Elsevier

Science of The Total Environment

Volume 621, 15 April 2018, Pages 1124-1141
Science of The Total Environment

Flood susceptibility assessment in Hengfeng area coupling adaptive neuro-fuzzy inference system with genetic algorithm and differential evolution

https://doi.org/10.1016/j.scitotenv.2017.10.114Get rights and content

Highlights

  • ANFIS was coupled with GA and DE for flood susceptibility modelling.

  • 195 flood events were used for the ANFIS training.

  • SWARA method was used to evaluate the relationship between floods and conditioning factors.

  • ANFIS-DE model shows the better result in flood prediction.

  • Very high flood susceptible area is up to 20 km2

Abstract

Floods are among Earth's most common natural hazards, and they cause major economic losses and seriously affect peoples' lives and health. This paper addresses the development of a flood susceptibility assessment that uses intelligent techniques and GIS. An adaptive neuro-fuzzy inference system (ANFIS) was coupled with a genetic algorithm and differential evolution for flood spatial modelling. The model considers thirteen hydrologic, morphologic and lithologic parameters for the flood susceptibility assessment, and Hengfeng County in China was chosen for the application of the model due to data availability and the 195 total flood events. The flood locations were randomly divided into two subsets, namely, training (70% of the total) and testing (30%). The Step-wise Weight Assessment Ratio Analysis (SWARA) approach was used to assess the relation between the floods and influencing parameters. Subsequently, two data mining techniques were combined with the ANFIS model, including the ANFIS-Genetic Algorithm and the ANFIS-Differential Evolution, to be used for flood spatial modelling and zonation. The flood susceptibility maps were produced, and their robustness was checked using the Receiver Operating Characteristic (ROC) curve. The results showed that the area under the curve (AUC) for all models was > 0.80. The highest AUC value was for the ANFIS-DE model (0.852), followed by ANFIS-GA (0.849). According to the RMSE and MSE methods, the ANFIS-DE hybrid model is more suitable for flood susceptibility mapping in the study area. The proposed method is adaptable and can easily be applied in other sites for flood management and prevention.

Introduction

The significance of water is well documented in the Chinese tradition through quotations such the one by the philosopher Laozi, who praises the importance of water: “The best of men are like water; water benefits all things”. Nonetheless, flood incidents are extreme water flow events with often catastrophic consequences. Floods are considered a natural hazard, and they cause major economic losses and seriously affect peoples' lives and health globally (Dottori et al., 2016). According to the United Nations Office for Disaster Risk Reduction (UNISDR), between 1996 and 2015, the number of casualties due to flood events was 150,061 globally, representing 11.1% of the global disaster mortality.

Flood occurrence is a complex and site-dependent phenomenon that has always intrigued scientists, compelling them to explore, analyse and better understand its mechanisms. Robins et al. (2009) defined surface stability, vegetation, soil growth, and diagnostic horizon data as the key factors for flood disaster evaluation (Robins et al., 2009). The reasons for flood occurrence and development can be distinguished between human and natural factors. Climate change is considered an important factor for extreme flood/drought occurrence (Rojas et al., 2012, Sampson et al., 2015). In that sense, Charlton et al. (2006) claimed that climate-induced alterations may influence land use and eventually increase flood risk, as the additional impermeable surfaces, drains and sewers favour rapid water transfer. Increased flow speed reduces the available response time and further increases the volume of the peak flow (Charlton et al., 2006). Arnell and Gosling (2016) evaluated the impact of climate change on global river flood risk, demonstrating that the latter would increase by approximately 187% by 2050, with Asia expected to be the most affected (Arnell and Gosling, 2016). Emerton et al. (2017) compared El Niño and La Niña events, which are also often grouped together under the term El Niño Southern Oscillation (ENSO). The extremes of the ENSO climate variability appear to influence river flow and flood recurrence at a global scale (Emerton et al., 2017). In their research, Wang et al. (2015) analysed the maximum values for 3-day precipitation, typhoon frequency and runoff depth as indices of flood development (Wang et al., 2015). The need to include additional parameters for rainfall accumulation is a widely accepted approach, as rainfall accumulation is a required but insufficient condition for inducing flash floods (Norbiato et al., 2008), as local hydrology critically controls flash flood creation. Initial soil moisture is an additional important factor, as is human intervention. The removal of vegetation results in soil erosion. Uncontrolled waste disposal minimizes the river's capacity and exacerbates the flood hazard. Urbanization also increases the flood risk as a major force altering the hydrological processes over a range of temporal and spatial scales (Suriya and Mudgal, 2012).

As stated by Dobler et al. (2012), flood susceptibility is a prerequisite for sustainable flood risk management, as it provides useful information about the appropriate strategies for adaptation and mitigation (Dobler et al., 2012). Flood susceptibility assessment generally builds on the latest Geographic Information System (GIS) tools to process and analyse complex planning strategies, decision making, and integrated management. GIS integrates maps in a visual manner that is query-based with statistical analyses. This explains its wide use in environmental management and modelling, including landslides, forest fires and flood hazards (Xiao et al., 2017). GIS has facilitated flood susceptibility studies and significantly improved their accuracy (Nunes Correia et al., 1998). The various parameters related to flood susceptibility may be used through heuristic approaches by assigning different weights depending on their relative importance. However, the recent practice of flood susceptibility modelling promoted statistical techniques via univariate or multivariate models, as they are considered more objective quantitative methods (de Walque et al., 2017).

In the current literature, one can find three main types of flood susceptibility assessment: i) hydrological models such as HYDROTEL (Aissia et al., 2012), SWAT (Oeurng et al., 2011), and WetSpa (Bahremand et al., 2007); ii) statistical and data-driven approaches such as the analytic hierarchy process (AHP) (Kazakis et al., 2015, Stefanidis and Stathis, 2013), frequency ratio (FR) (Tehrany et al., 2015a), logistic regression (LR) (Ettinger et al., 2016, Nandi et al., 2016), weights-of-evidence (WOE) (Tehrany et al., 2014a), and fuzzy logic (FL) (Pulvirenti et al., 2011); and iii) non-linear machine learning algorithms such as artificial neural network (ANN) (Kia et al., 2012), decision tree (DT) (Tehrany et al., 2013a), support vector machine (SVM) (Tehrany et al., 2015c), k-nearest neighbor (KNN) (Liu et al., 2016), and random forest (Wang et al., 2015). The aforementioned methods have successfully provided flood susceptibility assessments, but they also face some limitations (Ward et al., 2015). For example, conventional hydrological models are not robust automated methods because the model design, building and parameter calibration is time consuming. Statistical and data-driven methods have an increased subjectivity as they require selecting the flood conditioning factors. Chen et al. (2015) applied a multicriteria decision analysis method to evaluate flood hazard in Japan and observed that low-quality data may lead to errors and low-quality results (Chen et al., 2015). In the process of modelling, non-linear machine learning algorithms may also lead to poor projections due to the large and inconsistent value ranges in the datasets (Tien Bui et al., 2016a).

Although flooding is a serious hazard for China, insufficient attention has been paid to flood hazard assessment. Winsemius et al. (2016) coupled climate, socio-economic and hydrologic models to research global future river flood risk. The result showed that countries in Southeast Asia may face increased flood risk, especially in the Yangtze river (Winsemius et al., 2016). Human life and economic losses in China have been significant throughout history (Zong and Chen, 2000). Yu et al. (2009) analysed historical floods on the Yangtze River, China and concluded that after 1950, the flood frequency increased and that anthropogenic activity (e.g., deforestation, diking, and lake-coast reclamation) is an important factor for flooding (Yu et al., 2009).

This paper investigates the main influencing factors for flood susceptibility in the Hengfeng area, China. Climate change, local geomorphology, topography and hydrology are analysed, along with human activities and intervention. Accordingly, the analysis aims to identify the key factors in flood susceptibility modelling. A GIS-based model has been developed to perform flood susceptibility mapping in Hengfeng County, China. It is coupled with a neuro-fuzzy inference system (ANFIS) that incorporates a genetic algorithm (GA) and differential evolution (DE) models while comparing their performance. The main difference between the proposed methodology and existing studies is the coupling of ANFIS to GA and DE, which –to the authors' knowledge – is implemented for the first time. The overall performance of the models was evaluated using the appropriate datasets and collected information. Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Receiver Operating Characteristic (ROC) curve, and area under the ROC curve (AUC) evaluated the model performance. High risk locations for flood and the projected magnitude and distribution of the flood hazard can support governmental planning and decision makers' work on devising flood management mitigation strategies and planning (Xiao et al., 2017). Finally, the robustness of the model was evaluated by comparing the flood susceptibility map with historical flood events.

Section snippets

Data and methods

The developed methodology follows the following five steps: i) selecting the study area; ii) data assembly, including collecting the flood inventory map and influencing factors; (iii) factor correlation using Step-wise Weight Assessment Ratio Analysis (SWARA); (iv) flood spatial modelling using the ANFIS-GA, and ANFIS-DE models; and (v) validation and model comparison (see Fig. 1).

Correlation between floods and conditioning factors using the SWARA model

Table 1 shows the correlation between floods and conditioning factors using the SWARA model. The results show that in the case of slope angle, the SWARA values are the highest for the class of 0°–5° (0.334), indicating that this class has the highest probability of flood occurrence. The results show that the highest SWARA values are calculated in the Northwest-facing (0.169) and east-facing (0.161) areas, whereas in the case of elevation, the SWARA values decrease as the altitude increases (the

Discussion

Flooding is a very complex process that is controlled by many environmental factors. For instance, climatic parameters strongly influence flood events. However, climate change determines a variable environment, where extreme weather events occur more rapidly inducing natural hazards such as floods To the best of our knowledge, extreme events in the Hengfeng area have not been sufficiently studied to date. The present study aims to cover this knowledge gap by using a spatial approach that

Conclusions

Flood susceptibility assessment is of the utmost importance to prevent human casualties and socio-economic losses. Flood management should be a priority, particularly in emerging economies with high developing targets, such as China. In this study, an innovative method for flood susceptibility mapping was developed by coupling an adaptive neuro-fuzzy inference system (ANFIS) with a genetic algorithm and differential evolution. Hengfeng County was selected for the application of these methods

Acknowledgements

We express our great thanks to Wei Ouyang, Associate Editor of the journal Science of the Total Environment and all reviewers, with their comments and suggestions, which improved the quality of our paper. National Natural Science Foundation of China (Project No.: 41431177, 41601413), Natural Science Research Program of Jiangsu (Project No.: BK20150975), Natural Science Research Program of Jiangsu (Project No.: 14KJA170001), National Basic Research Program of China (Project No.: 2015CB954102),

References (136)

  • W. Chen et al.

    Landslide spatial modeling: introducing new ensembles of ANN, MaxEnt, and SVM machine learning techniques

    Geoderma

    (2017)
  • W. Chen et al.

    Spatial prediction of landslide susceptibility using an adaptive neuro-fuzzy inference system combined with frequency ratio, generalized additive model, and support vector machine techniques

    Geomorphology

    (2017)
  • W. Chen et al.

    A comparative study of logistic model tree, random forest, and classification and regression tree models for spatial prediction of landslide susceptibility

    Catena

    (2017)
  • A. Dehnavi et al.

    A new hybrid model using step-wise weight assessment ratio analysis (SWARA) technique and adaptive neuro-fuzzy inference system (ANFIS) for regional landslide hazard assessment in Iran

    Catena

    (2015)
  • C. Dobler et al.

    Assessment of climate change impacts on flood hazard potential in the Alpine Lech watershed

    J. Hydrol.

    (2012)
  • F. Dottori et al.

    Development and evaluation of a framework for global flood hazard mapping

    Adv. Water Resour.

    (2016)
  • J. Du et al.

    Assessing the effects of urbanization on annual runoff and flood events using an integrated hydrological modeling system for Qinhuai River basin, China

    J. Hydrol.

    (2012)
  • S. Ettinger et al.

    Building vulnerability to hydro-geomorphic hazards: estimating damage probability from qualitative vulnerability assessment using logistic regression

    J. Hydrol.

    (2016)
  • T. Fawcett

    An introduction to ROC analysis

    Pattern Recogn. Lett.

    (2006)
  • D.S. Fernández et al.

    Urban flood hazard zoning in Tucumán Province, Argentina, using GIS and multicriteria decision analysis

    Eng. Geol.

    (2010)
  • I.C. Fuller

    Geomorphic impacts of a 100-year flood: Kiwitea Stream, Manawatu catchment, New Zealand

    Geomorphology

    (2008)
  • J.M. García-Ruiz et al.

    Flood generation and sediment transport in experimental catchments affected by land use changes in the central Pyrenees

    J. Hydrol.

    (2008)
  • N.K. Goel et al.

    A derived flood frequency distribution for correlated rainfall intensity and duration

    J. Hydrol.

    (2000)
  • E. Hagen et al.

    Reverse engineered flood hazard mapping in Afghanistan: a parsimonious flood map model for developing countries

    Quat. Int.

    (2010)
  • B. He et al.

    Age and duration of the Emeishan flood volcanism, SW China: geochemistry and SHRIMP zircon U–Pb dating of silicic ignimbrites, post-volcanic Xuanwei Formation and clay tuff at the Chaotian section

    Earth Planet. Sci. Lett.

    (2007)
  • L.T.K. Ho et al.

    Micro-landform classification and flood hazard assessment of the Thu Bon alluvial plain, central Vietnam via an integrated method utilizing remotely sensed data

    Appl. Geogr.

    (2011)
  • R. Jafari-Marandi et al.

    Fluid genetic algorithm (FGA)

    J. Comput. Des. Eng.

    (2017)
  • A.L. Kay et al.

    RCM rainfall for UK flood frequency estimation. II. Climate change results

    J. Hydrol.

    (2006)
  • N. Kazakis et al.

    Assessment of flood hazard areas at a regional scale using an index-based approach and Analytical Hierarchy Process: application in Rhodope–Evros region, Greece

    Sci. Total Environ.

    (2015)
  • A.D. Knighton

    Downstream variation in stream power

    Geomorphology

    (1999)
  • K. Liu et al.

    Coupling the k-nearest neighbor procedure with the Kalman filter for real-time updating of the hydraulic model in flood forecasting

    Int. J. Sediment Res.

    (2016)
  • J.A. López-Tarazón et al.

    Rainfall, runoff and sediment transport relations in a mesoscale mountainous catchment: the River Isábena (Ebro basin)

    Catena

    (2010)
  • Z.Y. Marchetti et al.

    NDVI patterns as indicator of morphodynamic activity in the middle Paraná River floodplain

    Geomorphology

    (2016)
  • N. Mathur et al.

    Comparison of adaptive neuro-fuzzy inference system (ANFIS) and Gaussian processes for machine learning (GPML) algorithms for the prediction of skin temperature in lower limb prostheses

    Med. Eng. Phys.

    (2016)
  • J. McCall

    Genetic algorithms for modelling and optimisation

    J. Comput. Appl. Math.

    (2005)
  • R. Mukherjee et al.

    Modified differential evolution with locality induced genetic operators for dynamic optimization

    Eur. J. Oper. Res.

    (2016)
  • P.K. Naik et al.

    Distinguishing human and climate influences on the Columbia River: changes in mean flow and sediment transport

    J. Hydrol.

    (2011)
  • S. Nedkov et al.

    Flood regulating ecosystem services—mapping supply and demand, in the Etropole municipality, Bulgaria

    Ecol. Indic.

    (2012)
  • D. Norbiato et al.

    Flash flood warning based on rainfall thresholds and soil moisture conditions: an assessment for gauged and ungauged basins

    J. Hydrol.

    (2008)
  • C. Oeurng et al.

    Assessment of hydrology, sediment and particulate organic carbon yield in a large agricultural catchment using the SWAT model

    J. Hydrol.

    (2011)
  • H.-J. Oh et al.

    Application of a neuro-fuzzy model to landslide-susceptibility mapping for shallow landslides in a tropical hilly area

    Comput. Geosci.

    (2011)
  • N. Onat et al.

    Analysis of wind climate and wind energy potential of regions in Turkey

    Energy

    (2011)
  • W. Phootrakornchai et al.

    Online critical clearing time estimation using an adaptive neuro-fuzzy inference system (ANFIS)

    Int. J. Electr. Power Energy Syst.

    (2015)
  • H. Rashid

    Interpreting flood disasters and flood hazard perceptions from newspaper discourse: tale of two floods in the Red River valley, Manitoba, Canada

    Appl. Geogr.

    (2011)
  • S.L. Reneau

    Stream incision and terrace development in Frijoles Canyon, Bandelier National Monument, New Mexico, and the influence of lithology and climate

    Geomorphology

    (2000)
  • M. Rezakazemi et al.

    H2-selective mixed matrix membranes modeling using ANFIS, PSO-ANFIS, GA-ANFIS

    Int. J. Hydrog. Energy

    (2017)
  • C.R. Robins et al.

    Comparison of flood hazard assessments on desert piedmonts and playas: a case study in Ivanpah Valley, Nevada

    Geomorphology

    (2009)
  • S. Rozalis et al.

    Flash flood prediction using an uncalibrated hydrological model and radar rainfall data in a Mediterranean watershed under changing hydrological conditions

    J. Hydrol.

    (2010)
  • B. Saeidian et al.

    Evaluation and comparison of Genetic Algorithm and Bees Algorithm for location–allocation of earthquake relief centers

    Int. J. Disaster Risk Reduction

    (2016)
  • R. Aalto et al.

    Episodic sediment accumulation on Amazonian flood plains influenced by El Nino/Southern Oscillation

    Nature

    (2003)
  • Cited by (319)

    View all citing articles on Scopus
    View full text