Date of Graduation
7-2026
Document Type
Dissertation
Degree Name
Doctor of Philosophy in Geosciences (PhD)
Degree Level
Graduate
Department
Geosciences
Advisor/Mentor
Aly, Mohamed
Committee Member
Dumond, Gregory
Second Committee Member
Tullis, Jason
Keywords
geospatial assessment, flash flood susceptibility, Saudi Arabia, Makkah, Analytic Hierarchy Process (AHP), Synthetic Aperture Radar (SAR), Random Forest machine learning
Abstract
Flash floods are among the most serious natural hazards affecting arid and mountainous urban regions. In these environments, short duration intense rainfall can produce rapid runoff across steep slopes, narrow valleys, and surfaces with low infiltration capacity. Makkah, Saudi Arabia, is highly exposed to this hazard because of its rugged terrain, dense urban development, changes to natural valley drainage paths caused by urban expansion, and the presence of sacred sites that receive millions of pilgrims and visitors each year. Reliable flood assessment in Makkah remains challenging because long-term rainfall records, streamflow measurements, detailed flood of past flood events, and mapped flood extents are limited. This dissertation addresses these challenges by applying integrated geospatial approaches to assess and map flash flood susceptibility and inundation in Makkah. The study uses three complementary methods: the Analytic Hierarchy Process (AHP), Random Forest machine learning, and Synthetic Aperture Radar (SAR). In the AHP analysis, six flood conditioning factors were used: elevation, slope, rainfall, distance from drainage network, land use/land cover, and geology. These factors were reclassified, weighted, and integrated in a Geographic Information System GIS environment to produce a flash flood susceptibility map. The final AHP weights showed that elevation, slope, rainfall, and distance from drainage network were the most influential factors. The resulting map classified the study area into low, moderate, and high susceptibility zones, with moderate and high susceptibility covering the largest portions of the area. The Random Forest model used the same conditioning factors together with a historical flood inventory to develop a data-driven susceptibility map. The model showed reliable classification performance and helped identify the most important predictors of flash-flood-prone areas, elevation, rainfall, slope, and distance from drainage network. Sentinel-1 SAR imagery was then used to map selected flood inundation events. SAR preprocessing, change detection, and unsupervised classification were applied to separate flooded and non-flooded areas and estimate inundation extent. Together, the results show that combining expert-based analysis, machine learning, and radar remote sensing provides a practical framework for flood assessment in data limited arid urban environments. This dissertation contributes spatial information that can support flood risk reduction, land use planning, emergency preparedness, and the mitigation strategies in Makkah.
Citation
ALOTAIBI, M. A. (2026). Geospatial Assessment of Flash Flood Susceptibility in Makkah, Saudi Arabia. Graduate Theses and Dissertations Retrieved from https://scholarworks.uark.edu/etd/6428
