New flood-mapping tool has the potential to save lives during deadly floods
iStock illustration
Flooding is the most common and most expensive natural disaster in the United States.
Nationally, it costs billions of dollars each year, in addition to the number of lives lost. In 2024 alone, there were 145 fatalities reported due to national floods, up from 79 the year before.
When floods strike, emergency responders depend on satellite maps to understand where water has spread and where help is needed most. But during major flood events, which are often accompanied by heavy cloud coverage, satellite imagery can be obscured, creating dangerous blind spots when decisions must be made quickly.
Researchers at Arizona State University have created a way to help.
Wenwen Li, a professor in the School of Geographical Sciences and Urban Planning, and her team have developed an AI-powered system called SMAGNet, a flood-mapping tool that will eventually help emergency responders act faster, potentially saving lives and property during deadly storms.
The model combines radar and optical satellite data to generate clearer flood maps, even when clouds block traditional imagery, and is designed to function under the imperfect, data-limited conditions that take place during real-world disasters.
“Before using AI, flood mapping typically relied on a single data source and rule-based methods to identify flooded areas,” Li said.
ASU News spoke with Li and graduate assistant Hyunho Lee about how the new flood-mapping tool works and how it can help save lives when the next flood strikes.
Note: Answers have been edited for length and/or clarity.
Question: What inspired your interest in flood-mapping research? Was there a moment — a devastating storm perhaps — that first pulled you into flood-mapping research?
WenWen Li: What first pulled me into flood research was the news coverage of Hurricane Katrina in 2005. Thousands of people lost their lives because of massive flooding. I had just started graduate school at the time, and I was deeply saddened by the devastating impact of the disaster. That was when a seed was planted — to use my geospatial knowledge and skills to help vulnerable communities and save lives.
Q: How critical are accurate flood maps during active rescue and evacuation efforts?
Hyunho Lee: Accurate flood mapping is critical during active rescue and evacuation efforts, as it allows emergency teams to efficiently allocate resources and prioritize the most affected areas.
While ground observations are limited in spatial coverage, satellite imagery can monitor floods over large areas in near-real-time, offering crucial situational awareness for decision-making.
Q: What types of flood-mapping tools are most commonly used today, and where are they lacking?
Li: Today, flood mapping most commonly relies on satellite-based methods, physics-based hydraulic models and simple digital elevation model-based approaches. These tools are widely used because they are physically grounded and supported by established workflows. However, they often struggle with scalability, rapid real-time forecasting and transferring models across regions without recalibration. In comparison, AI-based approaches can integrate multiple data sources, learn complex nonlinear patterns and generate faster, large-scale predictions.
Q: Your model focuses on using AI and satellite imagery to map floodwater. How is a tool like this created?
Lee: Creating this tool starts with gathering SAR and optical satellite imagery from past flood events as training data. We then train an AI model to recognize flooded areas by learning patterns from these images. The model is validated and tested to measure its accuracy.
Finally, the AI model can rapidly analyze new satellite feeds to generate timely, large-scale maps that provide critical information for disaster management. The model accuracy is ensured by intelligently extracting complementary information from different data sources under varying weather and data availability conditions using our proposed multimodal deep learning strategy.
Q: SMAGNet integrates multiple types of satellite data. What motivated you to take this approach?
Lee: Many flood-mapping studies use both synthetic sperture radar (SAR) and multispectral satellite imagery because they offer complementary strengths — SAR can see through clouds and operate in all weather, while multispectral data provides detailed information about water and land surfaces. With more satellites now in orbit, combining these data sources can significantly improve flood detection accuracy. However, in real disaster situations, different data types are not always available at the same time due to satellite schedules and coverage limits. This motivated our study to develop AI models that can adapt to varying data availability while still producing reliable flood maps.
Q: When will it be available for use?
Li: Our AI model has already been made open source on GitHub to support broader applications and reproducibility. We are also working diligently to expand the model for deployment in real-world environments by addressing challenges such as model transfer across geographic regions and ensuring accuracy in data-scarce areas, a common challenge in disaster-mapping scenarios.
In addition, we are actively collaborating with government agencies and industry partners, including NASA and IBM Research, to advance the next generation of AI models for disaster mapping.
More Environment and sustainability
Rooting for change: Climate-focused pep rally brings Maryvale community together
This past weekend in the West Phoenix community of Maryvale, a brass band blared, Maryvale High School cheerleaders and a theater troupe fired up the crowd, and a pair of young trees rolled through…
Science of seismic proportions
Arizona’s deserts might not be situated on a tectonic plate boundary where geographical activity can cause earthquakes, but that doesn’t make the state immune from the disturbance of tremors.It has…
Colorado River basin under increasing strain; will cutting back on water use be enough to help?
As drought conditions intensify across the American West and the impact of climate change accelerates, cities such as Phoenix, Denver and Las Vegas may face a sobering reality. While it…