Estimation of rainfall-induced landslides using ANN and fuzzy clustering methods: A case study in Saeen Slope, Azerbaijan province, Iran Y. Alimohammadlou a, ⁎, A. Najafi b , C. Gokceoglu c a Department of Civil Engineering, Zanjan Branch, Islamic Azad University, Zanjan, Iran b Department of Industrial Engineering, Zanjan Branch, Islamic Azad University, Zanjan, Iran c Hacettepe University, Geological Engineering Department, Beytepe, 06800 Ankara, Turkey abstract article info Article history: Received 29 March 2013 Received in revised form 27 February 2014 Accepted 2 April 2014 Available online xxxx Keywords: Landslide Saeen Slope Rainfall Prediction Artificial neural network Fuzzy clustering As is the case around the world, Azerbaijan province in northwestern Iran experiences numerous landslides that occur following intensive precipitation periods. These landslides damage many aspects of human life as well as the natural environment, and hence it should be evaluated accurately. However, one of the main challenges of landslide studies is the estimation of the periods between potential landslides, which would provide information that is useful for the development of warning systems and/or mitigation measures. The aim of the present study is to propose a novel approach utilizing artificial neural network and fuzzy clustering methods for landslide frequency estimation. This study also investigates the 2005 Saeen, Iran landslide triggered by prolonged heavy rainfall that affected groundwater levels, and introduces a methodology to estimate the date range of the next probable landslide. Based on the interpretation of the triggering factor and failure mechanism, the Saeen landslide was induced by the prolonged rainfall behavior and resultant deep infiltration of water between the years 2002 and 2005. During this period, the maximum rainfall values were observed in April of each year, and then followed by decreased rainfall to a minimum value in June and August. The results of this investigation revealed that the failure probability will likely increase in the next precipitation periods and the saturation rate will be high in August and September of 2017 and 2018, resulting in landslides. In conclusion, this method is only used for the heavy precipitation as the triggering factor to estimate and analyze the next potential landslide. The information derived from this method will establish a time window for future failure, where the other slope-stability factors can be evaluated and then utilized to set up more accurate and reliable networks for further investigations. © 2014 Elsevier B.V. All rights reserved. 1. Introduction Landslides are known as one of the most important natural hazards to human habitations throughout the world. This event occurring in vul- nerable regions and civilization centers can result in serious damage to various aspects of human life. In addition to a high rate of mortality, landslides cause damage or destruction in urban and industrial centers (Yalcin, 2011). Landslides in the United States cause approximately 3.5 billion US$ (year 2001 dollars) in damage, and kill between 25 and 50 people annually (Highland, 2004), thus it is one of the most costly disasters worldwide. Similarly, Japanese annual losses are reported to be between 4 and 6 billion US$ (Herath and Wang, 2009). Different aspects of landslide phenomena have been investigated such as those concerned with causes, triggering factors and their impact priority (Cascini et al., 2011; Zezere et al., 1999), respective parameters to assigning slope sensitivity (Gullà et al., 2008), development of warning systems (Dai et al., 2002), providing landslide susceptibility mapping for use in urban development programs (Yao et al., 2008), slope stabilization and mitigation measures applied to reactivated land- slides based on Factors Prioritization (Kwong et al., 2004), and finally case studies in numerous areas and their failure scenarios (Anbarasu et al., 2010; Tang et al., 2011). However, investigations of the landslide estimation and the analysis of the probability of the future occurrences are of primary concern for the current study. Many individual and complex methods have been applied in order to analyze the probability of slope failure and evaluate landslide stabi- lization up to now. Some of these methods used a comprehensive criterion such as Factor of Safety (FS) which requires geotechnical and physical data on materials (e.g. Fellenius, 1936; Bishop, 1955; Janbu's Rigorous and Simplified Methods (Janbu, 1957); Morgenstern and Price Method (Morgenstern and Price, 1965); and Spencer method (Spencer, 1967)). In addition, there are some other prediction methods to estimate slope stability by using geological and geomorphological slope characteristics (e.g. Lee et al., 2008); however, there are serious Catena 120 (2014) 149–162 ⁎ Corresponding author. Tel./fax: +98 241 4260063. E-mail address: y.alimohammadlou@azu.ac.ir (Y. Alimohammadlou). http://dx.doi.org/10.1016/j.catena.2014.04.009 0341-8162/© 2014 Elsevier B.V. All rights reserved. Contents lists available at ScienceDirect Catena journal homepage: www.elsevier.com/locate/catena