Haddad et al. (2025) From ancient to recent floods: advances in flood susceptibility modeling and vulnerability, Makkah, Saudi Arabia

Identification

• Journal: Natural Hazards
• Year: 2025
• Date: 2025-12-22
• Authors: Bosy A. El Haddad, Ahmed M. Youssef, Hamid Reza Pourghasemi
• DOI: 10.1007/s11069-025-07871-3

Research Groups

• Geology Department, Faculty of Science, Sohag University, Sohag, Egypt
• Department of Soil Science, College of Agriculture, Shiraz University, Shiraz, Iran

Short Summary

This study develops flood susceptibility and vulnerability maps for Makkah City, Saudi Arabia, by testing four machine learning algorithms (Logistic Regression, Support Vector Machine, Extreme Gradient Boosting, and an ensemble model) with 12 flood-conditioning factors, finding the ensemble and XGB models to be most accurate and highlighting significant urban encroachment into high-risk flood zones.

Objective

• To develop a flood susceptibility map for Makkah City to mitigate the destructive effects of floods.
• To test the performance of four machine learning algorithms (Logistic Regression, Support Vector Machine, Extreme Gradient Boosting, and an ensemble model) for flood susceptibility mapping.
• To produce flood vulnerability maps by superimposing anthropogenic activities (urban and infrastructural) over the best susceptibility model.

Study Configuration

• Spatial Scale: Makkah City and its surrounding wadi catchments (Wadi Nu’man, Wadi Uranah, Wadi Fatimah) within Makkah Province, Saudi Arabia. The Makkah metropolitan area has a drainage area of 6637 square kilometers. Elevations range from 82 meters to over 1603 meters above sea level.
• Temporal Scale:
  • Historical flood records: 638 to 2024 (14 centuries).
  • Rainfall data: 1966 to 2023.
  • Urban expansion analysis: Landsat images from 1984, 1990, 1992, 1995, 2000, 2004, 2010, 2013, 2016, 2020, and 2024.
  • ALOS Global Digital Surface Model (AW3D30): 2006 to 2011.
  • UAV images: 2024.

Methodology and Data

• Models used:
  • Logistic Regression (LR)
  • Support Vector Machine (SVM)
  • Extreme Gradient Boosting (XGB)
  • Ensemble model (stacking of LR, SVM, and XGB)
• Data sources:
  • Geospatial Data: ALOS Global Digital Surface Model (AW3D30) with 30 m spatial resolution; Landsat 5, 7, and 8 images (1984-2024) from USGS; Google Earth images; OpenStreetMap (for road extraction); Geologic maps (scale 1:250,000).
  • Hydrological/Meteorological Data: Rainfall data from 16 rain gauges (1966-2023) from the National Center for Meteorology (NCM) and the Ministry for the Environment, Agriculture, and Water (MEAW).
  • Flood Inventory Data: 556 flood and 556 non-flood datapoints; historical flood records (638-2024) from Makkah municipality and General Directorate of Civil Defense; field observations and Unmanned Aerial Vehicle (UAV) images (2024) of inundated areas.
  • Flood Conditioning Factors (12 factors): Elevation, distance to stream, slope, aspect, lithology, landforms, relative slope position, topographic wetness index (TWI), slope length, rainfall, plan curvature, and profile curvature.
  • Software: ArcGIS Pro 3.3, SPSS v.26, R statistical software (xgboost package), QGIS.

Main Results

• Multicollinearity analysis confirmed no collinearity issues among the 12 flood-conditioning factors (Variance Inflation Factor values below 5.0, Tolerance above 0.1).
• The XGB algorithm and the ensemble model demonstrated the highest predictive performance, with Area Under the Curve values of 0.947 and 0.963, respectively, indicating excellent predictive capability.
• Logistic Regression and Support Vector Machine models showed good predictive capability, with Area Under the Curve values of 0.828 and 0.883, respectively.
• Variable importance analysis identified distance from streams, slope angle, topographic wetness index, rainfall, and landforms as the most significant factors affecting flood prediction (each above 10% importance).
• More than 15% (XGB) and 20% (ensemble) of the study area are classified as highly susceptible to flooding, primarily located downstream of Wadi Fatimah, Wadi Nu’man, and Wadi Uranah.
• Urban and road encroachment into high and very high flood-prone zones significantly increased from 1984 to 2024: urban areas increased from 30.8% to 70.9%, and roads from 58.3% to 68.1%.

Contributions

• Provided comprehensive flood susceptibility and vulnerability mapping for Makkah City using an integrated approach of machine learning algorithms and geospatial tools.
• Identified the most influential flood-conditioning factors specific to the Makkah region.
• Quantified the significant increase in urban and infrastructural encroachment into high-risk flood zones over four decades (1984-2024).
• Developed a robust flood susceptibility model (ensemble and XGB) that can be integrated into decision-making for sustainable urban planning and flood risk management in data-scarce arid regions.
• Proposed specific mitigation strategies and infrastructure recommendations for Makkah's wadis to enhance flood resilience.

Funding

• No Fund.

Citation

@article{Haddad2025From,
  author = {Haddad, Bosy A. El and Youssef, Ahmed M. and Pourghasemi, Hamid Reza},
  title = {From ancient to recent floods: advances in flood susceptibility modeling and vulnerability, Makkah, Saudi Arabia},
  journal = {Natural Hazards},
  year = {2025},
  doi = {10.1007/s11069-025-07871-3},
  url = {https://doi.org/10.1007/s11069-025-07871-3}
}