Flood Susceptibility Mapping in Ghare Ghom watershed
Aims: Floods are among the primary natural hazards that cause economic and social damage. The Ghare Ghom watershed in Razavi Khorasan province is prone to dangerous floods due to its climatic conditions, topography, and physiographic features. Most rivers in the basin are seasonal, leading to occasi...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | fas |
| Published: |
Hakim Sabzevari University
2025-08-01
|
| Series: | مطالعات جغرافیایی مناطق خشک |
| Subjects: | |
| Online Access: | https://jargs.hsu.ac.ir/article_208145_382cd95441c4cf364e467932224e4f38.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Aims: Floods are among the primary natural hazards that cause economic and social damage. The Ghare Ghom watershed in Razavi Khorasan province is prone to dangerous floods due to its climatic conditions, topography, and physiographic features. Most rivers in the basin are seasonal, leading to occasional severe flooding. This study assessed flood susceptibility mapping in this region.Materials & Methods: Machine learning models, including the Random Forest (RF) and Classification and Regression Tree (CART), were used for flood susceptibility mapping. Of the 117 recorded flood events, 70% were used for training and 30% for validation. Influential factors included land use, slope degree, geology, distance from river, digital elevation model, slope direction, geomorphology, soil type, plan and profile curvature, rainfall, and the Topographic Wetness Index (TWI). Model performance was evaluated using the Area Under the ROC Curve (AUC) and the Tolerance Index.Findings: The RF model results show that distance from the river, rainfall, and altitude had the greatest impact on flood sensitivity. In the CART model, altitude, distance from the river, and rainfall were the most influential factors. The CART model classified the area into very low, low, medium, high, and very high susceptibility classes as 9%, 29.8%, 21.9%, 32.4%, and 6.9%, respectively. For the RF model, these values were 11.48%, 24.8%, 28.7%, 24%, and 11%.Conclusion: The Area Under the Curve (AUC) was 0.91 for the CART model and 0.87 for the RF model. The prediction rate was 0.88 for CART and 0.83 for RF. These results indicate that the CART model performed better than the RF model in predicting flood susceptibility.Innovation: Flood susceptibility mapping and identifying influential factors using machine learning are key contributions of this study. The results support planners and policymakers in managing flood risks and reducing future economic and social losses in the region. |
|---|---|
| ISSN: | 2228-7167 2981-1910 |