Flash Flood Hazard Assessment using Hydrological and Machine Learning Models with Multi-Satellite-Based Precipitation and d4PDF Climate data sets in Saudi Arabia

Abstract

Floods are among the most prevalent natural hazards globally, causing significant damage and loss of life. The primary tool for flood risk assessment and damage mitigation is hydrologic early warning systems, which predict flood events using rainfall observations from ground stations. However, challenges such as data scarcity, spatial disruption, network density issues, and measurement errors in ground stations have necessitated using satellite-based rainfall products to address these limitations. Satellite-based precipitation estimates (SbPEs) and reanalysis precipitation datasets (RPDs) have become increasingly accurate, offering improved spatial and temporal resolution beneficial for hydrological applications and rainfall-runoff modeling. The challenges posed by climate change, particularly the increasing frequency of extreme events like floods, further complicate flood risk assessments, especially in regions like Saudi Arabia. The main challenge in assessing rainfall and runoff in Saudi Arabia is the limited availability and accuracy of observational data (1) rainfall station and monitoring that can be measured with reliable and less uncertain data with low maintenance cost. In addition, the extreme climatic variability under climate change impact with dam management and mitigation measures in KSA (2) the identification of suitable locations for new dams where they can effectively serve multiple purposes, including groundwater recharge, flood control, water storage, and protection. Therefore, the main objectives of this research thesis are to (1) Examine both the hydrological model and Machine learning to develop the flash flood susceptibility maps, (2) Deep understand the Spatiotemporal climatic variability of extreme storms, (3) To evaluate SBP rainfall data and Re-analysis Rainfall data over KSA, and (4) To develop Flood Hazard Map based on SBP and hydrological modeling including distribution of dams locations and purposes. This thesis consists of seven chapters that explain the flood risk assessment from different sources of rainfall data and culminate in the analysis of extreme rainfall in the present and future, using hydrological models and machine learning techniques to find the hazard mapping over Saudi Arabia. The introduction and the literature review about the rainfall-runoff processing under climate change in Saudi Arabia were presented [Ch.1, 2]. Afterward, a set of investigations was conducted through various analyses and modeling to find out the flash flood hazard and susceptibility map [Ch.3, 4, 5, 6]. Finally, the recommendation and future prospects were proposed with research conclusions in [Ch.7].