NAIROBI, Kenya — Kenya is among several countries grappling with severe flooding that has caused deaths, displacement, and widespread destruction of property in recent weeks.
According to the National Police Service, at least 62 people have died following torrential rains and flash floods across the country. The highest number of fatalities has been recorded in Nairobi with 33 deaths, followed by 17 in the Eastern region and seven in the Rift Valley.
Authorities have urged residents to remain vigilant and follow early weather alerts issued by the Kenya Meteorological Department as heavy rains continue to pound several regions.
AI-powered flood prediction tool
As governments and humanitarian agencies respond to flooding disasters worldwide, researchers at Google have developed a flash flood forecasting tool designed to predict the risk of urban flooding up to 24 hours in advance.
The system, available through Google’s Flood Hub, uses satellite data and artificial intelligence to estimate flood risks and map flooded areas.
According to Google, the platform generates Synthetic Aperture Radar (SAR) images from satellite data and processes them using algorithms that calculate the extent of flooded areas.
The technology company said it adopted an AI-driven approach by integrating mapping capabilities from Google Maps and machine learning tools from Gemini to estimate areas likely to be affected by flash floods.
Analysing decades of flood data
Google also introduced a new methodology known as Groundsource, which analyses historical public reports on flooding.
Using this approach, the company examined data spanning several decades and identified more than 2.6 million historical flood events across over 150 countries.
“It then used Google Maps to determine precise geographic boundaries for each event to create a dataset focused on flash floods,” the company said.
Limitations of satellite-based predictions
Despite the technological advancements, Google acknowledged that the system may not always be fully accurate.
The company said satellite imagery can sometimes make it difficult to distinguish between flooded and non-flooded areas.
“It’s difficult to recognise flooded and non-flooded areas from a satellite image, so there may sometimes be errors,” Google said.
In some instances, flood maps may also be unavailable if satellite imagery is missing or if available images fail to generate a qualified map.
Google explained that satellite data used for flood monitoring can take several hours to enter its systems, and in some cases, several days before a usable image is generated.
If the available satellite image becomes older than 72 hours, the platform automatically stops displaying flood maps until updated imagery is obtained.
