پهنه‌بندی حساسیت خطر سیل در حوضه آبخیز رودخانه کشکان با استفاده از دو مدل WOE و EBF مقالة

سیل یکی از خطرناک‌ترین و شایع‌ترین حوادث طبیعی است که در سراسر جهان اتفاق می‌افتد. در استان لرستان و در حوضه آبخیز رودخانه کشکان نیز، این بلای طبیعی یکی از پرتکرارترین حوادثی است که هرساله اتفاق می‌افتد و علاوه بر خسارات مالی فراوان جان برخی از مردم منطقه را نیز به کام مرگ می‌برد. در فروردین سال 98 سیلی که اتفاق افتاد، یکی از بی‌سابقه‌ترین سیل‌های ثبت شده در 200 سال اخیر بود؛ بنابراین، توسعه نقشه‌های حساسیت سیل و تهیه حریم رودخانه، برای شناسایی مناطق سیلاب در حوضه آبخیز برای بهبود مدیریت و تصمیم‌گیری سیل ضروری است. هدف اصلی این مطالعه ارزیابی عملکرد تابع شواهد وزنی و تابع شواهد قطعی برای تهیه نقشه حساسیت سیلاب در حوضه آبخیز رودخانه کشکان است. این مقاله با استفاده از وزن عوامل یا شواهد وزنی (WoE) و مدل شواهد قطعی (EBF) بر اساس سیستم اطلاعاتی جغرافیایی (GIS) در حوضه رودخانه کشکان نقشه حساسیت منطقه به سیلاب را بدست آورده است، که در آن موقعیت جغرافیایی 260 نقطه سیل گیر مشخص شده در منطقه به صورت تصادفی به یک گروه متشکل از (۷۰ درصدی) برای واسنجی و یک گروه (۳۰ درصدی) برای اعتبارسنجی تقسیم شدند. هر دو مدل 14 فاکتور موثر در ایجاد سیل را مورد توجه قرار داده‌اند که عبارتند از: شیب، جهت شیب، زمین‌شناسی، جنس خاک، کاربری اراضی، شاخص رطوبت توپوگرافی (TWI)، توان آبراهه (SPI)، بارش، فاصله از رودخانه، فاصله از جاده، پوشش گیاهی (NDVI)، انحنای شیب (Curvatior)، تراکم آبراهه و مدل ارتفاعی رقومی منطقه. همچنین برای اعتبار سنجی، نتایج مدل‌ها از منحنی راک استفاده شد. بر اساس مدل EBF، 32/47 درصد از سطح منطقه در کلاس خطر متوسط تا خیلی زیاد قرار دارد و بر اساس مدل WOE حدود 8/52 درصد از سطح منطقه در کلاس خطر متوسط تا خیلی زیاد قرار دارد. به‌منظور اعتبارسنجی نقشه‌های پیش‌بینی پتانسیل سیل‌خیزی، از منحنی ROC استفاده شد. از بین دو مدل WOE و EBF، بیشترین صحت به مدل‌ EBF (875/0) اختصاص داشت؛ بنابراین در زمینه پتانسیل‌یابی خطر سیل‌خیزی، مدل EBF نسبت به مدل WOE دارای عملکرد بهتری است.

1. Introduction Among natural disasters, floods are of the most devastating. The accurate assessment of its dangers has failed at various scales due to lack of information and knowledge about flood losses. During heavy rains, flow in a river increases rapidly and the water level rises above normal and covers the floodplain and its surroundings and causes floods. Catastrophic floods occur in the Kashkan River basin every year. Due to the abundance of floods in this region, flood sensitivity maps should be considered to prevent loss of life and economic damage. From 1955 to 2011, the Kashkan River was faced with 16 floods with a volume of more than 1000 cubic meters per second, which was one of the most severe floods in 2019 in this river. The April 2019 flood was one of the most unprecedented floods that’s recorded in the last 200 years; therefore, the development of flood sensitivity maps and river area preparation to identify flood areas in the watershed is essential to improve flood management and decision making. The main purpose of this study is to evaluate the performance of the weighted evidence function and the definitive evidence function to prepare a flood susceptibility map in the Kashkan River watershed. 2. Study Area Kashkan River is one of the most important and full of water branches of Karkheh river which collects the waters of a large area of Lorestan province. This river joins Seymareh river in the southwest of Poldokhtar in an area called Kol-e-Sefi and forms Karkheh river. The length of Kashkan river is about 270 km and its catchment area upstream of Kashkan-Poldakhtar station is 9400 square kilometers. The main drainage system of the Kashkan river first consists of the join together branches of the Hero and Aleshtar doab rivers in the northeast and north of the basin, respectively, from east to west and north to south. These tributaries which originate from the high and snow capped Garin and Misho paro mountains along the main route joins other sub-rivers such as Cham Zakaria, Khorram Rud, Chulhol, and Madian Rivers. The largest recorded flood in the province was the flood on April 1, 2019 in Kashkan River with a peak discharge of 6500 cubic meters per second. The amount of damage caused by it in different sectors was estimated 11 thousand billion Tomans (report of the provincial crisis headquarters 2019). 3. Materials and Methods The geographical location map of 260 flood points in the region was prepared based on the information of Lorestan Regional Water Company. This set was randomly divided into two groups including training group data (70%) and validation group data (30%). In this study several factors used including slope percentage, slope direction, height, distance from river, river density, distance from road, land use, soil type, slope curvature, lithology, vegetation, rainfall, NDVI, stream capacity index (SPI) And topographic wet index (TWI) were considered for zoning of flood risk potential. It should be noted that the classification of the layers was done according to the conditions of the region and experts’ opinion. The tools used in this study were divided into two parts: physical tools including DEM maps, geological maps, land use and waterway, and conceptual tools including ArcGIS and Excel software. Two methods, WOE and EBF, were used. The WOE method or evidence weight method is a data-driven method used to combine data sets and is based on the use of a linear form of the Bayesian probability model to estimate the relative importance of evidence using statistics. The EBF method includes several parameters Bel or degree of confidence, Dis degree of uncertainty, Unc or degree of uncertainty, and Pls degree of reasonableness in the range between zero and one. The main part of this theory is presented by the degree of certainty and the degree of reasonableness, so the degree of reasonableness is greater than or equal to the degree of certainty. Data extracted from EBF not only estimates the spatial correlation between effective factors and flood occurrence but also the spatial correlation between classes of each effective factor. 4. Results and Discussion According to the model of gravitational evidence and the function of conclusive evidence, southeast direction, slope of 0-5%, altitude of 448-900 m, soil type (Inceptisol/Vertisol), distance from the river (0-50), river density (0.43-0.69), residential use, rain ( 348-450mm), concave slope curvature, Lithology (OMq), TWI (high floor), SPI (very low), NDVI (very low) and the distance from the road (200-150 meters) have the greatest impact on creating flood potential. The maps obtained from the EBF model include degrees of certainty (Bel), uncertainty (Dis), unreliability (Unc) and probability (Pls) in the range between zero and one, and the sum of the values of certainty, uncertainty and unreliability is 1. As can be seen 47/32% of the study area is in the medium to very high risk class. Among the studied WOE and EBF models, the highest accuracy was attributed to the EBF model (0.875); Therefore, in terms of flood risk potential, the EBF model performs better than the WOE model. 5. Conclusion In this study, two models, EBF and WOE, were used to predict flood risk potential. The largest area of areas with very high potential was observed in the WOE model. The results of the study showed that flooding is the result of various environmental and human factors. Based on the presented flood risk forecast map, appropriate management measures can be taken to reduce flood damage and casualties. Areas that have a very high potential for flooding according to the final plan should be given more attention while planning.

این مقاله با استفاده از وزن عوامل یا شواهد وزنی (WoE) و مدل شواهد قطعی (EBF) بر اساس سیستم اطلاعاتی جغرافیایی (GIS) در حوضه رودخانه کشکان نقشه حساسیت منطقه به سیلاب را بدست آورده اسـت ، که در آن موقعیت جغرافیایی ٢٦٠ نقطه سیل گیر مشخص شده در منطقه به صورت تصادفی به یک گروه متشـکل از (٧٠ درصدی) برای واسنجی و یک گروه (٣٠ درصدی) برای اعتبارسنجی تقسیم شـدند. روش های مختلفی از جمله روش تحلیل ســلســله مراتبی (رضــوی ترمه و ملک ، ١٣٩٦)، روش ارزیابی چندمعیاره رویکرد WLC (رحمانی و همکاران ، ١٣٩٤)، مدل رگرسیون لجستیک (خسروی و همکاران ، ١٣٩٥)، فازی-تحلیل سلسله مراتبی (اسماعیلی و همکاران ، ١٣٩٦)، مدل تابع عامل مستند محور٣ (رضوی ترمه و ملک ، ١٣٩٦)، رویکرد خوشه بندی فازی (چن و همکاران ، ٢٠١١)٤، مدل ترکیبی شاهد وزنی ٥ (شفاپور تهرانی و همکاران ، ٢٠١٤)، ماشین بردار پشـتیبان ٦ (شفاپور تهرانی و همکاران ، ٢٠١٤)، روش ارزیابی چندمعیاره فضایی ٧ (آریان پور و جمالی، ٢٠١٥)، شاخص آماری ٨ (خسروی و همکاران ، ٢٠١٦)، فاکتور وزنی ٩ (خسروی و همکاران ، ٢٠١٦)، نسبت فراوانی (رحمتی 1 Grahn and Nyberg 2 Chapi et al 3 Evidential belief function (EBF) 4 Chen et al 5 Weights-of-evidence (WoE) 6 Support vector machine (SVM) 7 Spatial Multiple Criteria Evaluation (SMCE) 8 Statistical index 9 Weighting factor و همکـاران ، ٢٠١٦) و مـدل ترکیبی هوش مصـــنوعی جـدید١ (چپی و همکاران ، ٢٠١٧) برای ارزیابی پتانســـیل سیل خیزی استفاده شده است .