Catastrophic floods interrupt the lives of over 40 million U.S. residents every year, killing dozens and causing tremendous damage to homes and businesses.
Predicting floods in time for people to protect their property and their lives is not always easy. In some cases, such as an approaching hurricane, the flooding in flood-prone areas is a near certainty. In other situations, though, it is more difficult to make an informed prediction. There is much more to it than just rainfall. Factors such as soil moisture, tree canopy spread and even solar radiation play a role in the likelihood of significant flooding.
Forecasters have used these observations to inform forecasting of individual events or to improve our understanding of certain processes. However, they have not been integrated into a systematic predictive analytics approach, which would make them easily interpreted and useful to decision makers. Forecasts informed by this data could lead to better decisions about emergency evacuations and road closures.
In some areas of the country, such as Louisiana, the need for accurate flood forecasting is paramount to ensuring public safety before, during, and after a storm hits. As a result, the University of Louisiana at Lafayette (ULL) and CGI have partnered on the Machine Learning-enabled Flood Forecasting Prototype project, a deep learning, data-driven flood forecasting system.
Today’s big data analytics [techniques have] gotten smarter and more powerful. With the support of CGI and in collaboration with the National Science Foundation CVDI program, we are excited to present a ML-enabled Flood Forecasting System. – Dr. Mohamed ElSaadani, Research Professor, University of Louisiana at Lafayette.
With an innovative lab-to-market approach through the public-private partnership, the prototype will transform existing capabilities for flood forecasting and warning systems. It brings the power of machine learning and artificial intelligence (AI) to the myriad environmental observations that various sensors-- including space-based systems, ground radar, field sensors and social media data--are continually collecting.
We are leveraging machine learning and AI to develop new methodologies for data integration and the discovery of complex inter-data relationships. “Today’s big data analytics [techniques have] gotten smarter and more powerful. With the support of CGI and in collaboration with the National Science Foundation Center for Visual and Decision Informatics program, we are excited to present a machine language-enabled flood forecasting system,” said Mohamed ElSaadani, a research professor at ULL.
For example, we have proven the ability to accurately predict levels of soil moisture using a deep learning model that analyzes variables such as precipitation, temperature, vegetation, plant canopy, surface water, and incoming long- and short-wave radiation. This is helpful in flood forecasting because soil moisture has very limited remotely-sensed and ground-based observations.
While our current work is concentrated on flood events, the data analytics approach is scalable and easily transferable to other applications. The methodologies we are developing are mainly for data integration and discovery of complex inter-data relationships that will use remotely-sensed and ground-based observations, as well as the latest machine learning and AI approaches. This includes real-time environmental data being disseminated by entities such as NASA, the National Oceanic and Atmospheric Administration, and the National Weather Service.
In line with President’s Management Agenda (PMA), this public-private partnership between ULL and CGI will provide Louisiana’s local authorities with the ability to warn residents of upcoming flood events faster and with greater accuracy, to take actions to protect homes and businesses where possible, and to ensure that people are on safe ground by the time floodwaters make roads inaccessible.
According to americanrivers.org, every dollar invested in developing flood mitigation solutions yields a five dollar return. The ROI potential is reinforced by the U.S. government’s continued investment in research and development, including allocations for flood mitigation projects—$150 billion dollars annually.
The advances in data science, big data, machine learning, artificial intelligence and other leading edge technologies are harnessed and realized through the PMA and its lab-to-market cross-agency priority goal that leverages the power of the public-private partnership.
The National Science Foundation’s Center for Visual and Decision Informatics program supports the collaborative CGI and ULL effort in developing the Machine Learning-enabled Flood Forecasting Prototype system.
Our commitment to seek opportunities to collaborate with universities, federal laboratories, and other research organizations remains a strategic focus for CGI Federal as we foster innovations. Through public-private-academic partnerships, we are able to bring solutions to the market more quickly—and more importantly, help deliver life-changing solutions that affect communities.
To learn more about CGI Federal's capabilities in AI and machine learning, visit our Intelligent Automation site.