Highlights from the Water-Futures meeting in Athens

For this event, the KWR delegation, led by our Principal Investigator Dragan Savić with the support of Lydia Vamvakeridou-Lyroudia, gave four presentations, reporting on the progress of our PhD students Lydia Tsiami, Konstantinos Glynis, Christos Michalopoulos and Dennis Zanutto.

Research highlights

During the first session on transitioning urban water systems under deep uncertainties, Lydia kicked off the meeting with her work on Reinforcement Learning for the phased design of the water infrastructure, showing how this approach enhances flexibility in planning these systems. Dennis followed up by emphasizing the importance of operations and management alongside infrastructure planning and showing his recent presentation at EGU (read more about this event here). After that, Konstantinos reported on his experiments on growing biofilms inside pipes, and Christos presented his PhD plan for using generative AI to produce future development scenarios to test adaptability and flexibility of network designs.

Effectively planning water distribution networks for the long-term is a challenging task for water utilities due to deep uncertainty. Therefore, incorporating more adaptive approaches, such as Reinforcement Learning, into the decision-making process is necessary. Preliminary benchmark tests showed the potential of this methodology.

Long-term planning of water distribution systems is challenged by deep uncertainties. Daily operations can help in dealing with them but their impact on the infrastructure design is has been poorly investigated. Preliminary results show increased performance when design and operations work together.

Biofilm accumulation in water distribution networks is a public health concern. Monitoring biofilm is crucial for appropriate management of networks. Experiments with real water pipes and drinking water quality taking place at KWR aim to address this niche.

Deep uncertainties are challenging effective long-term planning of drinking water distribution systems. Scenarios generation is helpful in exploring possible future system states for better planning. Generative AI can help connect the dots.

In the other session, we saw great works from our partners on machine learning, the concept of fairness, public engagement, and more. The common topic of drinking water makes these seemingly unrelated concepts work together harmoniously.

Insights from leading experts: external talks

Three external talks have also enriched this meeting. We had the pleasure of following a private lecture by Prof. Costis Daskalakis, Professor of Computer Science at the renowned MIT (Massachusetts Institute of Technology, USA). With his presentation, he gave us a glimpse of the context and the challenges of training multiple competing Reinforcement Learning agents in the same environment and how game theory plays a crucial, yet tricky, role in these cutting-edge research problems. As the role of RL becomes more and more prominent in water systems (as Lydia’s contribution showed), seeing this work was not only incredibly informative but also very pertinent to our project. Water Futures aims to be at the forefront of technological advancements in all fields to design the next generation of drinking water systems.

Prof. Patrick Reed from Cornell (USA) opened the afternoon session and kept us engaged with a presentation on the planning and operations of water supply systems. There is no need to stress the importance of this presentation as it touches on precisely the same topics addressed by Water Futures, with particular relevance and synergy to Dennis’s work. For example, multiple objectives and the interaction between short-term operation and long-term infrastructure interventions are common traits between these types of problems.

The third invited talk was by Prof. Allison Lassiter, who is an Assistant Professor of City and Regional Planning at the University of Pennsylvania (USA). She introduced us to the fragmented panorama of the water utilities in the USA, which is very different from the centralized system of the Netherlands and, therefore, subject to quite a few different challenges, including the lack of capacity to deal with substantial challenges by many small utilities. An example is the absence of funding for the small water utilities (less than 500 people served) and the consequent limited planning ahead. A commitment of the Water Futures project is to apply some of the results of the research to a real case study, and Prof. Lassiter’s presentation was particularly relevant in this context. She introduced us to the Tampa Bay water system (Florida, USA) and its unique challenges, which make it a suitable option to enrich our portfolio of case studies, currently focusing on the Netherlands and Cyprus. Prof. Lassiter spent some time with Water Futures researchers in Greece and Cyprus as a result of the collaborative agreement between the USA National Science Foundation (the funder of her research project) and the European Research Council (the funder of the Water  Futures project) and came back to share her experiences and research challenges.

Collaboration and achievements: looking ahead

Finally, this meeting marks the completion of our first output in which all four groups came together to provide a solution for the Battle of Water Demand Forecasting competition that was recently submitted. We presented the challenges and results of this joint work to colleagues who were not involved firsthand. The idea to work together and participate in this competition as a Water-Futures team originated in the previous meeting in Bielefeld. It is the perfect example to highlight the importance of synergistic work and the value of these recurrent meetings in developing European projects. Thanks to the added benefit of the in-person discussions, ideas are generated and collaborations are initiated quickly.

The results of the Battle will also be presented at the upcoming WDSA/CCWI conference in Ferrara next July. This event will see strong participation from KWR, with our Water Futures PhD presentations alongside those of many other colleagues and the kenote talk by Mirjam Blokker.

Once again, a thank you to Athens University of Business and Economics for hosting and in particular Ae4ria for providing the photos and organizing the meeting.