Applications of Machine Learning and Geospatial Technology to Assessing Urban Flood Susceptibility: A Case Study of Wadi Hanifa, West of Riyadh City

In many parts of the world, cities are growing faster than the land beneath them can handle. When heavy rain falls on a landscape covered in concrete and asphalt, the water cannot soak into the ground as it once did. Instead, it rushes across the surface, gathering speed and volume until it overwhelms streets and basements. This is the essence of urban flooding, a crisis that is becoming more frequent as climate change brings more intense storms and rapid construction seals off natural drainage. To protect people and property, city planners need to know exactly where water will pool and how deep it might get. For decades, experts have relied on maps and historical records to guess these risks, but a new approach is changing the game. By combining satellite eyes that can see through clouds with powerful computer programs that learn from patterns, scientists can now predict flood risks with a level of precision that was previously impossible. This shift allows communities to move from reacting to disasters to preventing them.

In the western part of Riyadh, Saudi Arabia, a researcher applied this modern toolkit to a specific challenge: the Wadi Hanifa basin. This valley, which cuts through the city's western and central districts, is a natural channel for water flowing from the nearby mountains. While it has historically supported agriculture and settlement, the rapid expansion of Riyadh has turned much of this valley into a dense urban area. The city's growth has created a situation where heavy rains, which are rare but intense, can quickly turn into dangerous flash floods. The researcher set out to create a detailed map showing exactly which neighborhoods are most likely to be inundated. They did not rely on guesswork or simple rules of thumb. Instead, they built a digital model of the entire basin, feeding it a vast array of information gathered from space and the ground.

The researcher gathered data from five different sources to build a complete picture of the landscape. They used satellite images to see the types of surfaces covering the ground, distinguishing between green vegetation, bare soil, and the hard, impermeable surfaces of buildings and roads. They pulled in radar images that can detect water even when the sky is cloudy, ensuring they could see the actual extent of past floods. They also incorporated a detailed 3D map of the terrain to understand the slope and shape of the land, along with decades of rainfall records to see where the heaviest storms have hit. Finally, they mapped the road networks, which can act as either barriers or channels for water. With all this information assembled, they turned to a machine learning algorithm, a type of computer program that learns by example. They showed the computer thousands of examples of areas that had flooded in the past and thousands of examples of areas that had stayed dry. The computer analyzed these examples to find the hidden patterns that connect the landscape to the water.

The computer learned that the risk of flooding in this region is driven by a few key factors. The most important factor turned out to be the density of the drainage network. Areas with a high concentration of streams and channels are more likely to flood because the water concentrates there, especially during heavy rains. The second most critical factor is the type of land cover. Neighborhoods with many buildings and paved roads, where water cannot soak in, face a much higher risk than areas with more soil or vegetation. Rainfall intensity was the third major driver, confirming that the sheer volume of water falling in a short time is a primary cause of the problem. The computer also learned that the roughness of the terrain matters; smoother, flatter areas allow water to spread out and pool, while rougher ground can slow it down. Other factors, such as the direction the land slopes or its exact elevation, played a role but were less decisive than the drainage and land use patterns.

Once the computer had learned these patterns, it applied them to the entire Wadi Hanifa basin to predict flood risks for every single spot on the map. The results were strikingly accurate. When the researcher tested the model against real-world data, it correctly identified flooded areas 90 percent of the time. The model successfully distinguished between places that would flood and those that would not, proving that the combination of satellite data and machine learning could capture the complex reality of urban flooding. The resulting map divides the basin into five categories of risk, ranging from very low to very high. The analysis revealed that about 6.1 percent of the basin is at very high risk, with another 14.2 percent facing high risk. These dangerous zones are not scattered randomly; they cluster in specific, densely populated neighborhoods.

The map identified several communities that require immediate attention. Areas such as Center Uraija, Uraija, Derab, Ukaz, Marwwah, Uraid, Twaeeq-Etaigah, and Hejrat Wadi Leban were flagged as having a very high susceptibility to urban flooding. These neighborhoods sit in low-lying areas where water naturally accumulates, and they are surrounded by extensive networks of roads and buildings that prevent water from draining away. The study highlights that these are not just theoretical risks; the model's predictions align with the actual locations where floods have occurred. For instance, when a major flood event happened in April 2026, the vast majority of the observed flood points fell exactly within the zones the model had marked as high or very high risk. This validation gives city officials a powerful tool for decision-making. They can now see exactly where to reinforce drainage systems, where to restrict new construction, and where to prioritize emergency response plans.

This research demonstrates that the future of flood management lies in integrating advanced technology with local knowledge. By using machine learning to analyze the complex interplay of rain, terrain, and human development, scientists can provide clear, actionable insights to protect growing cities. The study does not claim to have solved the problem of flooding forever, but it has provided a reliable method for identifying the most vulnerable spots. For the residents of Riyadh and other rapidly urbanizing cities, this means a shift from hoping for the best to preparing for the worst with confidence. The map serves as a guide, showing where the water will go so that people can build better, plan smarter, and stay safe.