Resilience framework for urban water supply systems planning

ABSTRACT

As the concept of resilience is becoming a criterion in planning, water utilities are seeking support and practical guidance to enhance their conventional risk-based planning processes. This paper presents a resilience framework for urban water supply systems planning during the transition towards integrated water resources management. Based on a synthesis of literature across engineering, ecological and social sciences, resilient system performance is defined using crossings of fail-safe and safe-fail thresholds as key indicators. System performance curves conceptually illustrate the capabilities withstanding, absorptive, restorative, adaptive, transformative, and anticipative (WARATA), during sudden and gradual disruptions. Sustainability goals are explicitly considered in the resilience framework, and the role of transformative and anticipative capabilities to facilitate transitions is discussed. Specifically, the desirability of physical design and predicted community consequences from performance impact and collapse can be used in resilience management to prioritize between fail-safe and safe-fail system capacities. Finally, key considerations for operationalizing the framework are summarized, including how issues related to social justice can be addressed when simulating performance and deriving metrics. While this paper focuses on urban water supply, the framework could be applied to other service-providing infrastructure where resilience-based planning supported by quantitative evidence is required to inform investments.

KEYWORDS:

• Resilience framework
• infrastructure planning
• urban water supply
• resilient system performance
• sustainable design

1. Introduction

The United Nations (2019) estimate that by 2050, close to 70 % of the world’s population will live in cities, of which many are already struggling with limited water resources of sufficient quality (McDonald, Weber, Padowski, et al., 2014; World Bank, 2018). Ensuring resilient urban water systems that can provide efficient, reliable, and equitable distribution to citizens is no minor task. Rapid urbanization will increase the stress on aging water infrastructure and, along with unsustainable practices and escalating impacts from climate change, place access to clean water as one of the main challenges of the future. The need to approach water resources in an integrated manner has been adopted by most of the water research community and explicitly made a high priority through the United Nations (2015b) Sustainable Development Goal 6.5 : ’By 2030, implement integrated water resources management at all levels, including through transboundary cooperation as appropriate’. This requires a paradigm shift in the planning of urban water systems, where a participatory systems approach is being encouraged to arrive at holistic and integrated management practices (Brown, Lund, Cai, et al., 2015; Hering, Waite, Luthy, et al., 2013; Ma, Xue, González-Mejía, et al., 2015).

Today’s urban water supply systems were constructed during the 20th century, and are normally characterized by a linear ’take, make, waste’ approach with centralized sources and treatment facilities, and a complex water distribution network. Aside from the unsustainable management of water they imply at the urban scale, there is consensus in literature that these systems are often not suitable for meeting future challenges. Among envisioned system design concepts, the utilization of treated wastewater and stormwater as decentralized supply sources is commonly promoted (Daigger, 2009; Gikas & Tchobanoglous, 2009; Libralato, Volpi Ghirardini, & Avezzù, 2012; Tsegaye, Missimer, Kim, et al., 2020; Wong & Brown, 2009). This is mainly to increase the sustainability of the water supply, but also because diversifying sources and being less dependent on centralized infrastructure is expected to increase system resilience (Leigh & Lee, 2019; Schramm & Felmeden, 2012).

Reliability, i.e., the probability of successful operation, is currently the main measure of infrastructure system performance, and risk management is the common approach in planning and design of urban water supply. While rationales for transitioning from conventional risk to resilience-based planning are extensive in academic literature (Blackmore & Plant, 2008; Butler, Ward, Sweetapple, et al., 2017; Chester, Underwood, Allenby, et al., 2021; Hollnagel, Woods, & Leveson, 2006; Linkov, Bridges, Creutzig, et al., 2014; S. Simonovic, 2016), water utilities lack support and practical guidance on how to use the concept of resilience in infrastructure planning. Overall, there is a clear demand for multi-level models and tools to optimize and prioritize among sustainable infrastructure solutions to better support decision-making (Koliou, van de Lindt, McAllister, et al., 2020), and none of the existing integrated urban water management tools incorporate resilience quantification (Mosleh & Negahban-Azar, 2021). Although there are many barriers to innovation in urban water supply such as the long-term planning horizon, high capital intensity and lifespan of current infrastructure, as well as institutional and physical compartmentalization, the adoption rates of proposed integrated design solutions would likely be higher if quantitative evidence existed to support their perceived resilience-enhancing benefits. Describing what we mean by resilience through a conceptual management framework is, however, a necessary first step in the transition towards resilience as a quantifiable metric of system designs in the planning process.

Hence, the objective of this paper is to present a framework for incorporating resilience in the planning of urban water supply systems. In section 2, a summary of how resilience has been approached and interpreted historically, including the latest perceptions and debates, is presented based on a review of literature within the engineering, ecological and social sciences. Section 3 discusses resilience as a metric of system designs, by summarizing common descriptions of resilient systems and comparing existing concepts in terms of design strategies they promote. The preference for performance-based approaches to quantify resilience is outlined, and consequently, in section 4, a resilience framework that emphasizes levels of performance and the importance of specific system capabilities is proposed. Causal relationships between variables related to sustainability goals, system capabilities, capacities, and level of resilience, are illustrated to explain how resilience can be managed in the planning process through design choices. The proposed resilience framework expands on previous approaches to incorporate transitions towards sustainable performance goals, and section 5 addresses the linkage to sustainability and the importance of transformative and anticipative system capabilities to inform planning. Finally, key considerations and recommendations for operationalizing the framework, i.e., using computer modeling to simulate and quantify resilient system performance, are summarized. This informs future work in collaboration with water utilities and other industry partners.

2. The concept of resilience

Resilience as a concept is not straight-forward to define or explain, and to date there exists no common understanding of resilience across or within academic disciplines. This section summarizes the history of the resilience concept, emphasizing the increasing focus on resilience research within the engineering, ecological and social sciences during the last couple of decades, and how it is currently approached within the water sector.

2.1. Early uses of the term

The word resilience originates from the Latin word ’resilire’ which means to ’bounce back’. Alexander (2013) outlined the long and diverse etymological history, stating that the first known scientific use of the word in English was by Francis Bacon in 1625 as he wrote about the rebound of echoes. In 1858, William Rankine applied the term in mechanics to describe the strength and ductility of steel beams, and about the same time it was also used to indicate ’recoil’ capacity in other mechanical applications such as watchmaking (Alexander, 2013; Blackmore & Plant, 2008; Chester, Underwood, Allenby, et al., 2021). The immediate and traditional interpretation of resilience is therefore an ability to return to normal condition after a disruption (Hosseini, Barker, & Ramirez-Marquez, 2016). While the inclusion of ductility in mechanical resilience, i.e., absorption by deformation, started expanding the concept, Alexander (2013) showed that ’the ability to bounce back’ prevailed as concept until the 20th century. In the 1950s, the term resilience started migrating into social sciences through child psychology and during the 1970s, it was adopted by ecological sciences through the analysis of ecosystem structure and function (Folke, 2006; Holling, 1973). The latter challenged the dominant stable equilibrium view favored across disciplines at the time. Since then, resilience research and debates have evolved mostly within the social and ecology domains, with engineering looking towards integrating aspects of resilience from these research communities in later years. Hosseini, Barker, and Ramirez-Marquez (2016) illustrated that psychology is the discipline with the highest number of resilience-related publications, however this field of research has been excluded in this paper focusing on the system level. The evolving perceptions of resilience within engineering, ecological, and social sciences, as well as significant milestones during the last 50+ years, are synthesized in the following section, and Figure 1 illustrates how the concept has gained traction in recent decades.

Figure 1. Evolving perceptions of resilience within the engineering, ecological and social sciences during the last 50+ years and historical milestones (papers, reports, alliances, networks, programmes, policies).

2.2. Evolving perceptions and historical milestones

Robustness and speed of return to static equilibrium after a disturbance, later termed engineering resilience by the ecological research community (Folke, 2006; Holling, 1996), was the prevailing perception of resilience when Holling (1973) positioned the concept of resilience as the persistence of relationships within an ecological system with multiple stability domains, i.e., a dynamic equilibrium. As per Holling’s definition, ecological resilience of a system is measured by the magnitude of disturbance that can be absorbed within a stability domain before the structure, i.e., the variables and processes that control behaviour, is changed due to instabilities flipping the system into another stability domain. The ideas presented in this paper, which is considered seminal in the field of resilience theory, were initially met with resistance from the research community. As a result, there was little progress beyond the traditional engineering definition for several decades.

Meanwhile, sustainability science was introduced in the late 80s based on the need to start considering future generations, as highlighted in the ground-breaking report ’Our common future’ by the UN World Commission on Environment and Development (1987). By definition, sustainable development meets the needs of the present without compromising the ability of future generations to meet their own needs. Since its introduction, sustainability science has been contrasted with the concept of resilience with regards to overlaps and distinctions, as will be discussed in more detail later.

Around 2000, when the support for Holling’s definition became substantial, the research consortium Resilience Alliance (https://www.resalliance.org) was formed to lead the development of resilience theory using adaptive cycle and panarchy as heuristic models to describe non-linear dynamics of social-ecological systems (Carpenter, Walker, Anderies, et al., 2001; Walker, Gunderson, Kinzig, et al., 2006). Herein, resilience thinking was portrayed as an intellectual framework for understanding how complex systems self-organize and change over time (Folke, Carpenter, Walker, et al., 2010; Walker, Holling, Carpenter, et al., 2004). It promotes the counter-intuitive idea of change being a requisite to persist, and embraces disturbances as opportunities to evolve and transition into more favorable development pathways. Despite this broader definition of resilience, known as social-ecological resilience, most work within ecology has concentrated on the narrower definition initially proposed by Holling, i.e., the capacity of a system to absorb shocks within a stability domain, such that its original function is maintained (Folke, 2006). However, as explained in the following, the adaptation aspect of resilience theory was quickly acknowledged within the engineering and social sciences, as well as in public policy.

Bruneau et al. (2003) from the seismic research community introduced community resilience to hazards as ’the ability of social units (e.g., organizations, communities) to mitigate hazards, contain the effects of hazard-related disasters when they occur, and carry out recovery activities in ways that minimize social disruption and mitigate the effects of future hazards’ (Opdyke, Javernick-Will, & Koschmann, 2017). Emphasis was for the first time placed on the role of social units in reducing impacts, consequences, recovery time, and future disaster vulnerabilities through adaptation (Koliou, van de Lindt, McAllister, et al., 2020). This triggered international strategies for disaster resilience through the Hyogo Framework for Action (HFA) 2005–2015 (United Nations, 2005) and the successor Sendai Framework for Disaster Risk Reduction 2015–2030 (United Nations, 2015a). The intergovernmental panel on climate change released a special report on disaster resilience (IPCC, 2012), and resilience to hazards was recognized as a component of sustainability in the United Nations (2015b) SDG framework through target 13.1: ’Strengthen resilience and adaptive capacity to climate-related hazards and natural disasters in all countries’. In recent years, the framework has been criticized for only measuring resilience to extreme events such as earthquakes, droughts and floods, and not considering persistent stress due to climate change and urbanization in the few indicators of resilience that are included (Assarkhaniki, Rajabifard, & Sabri, 2020).

Around the same time as community resilience was introduced, the field of resilience engineering (RE) started evolving from discussions on how to overcome limitations of traditional risk assessment and system safety by building adaptive capacity to unpredictable disturbances (Woods & Wreathall, 2003). Through borrowing of concepts and principles from other fields, RE is presented as a paradigm in safety management where reliability has been the focus historically (Righi, Saurin, & Wachs, 2015). Yodo and Wang (2016) defined RE as the practice of how to engineer resilience into the design to cope with system complexity and unforeseen failure modes, i.e., through proactive resilient processes rather than through reactive barriers and defences (Hollnagel, Woods, & Leveson, 2006). This acknowledgement of dynamic system management to handle surprises presented a change in how resilience is framed within the engineering domain, but as addressed by Chester, Underwood, Allenby, et al. (2021), most research from the engineering community still focuses on robustness and rebound, i.e., the static system management ’engineering resilience’ perspective, and mainly on challenges related to extreme events.

The sole focus on extreme events is also reflected in the US National Infrastructure Advisory Council’s (2009) final report and recommendations for critical infrastructure resilience. Recognizing that protection and resilience are complementary elements of risk management strategies, they defined infrastructure resilience as ’the ability to absorb, adapt to, and/or rapidly recover from a potentially disruptive event’. This is in line with the 2012 US National Academy of Sciences (NAS) definition of resilience as ’the ability to prepare and plan for, absorb, recover from, and more successfully adapt to adverse events’ (Linkov, Bridges, Creutzig, et al., 2014) A decade later, based on the call to strengthen critical infrastructure in the Sendai Declaration, the American Society of Civil Engineers (ASCE) published a manual of practice to guide the transformation towards hazard-resilient infrastructure (Butry, Davis, Malushte, et al., 2021). Here, they adopted the Presidental Policy Directives (PPD)-21 definition of resilience from 2013 to be used by federally sponsored research. It reaches beyond extreme events by stating that resilience is ’the ability to prepare for and adapt to changing conditions and to withstand and recover rapidly from disruptions’ (Koliou, van de Lindt, McAllister, et al., 2020).

2.3. Current focus and discourse

The occurrence of several natural disaster events has led to an increased focus on community resilience during the last decade. Policy-makers have adopted resilience as an ideal for society, but in recent years the focus in academic literature has shifted towards resilience as a ’formal’ non-normative concept representing the ability to keep or enhance certain still-to-be specified functions (Doorn, Gardoni, & Murphy, 2019). This positions resilience in a broader context than climate change adaptation and disaster management, where challenges related to sustainability can be addressed through active management of resilience (Chelleri, Waters, Olazabal, et al., 2015; Elmqvist, Andersson, Frantzeskaki, et al., 2019). In line with this, urban resilience was recently defined by Elmqvist, Andersson, Frantzeskaki, et al. (2019) as ’the capacity of an urban system to absorb disturbance, reorganize, maintain essentially the same functions and feedbacks over time and continue to develop along a particular trajectory’. In practice, cities are encouraged to improve the ability of institutions, the built environment, and communities to cope with and adapt to inevitable disruptions and change (Meerow, Pajouhesh, & Miller, 2019), and the Resilient Cities Network (https://resilientcitiesnetwork.org) was recently launched to support cities with their strategies. Acknowledging that a community’s resilience is dependent on both the resilience of its infrastructure and the individuals’ ability to deal with disruptions (Doorn, Gardoni, & Murphy, 2019), researchers are calling for integrated ’system-of-systems’ efforts that address the complex interactions between physical, social and economic infrastructures that jointly enable community resilience (Koliou, van de Lindt, McAllister, et al., 2020).

A current discourse within the social science literature relates to whether resilience is a suitable concept for addressing social justice issues, or if it merely promotes status quo by failing to pay sufficient attention to power and politics (Meerow, Pajouhesh, & Miller, 2019). These authors reviewed the first ten North American city resilience strategy plans that were created through the Rockefeller Foundation’s 100RC programme running between 2013 and 2019, and uncovered varying emphasis on social equity questions. Overall, focus has been on the distributional dimension, i.e., equitable distribution of goods, services and opportunities, with little attention to acknowledging and respecting different groups (recognitional dimension), and equitable participation in decision-making processes (procedural). As such, the importance of asking ’resilience for whom, what, when, where, and why?’ is being stressed within urban resilience planning (Meerow & Newell, 2019). Romero-Lankao, Gnatz, Wilhelmi, et al. (2016) emphasized the importance of envisioning desirable futures and re-imagining our relationship with the environment as part of this process. This way, the long-term goal of sustainable transformation can be operationalized through resilience management (Chester, Underwood, Allenby, et al., 2021). Doorn, Gardoni, and Murphy (2019) explained why a major challenge going forward relates to developing models that integrate aspects of infrastructure resilience from the engineering literature with community resilience from social science literature.

2.4. The water sector

Within the water sector, there has been a significant increase in publications addressing resilience, but as shown by Rodina (2019) the concept is still vaguely articulated across subsectors. Researchers are also investigating what resilient urban water services imply in practice through stakeholder surveys, workshops and expert interviews (Johannessen & Wamsler, 2017; Laitinen, Kallio, Katko, et al., 2020). Results point towards the importance of institutional management and human agency to achieve resilience, in addition to the technical aspects. In agreement, the City Water Resilience Framework (CWRF) was recently created to help enhance a city’s institutional and decision-making process, referred to as resilience capacity (Saikia, Beane, Garriga, et al., 2022). It is part of the City Water Resilience Approach, integrating the assessment, planning, design and implementation of resilient water systems into a holistic step-wise process (Arup, 2019). Although the CWRF describes elements of resilient infrastructure performance, it is a governance level framework that cities can use to measure qualitative indicators of water service resilience. Tools that inform investment decisions for infrastructure design are needed to help reach sub-goals identified in such frameworks.

Water systems engineering, the water sector with the largest proportion of papers, is predominated by the conventional engineering definition of resilience, i.e., reliability, recovery, or ability to ’bounce back’ from disruptions without considering the potential need to adapt (Rodina, 2019). However, the shift towards holistic and integrated water resources management requires a broader view of resilience that considers cross-scalar interactions and complex system dynamics (Brown, Lund, Cai, et al., 2015; Hering, Waite, Luthy, et al., 2013; Ma, Xue, González-Mejía, et al., 2015). Incorporating aspects from ecological definitions of resilience is therefore desirable going forward (Blackmore & Plant, 2008; Rodina, 2019). The concept of resilience can also be applied to guide transitions towards integrated urban water management (Butler, Ward, Sweetapple, et al., 2017; Johannessen & Wamsler, 2017). This idea largely remains at the conceptual level and resilience thinking still awaits operationalization in infrastructure planning for urban water systems (Chester, Underwood, Allenby, et al., 2021; Makropoulos, Nikolopoulos, Palmen, et al., 2018).

3. Towards resilience as quantifiable system design metric

Beyond its original meaning to ’bounce back’, resilience is a relatively new concept within the engineering discipline, with opportunities to influence how it is approached in the system design process (Hosseini, Barker, & Ramirez-Marquez, 2016). This section first explains the short-comings of risk-based planning and how shifting to resilience-based planning is believed to improve the performance of urban water systems. Commonly perceived resilient system design characteristics are then summarized, and the preference of performance-based approaches to quantify resilience in modeling is explained. Finally, existing concepts of resilience are compared in terms of emphasis on fundamental system capabilities, and the design strategies they promote using a fail-safe versus safe-fail performance lens.

3.1. Risk- versus resilience-based planning

In the risk-based planning process, a risk assessment contains the evaluation of what can go wrong, i.e., the hazards/threats, and the likelihood and consequences of each scenario (Kaplan & Garrick, 1981). As explained by S. Simonovic (2016), the consequence is a combination of exposure and vulnerability, where exposure is the relative location of assets, and vulnerability is the susceptibility of these assets taking damage from the hazard. In engineering practice, the product of probability and consequence is used to quantify the perceived risks and further to prioritize among them for mitigation options to reduce the likelihood of disruptive events and the potential consequences (Hosseini, Barker, & Ramirez-Marquez, 2016). Butry, Davis, Malushte, et al. (2021) described the concept of risk concisely as relating to the probabilities of unacceptable consequences due to hazards.

Risk-based planning has often proved successful in the design of urban water supply by guiding prevention and protection strategies. However, clear challenges to this conventional management approach have been raised in academic literature. First, it is time-consuming to perform risk assessment of all system components and mitigation quickly becomes expensive (Butler, Ward, Sweetapple, et al., 2017; Linkov, Bridges, Creutzig, et al., 2014; Yodo & Wang, 2016). Second, it is impossible to foresee, protect and prevent all undesired scenarios due to the uncertainty of disturbances and complicated interconnections in coupled human-nature systems (Hosseini, Barker, & Ramirez-Marquez, 2016; Shin, Lee, Judi, et al., 2018). Third, conventional risk management often leads to rigid systems with efficiency-driven, top-down control with low levels of adaptive capacity to deal with unexpected failures (Walker & Salt, 2012). Finally, risk management is not considered suitable for addressing interactions among multiple systems operating at different temporal and spatial scales (Anderies, Folke, Walker, et al., 2013; Linkov, Bridges, Creutzig, et al., 2014).

As opposed to risk-based planning that aims to reduce the vulnerabilities in a system by hardening specific components or create redundancy, resilience-based planning focuses on the whole-of-system behaviour and identifies critical functionalities (Blackmore & Plant, 2008; Linkov, Bridges, Creutzig, et al., 2014). It is not meant to substitute risk management in the design process, but rather complement it as achieving reliability is the first necessary step towards achieving resilience. While risk management helps in preparing and planning for foreseeable adverse events by reducing the likelihood of failure and following consequences (Linkov, Bridges, Creutzig, et al., 2014), resilience extends the process by considering the temporal capacities of a system that comes into play when design conditions are exceeded (Butler, Ward, Sweetapple, et al., 2017). Shifting the focus towards resilience in the planning of urban water systems therefore implies acknowledging the need for system designs that not only prevent but can also overcome failure.

3.2. Describing resilient system designs

While there exists no unique insight on how to define and measure resilience for a system, Hosseini, Barker, and Ramirez-Marquez (2016) showed that there are several similarities or trends across academic disciplines. Resilient systems are often described in terms of characteristics that are perceived as resilience-enhancing, for instance redundancy and resourcefulness, which are the ’means’ of the 4 R’s of community resilience first introduced by Bruneau, Chang, Eguchi, et al. (2003). Diversity, flexibility, redundancy and inclusiveness are widely cited factors for achieving urban resilience (Meerow & Newell, 2019; Zuniga-Teran, Gerlak, Mayer, et al., 2020), and flexible response, distributed decision-making, modularity and redundancy were suggested by Linkov, Bridges, Creutzig, et al. (2014) as strategies to build infrastructure system resilience. Studies of social-ecological systems have also indicated that design characteristics such as diversity, openness, reserves, tightness of feedbacks, modularity and redundancy would provide high levels of resilience (Biggs, Schlüter, Biggs, et al., 2012; Walker & Salt, 2012).

Definitions, however, tend to focus on response mechanisms of the system when undergoing change and, according to Butry, Davis, Malushte, et al. (2021), a resilient infrastructure system is often judged by the ability to withstand disruptions and recover from impacts. Yodo and Wang (2016) also positioned reliability and recovery as the key designable attributes of resilience. These relate to robustness and rapidity, the ’ends’ of the 4Rs of community resilience (Bruneau, Chang, Eguchi, et al., 2003), where the former is commonly used by engineers as a measure of ’designed resilience’ (Walker & Salt, 2012). While reliability-based designs have been vastly investigated, research on design strategies to enhance recovery is very limited despite the importance for realizing resilient performance (Yodo & Wang, 2016). The underlying assumption in the emerging field of Resilience Engineering is that resilience can be engineered into a system to support the use of adaptive capacity in the recovery phase, although it remains an open research question (Righi, Saurin, & Wachs, 2015; Woods, 2015). A forerunner is the tool ResilSIM developed for decision-makers to evaluate the impact of adaptation options on community resilience to urban flooding (Irwin, Schardong, Simonovic, et al., 2016). The work is based on defining the 4Rs in terms of system performance, and conceptually relating them to a resilience curve (S. P. Simonovic & Arunkumar, 2016).

Overall, the emphasis and meaning of each response mechanism and the perceived system features that contribute to their capacity are not streamlined in literature. To address this, Shin, Lee, Judi, et al. (2018) performed a systematic review of existing quantitative approaches for measuring resilience of water infrastructure systems, and derived withstanding, absorptive, restorative and adaptive as fundamental capabilities. Resilience was further defined as ’a comprehensive ability of a system to sustain its performance within an acceptable level by combating disruptions with the fundamental capabilities in timely and efficient ways, after failures of the system, for a given uncertain environment’ (Shin, Lee, Judi, et al., 2018).

3.3. Performance-based approaches to quantify resilience

Existing quantification methods of infrastructure resilience tend to be dependent on hazard and systems, and there is a need to arrive at a generally applicable standardized approach for comparing level of performance (Yodo & Wang, 2016). Shin, Lee, Judi, et al. (2018) highlighted that several of the existing quantification approaches are based on surrogate measures of reliability or apply property-based methods. For example, structural network measurements from Graph Theory have often been applied in water infrastructure system analysis to explicitly evaluate redundancy and connectivity of supply routes as surrogate quantities for resilience (Creaco, Franchini, & Todini, 2016; Yazdani & Jeffrey, 2012). While correlations exist between structural properties and some of the multiple aspects of system performance that resilience entails, Meng, Fu, Farmani, et al. (2018) showed that this approach can be misleading due to potential trade-offs among indicators. The relationship between system properties and resilient performance is a topic of ongoing research (Mugume, Diao, Astaraie-Imani, et al., 2015), and Butler, Ward, Sweetapple, et al. (2017) stressed the importance of distinguishing system characteristics from performance metrics. Further, Shin, Lee, Judi, et al. (2018) emphasized that property-based measures do not have the capability to provide proof of how the system will perform during disruptions as they exclude baseline comparisons and threshold functionality. Performance-based modeling approaches are recommended since they are capable of presenting the complex dynamic system responses to various failure modes (Butry, Davis, Malushte, et al., 2021; Diao, Sweetapple, Farmani, et al., 2016; Pagano, Sweetapple, Farmani, et al., 2019). Also, Opdyke, Javernick-Will, and Koschmann (2017) explained how the isolation of infrastructure from the community it is intended to serve has led to the abstraction of infrastructure resilience, making the inclusion of human aspects in modeling important to better understand the linkages with community resilience. Finally, Doorn, Gardoni, and Murphy (2019) clearly stated that an infrastructure’s contribution to human well-being should be considered when specifying required functionality by defining minimum levels that are acceptable or tolerable. This would increase system complexity further and make property-based methods even less suitable for comparing resilient performance.

3.4. Resilience concepts translated into design strategies

The capabilities withstanding, absorptive, restorative and adaptive were proposed by Shin, Lee, Judi, et al. (2018) as fundamental to assess resilient performance of water infrastructure systems. Withstanding capability aims to maintain normal functionality under system disruptions, absorptive, to minimize system damage, restorative, to quickly recover to an acceptable state, and adaptive, to deal with conditions from uncertain disruptions through change (Shin, Lee, Judi, et al., 2018). As summarized in Figure 2 and explained in detail below, the emphasis on each of these system capabilities and the implications when translated into overall design strategies varies among the concepts of resilience outlined in section 2.

Figure 2. Design strategies of risk management and concepts of resilience, derived from comparison to the fundamental resilient water system capabilities defined by Shin, Lee, Judi, et al. (2018). Outer circles indicate which capabilities are covered, illustrating overlap with and extension beyond conventional risk management.

Risk management strives to provide withstanding and absorptive capabilities through mitigation efforts that build robustness and redundancy to avoid potential impacts from predictable disturbances. The overall aim is to avoid any type of system failure, meaning a fail-safe design strategy is undertaken. Engineering resilience also promotes fail-safe designs within a defined range of disturbances by focusing on efficiency, constancy and predictability (Anderies, Folke, Walker, et al., 2013; Holling, 1996). However, since focus is also on recovery speed after a disturbance, some level of failure is implied for a duration of time. The engineering resilience approach therefore emphasizes withstanding capacity to prevent failure, and restorative capacity to overcome failure, meaning both fail-safe and safe-fail design strategies to predictable disturbances as illustrated in Figure 2.

The single-equilibrium steady-state approach in risk management and engineering resilience requires defining which system configurations and disturbances are of interest for each specific scenario to build resilience. When resilience ’of what to what’, i.e., resilience of some specific part of the system to a specific disturbance is clearly defined, it is referred to as specified resilience within ecological literature (Walker & Salt, 2012). The authors also explained how ecological resilience, in its most narrow definition, resembles specified resilience by focusing on persistence, i.e., robustness to change, ensuring all parts of the system keep functioning as before. However, ecological resilience is a multiple equilibria approach that also manages unknown disturbances through absorption to prevent the state of the system from reaching a threshold that can flip the system into another stability domain (Folke, 2006). It aligns best with the fundamental system capability to immediately absorb disruptions and minimize damage from any disturbance, and thus a fail-safe strategy would be pursued in system design. Anderies, Folke, Walker, et al. (2013) described a general resilience view as the ability to build and increase the capacity for learning and adaptation, as reflected in the social-ecological definition of resilience (Folke, 2006). Adaptability is naturally linked to resourcefulness in human-nature systems and emphasizes capacity to change to reduce the consequences of failure when the withstanding and absorptive capacities have been depleted. Social-ecological resilience is therefore illustrated as an expansion of ecological resilience towards safe-fail or ’safe-to-fail’ design strategies in Figure 2.

Although different terminology, the concepts graceful extensibility and sustained adaptability, defined by Woods (2015) as part of Resilience Engineering (RE), coincide with the focus on adaptability in social-ecological resilience. In ecological literature, adaptability is defined as the capacity of a system to manage resilience by avoiding the crossing of thresholds, facilitating crossing back into a desired regime, or by moving thresholds to create a larger safe operating space (Walker & Salt, 2012). In RE, graceful extensibility is defined as how a system extends performance or brings extra adaptive capacity to bear when surprise events challenge its boundaries, and sustained adaptability as the ability to adapt to future surprises as conditions continue to evolve (Woods, 2015). The four core concepts of resilience within RE: robustness, rebound, graceful extensibility and sustained adaptability, thus align with withstanding, restorative and adaptive system capabilities as seen in Figure 2.

Definitions of infrastructure resilience tend to specify absorptive, restorative and adaptive capabilities only, however in the recently published ASCE manual of practice (Butry, Davis, Malushte, et al., 2021), withstanding capability seems to be incorporated within the absorptive capability. Thus, there is complete overlap with the four fundamental capabilities in Figure 2, representing both fail-safe and safe-fail design strategies. Further, anticipative capability is added by Butry, Davis, Malushte, et al. (2021) to reflect the ability to rapidly sense and identify hazards through surveillance and monitoring systems before they manifest.

Finally, community resilience mainly focuses on the human resources that can be tapped into after a hazardous event to reduce consequences, speed up recovery, and adapt in a way that reduces future vulnerabilities. This means a safe-fail design strategy is emphasized, aligning with the fundamental capabilities restorative and adaptive in Figure 2. Social-ecological resilience also speaks of transformability, i.e., the capacity of a system to become a different system (Walker & Salt, 2012), which has become a key attention within urban resilience. As discussed later, it may not be a prerequisite for resilience, but transformative capacity would be necessary to deal with system performance collapse and to guide sustainable development, i.e., linking the resilience management approach to the overarching sustainability goal as illustrated in Figure 3.

Figure 3. Resilience framework showing levels of performance when the system is exposed to disruptions, the related design dimensions and contributing capabilities, and the linkage with sustainable development through system transformation.

4. Framework for resilience-based planning

There is increasing focus on resilience-based planning to deal with infrastructure challenges, but practical implementation has been slow and the requested shift towards resilience as a quantifiable metric of system designs is still ongoing (Butler, Ward, Sweetapple, et al., 2017; S. Simonovic, 2016). In this section, a resilience framework is presented based on an operational definition of resilience that emphasizes different levels of system performance. The framework is explained using system performance curves that illustrate the importance of specific capabilities during sudden and gradual disruptions. Finally, a description of how the framework can be used to actively manage system resilience during planning is included. The framework applies to any infrastructure system that provides a service, however, the practical examples in this paper relate to urban water supply systems.

4.1. Operational definition of resilience

Based on a review of literature, synthesis of existing concepts of resilience, and the previously derived fundamental capabilities of water infrastructure systems (Shin, Lee, Judi, et al., 2018), an operational definition of system resilience is proposed as

the ability to provide fail-safe performance through withstanding and absorptive capabilities, and safe-fail performance through restorative and adaptive capabilities, where the need to transform is informed by anticipative capabilities.

Descriptions of the capabilities (WARATA) and their toolkit, i.e., actions in practice, can be found in Table 1 along with examples for urban water supply systems. The proposed resilience framework is illustrated in Figure 3, indicating levels of system performance and related design dimensions with their contributing system capabilities. A focus on predicting and managing disruptions rather than the initial threat is adopted from Butler, Ward, Sweetapple, et al. (2017) to enable the effects of unknown threats to be accounted for in modeling. Multiple threats can result in the same failure mode, so this ’middle-based approach’ would capture more scenarios in one analysis (Butler, Ward, Sweetapple, et al., 2017). Resilient performance during both sudden and gradual disruptions is considered, where the latter plays a central role in directing sustainable development as will be discussed later. When applying the operational definition of resilience to urban water supply systems, performance can be measured as water supplied to end users. Aligning with the resilience framework, performance is then impacted when water demand is not satisfied by the physical design dimension, and collapsed if water demand is not satisfied by the institutional design dimension.

Table 1. Essential capabilities (WARATA) and their toolkit in practice for providing resilient system performance.

Capability	Description	Toolkit	Examples for water supply
Withstanding*	Capability to withstand system disruptions and maintain normal functionality under the disruptions.	Increaserobustness.	Hardened pipes, capacity expansion.
Absorptive*	Capability to immediately absorb the disruptions and minimize system damage.	Add buffers, redundancy.	Twinned pipes, back-up generator, piped source options.
Restorative*	Capability to quickly recover to the normal or acceptable state.	Repair readiness.	Operator capacity to repair a failed pipe.
Adaptive*	Capability to adapt to the changing conditions and uncertain disruptions.	Prioritize, distribute resources, implement programs.	Water restrictions, temporary servicing, emergency supply distribution (reactive), conservation and demand management (proactive).
Transformative	Capability to transform into a different system when critical performance thresholds are or will be crossed due to disruptions.	Restructure, reorganize, educate.	New piped water source, changed resource availability.
Anticipative	Capability to predict disruptions, system performance and its desirability.	Monitor, acquire data, model, learn.	Streamlined technology, connected information systems, computer modeling, interactive tools.

[*] As originally described by Shin, Lee, Judi, et al. (2018).

4.2. Fail-safe system performance

The first part of the definition deals with capabilities to ’stay safe from failure’ and can be interpreted as the physical dimension of resilient design. This fail-safe aspect of resilience aimed at avoiding disruptions and minimizing impact on performance when they occur is, according to Yodo and Wang (2016), more straightforward to implement in the design. Impacts on system performance from a potential sudden disruption are shown in Figure 4, where 3 of the 4 R’s (Bruneau, Chang, Eguchi, et al., 2003), which S. P. Simonovic and Arunkumar (2016) conceptually related to system performance and resilience curves, are included. Line a) indicates the relationship between the withstanding system capability to a predicted disruption and the design feature robustness pursued in risk management to achieve reliable designs. Examples of building withstanding capacity as a proactive response to sudden disruptions is the hardening of pipes to avoid pipe failures, or expanding capacity of the supply source as a proactive response to avoid water supply deficit due to an increasingly stressed water resource. Capacity expansion means pushing the fail-safe threshold in Figure 5 out in time and following the fail-safe performance line, which would be rising if the predicted water supply deficit was also caused by demand growth.

Figure 4. Conceptual system performance curve when exposed to a sudden disruption. a) No impact from predicted disruption. b) Impact from predicted disruption. c) Impact from unpredicted disruption.

Figure 5. Conceptual system performance curve when facing a predicted gradual disruption. Magnification of sudden disruption included to illustrate time scales, and of critical safe-fail threshold to show details.

The design feature redundancy is intuitively related to absorptive capabilities, and immediately available options such as twinned pipes in case of a pipe failure or back-up generator in case of a power outage, would contribute to absorbing the potential impact from a predicted disruption. If completely absorbed throughout the system, the fail-safe threshold in Figure 4 would be pushed out in time and line a) would be followed; if partly absorbed, line b) would be followed. For unpredicted disruptions, the absorptive capacity would be more difficult to incorporate proactively in the design, and the response would be reactive depending on chance. This is illustrated by the difference in impact on system performance in lines b) and c). Also, since absorptive capability is an immediate system response, it is of less relevance to gradual disruptions and therefore omitted on the right hand side of the framework in Figure 3. Finally, it is assumed that gradual disruptions are always predicted and would be dealt with proactively as part of long-term planning.

4.3. Safe-fail system performance

The second part of the definition deals with capabilities to ’fail safely’ and can be interpreted as the institutional dimension of resilient design. Impact on system performance occurs when the fail-safe threshold is crossed, and focus is on minimizing consequences through restorative and adaptive capabilities. The exact institutional response depends on whether the disruption was predicted and whether the resulting impact is sudden or gradual, as illustrated in the framework in Figure 3. The anticipative capacity of the system would answer both questions through streamlined technology and connected information systems, and predict the crossing of fail-safe and safe-fail thresholds through modeling.

In the case of impact from an unpredicted sudden disruption, see line c) in Figure 4, the distribution of emergency supply and repair resources are examples of reactive adaptive and restorative responses. As illustrated, reactive adaptive capacity can be interpreted as a temporary measure to fill the gap from impact on piped water supply, i.e., satisfy water demand, while restoration is being undertaken. If the sudden disruption was predicted, a faster and more coordinated response is expected, see line b) in Figure 4. An example would be operators already trained in quickly repairing a broken pipe, where, in the meantime the community would have to adapt to restrictions on water supply or temporary servicing solutions as part of reactive adaptive capacity.

Restorative capability can be measured relatively straight-forwardly as system rapidity, i.e., the time between the first impact on baseline performance and its recovery, as illustrated in Figure 4. Reactive adaptive capability is dependent on restorative capacity in the sense that if it is depleted before the piped water supply has been restored, safe-fail threshold is crossed and performance collapsed. Transformative capabilities would then be required to satisfy water demand, as illustrated in Figure 4. Examples are improvised water supply solutions through the deployment of water trucks or establishing water distribution points by utilizing supplementary water sources. Educating additional human resources to assist with these solutions or to prepare for future disruptions, may be a resulting action from the collapse. While performance is ultimately achieved, i.e., water demand satisfied, the permanent change in availability and location of resources means the system has gone through an abrupt transformation. Instead of restoring the pre-impact system state as part of safe-fail performance, transformative capability permanently changes the system.

If the fail-safe threshold in Figure 5 is crossed, proactive adaptive capacity would determine the safe-fail performance of the system. For example, in the case of a predicted steadily growing or seasonal deficit in piped water supply, the implementation of demand management programs with appropriate conservation incentives is a proactive adaptive option that can be tapped into to counter the impact on system performance. The safe-fail threshold in Figure 5 indicates at what point in time conservation capacity would be depleted, and thus increased conservation at demand nodes would push this threshold out in time. To avoid future collapse in performance, i.e., water demand not satisfied, a gradual transition to new piped water supply source needs to commence latest at this point in time, which would be informed by the anticipative capability. Directed transformation, illustrated by dashed green lines in Figure 5, can therefore overlap with proactive adaptation, and the opportunity to encourage systemic chance is discussed in more detail later. However, if the predicted gradual disruption is too severe to be dealt with in time by a combination of proactive adaptive capacity and directed transformation, or has simply not been managed proactively, thresholds for proactive and reactive safe-fail capacities would inevitably be crossed (magnified in Figure 5). This means system performance is collapsing, and abrupt transformations must occur for water demand to be satisfied.

4.4. Resilience management

The proposed framework can be used to guide decisions in the planning process of infrastructure systems providing a service. In practice, resilience management means increasing or decreasing distance to thresholds, and facilitating transformations when needed. The causal loop diagram (CLD) in Figure 6 illustrates the relationships between variables related to sustainability goals (desirability of physical design and community consequences from disruptions), system capabilities, capacities, and level of resilience. Positive relationship arrows are indicated with a plus sign, i.e., increasing the variable would result in an increase in the connected variable or, in cause and effect terms; ‘if a cause increases, the effect increases above what it would otherwise have been’ (Sterman, 2000). Negative relationship arrows are indicated with a minus sign, and form several balancing feedback loops (indicated with a B) in Figure 6. These can be leveraged in management to guide transformations of systems with undesirable physical designs, and to prioritize between design dimensions of resilient performance based on predicted community consequences from impact and collapse.

Figure 6. Causal loop diagram showing relationships (all arrows) between variables related to sustainability goals, system capabilities, capacities, and level of resilience. Positive (+) and negative (-) relationships, and balancing feedback loops (B). Management and design choices (dashed arrows) in a sustainability decision-making context.

Relationships shown as dashed arrows are management and design choices, and a sustainability decision-making context is implied in terms of maximizing or minimizing the variables. The overall objective is to increase the resilience of systems with desirable physical designs through fail-safe and safe-fail capacity enhancement. Although the CLD helps visualize the components and relationships involved, simulation of system performance using computer modeling would be necessary to inform decisions, ideally based on a cost versus equitable resilience gain analysis of options. This aspect falls under anticipative capability in the resilience framework, along with determining the desirability of the physical design dimension when using sustainability as an overarching goal. Both are discussed in more detail later.

4.4.1. Strengthen system resilience or facilitate directed transformations

Resilience of systems with desirable physical designs (built infrastructure), can be strengthened by enhancing capabilities that increase fail-safe and safe-fail capacities, see Figure 6. For example, if the piped supply source is considered sustainable, expanding its volume capacity, i.e., increasing the capability to withstand a gradual disruption, could be justified. Another option is to increase the proactive adaptive capacity through conservation incentives to further increase its sustainability. This would provide safe-fail performance in case withstanding capacity reaches efficiency limits in the future. Also, providing safe-fail performance during sudden impacts would ideally be prioritized by enhancing restorative and reactive adaptive capacities. This would avoid abrupt transformations that can lead to either positive or negative impacts on desirability, thus indicated with both plus and minus signs in the relationship arrow in Figure 6.

However, if the piped supply source is not considered sustainable in the long term, directed transformations to more sustainable pathways might be preferred. Balancing feedback loop B1 illustrates how directed transformation to more desirable system designs can be facilitated by only increasing transformative efforts. Although abrupt transformations are considered riskier than directed transformations, collapse can be provoked by down-prioritizing reactive adaptive capacity to, for instance, harvest community support for costly systemic changes. In general, the distance to thresholds can be increased to enhance resilience of existing design or decreased to facilitate change, where the latter is viewed as a complex socio-political decision to transition when a system is in proximity of critical thresholds (Chelleri, Waters, Olazabal, et al., 2015).

4.4.2. Prioritize design dimensions based on community consequences

Although the resilience of systems with desirable physical designs should ideally be strengthened through both design dimensions, in practice, limited resources require prioritizations. In addition to the cost component, predicted community consequences from performance impact and collapse can be used to decide between enhancing fail-safe and safe-fail capacities in the design phase. For instance, essential services such as hospitals and emergency shelters might be designated a higher importance in terms of availability of piped water supply, i.e., fail-safe performance. This means the physical fail-safe design dimension should be emphasized at these system nodes by increasing withstanding and absorptive capacities to sudden disruptions. Balancing feedback loop B2 illustrates how strengthening these capabilities would increase the fail-safe capacity, and therefore reduce the chance of impact and related consequences. In the case of impact occurring however, restorative capacity would at least ensure a speedier recovery and therefore less community consequences from the impact on performance, as depicted in feedback loop B3.

Restorative capability is a component of the institutional safe-fail design dimension and therefore also related to avoiding performance collapse. A hypothesis is that vulnerable populations are less likely to be prepared with personal emergency supply options, so a collapse in system performance would have more dire consequences for them. System nodes serving these groups should therefore be prioritized in terms of access to emergency supply and in the allocation of repair resources to ensure equitable resilience management. Feedback loops B4 and B5 show how the enhancement of reactive adaptive and restorative capacities would increase the safe-fail capacity, and as such reduce the chance of performance collapse and following community consequences. This is especially important as a short-term measure for addressing structural racism that historically has led to underserved groups of society in terms of reliable piped water supply.

5. Discussion

The resilience framework presented in this paper draws upon and expands on previously proposed approaches to explicitly incorporate transitions towards sustainable performance goals, as illustrated in Figure 3. The rationale behind and importance of transformative and anticipative system capabilities are discussed in the following sections, as well as key considerations for operationalizing the framework.

5.1. Including sustainability in the resilience framework

The debate on whether it is beneficial to combine the field of sustainability science with resilience theory has been ongoing for decades. Contrasting elements of the two concepts, Redman (2014) argued that a clear distinction is that sustainability prioritizes outcomes and strategies to attain these, whereas resilience prioritizes process through adaptive capacity and open-ended transitions to maintain system dynamics. Others have posed that while a system must be resilient to be sustainable, a resilient system is not necessarily sustainable (Butry, Davis, Malushte, et al., 2021; Doorn, Gardoni, & Murphy, 2019). For instance, Elmqvist (2017) pointed out that diversity and redundancy are often seen as key features of resilience, which may run counter to the common sustainability goal of having an efficient system. Carpenter, Walker, Anderies, et al. (2001) suggested that since a resilient state can be desirable or undesirable, sustainability should be used as an overarching goal to determine which system states are desirable. Similarily, Elmqvist, Andersson, Frantzeskaki, et al. (2019) proposed that sustainability should be viewed as a normative concept representing the vision for society, and resilience as a non-normative deconstructable attribute of a system that can be enhanced or reduced depending on the desirability of current state. As such, a sustainability decision-making context can be implied when managing resilience (Anderies, Folke, Walker, et al., 2013), where resilience is key for achieving desirable performance. This is often reflected by the conventional ’three pillars’ of sustainability, e.g., ’the degree to which the system maintains levels of service in the long-term whilst maximising social, economic and environmental goals’ (Butler, Ward, Sweetapple, et al., 2017). Reconciling the environment, social equity, and economic demands can be mutually reinforcing but also conflicting (Butry, Davis, Malushte, et al., 2021). Using resilience-thinking to help center focus and determine information needed to sustainably manage the valuable functions of a system can therefore be a useful problem-framing approach (Walker & Salt, 2012). Chester, Underwood, Allenby, et al. (2021) posed that resilience theory is well-positioned to support building capacities to adapt and transform and should be used as a guiding framework for infrastructure to ensure that needs are met in the future.

The system’s withstanding, absorptive, restorative and adaptive capabilities in Table 1 determine the fail-safe and safe-fail performances of an infrastructure system, which again determine the social, economic, and environmental dimensions of sustainability. This aligns with the idea that while resilience deals with somethings’ impact on the infrastructure, sustainability deals with how the infrastructure impacts something (Bocchini, Frangopol, Ummenhofer, et al., 2014). Active management of resilience to increase physical system desirability and reduce community consequences, as illustrated in Figure 6, therefore requires complex long-term processes involving economic, social and environmental dimensions (Chelleri, Waters, Olazabal, et al., 2015; Elmqvist, Andersson, Frantzeskaki, et al., 2019). Through frameworks and planning tools that integrate the two concepts, system designs that also meet sustainability requirements, i.e., desirable resilience, can be identified and guide structural transformations of urban systems (Butry, Davis, Malushte, et al., 2021; Elmqvist, Andersson, Frantzeskaki, et al., 2019; Romero-Lankao & Gnatz, 2016).

5.2. The importance of transformative system capabilities

In the context of urban resilience, it is argued that the long-term goal of sustainable transformation should be operationalized through resilience management (Chelleri, Waters, Olazabal, et al., 2015). Elmqvist, Andersson, Frantzeskaki, et al. (2019) stated that, rather than focusing on multiple stable states, urban systems should be described in terms of having multiple possible development pathways or trajectories. One can either strengthen resilience of a pathway to avoid abrupt transformations or facilitate directed transformations to more desirable pathways. Preferably, systemic changes through reorganization, restructuring and education are planned long-term transformative processes that interact and overlap with proactive adaptation as illustrated by Chelleri, Waters, Olazabal, et al. (2015). Abrupt transformations are more focused on providing service and therefore less likely concerned with adhering to sustainable development pathways (Elmqvist, Andersson, Frantzeskaki, et al., 2019). However, although Johannessen and Wamsler (2017), as expected, found that system collapse has a negative perception among stakeholders, the resulting reorganization afterwards is also seen as potentially important for pushing transitions into more sustainable urban water services (Blackmore & Plant, 2008). This aligns with the ’Build Back Better’ focus in the United Nations (2015a) Sendai Framework for disaster risk reduction and the newer ’bounce forward’ interpretation of resilience (Romero-Lankao, Gnatz, Wilhelmi, et al., 2016), and means that in some cases collapse might be necessary although human beings are naturally opposed to it.

In the proposed resilience framework, transformative capability is defined as the ability to transform into a different system when critical performance thresholds are or will be crossed due to disruptions, see Table 1. For urban water supply, the system has evolved into a new system when there is a permanent change in significant physical infrastructure components such as piped supply source, or in the availability of resources. Such transformations may also positively influence future capacities and by that increase the level of system resilience. For example, the investment in local sustainable supply sources as a transformative response to predicted water supply deficit and future collapse, would also improve absorptive capacity to handle disruptions in central supply. This is why decentralization is commonly listed as a system feature for absorptive capacity, for instance in Butry, Davis, Malushte, et al. (2021). Further, the deployment of water trucks or training of additional human resources as part of an abrupt transformation, would likely improve adaptive capacity to future sudden disruptions.

As opposed to restorative capability that restores status quo, transformative capability can be viewed as recovery that reduces future system vulnerabilities through restructuring, reorganization and education (Doorn, Gardoni, & Murphy, 2019; Koliou, van de Lindt, McAllister, et al., 2020). The ASCE manual for hazard-resilient infrastructure does not mention transformative capability explicitly, but reorganization and restructuring are listed as part of adaptive capacity as a second line of defense in between absorptive and restorative (Butry, Davis, Malushte, et al., 2021). Chelleri, Waters, Olazabal, et al. (2015) defined three partly overlapping stages of urban resilience related to short-, medium- and long-term perspectives, i.e., recovery from shocks, adaptability to actual or expected changes, and structural transformation into new regimes. Herein, adaptation is about moving thresholds to make the system persist within the same regime while longer-term transformations are the alteration of fundamental attributes of the system. Lewin, Rossi, Soultani, et al. (2023) discussed the role of trigger points and adaptation in strategic thinking about infrastructure resilience, expressing adaptability as the ability to modify the system through remedial actions that mitigate or restore performance after impact. Rather than transformations, new adaptive pathways are marked based on the degree of change required to continue providing service. Determining which toolkit falls within each capability is not straightforward, and the resilience framework using WARATA capabilities aligns more with the urban resilience frameworks’ view on adaptation and transformation (Chelleri, Waters, Olazabal, et al., 2015; Elmqvist, Andersson, Frantzeskaki, et al., 2019). In essence, any structural transformation that results from a recovery process would be referred to as an expression of the transformative system capability in the proposed resilience framework.

5.3. The role of anticipative capabilities

Anticipative capability is defined as the ability to predict disruptions, system performance and its desirability, see Table 1. While the prediction of disturbances is a central component in risk management, focus is shifted to predicting potential disruptions in the proposed resilience framework. This means predicting a pipe break happening rather than the cause of it, e.g., a seismic event, aging, etc. In this context, predictability does not mean predicting exactly when a failure will occur, but knowing it is a possibility, and potentially having a probability estimate for it. Streamlined technology and connected information systems for monitoring and data gathering are necessary to enable this for both sudden and gradual disruptions to urban water supply.

When it comes to predicting system performance, emphasis should be placed on understanding thresholds, determining their relative location and how to deal with them as crossing them can bring both positive and negative consequences (Walker & Salt, 2012). Further, the multidimensional nature of resilience management implies that resilience strategies would coexist in practice, emphasizing the importance of uncovering potential trade-offs among the different capabilities due to interconnections across temporal and spatial scales (Chelleri, Waters, Olazabal, et al., 2015). This can be informed by computer models simulating fail-safe and safe-fail performances during disruptions, as it would reflect the physical and institutional dimensions of resilient design and their contributing capabilities. By carefully examining the social, economic and environmental impacts of the physical design dimension, its desirability can also be established to direct transformation towards sustainable development goals. Anticipative capability therefore has a central role in the resilience framework through information technology, computer modeling and interactive tools.

5.4. Operationalizing the framework

For resilience-based planning to become reality, resilience needs to move beyond conceptualization to measurable system characteristics with established performance targets (Butry, Davis, Malushte, et al., 2021). In the proposed resilience framework, the crossing of thresholds can be used as performance metrics at system demand nodes. For urban water supply systems, this means establishing whether there is or would be a deficit in piped water supply during scenarios, and if so, whether water demand would still be satisfied through safe-fail performance. Since safe-fail and transformative capacities mainly reside in institutions and individuals, techniques such as agent-based modeling can readily assist to arrive at a holistic representation of system performance (Blackmore & Plant, 2008; Butry, Davis, Malushte, et al., 2021).

The performance measure ‘water supplied to end users’ aligns with Poulin and Kane (2021)’s definition of productivity performance measures. Performance targets that are variable and adaptive can therefore be facilitated in modeling, reflecting that demand naturally changes over time and responds to scenarios. For instance, when facing a sudden disruption, the critical safe-fail threshold of resilient performance would be crossed if the time to restore piped water supply exceeds time to deplete emergency supply options, but factoring in a potentially lower demand level for the given scenario would increase the distance to threshold. In practice, acceptable levels of service for end users would be influenced by frequency of events, trust in service provider, as well as system comparisons and interdependence (Lewin, Rossi, Soultani, et al., 2023), and should also consider individuals’ capacity to handle adverse situations (Doorn, Gardoni, & Murphy, 2019).

Meerow, Pajouhesh, and Miller (2019) explained how resilience tends to be highly unequal for different groups of a community, and herein lies the opportunity to place infrastructure services within a broader context to ensure that historical inequities and structural racism embedded in the system are not further locked into place (Chester, Underwood, Allenby, et al., 2021). Figure 6 illustrates how the predicted community consequences from system performance impact and collapse can be used to inform management and design choices in planning. In modeling, prioritization scores could be used to allocate emergency and repair resources and then evaluate impact on safe-fail capacity at system nodes. Another option is to weight the importance of fail-safe and safe-fail capacities when aggregating the spatially distributed performance metrics at demand nodes into one system performance curve.

Further, a scalar measure, or an ensemble of so-called ’summary metrics’, needs to be derived from the system performance curve to quantify resilience. These curves are often referred to as ’resilience curves’, where the impacted area over the curve represents loss in resilience and thus resilience is quantified using the integral (Yodo & Wang, 2016). However, Poulin and Kane (2021) suggests the process of quantifying resilience should be done in collaboration with stakeholders to best reflect their goals. For instance, by combining a threshold-based metric with the common integral-based, the relative distance from critical safe-fail threshold can be emphasized. Summary statistics, contour diagrams or scenario-dependent weighting are methods that can further be applied to derive a single value representing level of resilience across scenarios. Here, scenario-dependent weighting is widely acknowledged but unexplored across literature according to Poulin and Kane (2021). It would likely require the use of probability to determine the importance of scenarios, keeping in mind that the difficulty in exhaustively including all possible events in the distribution was among the rationales for moving beyond risk-based planning.

6. Conclusion

It is becoming increasingly evident that resilient performance needs to be prioritized in the planning and design of urban water supply systems. Reviewing how existing concepts of resilience compare to risk management and emphasize fundamental capabilities of resilient water infrastructure systems (Shin, Lee, Judi, et al., 2018), a framework for incorporating resilience in the planning process is proposed. The operational definition of system resilience is derived using a performance-based approach, and translating system capabilities into fail-safe and safe-fail design strategies. It is supported by conceptual performance curves illustrating system resilience during sudden and gradual disruptions. The importance of transformative and anticipative system capabilities is emphasized along with the previously defined withstanding, absorptive, restorative and adaptive. In addition to specifying when and where the various system capabilities (WARATA) become important, the framework conceptually illustrates how to incorporate sustainable development into resilience management. This is an increasingly encouraged path to guide transitions in the urban century, and the causal relationships between variables related to sustainability goals, system capabilities, capacities, and level of resilience are explained.

In long-term planning of urban water supply, management decisions and design interventions would ultimately be prioritized based on a comparison of desirable resilience gains, and costs. Sustainable urban water system concepts that are difficult to justify cost-wise, would therefore benefit from having quantitative evidence that supports resilience-enhancing prospects. When integrated in predictive analysis tools that simulate system performance and derive metrics, the proposed resilience framework can help explore the desirability of existing water supply systems and alternative physical and institutional configurations. For instance, simulations of water supply sources and human resources at different spatial locations can help uncover resilient development trajectories to guide transformative efforts proactively. Through envisioned end-user tools similar to ResilSIM (Irwin, Schardong, Simonovic, et al., 2016), the framework can also assist with understanding emergency preparedness and improve response planning based on predicted community consequences from system performance impact and collapse (Koliou, van de Lindt, McAllister, et al., 2020). Herein, the capacity of humans to deal with adverse conditions needs to be considered in addition to the functioning of built infrastructure (Doorn, Gardoni, & Murphy, 2019). Finally, while the examples in this paper are focused on water supply systems in particular, the framework could be applied in modeling any infrastructure system that provides a service, with the aim of translating descriptions of how resilient performance occurs into prescriptions supported by quantitative evidence at the design stage (Righi, Saurin, & Wachs, 2015; Yodo & Wang, 2016). Details of the quantification approach should preferably be determined in collaboration with relevant industry partners (Poulin & Kane, 2021), and for urban water supply, these could include system and emergency planners from water utilities and water supply experts from consulting companies.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The authors confirm that the data supporting the findings of this study are available within the article.

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

We acknowledge the support of the Natural Sciences and Engineering Research Council of Canada (NSERC), [funding reference number RGPIN-2023-05266].

Notes on contributors

Aina Crozier

Aina Crozier is a PhD candidate in Civil Engineering at the University of British Columbia (UBC), with an M.Sc. in Energy and Environmental Engineering from the Norwegian University of Science and Technology. In addition to her Master’s thesis, which entailed the design and dynamic modeling of an offshore wind turbine support structure, she has practical experience with modeling and data analysis from the energy industry and as a Water Action Plan Researcher at UBC. Her PhD research focuses on resilience quantification with application to design choices in urban water supply planning.

Barbara J. Lence

Barbara J. Lence is a Professor in the Department of Civil Engineering at the University of British Columbia, and until recently, Director of the UBC Master of Engineering Leadership in Integrated Water Management. She received her doctorate from the University of Illinois at Urbana-Champaign for research on evaluating the engineering, environmental, and economic impacts of water pollution control policies. She has extensive experience with numerical modeling, optimization and decision analysis of civil engineering systems. Current research focuses on design and operational strategies of water distribution networks, identifying reliable water quality management systems in the face of uncertainty, and climate change adaptation.

Steven V. Weijs

Steven V. Weijs is an Associate Professor in the Department of Civil Engineering and an instructor in the Master of Engineering Leadership in Integrated Water Management program at the University of British Columbia. He holds an M.Sc. and Ph.D. in Civil Engineering and Water Resources Management from TU Delft (Delft University of Technology) in the Netherlands. His past work addressed mountain hydrology, water system control, forecast verification and model complexity control using information theory. His current work involves monitoring network optimization in water resources management, and the dynamics of information in a systems-based approach to water-related risks.