a model and tested in different fields, such as the study of mass movements, mineral research and mapping underground springs.
ANP Method:
Network analysis process is one of the MCDM techniques and in behalf of the compensation models. This model of hierarchical control, clusters, elements, and elements is an interaction between clusters. In general, the implementation of network analysis process model can be described in four stages as follows. 1. Build the model and create a lattice structure 2 binary comparison and determine the priority vectors that are actually similar to those in the analytic hierarchy process in the network analysis process repeated measures conducted in the screw. 3. The formation of primary and uneven super matrix. 4. The weighted matrix (5) formed Super eventually formed the super matrix.
weight of evidence(WofE):
weight of evidence(WofE) is a statistical method based on probability theory. This model represents relationships between an event landslides occurred and causative factors predisposing factors landslide estimates. In 1988, the model was used for mineral exploration. It is used to assess the landslide susceptibility. If causal factors (predisposing factors landslide) to B, class B i and landslides occurred in each parameter to consider S, then the Bayesian theory to calculate the conditional probability of landslides (S) in certain class (B i), can be used the following equation:

    In this study, using 10 natural and human parameters (lithology, distance from fault, distance from the river, rainfall, land slope, slope, land use, density of vegetation(NDVI) and sediment transport index(STI) stream power index (SPI) and topographic wetness index (TWI) zonation action has been taken.
Discussion of Results
According to maps and information provided by a layer formation, and are old alluvial terraces conglomerate, the greatest impact and the weight to be allocated. Tuff, silt-ignimbrite alternating with layers of sandstone and conglomerate is the next category. Pyroxene and andesite rock formations as well as the other layers of the lowest weight and effect in landslides in the region. Among the different classes of land, pasture medium with the most weight had the greatest impact on landslide occurred that can focus in relation to human activities and population centers in these areas. Among the different tilt directions, directions north and northwest and the southern and south-eastern most minimal impact on land sliding. Also, the slope of 15-5 and 25-15 have the highest impact and landslides have weight. Class 270- 125 meters away from the river layer also have the greatest impact on landslides.
Conclusion
In this study, potential zoning landslides type South Bank Ahar drainage basin to the dam of the village of Nasirabad Khan, was evaluated. To do this, set the parameters of natural and man and landslides type occurred in the area (70 percent landslide) and ANP and Bayes' theorem is used to combine the two models. According mapped the western and southern parts of the North East region in terms of range and landslides has the greatest potential landslides. According to Table 3, 29/17% of the land as a regional study area with high potential for landslides have been considered. To assess the final map of 30% was used for regional landslide to the accuracy of the model used for zoning. Based on the results obtained from the assessment and placement of a significant percentage of the landslide in the floor with high sensitivity and very much (57/05) the accuracy of the model used in the occurrence of landslide susceptibility zoning acceptable and good.

کلیدواژه‌ها [English]

  • Landslide
  • Bayes' theorem
  • Network analysis process
  • Aharchay
  • Zoning