ارزیابی حساسیت زمین‌لغزش با استفاده از مدل جدید ترکیبی الگوریتم مبنا (مطالعه موردی: شهرستان کامیاران، استان کردستان)

نوع مقاله : مقاله پژوهشی

نویسندگان

1 دانش آموخته دکترای ژئومورفولوژی از گروه جغرافیای طبیعی ،دانشگاه محقق اردبیلی، اردبیل، ایران.

2 استاد گروه جغرافیای طبیعی (ژئومورفولوژی)، دانشگاه محقق اردبیلی، اردبیل، ایران.

3 استاد گروه جغرافیای طبیعی (ژئومورفولوژی)، دانشگاه تبریز، تبریز، ایران.

4 کارشناس آموزشی و پژوهشی گروه مرتع و آبخیزداری، دانشکده منابع طبیعی، دانشگاه کردستان

10.22034/gmpj.2021.131015

چکیده

زمین­لغزش­ها به عنوان یکی از مخرب­ترین پدیده­های طبیعی محسوب می­شوند. به دلیل تهدید آن­ها، باید یک نقشه جامع حساسیت زمین­لغزش برای کاهش آسیب­های احتمالی به افراد و زیرساخت­ها تهیه شود. کیفیت نقشه­های حساسیت زمین­لغزش تحت تأثیر بسیاری از عوامل، از جمله کیفیت داده­های ورودی و انتخاب مدل­های ریاضی است. هدف اصلی این پژوهش ارائه یک مدل ترکیبی جدید داده­کاوی به نام Rotation Forest - Functional Trees (RF-FT) که یک رویکرد هوشمند ترکیبی از دو تکنیک یادگیری ماشین مدل Functional Trees (FT) و تکنیک طبقه­بندی مدل Rotation Forest (RF) برای ارزیابی حساسیت زمین لغزش­های اطراف شهر کامیاران واقع در استان کردستان می­باشد. در ابتدا، بیست و یک عامل مؤثر بر وقوع زمین­لغزش­های منطقه مورد مطالعه شامل درجه شیب، جهت شیب، ارتفاع، انحنای شیب، انحنای عرضی شیب، انحنای طولی شیب، تابش خورشید، عمق دره، شاخص قدرت جریان، شاخص نمناکی توپوگرافی، شاخص طول دامنه، کاربری اراضی، تراکم پوشش گیاهی، فاصله از گسل، تراکم گسل، فاصله از جاده، تراکم جاده، فاصله از آبراهه، تراکم آبراهه، همباران و لیتولوژی  به همراه نقشه پراکنش زمین­لغزش با 60 نقطه لغزشی برای جمع­آوری داده­های آموزشی و آزمون جمع­آوری شدند. سپس، بر اساس شاخص Information Gain Ratio هفده عامل مؤثر از بین آن­ها انتخاب و جهت مدل­سازی به کار گرفته شدند. در مرحله بعد  مدل هیبریدی RFFT برای ارزیابی حساسیت زمین­لغزش با استفاده از مجموعه داده­های آموزشی ساخته شد. عملکرد مدل پیشنهادی RFFT با استفاده از چندین پارامتر آماری از جمله حساسیت، شفافیت، صحت، مجذور مربعات خطا، منحنی نرخ موفقیت و سطح زیر این منحنی مورد ارزیابی قرار گرفت.
 

کلیدواژه‌ها


عنوان مقاله [English]

Landslide susceptibility assessment using a novel ensemble algorithm based model (Case Study: Kamyaran city, Kurdistan province)

نویسندگان [English]

  • Bahareh Gasemyan 1
  • Mousa Abedini 2
  • Shahram Roostai 3
  • Ataalah Shirzadi 4
1 mohaghegh ardabili university
2 professor in Geomorphology, University of Mohaghegh Ardabili
3 Department of Physical Geography, University of Tabriz
4 Department of Rangeland and Watershed Management, Faculty of Natural Resources, University of Kurdistan, Sanandaj, Iran
چکیده [English]

Landslides are considered one of the most destructive natural phenomena. Landslides are dangerous natural hazards. Because of their threat, a comprehensive landslide susceptibility map should be produced to reduce the possible damages to people and infrastructure. The quality of landslide susceptibility maps is influenced by many factors, such as the quality of input data and the selection of mathematical models. The main purpose of this study is to presentation, a novel hybrid model namely Rotation Forest based Functional Trees (RFFT), which is a hybrid intelligent approach of two state of the art machine learning techniques of Functional Trees (FT) classifier and Rotation Forest (RF) ensemble, for landslide susceptibility Assessment prediction in Kamyaran city located in Kurdistan Province, Iran. At first, twenty-one factors affecting the occurrence of landslide in the study area including Slope angle, Aspect, Elevation, Curvature, Plan curvature, Profile curvature, Radiation, Valle depth(VD), stream power index (SPI), topographic wetness index (TWI), combination of length-angle of slope (LS), Land use, NDVI (normalized vegetation index), Distance to Faults, Faults density, Distance to Road, Road density, Distance to River, River density Lithology and Rainfall with total of 60 landslide locations have been collected for generating training and testing datasets. Then, based on the Information Gain Ratio Index, eight effective factors were chosen and used for modeling. Performance of the proposed RFFT model was evaluated using some statistical-based measures such as sensitivity, specificity, accuracy, RMSE and area under the ROC curve (AUROC). The results showed that the proposed model performed well in this study (AUC = 0.891), and it improved significantly the performance of the FT base classifier (AUC = 0.819). Therefore, it can be concluded that the proposed RFFT model should be used as a great alternative method for better landslide susceptibility assessment in landslide prone area.Landslides are considered one of the most destructive natural phenomena. Landslides are dangerous natural hazards. Because of their threat, a comprehensive landslide susceptibility map should be produced to reduce the possible damages to people and infrastructure. The quality of landslide susceptibility maps is influenced by many factors, such as the quality of input data and the selection of mathematical models. The main purpose of this study is to presentation, a novel hybrid model namely Rotation Forest based Functional Trees (RFFT), which is a hybrid intelligent approach of two state of the art machine learning techniques of Functional Trees (FT) classifier and Rotation Forest (RF) ensemble, for landslide susceptibility Assessment prediction in Kamyaran city located in Kurdistan Province, Iran. At first, twenty-one factors affecting the occurrence of landslide in the study area including Slope angle, Aspect, Elevation, Curvature, Plan curvature, Profile curvature, Radiation, Valle depth(VD), stream power index (SPI), topographic wetness index (TWI), combination of length-angle of slope (LS), Land use, NDVI (normalized vegetation index), Distance to Faults, Faults density, Distance to Road, Road density, Distance to River, River density Lithology and Rainfall with total of 60 landslide locations have been collected for generating training and testing datasets. Then, based on the Information Gain Ratio Index, eight effective factors were chosen and used for modeling. Performance of the proposed RFFT model was evaluated using some statistical-based measures such as sensitivity, specificity, accuracy, RMSE and area under the ROC curve (AUROC). The results showed that the proposed model performed well in this study (AUC = 0.891), and it improved significantly the performance of the FT base classifier (AUC = 0.819). Therefore, it can be concluded that the proposed RFFT model should be used as a great alternative method for better landslide susceptibility assessment in landslide prone area.Landslides are considered one of the most destructive natural phenomena. Landslides are dangerous natural hazards. Because of their threat, a comprehensive landslide susceptibility map should be produced to reduce the possible damages to people and infrastructure. The quality of landslide susceptibility maps is influenced by many factors, such as the quality of input data and the selection of mathematical models. The main purpose of this study is to presentation, a novel hybrid model namely Rotation Forest based Functional Trees (RFFT), which is a hybrid intelligent approach of two state of the art machine learning techniques of Functional Trees (FT) classifier and Rotation Forest (RF) ensemble, for landslide susceptibility Assessment prediction in Kamyaran city located in Kurdistan Province, Iran. At first, twenty-one factors affecting the occurrence of landslide in the study area including Slope angle, Aspect, Elevation, Curvature, Plan curvature, Profile curvature, Radiation, Valle depth(VD), stream power index (SPI), topographic wetness index (TWI), combination of length-angle of slope (LS), Land use, NDVI (normalized vegetation index), Distance to Faults, Faults density, Distance to Road, Road density, Distance to River, River density Lithology and Rainfall with total of 60 landslide locations have been collected for generating training and testing datasets. Then, based on the Information Gain Ratio Index, eight effective factors were chosen and used for modeling. Performance of the proposed RFFT model was evaluated using some statistical-based measures such as sensitivity, specificity, accuracy, RMSE and area under the ROC curve (AUROC). The results showed that the proposed model performed well in this study (AUC = 0.891), and it improved significantly the performance of the FT base classifier (AUC = 0.819). Therefore, it can be concluded that the proposed RFFT model should be used as a great alternative method for better landslide susceptibility assessment in landslide prone area.

کلیدواژه‌ها [English]

  • Landslide
  • Hybrid Model
  • Information Gain Ratio Index
  • Kamyaran
  • پورقاسمی، حمیدرضا، 1389، پهنه­بندی حساسیت زمین­لغزش با استفاده از مدل احتمالاتی وزن واقعه، مهندسی فناوری اطلاعات مکانی، سال یکم، شماره نهم، ص80 –
  • شیرزادی عطااله، سلیمانی کریم، حبیب­نژاد محمود، کاویان عطااله، چپی کامران، 1396، معرفی یک مدل جدید ترکیبی الگوریتم مبنا به منظور پیش­بینی حساسیت زمین­لغزش­های سطحی اطراف شهر بیجار، جغرافیا و توسعه، شماره 46، صفحات 246 –
  • طالبی علی،  گودرزی سحر، پورقاسمی حمید رضا، 1396،  بررسی امکان تهیه نقشه خطر زمین لغزش با استفاده از الگوریتم جنگل تصادفی (محدوده ی موردمطالعه: حوزه آبخیز سردارآباد، استان لرستان)، مجله مخاطرات محیط طبیعی.
  • مرادی حمیدرضا، محمدی مجید، پورقاسمی حمید.رضا، 1391، حرکات دامنه­ای با تأکید بر روشهای کمی تحلیل وقوع زمین­لغزش. انتشارات سمت، ص 209.
  • Abedini M, Ghasemyan B, Rezaei Mogaddam M H, 2017, Landslide susceptibility mapping in Bijar city, Kurdistan Province, Iran: a comparative study by logistic regression And AHP models, Environ Earth Sci, 76:308, DOI 10.1007/s12665-017-6502-3
  • Benda, L., Dunne, T. 1997. Stochastic forcing of sediment supply to channel networks from landsliding and debris flow, Water Resources Research, 33: 2849–2863.
  • Chen W, Xie X, Wang J, Pradhan B, Hong H, Bui DT, Duan Z, Ma J. 2017. A comparative study of logistic model tree, random forest, and classification and regression tree models for spatial prediction of landslide susceptibility. Catena. 151:147–160.
  • Dahal, R.K., Hasegava, Sh., Nonoura, A., Yamanka, M., Dhakal, S., Pauudyal, P., 2008. Predictive Modeling of Rainfall-Induced Landslide Hazard in the Lesser Himalaya of Nepal Based on Weights of Evidence, Geomorphology, Vol.102, NO.3-4, and PP: 496-510.
  • Gama, J.: Functional trees for classification. In: Proceedings IEEE International Conference on Data Mining, 2001, ICDM 2001, pp. 147–154.
  • Kavzoglu, T., Colkesen, I.: 2013, an assessment of the effectiveness of a rotation forest ensemble for land-use and land-cover mapping. Int. J. Remote Sens. 34, 4224–4241
  • Lan H, Frank E, Hall M, 2011, Data mining: Practical machine learning tools and techniques. Morgan Kaufman, Boston.
  • Omar F. Althuwaynee I Biswajeet Pradhan I Hyuck-Jin Park I Jung Hyun Lee. ,2014, A novel ensemble decision tree-based CHi-squared Automatic Interaction Detection (CHAID).
  • Ozcift, A., Gulten, A.: 2012, robust multi-class feature selection strategy based on rotation forest ensemble algorithm for diagnosis of Erythemato-Squamous diseases. J. Med. Syst. 36, 941–949.
  • Ozcift, A., Gulten, A.: 2011, Classifier ensemble construction with rotation forest to improve medical diagnosis performance of machine learning algorithms. Comput. Methods Programs Biomed. 104, 443–451.
  • Pham, B.T., Bui, D.T., Dholakia, M.B., Prakash, I., Pham, H.V., Mehmood, K., Le, H.Q. 2016, A novel ensemble classifier of rotation forest and Naïve Bayer for landslide susceptibility assessment at the Luc Yen district, Yen Bai Province (Viet Nam) using GIS. Geomat. Nat. Hazards Risk, 1–23.
  • Pradhan Biswajeet ,2013, a comparative study on the predictive ability of the decision tree, support vector machine and neuro-fuzzy models in landslide susceptibility mapping using GIS, Computers & Geosciences 51, PP: 350–365.
  • Quinlan JR, 1993, C4.5: programs for machine learning. Morgan Kaufmann, San Mateo, CA, USA.
  • Rodriguez JJ, Kuncheva LI, Alonso CJ. 2006. Rotation forest: a new classifier ensemble method. Pattern Anal Mach Intell IEEE Trans. 28:1619–1630.
  • Shirzadi Ataollah, Dieu Tien Bui,Binh Thai Pham, Karim Solaimani, Kamran Chapi, Ataollah Kavian, Himan Shahabi, 2017. Shallow landslide susceptibility assessment using a novel hybrid intelligence approach, Environmental Earth Sciences 76:60. Pp 1-18.
  • Tsangaratos P, Ilia I, 2015, Landslide susceptibility mapping using a modified decision tree classifier in the Xanthi Perfection, Greece Landslides: 1-16 doi: 10.1007/s10346-015-0565-6.
  • Yesilnacar E, Topal T,2005,Landslide susceptibility mapping: comparison of logistic regression and neural networks in a medium scale study, Hendek region (Turkey) Engineering Geology 79:251-266.