Flood susceptibility mapping with ensemble machine learning: a case of Eastern Mediterranean basin, Türkiye


Stochastic Environmental Research and Risk Assessment, vol.37, no.11, pp.4273-4290, 2023 (SCI-Expanded) identifier

  • Publication Type: Article / Article
  • Volume: 37 Issue: 11
  • Publication Date: 2023
  • Doi Number: 10.1007/s00477-023-02507-z
  • Journal Name: Stochastic Environmental Research and Risk Assessment
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, CAB Abstracts, Compendex, Environment Index, Geobase, Index Islamicus, Pollution Abstracts, zbMATH, Civil Engineering Abstracts
  • Page Numbers: pp.4273-4290
  • Keywords: Artificial neural network, Decision trees, Flood susceptibility mapping, Gradient boosting trees, Support vector machines
  • Hacettepe University Affiliated: Yes


In recent years, with the advances in remote sensing and geospatial technology, various machine learning algorithms found applications in determining potentially flooded areas, which have an important place in basin planning and depend on various environmental parameters. This study uses ensemble models of decision trees (DT), gradient boosting trees (GBT), support vector machines (SVM) and artificial neural network (ANN) algorithms to generate flood susceptibility maps of the Eastern Mediterranean Basin located in the Eastern Türkiye where intense short-duration rainfall causes flash floods with devastating effects on the densely populated coastal region and agricultural areas. Results of test-set analyses showed that all algorithms were more successful with ensemble models included compared to the models alone. Among the ensemble models created, the ensemble ANN model substantially increased performed best when sued with training and test sets. It was observed from the flood susceptibility maps that the flood areas formed by ensemble models were more distributed than those created by a single machine learning algorithm, and with the help of ensemble models, the distribution of the parameters affecting the floods at the flood points more elucidating. Furthermore, the McNemar test was applied to assess the differences between the predictions of the generated models using test data. It is concluded, in general, that ensemble models created with the ANN and GBT algorithms can help decision-makers in identifying flood susceptibility areas.