پهنه بندی و ارزیابی خطر زمین لغزش با استفاده از مدل های عامل اطمینان، تراکم سطح و هیبریدی قضیه بیز(مطالعه موردی: حوضه بقیع ، نیشابور)

نوع مقاله : مقاله پژوهشی

نویسندگان

حکیم سبزواری

10.22034/gmpj.2021.251394.1215

چکیده

شناسایی نواحی مستعد وقوع زمین لغزش‌ها یکی از اقدامات اولیه در مدیریت و کاهش خسارات ناشی از این پدیده‌ها محسوب می شود. در این پژوهش، حساسیت زمین لغزش با استفاده از سه مدل عامل اطمینان، تراکم سطح و هیبریدی پهنه‌بندی گردید و مناسب‌ترین مدل معرفی شد. جهت این مطالعه با از تصاویر ماهواره‌ای‌‌ETM+، نقشه‌ زمین شناسی 25000/1، نقشه توپوگرافی 50000/1، نقاط لغزشی مشخص گردید. در این پژوهش با بررسی 19 پارامتر موثر در رخداد زمین لغزش شامل لیتولوژی، کاربری اراضی، شیب، جهت شیب، ارتفاع، شاخص رطوبت توپوگرافی‌(TWI)، شاخص انحنای سطح، شاخص انحنای مقطع، فاصله از گسل، فاصله از جاده و فاصله از آبراهه مورد بررسی قرار گرفتند. پس از تهیه لایه‌های اطلاعاتی در محیط نرم افزار ArcGIS 10.4 به عوامل موثر با استفاده از نظرات کارشناسی اقدام به تهیه وزن کلاس‌ها و در نهایت تهیه نقشه‌های پهنه بندی حساست زمین لغزش با استفاده از روش‌های آماری تراکم سطح، فاکتور اطمینان و هیبریدی بیز در پنج کلاس ریسک خطر خیلی کم، کم، متوسط، زیاد و خیلی زیاد طبقه بندی گردید‌. جهت صحت سنجی روش‌ها از شاخص‌های جمع کیفی‌(QS)، دقت‌(P) و شاخص نسبت تراکمی‌(DR)‌ استفاده شده است. مقادیر شاخص‌های مجموع کیفیت و دقت که بیانگر کارایی مدل‌ها در پهنه بندی حساسیت زمین لغزش می‌باشند، به ترتیب برای مدل‌های تراکم سطح (21/0- 04/0)، فاکتور اطمینان (39/0- 09/0) و هیبریدی (88/0- 33/0)، به دست آمد. مقادیر بالای شاخص‌ها در مدل هیبریدی بیانگر کارایی بیشتر این روش نسبت به روش‌های تراکم سطح و فاکتور اطمینان در تهیه نقشه پهنه بندی می‌باشد.

کلیدواژه‌ها


عنوان مقاله [English]

Zoning and landslide risk assessment using reliability, surface density and hybrid models of Bayesian theorem (Case study: Baqi basin, Neishabour)

نویسندگان [English]

  • Mahnaz Naemi tabar
  • Mohamad Ali Zangane Asadi
  • Mokhtar karami
چکیده [English]

Introduction

Landslides are among the geomorphic hazards that occur due to the interaction of various environmental factors, especially in mountainous areas with special morphometric features (Shirani, 2018, 96). The existing spatial relationships between landslide occurrence and environmental factors are key elements in the study of landslide sensitivity (Samotra et al. 1, 2015, 308). Since it is not possible to predict the time of landslides, the identification of landslide-prone areas and the zoning of these areas based on the risk potential reveal the importance of these studies (Ragma et al, 2014).

Research Methods

Statistical method of level variable density

In this method, landslide density was calculated for each of the parameters using equations 1 and 2, and then a landslide zoning map was prepared (Lee et al. 2001, 1095).

Equation 1; Darea = (Npix (Sxi))/(Npix(Xi))

Equation 2; Warea = 1000 [𝐷𝑎𝑟𝑒𝑎 −(ΣNpix(Sxi))/(ΣNpix(Xi))]

Statistical method of two-factor confidence factor

In this study, in order to evaluate the correlation between landslides and selected factors, the weight values obtained from the reliability factor model in the form of two-variable statistical analysis have been used. The weights calculated by this model were also used to prepare and convert factor maps (classes with a negative weight of zero and classes with a positive weight of one) to enter the conditional independence test.

Equation 3; F ={ (if PPs≥PPs)/(if PPa< p Ps) ((PPa-PPs)/(PPa(1-PPs) ))/((-PPa-PPs)/(PPs(1-PPa)))}

PPa: The ratio of the number of sliding pixels in a class to the total pixels of that class, PPs is the ratio of the total sliding pixels in the area to the total map pixels. With the help of this formula, each class is evaluated as 1- and +1. If the value of the relevant class is positive, it indicates that the landslide probability is high, and if the value of the relevant class is negative, it means that the landslide reliability is low, and if the value of that class is zero, it means that there is not enough information about the variable. Therefore, there is uncertainty about the occurrence of landslides.

Bayesian theorem model or weight of evidence

The evidence weight model is based on a series of statistical formulas. In this model, modeling and forecasting work is done based on two categories of information. Effective parameters in the occurrence of a phenomenon (landslide) or causal factors (landslide predisposing factors), the level of occurrence of the phenomenon in the past (landslides occurred). In this case, the probability of landslides in the future depends on the conditions that existed when it occurred in the past (Piacanti et al. 2, 2012: 200). This model is defined according to Equation 4. In this regard, consider causal factors (landslide predisposing factors) B, classes of any relation Bi, landslides occurred S Bayes theory to calculate the conditional probability of landslide (S) in a given class (Bi ) Is defined as follows.

Equation 4; P(S/‌Bi) = (P( Bi/S)×P(S) )/(P(Bi))

Probability of occurrence of S event in the study area (AS), P (Bi) Probability of occurrence of class Bi in the study area (AS), P (Bi / S) Probability of occurrence of Bi event provided that event S occurred, P (S / Bi), the probability of occurrence of event S provided that the event Bi occurred.

The conditional probability of a landslide occurring when class Bi has not occurred is defined as Equation (5).

Equation 5; P(S /Bi^)= (P(Bi^/S)×P(S))/(P(Bi^))

Discussion and findings

The results of parameter prioritization showed that lithology, land use, slope and slope direction factors had the greatest impact on the occurrence of landslides and surface curvature factors, topographic wetness index, distance from waterway, land use, vegetation index, altitude Fault distance, road distance, NDVI and cross-sectional curvature are in the next ranks. To obtain the weight of each of the factor classes, the landslide distribution map is combined with the maps of the factors affecting the landslide event and the slip density in each of the factor classes affected by the landslide. Was obtained. The results of the subclasses of factors affecting landslides indicate that in the lithology factor, Bahram lime in the surface density, reliability and hybrid models with scores (0.516, 0.096, 21.33), respectively. , Land use, orchards in surface density model (0.311), reliability factor (0.044) and hybrid with weight (48.99), slope more than 45 degrees in surface density model (0.413), reliability factor ( 0.366) and hybrid (71,259), slope direction, southeast direction in surface density model (0.455), confidence factor (0.076) and hybrid (1.0985) have the greatest impact on the occurrence of land There has been a landslide in the area.

Conclusion

. Evaluation of accuracy (P) and total quality (Qs) showed that the hybrid model has better performance than the surface density and reliability models in identifying landslide prone areas. The value (Qs) indicates the desirability of the model performance in predicting landslide risk in the study area. The value of this index for different models is in the range of zero to one hundred. The higher this value, the better the performance of the landslide risk zoning method or model. The results of the index (Dr) showed that in all three models, they performed well in identifying high-risk classes compared to low-risk classes. Among the influential parameters of lithology factor, land use, slope and slope direction had the most impact on landslide occurrence. Unlike other models, in addition to zoning of safety zones, the hybrid model has the ability to zoning of uncertainty zones to a small extent and can be introduced as an optimal method for similar areas. In order to control vulnerable areas, it is recommended to prevent construction in the area of roads, conversion of forests into orchards, pastures and meadows, interference of human factors and terraces and construction of dams.

کلیدواژه‌ها [English]

  • Zoning
  • Landslide
  • Surface compaction method
  • Confidence factor method
  • Bayesian hybrid method