Humans rely on water. Yet almost one third of the world population lacks access to safely managed drinking water services, while over 50% cannot rely on adequate sanitation. Floods and other water-related disasters account for 70% of all deaths related to natural disasters. Furthermore, urban water systems are facing growing pressure from climate change and increasing demographics, forcing cities to devise new approaches to ensure water supply, sanitation and flood risk management.
At the same time digitalisation is taking over the water sector by storm. In the future, we will rely more and more on data, as sensors are becoming cheaper. Internet of Things devices and remote sensing are bringing ‘big data’ into the water sector.
To leverage big data at best, urban water systems require AI models that can account for their underlying networked structure. Unfortunately, traditional AI techniques, such as those used in computer vision, speech recognition, and language processing, are largely unsuitable for modeling the physical processes in water networks or those governing flooding events.
A mathematical solution to this problem might lie in Graph Neural Networks (GNNs), an extension of Deep Learning to non-Euclidean data, such as graphs. In the last five years, the scientific community has increased efforts to devise GNNs solutions, and obtained promising results in chemistry, biology, finance, and social sciences.
At AidroLab, we believe GNNs can be successfully applied to urban water systems as well. The methods and tools developed by the lab will enhance the adaptability and resilience of urban water systems, support effective flood control and help decision makers and first responders.