Flood hazards susceptibility mapping using statistical, fuzzy logic, and MCDM methods


AKAY H.

SOFT COMPUTING, vol.25, no.14, pp.9325-9346, 2021 (Peer-Reviewed Journal) identifier identifier

  • Publication Type: Article / Article
  • Volume: 25 Issue: 14
  • Publication Date: 2021
  • Doi Number: 10.1007/s00500-021-05903-1
  • Journal Name: SOFT COMPUTING
  • Journal Indexes: Science Citation Index Expanded, Scopus, Academic Search Premier, Applied Science & Technology Source, Compendex, Computer & Applied Sciences, INSPEC, zbMATH
  • Page Numbers: pp.9325-9346
  • Keywords: Bivariate statistical models, Flood hazards susceptibility, Fuzzy logic model, Hybrid methods, Multicriteria decision making methods, MULTICRITERIA DECISION-MAKING, EVIDENTIAL BELIEF FUNCTION, MACHINE LEARNING-MODELS, WEIGHTS-OF-EVIDENCE, FREQUENCY RATIO, HYDROLOGIC PARAMETERS, LOGISTIC-REGRESSION, GENETIC ALGORITHM, INFERENCE SYSTEM, NEURAL-NETWORKS

Abstract

In this study, the flood hazards susceptibility map of an area in Turkey which is frequently exposed to flooding was predicted by training 70% of inventory data. For this, statistical, and hybrid methods such as frequency ratio (FR), evidential belief function (EBF), weight of evidence (WoE), index of entropy (IoE), fuzzy logic (FL), principal component analysis (PCA), analytical hierarchy process (AHP), technique for order preference by similarity to an ideal solution (TOPSIS), and VlseKriterijumska optimizacija I Kompromisno Resenje (VIKOR) were adapted. Values at both 70% and 30% of inventory data from the generated maps were extracted to validate the training and testing processes by receiver operating characteristics (ROC) analysis and seed cell area index (SCAI). Sensitivity, specificity, accuracy, and kappa index were calculated from ROC analysis, and SCAI was computed from the classification of map by natural break method and flood pixels in that classification. Since the predicted results of the methods applied did not point out the same model for each criterion, a simple method was selected to determine the most preferable method. Analysis showed that, IoE model was found to be the best model considering the ROC parameters, while PCA and AHP methods gave more reliable results considering SCAI. This study may be considered as a comprehensive contribution to the hybridization methods in predicting accurate flood hazards susceptibility maps.