Evaluation of pre- and post-fire flood risk by analytical hierarchy process method: a case study for the 2021 wildfires in Bodrum, Turkey

Yilmaz O. S., AKYÜZ D. E., Aksel M., DİKİCİ M., Akgul M. A., Yagci O., ...More

LANDSCAPE AND ECOLOGICAL ENGINEERING, vol.19, no.2, pp.271-288, 2023 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 19 Issue: 2
  • Publication Date: 2023
  • Doi Number: 10.1007/s11355-023-00545-x
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, PASCAL, Periodicals Index Online, Aquatic Science & Fisheries Abstracts (ASFA), BIOSIS, Environment Index, Civil Engineering Abstracts
  • Page Numbers: pp.271-288
  • Istanbul University Affiliated: No


Wildfires are regarded as one of the devastating natural disturbances to natural ecosystems, and threatening the lives of many species. In July 2021, a wildfire took place in the Mediterranean region of Turkey in multiple areas. In Bodrum, a town with high touristic value and attraction, approximately 17,600 hectares of forest have been affected by the wildfire. In this study, the fire-affected areas were determined using an analytical hierarchy process (AHP) and geographical information system (GIS). Rainfall, slope, distance from the stream, pre- and post-fire land use and land cover, elevation, curvature, topographic wetness index, and lithology were selected as the governing variables for the AHP model. The contribution of each variable was determined from the literature. Based on the model, it was found that the area with a very high flood risk increased from 8.6 to 18.4%, implying flood risk in a particular region doubled following the wildfire. Immediately after the forest fire, floods occurred in Mazikoy in the region and its surroundings. The model accuracy was tested by using randomly selected 61 points in and around the flooded area. The model accuracy was quantified by the receiver operating characteristic (ROC) curves method. Pre- and post-fire areas under curve (AUC) values were found 0.925 and 0.933, respectively, which implies that the prediction ability of the model is acceptably accurate. The study revealed that the model could quantify the increased flood risk for vulnerable areas after a forest fire. Such knowledge may aid local authorities in determining the priorities of the precautions that need to be taken after a forest fire.