With so many environmental issues arising, how can we even begin trying to solve them all? Perhaps, we can start by asking AI. Researchers at Tohoku University used Artificial Intelligence (AI) to try and solve the deeply complex and multi-faceted environmental issues in today's society. Their study highlights innovative approaches where AI can pinpoint viable action plans, representing the most promising solutions available to us.
"Our study reveals the breakthrough value of technologies such as machine learning in material screening, performance prediction, real-time prediction, global distribution simulation of pollutants, and health risk management," explains Professor Hao Li (WPI-AIMR).
They focused on five major fields: water pollution treatment, air pollution control, solid waste disposal, soil remediation, and environmental health. For example, AI can be used to develop strategies to improve our water treatment techniques or to predict which materials are the most effective at removing pollutants such as greenhouse gases from the air. AI-driven material screening and process optimization can reduce the cost of pollution treatment, improve the efficiency of resource recycling, and improve our immediate surroundings.
"Some of these issues have so many factors that it can be difficult for humans to figure out alone," explains Li, "For example, some pollutants may actually be more or less toxic to humans, depending on their interactions. These predictions are far from simple."
Their analysis provides key support for formulating public health policies and ensuring food and drinking water safety, ultimately helping to build a more sustainable, resilient, healthy society.
The researchers recognize that bottlenecks to wide adoption of this AI-based strategy exist - such as data scarcity, overfitting of small-sample models, and uneven geographical distribution of observational data. Afterall, AI is more refined when it has access to larger databases. The team proposes a shared "Digital Catalysis Platform" which integrates cross-media data processing with the embedding of domain prior knowledge. A platform such as this could help provide a technical framework for the large-scale application of AI in the environmental field.
Therefore, the team plans to build a cross-media environmental database and develop solutions to solve the problem of overfitting with small samples. As the project moves forward, they also plan to collaborate with global research institutions to establish a standardized data collection and sharing platform to promote large-scale application verification of AI in environmental governance.
The findings were published in Environment International on September 12, 2025.