Citation

Avand M, Khazaei M, Ghermezcheshmeh B. Int. J. Disaster Risk Reduct. 2023; 96: e103910.

Copyright

(Copyright © 2023, Elsevier Publishing)

DOI

10.1016/j.ijdrr.2023.103910

PMID

unavailable

Abstract

Watersheds have been heavily affected by natural and human stresses in recent years, and their ability to recover and adapt to changed conditions depends on the resilience of the watersheds. Flood is one of these tensions, and to reduce the damages caused by it, it is necessary to identify vulnerable areas. This study aims is to evaluate the resilience of flood-prone sub-watersheds in the Beshar basin the Kohgiluyeh-Boyerahmad province of Iran. For this purpose, flood risk areas were determined using three machine learning models (MLMs), including random forest (RF), generalized linear model (GLM), and artificial neural network (ANN). Three models were evaluated based on criteria such as the ROC curve and Kappa coefficient, and the most accurate model was selected to identify areas at risk and complete the resilience questionnaire by residents. Social, economic, policy, and infrastructure criteria and 24 important and influential items were used to measure resilience. Different statistical methods were used to analyze the questionnaires and determine the resilience of different sub-basins. The results showed that the RF model (AUC = 0.96) is more accurate than the other two models. The flood risk map also showed that the very low-risk class had the largest area (2722 km2, 86% of the total study area). Also, the resilience results showed a decrease in the mean resilience scores after 2006 compared to before 2006. The results of the spatial changes of the resilience of different sub-watersheds in these two periods showed that in the first period, 4, 9, 13, and 16 sub-watersheds are in the low resilience class and, 10 and 17 sub-watersheds are in the high resilience class. Also; after 2006, 3, 4, 9, and 21 sub-watersheds were placed in the low resilience class and 10, 13, and 14 sub-watersheds were in the high resilience class, which has had changes compared to the previous period.

Language: en

Keywords

Beshar watershed; Flood hazard; Machine learning model; Resilience and vulnerability