*2.6. New Design Concepts and Strategies*

The availability of new ubiquitous data, advanced analytics and more integrated modeling frameworks is allowing the sector to perform more realistic stress-tests of water infrastructure (in its

physical and cyber-physical sense) to help improve its performance under uncertainty. This activity is currently pushing the discipline's methodological boundaries into developing and applying novel design concepts driven to a large extent by cities worldwide demanding realistic risk managemen<sup>t</sup> under uncertainty within a context of limited new investments (see for example the 100 Resilient Cities network supported by the Rockefeller Foundation [86]). These e fforts are, recently, centered mostly around the challenging concept of resilience and the development of methods, metrics and tools to assess the resilience of urban water systems. Notable examples include models and tools developed by Irwin et al. [87], Butler et al. [88], Klise et al. [89], Makropoulos et al. [8], Kong et al. [90] as well as Sweetapple et al. [91]). Although a discussion on resilience per se is outside the scope of this paper, we note that this growing body of work, focusing on the highly interdisciplinary and multi-stakeholder context of resilience [92] is an important manifestation of the sociotechnical nature of hydroinformatics. The need to understand resilience emphasizes the role of hydroinformatics as an interface between science and policy, between water systems and urban processes as well as between technology, society and the environment.

## **3. Sky Is (Not) the Limit**

This overview of some of the most exciting developments in hydroinformatics today, may give the impression that most of the important tasks are behind us. This, however, could not be further from the truth. As the discipline is, by definition, linked to and influenced by developments in the dynamically evolving IT sector, with every new development come new challenges and also new opportunities. Although the details of what can happen next are by virtue of this dynamic evolution, hard to predict, some of the most important trends are already visible. In an e ffort to summarise these future trends, four activity lines towards a hydroinformatics roadmap have been proposed below:

### *3.1. Tapping into the New Data Landscape*

The proliferation of smart systems (including developments in the smart city and more generally the IoT arena) mean that data become more ubiquitous—although work on novel water quality sensors is still needed (see ideas on using graphene for heavy metal detection [93]). However, as more data from di fferent sources become available the issue of standardization becomes vital. This is because standardization allows the pulling together and combined exploitation of data coming from di fferent sources and di fferent data providers, both within a utility but also potentially across multiple utilities, reaching the critical mass of data required to categorize water data as big data and, in turn, unlock the true potential of big data analytics. As such, data standardization, in terms, for example, of metadata, standardized markup languages (like the Open Geospatial Consortium's (OGC) WaterML [94], controlled vocabularies and ontologies [95–97] inevitably play a key role in bringing information and analytics together. Due to their importance in an IoT and related telecommunications contexts, the most successful of these standardization e fforts will probably not be initiated within the water domain per se, but rather within smart city, smart home and smart industry contexts, growing towards water, energy and other utility sectors. A case in point is the work by the European Telecommunications Standards Institute (ETSI) and its Smart Appliance REFerence (SAREF) ontology [98], which is currently being expanded [99] towards energy and water, with obvious implications for smart water meters, smart(er) water consuming devices and domestic water demand forecasting and management. Another important development in this field, worth highlighting is FIWARE [100], a curated framework of open source platform components that aims to accelerate the development of smart solutions, including transport, energy, as well as more integrative smart city solutions. FIWARE has already been used to develop interesting examples of interoperability for smart agricultural water managemen<sup>t</sup> [101] and is now expanding [102] also towards urban water managemen<sup>t</sup> at di fferent scales. Data quality control and validation (potentially in a distributed way, closer to the data collection itself, see for example developments in edge analytics [103]) and improvement of data access (including data sharing and open data [104]) is also expected to be at the heart of the next steps in hydroinformatics.

With this critical milestone completed, the industry may be able to exploit new developments that allow the industry to ge<sup>t</sup> new insights out of large, heterogeneous databases and leverage progress on AI, such as deep learning [105], from the ICT sector, to extract information, develop more accurate forecasts and o ffer customized services to end users. New opportunities a fforded by leveraging the power of AI on larger (and more real time) water datasets, include discovering new causal relationships from data already collected to improve predictive ability, e.g., in infrastructure maintenance, water demand managemen<sup>t</sup> or emergency response. It may also allow for progress into data assimilation techniques that couple models to field data in real time. Field data from di fferent sources and with di fferent uncertainties is expected to be used in combination with models, thus greatly increasing current abilities for pro-active managemen<sup>t</sup> of water systems. This new data may also increasingly come from the customer/citizen side, where data crowd-sourcing tools will play an increasing role in collecting real time information [25] as well as in gauging public opinion towards water relevant issues (e.g., water reuse attitudes mined from micro-blogs [106]). These (significantly increased) data streams may range from data collected by smartphone embedded sensors, to information posted on social media, to data collected by, soon to be available, autonomous vehicles—cross referenced and linked to open environmental data, utility sensors and remote sensed information from new satellite networks (like NASA's Surface Water and Ocean Topography (SWOT) mission scheduled to start by 2021 [107].

### *3.2. Getting More Out of Existing Models*

This activity line, is expected to provide the sector with more advanced optimization (including smart model calibration under uncertainty and noise), new ways of model integration (with databases and other models) as well as with real time data (including IoT sensors) to form digital twins of utilities. The concept of digital twins, where the data from the IoT sensors are seamlessly linked with asset managemen<sup>t</sup> information and both support and are supported by models of the system's operations, recalibrated and updated in real time, across the complete value chain from water resources to customers, is expected to become possible in the near future. This ambition, of a complete integrated digital picture of a water utility may appear far-fetched at this time, but is a future in the making, judging from the interest and investment already underway in forward looking cities, such as Amsterdam [108] and its water utility (Waternet). Necessarily, this process shifts online much of the computing infrastructure for water utilities, with cloud computing for water services and software-as-service becoming the norm. This trend, however, is not without its challenges as is discussed in the following sections.

### *3.3. Planning for More Resilient (Cyber-Physical) Systems and Services*

Armed with new data and models, the sector may also work more on model integration and higher abstraction level modelling/model coupling, where whole system strategic models—potentially linked to digital twins—can be used as real-time control, forecasting and scenario planning tools in a collaborative and inclusive way.

This direct coupling between the physical system and related infrastructure and the controlling cyber layer (from sensors to models to actuators) is expected to a fford new opportunities for increased efficiency of water infrastructures throughout their lifetime, from design to building to operating. It would allow, for example, their real time control, with data from multiple sensors being continuously integrated within living models of the physical environment and the infrastructure. Furthermore, it would enable moving significant parts of these calculations to the edge [103], enabling precise and pro-active actuation of pumps, valves, sluice gates, for applications, such as flood forecasting and control [109,110], managing combined sewer overflows [111] and urban water managemen<sup>t</sup> in general [112].

In this context of ever increasing integration between the physical and the cyber sides of water infrastructure, a growing focus on cyber-physical systems risk assessment and threat modelling (e.g., [71,72]), is expected to become more central in water company preoccupations. Cyber-physical modelling can help the sector manage emerging cyber-physical risks, especially in the context of digital twins. In the same vein, it is suggested that work on modelling cascading e ffects between water systems and other infrastructures may also move from the research environment [113] to the operational environment of the sector. The move may also involve other water and crisis managemen<sup>t</sup> stakeholders at national and international levels.
