Global Environmental Change Mapping narratives of urban resilience in the global south

In the context of global environmental change much hope is placed in the ability of resilience thinking to help address environment-related risks. Numerous initiatives aim at incorporating resilience into urban planning practices. The purpose of this paper is to open up a conversation on urban resilience by unpacking how diverse science methods contribute to the production of di ﬀ erent narratives of urban resilience mobilizing di ﬀ erent experts and forms of evidence. A number of scholars have cautioned against uncritical approaches to resilience and asked what resilience means and for whom, also pointing out the normative dimension of the concept. Building on this emerging scholarship we use insights from science and technology studies (STS) and critical social sciences to look at the knowledge infrastructures and networks of actors involved in the development of resilience strategies. Drawing on ﬁ eldwork in Manila, Nairobi, and Cape Town, we map di ﬀ erent narratives of urban resilience identifying the ways in which science serves to legitimate or alienate particular perspectives on what should be done. We discuss the multiple roles that science methods have for resilience planning. Whereas urban resilience is often portrayed as consensual, we show that a range of narratives, with diverse socio-material implications, exist at the city level. In this way we unearth the con ﬂ ict that lies beneath an apparent consensus for resilience policy and outline future research directions for urban sustainability.


Introduction
The resilience agenda is being pushed forward in urban governance, in both developed and developing countries. The Sustainable Development Goals (SDGs) directly call for 'inclusive, safe, resilient and sustainable' cities (SDG 11). In this context, urban planning is a central component of global ambitions to deliver climate change adaptation and disaster risk reduction and emerging urban policy agendas have galvanised around the notion of resilience (Moglia et al., 2018;Pelling et al., 2018;Elmqvist et al., 2014). This movement has placed intuitive, lay interpretations of resilience at the level of policy discourse, somewhat separating this from more precisely defined technical applications within socio-ecological systems, engineering and disaster risk reduction. This has been successful in providing a logical framing to help organise risk management as part of integrated development planning. Through this process, resilience has become a discursive field deployed to represent the city and particular visions of its future, often independent of the technical scrutiny the term has previously benefited from.
If adaptation is about shaping the future through judgements on what to enhance, retain and discard, then resilience helps set the frame of reference to legitimate these decisions. Resilience narratives frame policy discussions, bound the aims of climate change adaptation and disaster risk management, and give legitimacy to specific forms of knowledgeand to those who hold and produce this knowledge (Owens et al., 2006;Goldstein et al., 2015). These narratives are constructed by dominant actors and countered by subordinate actors to shape the possible for urban futures and are associated with clear practical and material implications Moench, 2015, Sandercock,2003). In this context, the dominant urban policy narrative often projects resilience as a consensual object, on which hardly anyone could disagree. For example, in the context of the 100 Resilient Cities programme supported by the Rockerfeller Foundation, urban resilience has been defined as: This is just one, albeit prominent, example of resilience presented as a common ground for otherwise diverse urban interests. To a certain extent this framing is perceived as an opportunity for those operating in interdisciplinary domains where multiple values can slow-down or block decision-making progress (Chelleri et al., 2015;Leichenko, 2011, Pelling,2010. Resilience made apolitical in this way can enable actors who do not necessarily share values and aims to reposition themselves and aspire for consensus. However superficial value neutrality does not mean that value conflicts are resolved, rather conflicts over values may be supressed and hidden (Simon and Randalls, 2016;Welsh, 2014). If so this may only delay and deepen tensions and potentially undermine the long-term functioning of integrated planning approaches that are thought essential for sustainable urban development (McEwen et al., 2017;Mitra et al., 2017;Bull-Kamanga et al., 2003). The meaning as well as practice of resilience is shaped by competing and unequally powerful actors in the city and beyond (Leitner et al., 2018;Wilson, 2012). Where diversity is not acknowledged debates over resilience will undermine the potential for more integrated policy and democratic decision-making. Consensus will be built on false foundations and may undermine trust between urban actors (Solecki et al., 2017). On the other hand, if urban resilience is negotiated, and even contested, through a process, it can help to surface these tensions and better situate and ground the focus of resilience goals and activities (Harris et al., 2017;Friend and Moench, 2013).
Critical readings of resilience have emerged, in particular, in the field of anthropology, critical geography and other constructivist paradigms (MacKinnon andDerickson, 2013, Leach,2008). A recurrent criticism directed towards resilience points towards the falsely apolitical aspect of the concept (Cretney, 2014;Béné et al., 2012;Coaffee et al., 2008). With regards to urban planning, several studies emphasize the need to look at what resilience means in practice (e.g. Kythreotis and Bristow, 2017;Wilkinson, 2012) and a smaller literature has begun to tease out the different deployments of urban resilience Coaffee, 2013). In this respect, White and O'Hare introduce a distinction between 'equilibrium resilience' and 'evolutionary resilience ' (2014). They suggest that the former facilitates a focus on infrastructure while the latter is more sensitive to change and social concerns. Resonating with this distinction, we differentiate here between conservative and transformative approaches (Pelling, 2010) to resilience and suggest that such distinction is useful to make sense of different resilience narratives. Scholars working with environmental and social justice approaches have also begun to uncover the latent ethical and justice questions around urban governance and resilience strategies (Allen et al., 2017;Ziervogel et al., 2017;Shi et al., 2016). A major point here is that the use of resilience is not neutral and the ways in which resilience policies are made deserves more attention. Resilience is deeply normative and the assumptions underpinning different perspectives have to be made explicit (Alexander, 2013, Brown,2014. This paper advances a field of critical research on the application of urban resilience by drawing on Science and Technology Studies (STS) and narrative analysis. This opens analysis onto the forms of knowledge that underpin and legitimise particular resilience narrative positions. The following section presents this analytical framework, followed by a methodological note and individual and comparative analysis of narratives in three cities: Cape Town, Manila and Nairobi where resilience programming is active and through which the value of a narrative lens is demonstrated. Cape Town and Nairobi are part of the Rockefeller 100 resilient cities programme.

Conceptual framework
Resilience has now become a buzzword and is being used by actors and organizations from both science and policy backgrounds, operating at different scales and with different purposes and meanings (Brown, 2014;. In this respect, while becoming a hegemonic framing at the policy level, resilience acts apparently as a boundary object (Brand and Jax, 2007) in urban politics, able to bring together actors and organizations with otherwise different agendas and interests: "Boundary objects are objects which are both plastic enough to adapt to local needs and constraints of the several parties employing them, yet robust enough to maintain a common identity across sites. They are weakly structured in common use, and become strongly structured in individual-site use. They may be abstract or concrete. They have different meanings in different social worlds but their structure is common enough to more than one world to make them recognizable, a means of translation. The creation and management of boundary objects is key in developing and maintaining coherence across intersecting social worlds." (Star and Griesemer, 1989:393) From this perspective resilience has the potential to allow heterogeneous actors and organizations to work together -stabilizing relations despite heterogeneity of concerns and interests between actors. Yet, boundary objects, due to their malleability, often hide important conflicts and disagreements (Forsyth, 2018, Borie andHulme,2015). They render collaboration possible but between actors who do not necessarily share the same aims: conflicts over values may be rendered invisible. This suggests that resilience can only be understood relationally. We need to enquire not about resilience per se, but about the resilience of whom to what Meerow et al., 2016a, Meerow and. Urban resilience processes are deeply political, often leading to disagreements and forcing trade-offs between values (McEwen et al., 2017;Cretney, 2014). At the same time, political, social, and ethical assumptions are rarely explicitly surfaced by resilience planning methods which tend instead to infer value neutrality and frame decision-making around tangible and quantifiable indicators and physical features (Eakin et al., 2017). Efforts to monitor and measure urban resilience itself are multiplying (e.g. Bozza et al., 2015;Cutter et al., 2008). A number of policy frameworks at different scale call for 'evidence-based' approaches for resilience planning. At a global scale, the Sendai Framework strongly encourages the use of scientific knowledge for disaster risk reduction. Welsh observes the development of an 'emerging form of governmentality through resilience ' (2014:16).
In this context, recent research has called for more attention to be dedicated to urban knowledge systems (Jon, 2018;Muñoz-Erickson et al., 2017). A number of scientific tools and technical devices, such as maps and Geographic Information Systems, are routinely used in urban planning and support the development of resilience strategies (Pelling, 2011, Godschalk,2003. Practices such as community vulnerability and risk assessment as well as resilience action plans pervade and also delineate how resilience is constructed. Yet science and technology are not value neutral, they have performative effects that need to be studied (Kitchin et al., 2012;Porter, 1996). Science is not a monolithic block and different epistemologies and forms of knowledges exist (Knorr-Cetina, 2007). With regards to resilience, much knowledge comes from natural science disciplines, with social science often limited to vulnerability assessments (Donovan and Oppenheimer, 2015). Quantitative tools are generally granted more authority than other forms of knowledge (Kovacic, 2018). Yet, rather than more physical science knowledge it has been argued that it is more social scientific knowledge that is needed to understand, for example, barriers to climate change adaptation (e.g. Hackmann et al., 2014;Lorenzoni and Whitmarsh, 2014). At the same time, a number of authors have questioned the dichotomy between lay and expert knowledge (Wynne, 1992) and emphasized the value of local knowledge, such as in the case of climate change adaptation (Naess, 2013).
Which forms of knowledge and epistemologies are mobilized also matter. Different forms of knowledge underpin different narratives, legitimating different actors and solutions. Resilience narratives emerging from art-based methods, for example, are unlikely to overlap neatly with those emerging from conventional methods, yet narratives have material implications (see Moglia et al., 2018;Heras and Tàbara, 2014). Knowledge itself is often contested, for example hazards maps and other land use documents, which are pictured as objective, can hide important controversies through the use of scale or selective depiction of component parts, or through the politics of their production (Haughton andWhite, 2017, Desportes andColenbrander, 2016). The use of science and scientific methods in the presentation of resilience is then mutablescience can be openly deployed as a critical tool, a veil to obscure values and a common language to facilitate communication. More than this, the borrowing of scientific methods and approaches confer to resilience narratives the appearance of scientific objective rationality. Analogous of advertising campaigns, the symbols of science can be deployed in communication strategies to project rigour, impartiality and systematic process. This further projects the image of resilience as an urban planning device that is value neutral and so apolitical.
This is not to argue for scientific methods to be left out of resilience planning. On the contrary scientific methodologies and approaches can bring transparency in decision-making, they can offer a common language and motivation to democratise decision-making as well as holding decision-makers to account. Much depends on the ways in which methods and their surrounding relationships are managed. Building on Stirling's metaphor, science can either 'open up' or 'close down' the range of policy options and possible futures (2008). To understand what resilience does (how it is used, what it means to different actors), one needs to analyse how it is constructed through particular knowledge-making practices, and by whom: whose disciplines and expertise shape resilience strategies.
In doing so we are particularly interested in the knowledge infrastructures underpinning those strategies, and in the ways that some knowledge practices are institutionalized by public authorities for resilience planning. Attending to the interactions between diverse actors (e.g. intergovernmental organizations, universities, think tanks) and the forms of knowledge they rely on is also important to understand how particular understandings of urban resilience circulate (see Bulkeley, 2005). The concept of knowledge infrastructure has been defined by Edwards as: "Robust networks of people, artefacts, and institutions that generate, share, and maintain specific knowledge about the human and natural worlds." (Edwards, 2010:17) Narrative analysis has been applied to a range of topics such as food security (Sonnino et al., 2016), biodiversity conservation (Hutton et al., 2005) and climate change (Hulme, 2009). A narrative can be understood as a technique to gain coherence (Nash, 2005). Following Haraway (1989), we are more particularly interested in the performativity of these narratives: they are not flat discourses but become embedded in practice and policy actions. In this sense, narratives resonate with the notion of the 'socio-technical imaginary' (Jasanoff and Kim, 2013) which highlights the co-production between science and societies. Particular narratives are underpinned by diverse forms of evidence and disciplines but little is known about the kind of resilience they actually support. Our results are based on qualitative research methods and analysis which are introduced in more details in the section below.

Mapping resilience narratives: a methodological note
The resilience narratives of three case study cities are presented below. Before introducing further our methods we introduce our study sites and explain the rationale for conducting fieldwork in Manila, Nairobi and Cape Town. All three cities have active resilience agendas deployed as part of responses to recognised challenges of risk associated with wide social and economic disparities including large populations forced to live in slums and informal settlements with inadequate land-tenure and service provision (Amin and Cirolia, 2018;Porio, 2011Porio, , 2014Shatkin, 2004). Disaster risk is manifest through chronic everyday events as well as episodic catastrophic disastersboth products of constrained development choices Ziervogel et al., 2016). In each city resilience agendas were championed by city government and civil society agencies working together, separately or in contestation providing a rich context to examine the deployment of resilience and its consequences for inclusive and integrated development paths to sustainable development.
The derivation of narratives in each city drew from 31 semi-structured interviews (12 in Manila, 8 in Nairobi, 11 in Cape Town) with public authorities at the municipal level, NGOs, and the corporate private sector. The interview guide and interviewee profiles are presented in supplementary material 1 . In addition to face to face interviews, we gathered and analysed documents relevant to our study including policy documents, institutional websites and newspaper articles. The material gathered was analysed using qualitative coding, paying particular attention to the following themes: conception of resilience, challenges with resilience, use of science and knowledge, responsibility and ethics. We then identified empirically different narratives of resilience in each city. These narratives are not necessarily mutually exclusive and can coexist in different places or policy documents and be mobilized by the same actor or organization depending on the situation. Some are supportive of each other while others are incompatible. In each place, the explicit use of the term 'resilience' also differs. For example, some individuals explicitly refuse to engage with the term that they perceive as already too biased towards powerful interests.
We applied the following analytical framework to describe each narrative: • Conception of resilience: Distinguishes between resilience understood predominantly to protect (conservative) or challenge (transformative) established development practices and processes, and whether policy preferences are biased towards short-or long-term solutions; • Mode of governance: We differentiate between three modes: bottomup, top-down, and co-production, to characterize the processes underpinning those narratives and the extent to which they are participatory and include the views of communities; • Actors: Identifies those individuals, groups and organizations whose conception of resilience falls under this particular perspective; • Use of science and technology: The ways in which scientific tools are mobilized focusing on the motives underpinning their use and on Stirling's metaphor (2008), whether they open-up or close down conversations; • Forms of knowledge used: An indication of the forms of knowledge perceived as relevant in each narrative. We differentiate between natural science knowledge, social science knowledge and multiple forms of knowledge.
• Examples: We give an example emerging from analysis to illustrate perspectives.
Building on this analytical frame we derived two figures that help visualize similarities and differences between the narratives. On Fig. 1(a) we cluster the narratives according to their predominant topical focuswhat are they about? -including modernization, social justice, environment and security. On Fig. 1 (b) we map narratives in relation to the governance approach they are associated with: top-down vs. bottom-up and conservative vs. transformative. Other visualizations could have been used but these two allow us to discuss two important aspects: the content of the narratives and the processes that underpin them.

Resilience narratives
This section demonstrates the richness in the range of narratives found in each city. Some narratives resonate across cities, for example, narrative M1 (Natural Hazards), predominant in Manila, has a counterpart, CT1 (Technocratic Resilience) in Cape Town. We chose to present narratives separately as each city has its own particular context and history, to emphasise diversity at the city level. They are not intended to be an assessment of how resilience strategies are implemented in these places. Rather, our interest is to map the multiplicity of resilience narratives, reflecting on their similarities and differences and on their socio-material implications.

Manila
We identified three competing resilience narratives operating in Manila characterised by a focus on enhancing resilience through Natural Hazards, Social Cohesion and a call for long-term planning (Planning Ahead).

M1 -Natural Hazards
Most visible in national policy documents (e.g. the National Disaster Risk Reduction Framework in the Philippines, see Shaw et al., 2017) this narrative emphasized resilience to natural hazards through preparedness and response. Geohazards maps identified areas at risk of floods, liquefaction, and landslides to delineate locations as 'safe' or notsecurity was the focus. Maps did not account for the social and economic aspects that mattered for people living in those places. Maintenance and reconstruction of infrastructure were an important concern and engineering teams were given key responsibilities. Technical devices were perceived as essential to monitor and predict weather events. For example, the city of Manila recently invested in a Command Centre equipped with a wide number of screens to visualize different parts of the city. The narrative focus on geo-hazards and the technical framing of risk management produced a narrow understanding of risk, going against other understandings of resilience which focused on the wellbeing of communities or/and on the longer term planning concerns.

M2 -Social Cohesion
This narrative was promoted in particular by civil society organizations and researchers and emphasized that, beyond security, the resilience of people and communities depends on their ability to maintain social networks. What mattered was to allow people to live close to livelihoods and families. The National Disaster Risk Reduction Framework in the Philippines identifies communities as a key component of resilience and the government recognized that local communities must be placed at the forefront of resilience planning. Yet, concurrently, the focus on security directly hampered the well-being of some of those communities, low income urban households in particular. For people living in areas identified as unsafe by government geohazards maps (M1) 'security' had sometimes been used in an instrumental manner by public authorities as a way to push to evict poor communities and justify relocation. These were also often blamed for contaminating the environment. While invoked by public authorities to take responsibility over disaster risk reduction, 'security' can also serve to avoid taking responsibility over poverty alleviation. These divergent views about resilience therefore crystallized more clearly over the issue of relocation: under M2 relocation is legitimate but preference is given for relocation on-site. Beyond the issue of relocation, people advocating for these perspectives argued that ultimately what matters is that people talk to each other and that bridges exist between different governance levels and areas.

M3 -Planning Ahead
A focus on short-term economic interests and on emergency response undermined investments in longer-term planning. This narrative emphasized the unequal distribution of resources in the city and a lack of anticipation in development planning to reduce risk. There was a desire for widespread human capital technical investmentfor example to make GIS available to City Planners to make possible a more systematic approach to risk identification and management by facilitating data management, project monitoring, and evidence-based decisionmaking. As emphasized by an architect working for Manila City Council: "We are also just beginning to work with GIS. It will definitely make our lives easier. The problem with City Hall as of now is that all departments are not integrated at all. The right hand does know what the left one is doing. Divisions have different data. We cannot resolve differing data and we don't know who is right." (IM3) So far most data, including planning maps, were only available in paper form and this limited the ability to plan, anticipate and predict. GIS would strongly benefit planning by making possible a more systematic approach. In particular, it would facilitate record keeping, monitoring and evidence-based decision-making. In this perspective, modernizing the city by equipping it with computers and GIS systems would allow better planning as this would facilitate bringing together comprehensive datasets. In addition to the lack of digital infrastructure, one problem was a lack of in-house expertise in at least some areas of the public sector.

Cape Town
Information technology was a fault-line in the deployment and ownership of resilience in Cape Town. This was associated with a City administration with high technical capacity and resource for building and analysing large population datasets and sophisticated GIS systems, represented enthusiastically in narratives around Technocratic and Coproduced Resilience, and more critically in a narrative seeking to move beyond reductionism: Emancipation.

CT 1 -Technocratic Resilience
Science, maps and spatial knowledge were used for planning purposes including 'Cityviewer', a sophisticated GIS system operated by the City of Cape Town. City officials could see the city virtually, with a high level of detail. This amount of data provided them with what bore some similarity with a "God's eye view" of the city or what STS scholars have termed a "view from nowhere" (Haraway, 1988). This describes an aspiration to provide objective knowledge whose validity is universaltrue everywhere no matter the context (e.g. Burgess, 2004, Shapin,1998). The approach developed by the City of Cape Town, at least in some departments, echoed those aspirations: of an all-encompassing, objective, picture of the city allowing rational management and control. Yet, the high degree of technology did not reflect the realities of living in informal settlements without access to services at risk from flooding and did not necessarily lead to better planning practices (see also Watson, 2014, Dierwechter, 2004. To some extent, technology was used to shortcut direct engagement and consultation with citizens. Maps could be interpreted as giving a summary of what is happening on the ground rather than trying to get officials to visit and interpret conditions locally. Officials might think that they see the city and do not need to sense it. This reflected an aspiration to build technical skills to solve problems rather than appreciating socio-technical realities. This also entailed a degree of depersonalization in exchange, as technology was used to support apparently democratising but simultaneously technocratic, top-down interventions, that disconnected local at risk poor communities from planning authorities. This use of technical devices in an attempt to suppress political debates has been widely documented elsewhere (e.g. Latour, 2004;Lupton and Mather, 1997). At an extreme this use of GIS and mapping reinforces and aggravates existing divides and inequalities.

CT 2 -Co-Producing Resilience
Knowledge and maps were extensively used with an explicit attempt to foster engagement and conversation between different actors and organizations via practices such as participatory mapping, qualitative mapping, participatory scenario planning and games. This narrative included both city and non-city officials, with NGOs such as the Community Organization Resource Centre (CORC) and the Development Action Group (DAG). Actors in this perspective often suggested that "The city is working for you", the motto of the City of Cape Town, ought to be "The city is working with you". They saw processes as being as important as outputs and mapping was often used to trigger a conversation such as in the case of this city official: "I try to find solutions using maps and drawing contours to identify places where houses could be moved to distribute services, and then new houses could be built. Sometimes I know what a solution could be but I need to let the community come up with it, and sometimes they come up with new points that I hadn't thought of. Dialogue is needed. We need to use both technical and local knowledge." (CI9) NGOs also used maps as advocacy tools to render visible things which some city officials, despite their modern facilities, could not see. How information is mapped facilitates other interpretations. For example, rather than presenting a map indicating the location of toilet facilities in a particular settlement, indicating who had access to them and when (e.g. for most women going out at night is unsafe so access will be bounded in time) changed how access was perceived. In addition to alternative maps, some NGOs routinely used GIS and participatory mapping in their activities, as in the case of CORC. Their approach was quite formalized and followed different steps: enumeration; data collection; analysis; presentation of results. They always gave knowledge (e.g. statistics and maps) back to communities. Sometimes civics knew those tools, maps for example, better than the City, and used them to engage with the City and challenge official views.
One of the limits of co-production identified in this narrative is that some datasets produced by non-city actors did not directly interface with official city datasets, being placed in different data repositories. This lack of interfacing can be explained by logistical reasons and concerns over data curation but also by the fact that not all datasets were perceived as equally credible. It is also worth underlining that there were numerous micro-politics regarding data-sharing and accessing datasets that depended on who one liaised with. An additional limit to coproduction is that many perceived the City of Cape Town as operating in a very top down manner.

CT3 -Emancipation
This narrative was explicitly mobilised as a critical response to the dominant (CT1) resilience narrative by NGOs, placemaking activists, and researchers operating outside the City of Cape Town. 'Emancipation' argues that dominant resilience framing served to maintain existing power relations in an unequal society. Some deliberately refused to use the term 'resilience', which they saw as already owned by powerful economic actors. The failed promise of resilience highlights a gap between people's expectations and city action connected to an overly technical framing of resilience. To move beyond this, there is a need to develop more bottom-up, participatory, planning practices (Parnell, 2002, Parnell andPieterse,2010). Actors advocating for this perspective often emphasized a need to make more room for qualitative data and stories, allowing multiple interpretations of resilience to coexist. With mapping, for example, this implied mapping social networks, to avoid reductionism, oversimplification, and decontextualization (e.g. Dovey and Ristic, 2017). Another example relates to the ways in which people navigated in the city. City maps showed cycling lanes but did not say anything about different people's experience when using cycling lanes at different times of the daythis plurality of experiences was seldom captured and therefore not used to build resilience. Art-space methods were also perceived as a fruitful way to move beyond technocratic and Western ways of thinking and planning to help residents develop a sense of place and generate creative ideas about their own future (Pieterse, 2006).

Nairobi
Resilience policy was undeveloped in Nairobi City County Government, despite its being made one of the Rockerfeller Foundation's 100 Resilient Cities, a formal policy agenda and public debate had yet to emerge. This challenge was reflected in the dominant narrative (Fragmentation) and through more optimistic future facing narratives built on Making the Invisible Visible and Holistic Thinking.

N1 -Fragmentation
In this narrative, lack of knowledge and data was recurrently emphasized and identified as a factor hampering the ability of urban planners and citizens groups to manage, plan and predict risk. It also undermined more systemic and holistic ways of thinking. Cumulative effects, for example regarding water supplies and flood risks, were not taken into consideration by contemporary development planning. This was emphasized in particular by an engineer working for an international company operating in partnership with local authorities: "For example, when there is a new development, its impacts on traffic are to be considered so there is a traffic impact assessment which has to be validated by the Nairobi Road Department but then what about storm water channels. Who is responsible for this? Who should assess the drainage system and the impacts the new development is likely to have on its capacity? Road and sewage systems are interconnected, but who is checking that when there is a new development project? There are multiple connections to assess but many of them are left out." (IN7) The vision and practice of resilience was then described as fractured by this interviewee and compartmentalized between different sectors with an unclear distribution of responsibilities. Proponents saw no clear process to tackle tensions between short-term priorities and long-term interests. For example, there was a tendency for planning decisions to prioritise the short-term and economic development while ignoring the longer-term effect on the soil and water supplies. As in M3, many expectations were placed in the potential for technical improvements, principally in access to GIS, to improve planning practices, monitor changes and facilitate record-keeping, but there was first a dire need to develop inclusive, up to date, datasets.

N2 -Making the Invisible Visible
Science and technology were used in an emancipatory manner by organised civil society; for example by Muungano wa Wanavijiji, a social movement of Kenya slum dwellers which uses household surveys to challenge official data or its absence and so to raise voice and the visibility of the urban poor. This was deployed to encourage public authorities to take responsibility in the most vulnerable areas of Nairobi including for example the Mukuru Special Planning Area (Urban ARK, 2017). Actors advocating for this perspective completed different kinds of tasks. They sometimes provided conventional maps for areas which were left out of official maps but also routinely used participatory mapping techniques, working with the communities, and provided alternative maps. The Kounkuey Design Initiative (KDI), an NGO operating in Kibera, the biggest informal settlement in Eastern Africa was one example. The data gathered by this organization were freely accessible and often served to trigger conversation with Nairobi City County. When providing conventional high resolution maps, such as hazard maps, KDI emphasized the value of including local knowledge, for example about flood risks (see also Mulligan et al., 2017). Their projects in Kibera were organized around the idea of 'productive public space' and combined expert and lay knowledges. In addition to GIS mapping, KDI also used digital tools and social media where people living in Kibera could share their stories. As emphasized by Slum Dwellers International (SDI), depending on where they work, data on informal settlements in particular could be completely missing from official registers and they had to produce it entirely. SDI formed relationships with local governments and shared data to encourage planning interventions.

N3 -Holistic Thinking
This narrative represented the vision of a resilient city in which there were numerous public spaces, parks and a well-connected public transport system. One in which public authorities saw the value of public and green spaces, their advantages in terms of public health rather than perceiving them as hampering economic growth and profit. There was a need to overcome the opposition beyond short term and long term by considering all the costs that public spaces and parks allowed avoiding. This meant, for example, being aware that the diminution of pollution levels also means less respiratory diseases, diabetes, and numerous positive externalities in term of public health. Advocates of the 'green city' were trying to make the case for this alternative imaginary of Nairobi, showing that it was worth investing in public spaces and green infrastructure. Maps were used to provide evidence that this alternative vision was possible (e.g. UN Habitat, 2016). This narrative advocates for a holistic vision of a resilient city. The biggest challenge is entrenched sectoral thinking, especially amongst public sector technocrats.

Discussion
The narrative evidence presented for Manila, Nairobi and Cape Town demonstrates the significance of narrative analysis for revealing value contestation in the city. Despite the fact that resilience was often pictured as consensual, a diversity of perspectives, sometimes contradictory, existed. These narratives were associated with different science methods and have implications for resilience planning and risk governance. Although a few narratives dominated in the different cities, looking across the cities to explore the similarities and differences between narratives is also useful. In the discussion we draw on the elements summarized in Table 1 and reflect on: (i) Relations between knowledge infrastructures and specific narrative positions in Cape Town, Manila, and Nairobi (ii) The multi-faceted roles of science and technology in resilience planning (iii) Implications for the governance of risk and resilience (iv) Ethical issues arising for resilience planning

Knowledge infrastructures in Cape Town, Manila and Nairobi
Knowledge infrastructures are useful to describe the range of actors whose knowledge is made legitimate and underpin the construction of specific narratives. They open up analysis into the relationship between knowledge and power in the city by helping to understand who is given responsibility and the division of labour between different actors. Although the three cities operated with different knowledge infrastructures, in all of them dominant narratives were those organized around technological capacity. In both Manila (M3) and Nairobi (N1) dominant narratives placed expectations for improved policy making and outcomes on enhanced technology for data collection and analysis (e.g. GIS system). Lack of resources and appropriate equipment at the municipal level were perceived as limiting this vision. More specifically, a striking observation in both cities was a fragmented landscape in terms of both distribution of data and expertise. Datasets were decentralised, often held by multiple organisations in diverse locations with much data not being shared, even between government agencies. Fewer datasets and technical human resource were available for public agencies administering low-income areas. Spatial analysis skills tended not to be held by the municipality but rather by higher levels of government, NGOs, consultants and universities. This led to a diffusion of power from the city region upwards to the municipality and outwards to non-state actors.
Where capacity gaps were recognised and relationships between actors were open this led to collaboration, for example in Nairobi where Slum Dwellers International and the Kounkuey Design Initiative generated and shared data on hazard and vulnerability with the municipality. In both Kenya and the Philippines, recent legislation mandating community involvement in risk management (Manila) and decentralisation of government functions (Nairobi) provided a context for continuing collaboration in the construction of knowledge infrastructures. In the Philippines this was motivated by a Department of the Interior memorandum recalling that in the Philippine Disaster Risk Reduction Management Act (2010) Local Government Units had to ensure the "safety and resiliency of communities to face the challenges of disasters" (DILG, Memorandum Circular, April 2016). In Kenya, the 2010 Constitution implemented a strategy of government decentralisation that had given more responsibility to local authorities and sought to enhance public participation. But while devolution processes were on-going, there was a lack of capacity and expertise at the local level and institutional features (such as joint management boards) had not been established making claims about collaboration aspirational more than operational. Interviewees working for the Nairobi City County or Manila City Hall often emphasized that they had growing responsibilities but not necessarily the means to achieve them.
In contrast to Nairobi and Manila, the City of Cape Town had invested significantly in technology and digital infrastructure. Technical expertise was concentrated though combined with a strong ethos of multi-stakeholder participation. All officers had access to a GIS platform and to a wide range of datasets which were regularly updated. The city employed close to 22,000 employees with 21 departments using GIS and roughly 300 using GIS in their everyday work (Steenekamp, 2016). Cape Town was pushing forward the development of a 'smart city' strategy. From the perspective of city residents at risk, however, the high capacity of the City could be alienating. The scale of inequality and political legacy of apartheid had generated expectations (e.g. poverty reduction, settlement upgrade) among the urban poor that had not been met. Knowledge infrastructure was dominated by a technical elite despite its desire for inclusivity. While this may have allowed better planning practices, this infrastructure also encouraged a very technocratic and top down approach to urban planning and hampered the development of more participatory processes (see also Kitchin, 2014).

The multi-faceted roles of science and technology in resilience planning
The role of science and technology in urban planning and resilience depends on how, and by whom, knowledge is mobilized (Fig. 1). Science can either open-up a conversation between different stakeholders or close it down (Stirling, 2008). Fig. 1 groups narratives by a) stated policy purpose and b) predominant science approaches. Almost paradoxically, the very different knowledge infrastructures in place in Cape Town and Nairobi limited the ability of officials to take responsibility. In Cape Town data and technology contributed to the separation of officers from local reality, in Nairobi a lack of data prevented the take up of responsibility.
With regards to a) where mobilizing science and technology in resilience planning was framed to serve modernization ideals (M1, M3, CT1, N1), associated with the idea of progress, with much emphasis placed on the ability of science for security to help anticipate, predict, plan, monitor and control. These approaches were focused and efficient but tended to close down alternative ways of doing and knowing within narrative accounts. A similar observation can be made for Natural Hazards (M1) that falls in the same quadrant, where technology was placed at the forefront of disaster risk reduction. On the other hand, where narratives were organized around social justice (M2, N2, CT2), science and technology were used to give voice to marginalized or vulnerable populations and through this to engage with public authorities and challenge their views. More explicit is CT3 which invoked the language of transformation to position resilience as an agenda for challenging established visions, administrative priorities and practices in the city. A particular variant is N3, where the point of science is to help make a space for natureto open up towards an alternative way of making the city that connects environmental and social sustainability to resilience.
Presented in this way narratives appear to diverge. As shown on across Fig. 1 a and b there is a gap between top-down approaches in b), which overlaps with concerns over modernization and security in a), and bottom up approaches in b), which are more oriented towards social justice and environmental concerns in a). As shown in b), those with social justice purposes also deploy more open science methods associated with transformative and bottom-up approaches; narratives championing engineering and natural hazards management as the purpose of resilience deploy more top-down command and control structures of science with conservative approaches aimed at improving the effectiveness of existing policy rather than opening discussion on alternative futures. However, when looking at more detail into city departments, some city officials already used science and technologies as engagement tools too (e.g. using games and participatory mapping). The potential for lay knowledge and participatory methodologies, or for multiple-values to influence formal planning processes was small, though for many actors this was an aspiration. This suggests that there are already some opportunities to develop more collaborative approaches for resilience planning and policy in all three cities.

Implications for the governance of risk and resilience
Ways of knowing resilience facilitate particular ways of governing and contribute to the distribution of roles and responsibilities to diverse actors (Jasanoff, 2004, Sheppard,1995. Science methods can be mobilized to implement both a transformative view of resilience, oriented towards social justice, or a conservative one, that maintains existing power relations and interests. Yet, as highlighted in Fig. 1 (a) and (b), the use of science and technology is never neutral for governance and policy and contributes to particular understandings of the problems at stake. It therefore matters whose perceptions or understandings of risks are included in resilience programming (Harris et al., 2017). This brings up the question of how compatible the different narratives are in each city. For Manila, there is clearly a contradiction between the predominant narrative M1 and M2 as the focus on natural hazards can go against concerns over social networks and livelihoods. This reflects dynamics in urban governance for Manila, where technical skills and capacity are unevenly distributed between and within city authorities and civil society. In Nairobi short term concerns over economic growth and sectoral approaches hamper more holistic visions such as the one promoted by N3 (Makworo andMireri, 2011, Oyugi andK'Akumu,2007). Capacity for urban planning is so constrained and unbalanced by the requirements of large private developments that while these development may meet global standards their impacts on surrounding risk through overland water flow, water extraction, traffic flow or consequences for surrounding land values and use are not included, so that resilience is separated from strategic urban planning capacity. Similarly, in Cape Town, the lack of systemic thinking and collaboration between different levels of governance and community actors has hampered successful management of flood risks (Ziervogel et al., 2016).
Risk management often follows technocratic, top-down, approaches, as in the case of Cape Town (CT1). This tension between topdown versus bottom-up approaches is recurrent and widely reflected in numerous debates on environmental change (Beck et al., 2014). There is an intimate connection here with the literature on the social construction of risks (Douglas and Wildavsky, 1983): a risk perceived as very important by some actors (e.g. climate change impacts) might not be perceived as such by others with different perspectives and priorities (e.g. poverty reduction). Although, the use of science and technology varies in different places and is socially and culturally shaped, understandings of risk emerging from the natural sciences, delineated according to geophysical criteria, are generally given priority over alternative understandings. Positivist disciplines are often perceived as more credible and authoritative, being able to speak 'truth to power', yet in contrast to this linear conception of science-policy relations many have emphasized the need to redistribute expertise and to recognize local knowledge, for example with regards to flood risks Lane et al., 2011).

Ethical issues surrounding resilience planning
The existence of different conceptions of resilience, and the contacts between different knowledge domains, raises a particular set of ethical issues, to which policymakers and planners involved in developing resilience strategies need to be sensitive. Three such issues are particularly salient. First, there is the risk of epistemic domination as a result of assumed 'expertise' on the part of one group or other (Lane et al., 2011). In order to counter this risk, this paper emphasises the validity and utility of different perspectives on resilience and the need to see each as expressing different forms of expertise rather than treating some views as epistemically privileged over others (Fricker, 2007).
Second, there is the need to recognise that different conceptions of resilience embed different assumptions about distributive justice and other substantive ethical positions, which need to be made explicit and transparent. Similar observations have been made in the field of biodiversity conservation (Martin et al., 2013). There is a rich literature in philosophy and elsewhere on the merits and problems with different models of distributive justicewhether, for example, rights-based, sufficientarian, egalitarian, or prioritarian (Cohen, 1995, Casal,2007Parfit, 1997;Otsuka and Voorhoeve, 2009). By exposing justice assumptions present in different conceptions of resilience, we are able to make progress on understanding and reconciling the normative standpoints of different stakeholders. Third, and relatedly, there is the need to ensure procedural justice in resolving conflicts between different conceptions of resilience. One way to do this is to create common ground (e.g. using deliberative spaces) where disagreements about technical, social and distributive justice issues can be fairly resolved, several narratives embody this call to enhance the use of deliberative space (e.g. CT3, N2).
There are opportunities, across all three cities, to make more inclusive the governance of risks and diversify the visions of each city's resilient future. The rich literature in political theory on deliberation, and on public reason, can provide a valuable perspective on how to take advantage of these opportunities in a fair and reasonable way (Dryzek, 2012, Gaus,2011. Overall, the divergence of views about resilience underlines the central importance of taking a justice-based approach both to academic discussion about resilience and to on the ground policymaking. Only by taking such an approach can we ensure that different perspectives get a fair hearing.

Conclusion
Building on narrative analysis and STS insights we have mapped different meanings that urban resilience can take in Manila, Nairobi and Cape Town. These narratives suggest that urban resilience has become a term-in-common, around which different interpretations backed up by specific agendas and priorities associate. One of the key limits of boundary objects is that they often hide important conflicts, an observation which is consistent with our narrative analysis on urban resilience. Often these differences are not brought to the surface until conflicts emerge later on. This is a missed opportunity for resilience to open up the value systems, development visions and project preferences of multiple actors. Using this opportunity would require different actors to come together to approach resolution in the visioning of resilience and the setting of priorities before detailed project proposals or policy agendas become sources of tension. This also suggests that we need to pay attention to the origins of resilience policies to understand the underlying political and value-based disagreements that are often obscured.
Little systematic work has studied the role of science and technology in shaping resilience policy trajectories. Yet some ways of knowing resilience entail particular ways of governing and distribute roles and responsibilities to different actors. Some views end up being privileged, alienating others. It therefore matters whose perceptions or understandings of risks are included in resilience programming, and how. We show that the use of science and technology is ambivalent. Knowledge can be mobilized both to encourage the implementation of transformative approaches to resilience, inclusive of concerns about social and environmental justice, as well as conservative approaches, driven by concerns over infrastructure and security. Importantly, our results show that the processes of knowledge production matter as much as the knowledge itself: participatory processes that empower different actors tends to produce more transformative outcomes. Fair and open processes are key to ensure that resilience is recognized as legitimate by the people whose future is shaped by these. Our results suggest that there are some opportunities, and demand, for more collaborative and relational approaches that cut across topdown/bottom-up dichotomies for resilience planning in all three cities. We also found a clear interest for more qualitative and creative approaches that can be used alongside more quantitative approaches. Knowledge and tools emerging from the natural sciences are often perceived as more credible and particularly attractivenumbers and maps circulate more easily than qualitative knowledge or local knowledge, which cannot be so easily decontextualized. Yet, this leaves out forms of knowledge that would allow other necessary values and understandings of resilience to be visible. This would help, for example, understand why some population would rather stay in areas labelled as 'unsafe' in geohazards maps. Resilience planning demands greater attention to issues of place, culture, justice, and identity, especially in contexts of deep poverty and inequality. In practice using methods such as deliberative mapping and creative methods, in addition to conventional tools, can help reveal diverse perspectives and values (e.g. Brown et al., 2017;Burgess et al., 2007). This can help develop more pluralist approaches that could map a range of policy options and different futures.
In addition to these implications for urban governance, this paper also suggests that fruitful conversations can be developed between urban studies and STS to better make sense of the challenges awaiting urban sustainability. So far resilience scholars have yet to engage with the knowledge aspects of resilience planning, and STS scholars have paid only marginal attention to urban processes (Farías, 2011;Farias and Blok, 2016). Building on emerging studies on urban knowledge systems (Muñoz-Erickson et al., 2017;Blok, 2013) a wider range of questions can be asked such as: What are the politics of resilience knowledge and expertise? How do different understandings of resilience circulate? Which networks of actors and organizations contribute to the globalization of some narratives while other remain marginal?' With regards to this study this will help understand why some narratives persist while other disappear or remain marginaland with what consequences for the ways in which resilient cities are constructed. This will also facilitate the identification and development of innovative and creative methods to encourage transformative and sustainable urban changes.
Studies. Maud's research looks at the roles of diverse forms of knowledge and expertise for environmental and risk governance. Having conducted her PhD research on biodiversity governance, she currently works on resilience and disaster risk reduction.
Mark Pelling is professor of Geography at King's College London. His research interests are in the institutions and social relationships that shape vulnerability and adaptation to natural disasters, and in the ways in which conflicting values and practices of development inform resilience and transformation in the face of environmental change.
Gina Ziervogel is associate professor at the University of Cape Town with 15 years of experience in the field of adaptation and vulnerability to global environmental change. Through this she has focused on a number of areas including municipal adaptation strategies, institutional barriers and enablers to adaptation, flood governance, water and food security.
Keith Hyams is reader at the University of Warwick and Deputy Director of the Interdisciplinary Ethics Research Groups. Keith's research interests are united by a concern with global challenges facing humanities. His current research includes questions over equity and ethics in adaptation to climate change and international development.