In recent years, the advantages of machine learning (ML) have been clearly demonstrated in research on urban water infrastructure (UWI) and has been applied in a wide range of applications. This review critically assesses the current quality of ML implementations in UWI by examining >100 articles from the recent literature, with a particular focus on common pitfalls throughout the development process. Most reviewed articles placed strong emphasis on performance benchmarking but provided limited reporting on key ML methodology implementation and deployment. Only around one third of the reviewed articles documented essential tasks such as feature scaling or automatic hyperparameter optimisation, despite their importance for performance and generalisation. Additionally, fewer than 25% reported explainability and uncertainty quantification techniques, making reported performance gains difficult to explain or operationalise. Furthermore, the lack of standardised documentation makes the extraction of relevant information about the methods, workflows, and steps needed for reproducibility difficult. Together, these issues negatively affect the reproducibility and reduce comparison and trust in the developed approaches. This is particularly important, as trust and confidence in ML-based decisions are key requirements for successful transformation of research into practice. To address these issues, a methodological and reporting checklist is provided to guide the design and integration of ML applications in UWI throughout the development process and to highlight common pitfalls. The review and the checklist can support developers, technicians, and operators in future ML applications, helping to raise awareness on reproducible ML implementations in UWI applications.

This study addresses complex multi-objective optimization challenges in large-scale, real-world water distribution networks (WDNs). The primary objectives are to improve a water quality index (water age) and network resilience by optimizing pipe diameters and network topology as decision variables. The proposed approaches leverage the non-dominated sorting genetic algorithm II (NSGA-II) producing Pareto optimal alternatives for water utility decision-makers. To enhance computational convergence runtime and solution quality of optimization, novel techniques are employed. These include advanced NSGA-II constraint handling, search space reduction, graph theory-based formulation of decision variables, constraints, and objective functions, as well as multi-stage and hydraulic-free optimization strategies. Furthermore, soft constraints are relaxed and integrated into Pareto decision-making spaces to provide a comprehensive, multi-criteria decision-making framework. The approaches are applied to a real case study, and the results demonstrate optimization performance improvements, with efficiency increasing by approximately 20% (in terms of convergence speed). Additionally, water age is reduced by 52% while achieving favorable results in the hydraulic and topological criteria. These findings highlight the effectiveness of the proposed methodologies in addressing WDN optimization challenges.

Water distribution networks (WDNs) are a vital component of urban water infrastructure. They transport water from production sites (sources) to spatially distributed consumers (sinks). Multiobjective optimization procedures are often used to minimize construction costs and at the same time maximize the resilience of such systems, which is usually a very computationally expensive task. Recently, highly efficient approaches based on complex network analysis (CNA) have been developed to solve this task more computationally efficiently. With CNA, very large WDNs can be optimized, considering network topology and demand distribution (using, e.g., demand edge betweenness centrality). However, existing CNA approaches do not consider network topography (i.e., height differences between sources and sinks). Comparing design solutions based on CNA with those found by evolutionary algorithms shows that the least-cost CNA design cannot compete with the latter. In this work, a hybrid approach is developed, where low-cost design CNA solutions are evaluated with a hydraulic solver (Epanet2), and subsequently the demand edge betweenness centrality distribution is iteratively altered for nodes with pressure deficits. This enhanced CNA-based optimization is tested on two different large case studies from the literature and shows promising results (2% cost increase). These solutions were obtained using significantly less computational effort (at least factor 1,000 faster), enabling solving very large WDN optimization problems (>150,000 decision variables).

Water distribution networks (WDNs) are vital parts of the urban infrastructure, and their construction, operation, and maintenance incur major investments. Therefore, many different approaches for optimizing WDNs exist. However, when it comes to large real WDNs, computational time becomes a significant factor, as the possible number of potential solutions grows exponentially. This paper discusses a highly efficient approach for Pareto-optimal design of WDNs based on complex network analysis (CNA). A real WDN with about 4,000 pipes (decision variables) was optimized first using a straightforward evolutionary algorithm approach with two objectives being cost minimization and resilience maximization. By systematically investigating topological features of the obtained Pareto-optimal solutions, insights into optimal networks are generated and a new design approach based on CNA is developed, which outperforms the results of the evolutionary algorithm. The proposed CNA approach is then successfully used to optimize a WDN with the same objectives where the evolutionary algorithm approach is computationally infeasible (semi-real case study with 157,040 decision variables).

The optimization of water networks supports the decision-making process by identifying the optimal trade-off between costs and performance (e.g., resilience and leakage). A major challenge in the domain of water distribution systems (WDSs) is the network (re)design. While the complex nature of WDS has already been explored with complex network analysis (CNA), literature is still lacking a CNA of optimal water networks. Based on a systematic CNA of Pareto-optimal solutions of different WDSs, several graph characteristics are identified, and a newly developed CNA design approach for WDSs is proposed. The results show that obtained designs are comparable with results found by evolutionary optimization, but the CNA approach is applicable for large networks (e.g., 150,000 pipes) with a substantially reduced computational effort (runtime reduction up to 5 orders of magnitude).

Urban water systems face severe challenges such as urbanisation, population growth and climate change. Traditional technical solutions, i.e., pipe-based, grey infrastructure, have a single purpose and are proven to be unsustainable compared to multi-purpose nature-based solutions. Green Infrastructure encompasses on-site stormwater management practices, which, in contrast to the centralised grey infrastructure, are often decentralised. Technologies such as green roofs, walls, trees, infiltration trenches, wetlands, rainwater harvesting and permeable pavements exhibit multi-functionality. They are capable of reducing stormwater runoff, retaining stormwater in the landscape, preserving the natural water balance, enhancing local climate resilience and also delivering ecological, social and community services. Creating multi-functional, multiple-benefit systems, however, also warrants multidisciplinary approaches involving landscape architects, urban planners, engineers and more to successfully create a balance between cities and nature. This Special Issue aims to bridge this multidisciplinary research gap by collecting recent challenges and opportunities from on-site systems up to the watershed scale.

In the urban water cycle, there are different ways of handling stormwater runoff. Traditional systems mainly rely on underground piped, sometimes named 'gray' infrastructure. New and so-called 'green/blue' ambitions aim for treating and conveying the runoff at the surface. Such concepts are mainly based on ground infiltration and temporal storage. In this work a methodology to create and compare different planning alternatives for stormwater handling on their pathways to a desired system state is presented. Investigations are made to assess the system performance and robustness when facing the deeply uncertain spatial and temporal developments in the future urban fabric, including impacts caused by climate change, urbanization and other disruptive events, like shifts in the network layout and interactions of 'gray' and 'green/blue' structures. With the Info-Gap robustness pathway method, three planning alternatives are evaluated to identify critical performance levels at different stages over time. This novel methodology is applied to a real case study problem where a city relocation process takes place during the upcoming decades. In this case study it is shown that hybrid systems including green infrastructures are more robust with respect to future uncertainties, compared to traditional network design.