AI and optimisation help utilities reduce flooding and sewer overflows by identifying where nature-based solutions deliver the greatest impact. 

Heavy rainfall is already causing sewer overflows and local flooding in many cities, leading to environmental impacts and costly damage. The Opti-SITE research project, funded by Interreg Öresund-Kattegat-Skagerrak, brings together utilities, universities and technology providers from Denmark and Sweden to develop new tools for smarter stormwater management.

 

By combining machine learning, optimisation and real-time radar data, the project helps utilities, consultants and decision-makers identify where to place solutions such as rain gardens, detention basins and ponds, and how to optimise existing and future sewer systems — both in long-term planning and during extreme events.

 

Challenge

Cities are experiencing more frequent rainfall events that exceed the capacity of existing sewer systems. This leads to combined sewer overflows, where untreated wastewater is discharged into the environment. Furthermore, stormwater quality and pollutants and their effect on receiving waters are getting increased attention, and utilities are facing stricter regulation.

 

Replacing or significantly expanding sewer infrastructure is often not feasible due to high costs and space constraints. As a result, utilities must optimise existing systems and carefully prioritise where to implement additional measures such as nature-based solutions.

 

However, it remains difficult to determine where these interventions will have the greatest effect across complex urban catchments. At the same time, most systems lack the ability to predict and respond to critical events before they happen, limiting opportunities to reduce damage and pollution.

 

Solution

The partners in Opti-SITE are developing a new generation of decision-support tools by combining machine learning with methods that balance factors such as cost, performance and environmental impact.

 

The project identifies key characteristics of areas where nature-based solutions are most effective and uses these to train advanced models, including graph neural networks. This makes it possible to evaluate a wide range of infrastructure scenarios and identify the optimal placement and sizing of solutions based on local conditions.

 

In parallel, Opti-SITE integrates improved radar rainfall data with fast, data-driven sewer models to enable nowcasting and early warning of combined sewer overflows. This allows utilities to predict where and when critical events will occur and take action, for example by adjusting system operation, closing vulnerable areas or reducing environmental impacts before discharges happen.

 

By linking planning and operational tools in one framework, the project enables a more proactive and coordinated approach to stormwater management.

 

Expected results

The project has already delivered a prototype machine learning model, including benchmarking and technical requirements, as well as workflows for generating training data and assessing stormwater quality.

 

Next steps include a finalised optimisation tool, integration of radar data into hydrodynamic models and workflows for implementing nowcasting systems that support real-time decision-making.

 

The solutions are being developed and tested in collaboration with utilities such as HOFOR and NSVA, including case areas in Copenhagen and Landskrona, ensuring that the research is grounded in real-world conditions.

 

The results are expected to be integrated into DHI’s CityFlow Live platform, enabling utilities, consultants and decision-makers to directly apply the research outcomes to reduce sewer overload, minimise discharges of untreated wastewater and improve climate resilience in practice.