Epistemic geographies of climate change

1 Field

Geographers of science have taken a particular interest in the constitution and nature of ‘the field’ as a particular kind of scientific space (e.g. Crang, 1998; Naylor, 2002; Powell, 2002, 2007b, 2008; Lorimer and Spedding, 2005; Matless and Cameron, 2006; Richards, 2011a; Forsyth, 2013). The field, as Kohler (2002) argues, occupies a curious position between laboratory and landscape (e.g. Melillo et al., 2011). It is in some senses a controlled space – selected perhaps for its representativeness or its boundedness, or its amenity for total surveillance. But it also evades control. It is inherently a part of much wider human or nonhuman systems, and thus resists attempts at segregation and insulation (Finnegan, 2008: 380). The field can, however, be a marker of authority – of empirical authenticity against the fabrications and abstractions of the laboratory (Forsyth, 2013; Latour and Woolgar, 1979), and as a source of entwined sensory, experiential and technical knowledge (Powell, 2007b; O’Reilly, 2016). In climate science, the field often disappears from view as the global climate system dominates scientific and political discourses (see Section IV). Yet despite the successful construction of a global atmosphere – through satellite imagery (Benson, 2012) and computer models (Edwards, 2010) – the field has offered resources for an empiricist streak in climate science, which sometimes conflicts with the hegemony of global, computerized simulation.

That the field is a contested source of epistemic authority is well illustrated by Sverker Sörlin’s work on mid-20th century glaciology, meteorology and climate change discourse. For a figure like Swedish glaciologist Hans Ahlmann, the glacier, as an object and site of field science, was the key to authoritative empirical knowledge of potential swings in climate, whether natural or anthropogenic (Carey, 2007; Carey et al., 2016). Ahlmann extended both the laboratory and the human body into the field, enacting a distinctly ‘Nordicist’ culture of field glaciology which integrated concerns for authoritative field observations with a form of scientific internationalism and a tentative receptiveness towards ‘local’ knowledges in Greenland and Iceland (Sörlin, 2011). In anxiously seeking to establish the glacier as a truth-spot (Gieryn, 2006), perhaps recognizing that his knowledge ‘could not escape its embeddedness in the field site’, Ahlmann sought to transcend heroism and self-sacrifice in field expeditions (Hevly, 1996; Carey et al., 2016) in favour of a ‘longue durée of observation’ built upon heterogeneous assemblages of instrumental installations, bodily work, ‘local’ knowledge and laboratory-like practices of precision (Sörlin, 2011: 84). Towards the end of the Second World War, Ahlmann established a new ‘micro-geography of authority’ with an observatory in the Tarfala Valley in Swedish Lapland, which would become ‘the ultimate material manifestation of Ahlmann’s constant interest in the institutionalization of precision and data gathering in the field’ (Sörlin, 2009: 241–2). As arguments about rising polar and global temperatures emerged in the 1940s, Tarfala became a site of empirical dissent to theories of anthropogenic causation, while the theoretical and computational meteorologist Carl-Gustaf Rossby reportedly took pride in having ‘not collected one single empirical observation’ during his long career in Scandinavia and the US (quoted in Sörlin, 2009: 248). ² Tarfala ‘became an internationally recognized field-based bulwark of climate scepticism’ (p. 247) which persisted among Stockholm glaciologists and geographers into the 1980s. Exercising no such ‘cult of local observations’ (p. 249), Rossby’s group, of which the first Intergovernmental Panel on Climate Change (IPCC) chair Bert Bolin was a part, turned their attentions instead to globally representative observations like those of the Mauna Loa CO₂ observatory, and to attempts to model the impacts of such observed changes on global climate (Bohn, 2011; Fleming, 2016). Myanna Lahsen (2013) makes related observations about the cultural histories of empirical, theoretical and computational climatology and their shaping of climate scepticism among US atmospheric scientists.

Sörlin’s history points to how the different geographies of field, laboratory and computer science have been shaped by different epistemic cultures – conflicting standards of evidence, sources of authority, and ways of relating theory to empirical data (Sundberg, 2006; Guillemot, 2010; Lahsen, 2013; see also Heymann, 2010; Rudiak-Gould, 2013) – thus illustrating a broader tension in 20th-century meteorology between ‘those who crunch the numbers, but never look outside’, and those ‘who are unimpressed by equations, but read the sky’ (Harper, 2003: 689). In these conflictual borderlands of field and computer lab, observation and theory, we can see how particular field sites exercise a hold over scientific ambition and imagination. Lahsen (2009) shows how the production of the Amazon rainforest as a workable field site has been shaped by conflicting epistemic and institutional cultures. Janet Martin-Nielsen’s history of scientific engagements with Eismitte – Greenland’s central point – illustrates how narratives of exploration, epistemic authority, political sovereignty and military strategy converge on a particular place, generating a unique ‘ecology’ of social and physical relations which has shaped the evolution of Arctic glaciology and meteorology (Martin-Nielsen, 2013). A comparable ecology can be observed on Antarctica, albeit one shaped by different geopolitical, epistemic and cultural trajectories (Yusoff, 2005; Turchetti et al., 2008; Howkins, 2011; O’Reilly, 2016).

Emerging work is examining the material constitution and practices of field sites in sub-disciplines such as dendroclimatology, from where tree samples are transported along an extended chain of reference (Latour, 1999: 70) from ecological locality, through the laboratory, to indices of regional or hemispheric change (e.g. Ramírez-i-Ollé, 2015; also Schinkel, 2016). Martin Skrydstrup uses Sloterdijk’s spherology (see Van Tuinen, 2009) to examine the architectural grammars of ice core drilling sites, interpreting Greenland’s NEEM station as ‘a modern micro-cosmos, anticipating and building within itself the very sense of a global ecumene to which its scientific work ultimately is addressed’ (Skrydstrup, 2016: 14). The project of producing scalar correspondence between the field and the global is not just epistemological, but cosmo-political.

The broader production of scalar correspondence has proceeded through uneven economies of recognition. Anthropologist Ben Orlove and colleagues argue that biophysical ‘regions’ like the Arctic and small oceanic islands have become particularly prominent parts of the geographical imagination of climatic change, as opposed to arguably equally vulnerable mountain or desert environments. The ‘distribution of concern about climate change impacts is historically situated and constructed rather than solely a reflection of environmental dynamics’, Orlove et al. (2014: 249) argue. Long histories of selective ‘recognition’, influenced by historical geographies of exploration and scientific fieldwork, colonial exploitation and postcolonial geopolitics, have all shaped this uneven geography of concern (Bravo and Sörlin, 2002; Carey, 2007), while contemporary scientific assessments and research programmes reinforce the scientific promotion or neglect of certain regions and processes (also Ford et al., 2011, 2016).

A growing body of work has focused on practices of knowledge-making around climate change adaptation, motivated by an interest in how the global claims of climate science are ‘re-localized’ in relation to local specificities, knowledges and politics (e.g. Bravo, 2009; Carey, 2010; Popke, 2016; Nightingale, 2016). Geographers have developed perspectives on ‘adaptation’ as a travelling idea. These have used concepts from STS and critical policy studies to examine the translation (or spatialization) of a certain ontology of adaptation and its enrolment into local political projects (Yamane, 2009; Weisser et al., 2014), such as reforestation projects and the forced resettlement of farmers in Rwanda (Gebauer and Doevenspeck, 2015; see also Bhatasara, 2015) and ‘best practice’ adaptation to climate change in Pacific islands (Webber, 2015).

‘Adaptation science’ (Moss et al., 2013) can be considered an epistemic ‘trading zone’ (Galison, 1997), where global climate simulation intersects with field-based research on socio-ecological systems and human vulnerabilities. As Castree (2015: 306) notes, this ‘grounded socio-environmental research’ is married with a desire ‘to transfer lessons about inquiry and intervention internationally’. It is a powerful site of co-production, where new knowledges are emerging and traveling alongside new political subjectivities and modes of governmental intervention (Eriksen et al., 2015). We might also consider adaptation science a ‘translation zone’ in Andrew Barry’s (2013b) sense; a political space where certain knowledges and practices may resist enrolment, and where interventions may become unexpectedly politicized. Geographers working in postcolonial and political ecology traditions have critiqued strands of adaptation science for its depoliticization of social vulnerabilities, for seeking to understand who rather than why and to develop strategies of intervention which bracket-out questions of power, inequality and social justice (e.g. Forsyth, 2003; Ribot, 2014). The notion of ‘resilience’ in particular has recently prompted lively debate across and beyond geographic fields (Cote and Nightingale, 2012; Brown, 2013; Welsh, 2013; Simon and Randalls, 2016), while others have critically examined the epistemic construction of adaptation through linear, reductionist conceptions of climate–society relationships (e.g. Hulme, 2011; Beck, 2011; Nielsen and Sejersen, 2012).

Further work which conjoins this concern for the discursive and cognitive aspects of adaptation science (Preston et al., 2015) with a critical engagement with the ‘field sites’ of adaptation research and practice (e.g. Popke, 2016) would add necessary empirical detail to Orlove et al.’s (2014) broad-brush observations of the uneven spatialities of climate change knowledges. Emerging work on the intersection of climate change narratives with the cultural geographies of landscape offers a promising entry point into thinking about climate change in its guises as intellectual artefact, material phenomenon, and ‘embodied and experiential process’ (Brace and Geoghegan, 2010: 296; also Leyshon and Geoghegan, 2012; Matless, 2014, 2016; Rice et al., 2015).

2 Assessment

Much scholarly attention to the politics of climate change knowledges has concerned practices of scientific assessment. The IPCC has been the pre-eminent focus (for an earlier review see Hulme and Mahony, 2010). Such institutions pose immediate challenges to the conventional vocabulary of geography of science. As its own website explains, the IPCC is ‘a huge and yet very tiny organization’ (IPCC, 2013). It is nonetheless ‘peopled and placed’ (O’Reilly, 2015: 108), consisting of government representatives, a small central staff, and hundreds of volunteer scientists who are convened for each sextennial assessment round, either in conference centre meeting rooms or in virtual, online spaces. But the IPCC also performs a form of ‘cultural erasure’ (O’Reilly, 2015: 108). Through stringent practices of review and even ‘audit’, the work of authors is made uniquely visible and scrutable. Knowledge claims are distilled to their global, consensus essences and re-circulated, with studious neutrality, as authoritative and global knowledge. Is it possible to study such an organization using the terms, for example, of the ‘socio-spatial’ of historians and geographers whose concern is the mutual constitution of the physical and social boundaries of scientific sites (Powell, 2007a: 313)?

We can at least begin by re-visiting the more geographically-sensitive work by students of the IPCC. Scholars have pointed, for example, to the continuing geographical imbalances in IPCC participation between countries (Kandlikar and Sagar, 1999; Ho-Lem et al., 2011), to a lack of recognition of ‘indigenous’ knowledge claims about particular places like the high Arctic (Ford et al., 2011), to a lack of governmental interest in the process in certain cases (Biermann, 2001), and to occasions of distrust on the part of certain scientific and political communities in the Global South with regard to problem framings and arguments which do not sit comfortably with alternative epistemic and normative commitments (Lahsen, 2004, 2007, 2009; Fogel, 2004, 2005; Mahony, 2014). The new knowledge claims and framings (Miller, 2000) produced by IPCC assessment processes can be seen as products of particular, situated commitments to forms of epistemic and social order. Russill (2016) has recently argued that the dominance of geophysical modelling and ‘trend detection’ practices, as opposed to more ecological approaches to climate-society relations, can be traced to 1980s US energy politics (also Howe, 2014). The dominance of ‘top-down’, global modelling strategies (Beck, 2011; Hulme, 2011; Nielsen and Sejersen, 2012) and of particular economic rationalities (Randalls, 2011a) subsequently coloured the way issues like carbon sequestration have been presented in IPCC documents. As Fogel (2004, 2005) and Blok (2010) argue, areas of the Global South have been presented as ‘blank’ spaces to be fed into the calculative regimes which underpin efforts to offset Northern emissions through mitigation projects in the South (see Bumpus and Liverman, 2008; Lovell and Liverman, 2010). New practices of carbon accounting are co-produced with re-territorializations of the carbon cycle, creating ‘a peculiar situation whereby a territorial substance…which contributes to a change in the operation of natural systems at a post-territoriality scale, is conceived of, classified, and managed through its association with the persistent territorialities of nation-states’ (Whitehead et al., 2007: 205; Lövbrand and Stripple, 2006; Twyman et al., 2015).

These powerful co-productions of the epistemic and normative content of climate knowledges have offered foci of contestation ‘among competing models of knowledge-making and governance’ (Jasanoff, 2010: 240). As if channelling Livingstone (2005b), Jasanoff (2010: 240) argues that ‘scientific facts bearing on the global environment never take root in a neutral interpretive field; they are dropped into contexts that have already been conditioned to produce distinctive cultural responses to scientific claims’. Reactions to the ‘climategate’ controversy which swirled around the IPCC in 2009/10 offered STS scholars an opportunity to test these claims. Emails hacked from the University of East Anglia appeared to show questionable scientific practices being employed in the communication of results, while a series of errors uncovered in the IPCC’s 2007 report added fuel to the media fire. Jasanoff (2011) argues that the uneven response to the controversy in different countries tells us something about how IPCC knowledge travels. While the IPCC and its scientist-authors came in for heavy criticism in the UK and the US, the episode received scant attention in German media (Beck, 2012). Seeking to go beyond explanations of this which would appeal simply to particular mass media cultures and conservative lobbying activities, Jasanoff (2011) argues that distinct ‘civic epistemologies’ were at play.

The notion of civic epistemology emerged from studies of regulatory science in the interactional tradition of co-productionist inquiry, and builds on the observation that ‘modern technoscientific cultures have developed tacit knowledge-ways through which they assess the rationality and robustness of claims that seek to order their lives; demonstrations or arguments that fail to meet these tests may be dismissed as illegitimate or irrational’ (Jasanoff, 2005: 255; see also Wynne, 2003; Miller, 2004b). These distinctive ‘knowledge-ways’ are embedded in apparently national cultures of public knowledge-making which are historically textured yet seemingly stable over time. An American tradition of pluralistic knowledge-making and ad hoc decision-making in adversarial national institutions can be contrasted with more depoliticized regulatory practices in continental Western Europe. Jasanoff (2011) sees in German civic epistemology a form of risk aversion which disavows the kind of scepticism expressed, especially in UK and US contexts, towards predictions of catastrophic climate change. In these two Anglophone cultures such predictions rub up against respective commitments to ‘common sense empiricism’ and to a form of quantitative objectivity which historically came to prominence amid political efforts to unite a sceptical and divided polity (Porter, 1995; see also Beck, 2012). Although we should be wary of essentializing ‘national cultures’, the concept of civic epistemology offers a means of adding geographical texture to understandings of how and why the boundaries between science and non-science, objectivity and subjectivity, are drawn where they are.

Institutions like the IPCC nonetheless pose challenges to the concept. They are transnational, and involve the participation of government representatives who may bear with them distinctive political or civic-epistemological commitments (Mahony, 2015). It is therefore important to attend to the agonistic processes by which knowledge framings are constructed within such institutional spaces. Peter Haas’s ‘epistemic communities’ model of how shared scientific knowledge is transmitted to policy-makers (Haas, 1992, 2004) arguably pays insufficient attention to the material and social processes by which ‘shared’ knowledge ‘comes to be shared in the first place’ (Miller, 2001: 248). The IPCC provides an ideal setting to study the practices and politics of collective reasoning (e.g. Adler and Hirsch Hadorn, 2014), and of how disputes between different modes of public knowledge-making are played-out. Following the fate of individual knowledge claims, such as particular predictions (O’Reilly et al., 2012; O’Reilly, 2015) or visual representations (Mahony, 2015), has proven a useful methodological opening. However, ethnographic work within organizations like the IPCC is challenging, due both to questions of access and to the dispersed, largely virtual nature of the organization. Few have been welcomed into relationships of ‘ethnographic complicity’ (Marcus, 1998: 105) – O’Reilly (2015), in collaborative relationships with IPCC authors, and Petersen (2006), attached to a government delegation, are two exceptions. Significantly, however, the IPCC has recently adopted a formal process for granting ethnographic access to social researchers (IPCC, 2015). Important opportunities are therefore arising for social scientists to study the epistemic and political complexities of climate knowledge-making inside the IPCC through close, multi-sited ethnographic engagement (Marcus, 1995; Krauss, 2009; Braun, 2006).

3 Conference

A wide range of literature on the production and circulation of climate science on public stages has been emerging (Yusoff and Gabrys, 2011), examining, for example, the place of film (Svoboda, 2016), print and online media (Carvalho and Burgess, 2005; Boykoff, 2011), the theatre (Bottoms, 2012), the courts (Hulme, 2010a; Peeters et al., 2016), museums (Cameron et al., 2013) and television satire (Brewer and McKnight, 2015) in the public adjudication of climate science’s facticity. In this section, we explore a particular type of stage in which geographers, along with political scientists, have recently shown growing interest as a particular site of social and political action – the conference (e.g. Craggs, 2014; Craggs and Mahony, 2014; Hodder, 2015). Carl Death (2010, 2011) has explored the role of ‘summit theatre’ in contemporary forms of advanced liberal government. His Foucauldian lens takes him beyond the critique of high-profile conferences as mere rhetorical smokescreens which obscure true structures of power and authority. Instead, he positions the performative and theatrical politics of conferences as a key element in the government of conduct through a form of ‘exemplary governmentality’ (cf. Rose, 1999). This perspective is instructive for new thinking on the role of the conference as a site of knowledge production and circulation.

Conferences, of both notionally scientific and political kinds, have long loomed large in the politics of climate (see Brenton, 1994; Agrawala, 1998; Orlove et al., 2015; Weisser and Müller-Mahn, 2016). For example, a large scientific conference held in the run-up to the ill-fated climate negotiations in Copenhagen in 2009 offered opportunities to examine the situated production and synthesis of scientific knowledge (Skrydstrup, 2009; Mahony 2013). As interdisciplinary exchanges proceeded, a scientific steering committee was tasked with synthesizing the discussions before presenting them to the Danish Prime Minister on the final day. The conventional rhythms of conference talk and summary response were disrupted by an atmosphere of political urgency – a shared affective telos which structured the material practices of conferencing (Schatzki, 2006; Weisser, 2014). Snap conversations in corridors and email threads joined the gathered-together paper abstracts in shaping and informing the synthetic claims of the steering committee. The presentation of the findings to Prime Minister Rasmussen was a highly choreographed affair. But discursive disruption ensued when Rasmussen disavowed considerations of uncertainty and the apparent destabilization of 2°C as a valid threshold of ‘dangerous’ climate change. The exchanges, filmed for the internet and circulated in transcript form (Baer and Kammen, 2009), were a public performance of, or contestation between, different idealizations of the relationship between science and politics (Skrydstrup, 2009; Mahony, 2013). Nervous laughs and supportive whoops emanated from the audience as Rasmussen pleaded for more certainty and ‘fixed targets’, interpreted by some as justified impatience with scientific reticence, but anathema to more cautious interpreters of climatic indeterminacies and science’s reach. This performance illustrates how at conferences, as at other sites of scientific speech and performance, ‘the setting sets limits on what can be spoken; the social space conditions what is heard’ (Livingstone, 2007: 75). The meeting at this particular conference of ostensibly different ‘social worlds’ (Clarke and Star, 2008) – science and politics – created ruptures in these tacit expectations of speech and discourse. Farbotko and McGregor (2010) make similar arguments about the ruptured divide between spaces of rationality and emotion at the subsequent climate negotiations in December of that year (COP15; see also Weisser and Müller-Mahn, 2016).

A focus on the contested and mutable meanings which become attached to climate science, pursued through textual analyses of scientific and political documents mined for implicit framings and normative commitments (Fløttum and Gjerstad, 2016), would be enhanced through conversation with Florian Weisser’s work on the documentality (Ferraris, 2013) of international environmental negotiations (Weisser, 2014). Documents can be considered as both effects and drivers of particular practices, material arrangements and political contestations, and thus as performative agents in the spatial practices of international diplomacy (see also Kuus, 2014; Craggs, 2014; Dittmer and McConnell, 2016; Dittmer, 2016). This would be to move beyond a conception of documents as mere vehicles of meaning (Riles, 2006; Asdal, 2015). Focusing on the materiality of documentary practices (cf. Schatzki, 2005, 2006) opens new questions about the spatialities of sites like international conferences (Phadke, 2015) and of the meetings, negotiations and performances which constitute organizations like the IPCC (e.g. Hollin and Pearce, 2015). Innovative methodologies such as ‘collaborative event ethnography’ (Campbell and Brosiusa, 2010; Davies et al., 2014) offer guidance on how to bridge the demands of detailed ethnographic attention and the sheer size and complexity of the high-profile ‘performance spaces’ (Livingstone, 2005a: 97; Wainwright and Williams, 2008) of public reasoning about climate change (cf. the digital methods of Venturini et al., 2014).

1 Locating the global

…these models – which predict global climate changes, man’s [sic] imminent decline or, at least, his uncomfortable future – fit conveniently into my pocket: A complete ocean with fishes, the atmosphere with clouds and the ability for turbulent behaviour, countrysides with vegetation, lakes, land and sea ice, the anthroposphere and a lot more existed in my handbag when I left the institute of meteorology. (Gramelsberger, 2006: 78)

General circulation models of the atmosphere, ocean and other ‘earth system components’ have become the dominant, perhaps hegemonic, way in which knowledge of future climates is constructed (Heymann et al., forthcoming). As Gramelsberger (2006) argues, climate models not only describe the world in a theoretical language of dynamic equations rendered in computer code, but also enact the world as an object of experimental research. As if in a Latourian laboratory, an abstract, purified climate system is coaxed into existence from a rarefied computer language and subjected to experimental manipulation, before difficult questions of evaluation, validation and credibility are explored in oftentimes very public ‘trials of strength’ (Latour, 1987).

Scholars of interactional co-production have argued that global climate models have reinforced, and been reinforced by, increasingly globalist forms of politics and spatial imagination (Shackley et al., 1998; Demeritt, 2001; Miller, 2004a; Hulme, 2010b; Gurevitch, 2014). Global mean temperature (GMT) has been the key index of observed and simulated anthropogenic climate change, as opposed to alternatives like radiative forcing or ocean heat content. While estimates of the equilibrium response of the climate system to a doubling of atmospheric carbon dioxide have remained remarkably stable (but still uncertain) over time (Van der Sluijs et al., 1998; IPCC, 2014), so too has GMT endured as the organizing metric of international climate politics: the ‘story of global climate has in many senses become the story of global temperature’ (Hulme, 2010b: 560). GMT has furnished an indexed storyline of change, become a focus of sceptical challenges, and become the locus of normative policy targets, such as the goal of limiting warming to 2°C (or 1.5°C) above pre-industrial temperatures (Randalls, 2010a; Morseletto et al., 2016). The 2°C target has nonetheless faced challenges as both a feasible target (Geden, 2015; Anderson, 2015) and as an accurate, just representation of the threshold of ‘dangerous’ climate change (Shaw, 2016) – challenges which were forcefully aired at the 2015 climate negotiations in Paris and related debates about whether integrated assessment models were smuggling in impossible assumptions about carbon sequestration in order to retain the virtual feasibility and political authority of temperature targets (Geden, 2015; Anderson and Peters, 2016).

GMT is the product of the ‘panoptic gaze’ of global data and models (Barnett et al., 2009), the power of which lies in its making visible of new, governable objects (Scott, 1998; Oels, 2005). It is a gaze which isolates and divides, separating global process from local experience (Head and Gibson, 2012; Jasanoff, 2010; Knox, 2014) while privileging certain synoptic processes and variables over other, more locally-relevant changes (Stott and Thorne, 2010; Bhatasara, 2015; Russill, 2016). Hulme (2014) cautions against the temptation to see GMT as a controllable variable to be purposefully manipulated, in silico or in vivo, with climate engineering technologies (see also Stilgoe, 2015). Like the forests constructed as empty spaces for the disposal of the North’s carbon (Fogel, 2005), a new imaginative geography of the global atmosphere (Doyle and Chaturvedi, 2010) risks enframing it as a space to be colonized by new forms of technologically mediated intervention (Hulme, 2014; Hamilton, 2015).

The fact that climate models can fit into a philosopher’s pocket (Gramelsberger, 2006) means that the mobility and (im)mutability of such tools has become a key element of the spatialization of climate science writ large, and of the physical and social environments of scientific practice (De Laet and Mol, 2000). Climate modelling has a particular geography rooted in post-war scientific collaborations, Cold War computational geopolitics and patterns of government and military funding for science (Doel, 2003; Masco, 2009; Howe, 2014). The global models which have so far entered the IPCC’s intercomparison programmes hail from the Global North, principally the US and Europe. Indeed, Edwards’ (2010: 171) statement that ‘the elite world of global climate simulation still includes no members from South or Central America, Africa, the Middle East, or Southern Asia’ remains true, although groups in India and Brazil are increasingly active. As climate models have developed, sub-modules and parameterization schemes – chunks of code representing particular small-scale processes – have circulated between modelling centres (Gramelsberger, 2006, 2010; Sundberg, 2007; Masson and Knutti, 2011; see also Jankovic, 2004). This ‘interbreeding’ raises epistemological questions about pluralism (Betz, 2009; Lenhard and Winsberg, 2010) and even democracy (Knutti, 2010) in the community of global climate models – are divergences in model projections a healthy sign of epistemic diversity? If so, which models should be trusted more? Models can be considered ‘boundary ordering devices’, regulating the flow of information and the management of uncertainty at the boundaries of different disciplines (Shackley et al., 1998; Lahsen, 2005; Sundberg, 2007), and of science, politics and decision-making more broadly (Shackley and Wynne, 1996; Van Der Sluijs et al., 1998; Lane et al., 2011; Daipha, 2015; Donovan & Oppenheimer 2015). They are emblematic cases not just of technical mutability and interpretive flexibility, but also of efforts to invest trust, credibility and authority in largely automated or ‘autonomous’ (Morrison, 1999) technologies of knowledge production (Van der Sluijs, 2002; Fine, 2009; Wynne, 2010; Hulme 2013; Daipha, 2015). As such, the social coordination of worldwide intercomparison programmes and of the resulting internationalization of climate simulation is worthy of geographical study (cf. Sundberg, 2010; Davies, 2013c)

Comparative perspectives on the epistemic cultures (Knorr-Cetina, 1999) of the elite climate modelling laboratories are in short supply. Simon Shackley’s 2001 study of climate modellers in the US and the UK is a notable exception (see also Krueck and Borchers, 1999). Shackley found the UK’s Met Office Hadley Centre to possess a ‘hybrid’ style of scientific work and policy engagement which was appreciably distinct from that of comparable US institutions (see e.g. Howe, 2014). The UK Government had in the late 1980s resolved to focus funding for climate simulation on one single, authoritative institution which would be close to and answerable to political actors (Mahony and Hulme, 2016). But climate modelling in the US developed in a number of relatively autonomous settings, with rival models, different research orientations and links to policymakers of variable strength (Shackley, 2001; on the UK context see also Shackley and Wynne, 1996; Shackley et al., 1998). The effects of these different ‘epistemic lifestyles’ (Shackley, 2001) can still be detected, for example in the Hadley Centre’s central role in producing national climate projections (now considered a ‘climate service’) for the UK (Hulme and Dessai, 2008; Tang and Dessai, 2012) and its influence on such practices in other national settings (Mahony, 2014).

In contrast, Howe (2014) recounts efforts at the National Center for Atmospheric Research (NCAR) in Boulder, Colorado, to inculcate a particularly collaborative, experimental research culture through architecture, with a building designed, like the NEEM station, to ease interpersonal interactions (Skrydstrup, 2016). NCAR’s veritable revolving door of visiting scientists can be read as a response to an environment wherein atmospheric scientists were wary of co-optation into Cold War research structures, with the institution’s relative autonomy and lively circulation of scientists a form of resistance to the increasingly centralizing tendencies of US science policy (see also Bassi, 2015). Such institutional cultures are formed through the intersection of various geographies: national, perhaps in the form of distinct civic epistemologies (Mahony and Hulme, 2016); regional, in the broader patterns of scientific exchange which constitute trans-local cultures of scientific practice (Livingstone, 2003; Finnegan, 2008; Naylor, 2010; cf. Beck et al., 2013; Fleming, 2016); and global, in the shape of infrastructural globalisms (Edwards, 2006) and attendant moves towards new forms of global, universalizing scientific practice (Miller, 2015).

2 Assembling the region

Questions remain on how to make sense of the links between such institutional geographies and the geographies which are the objects of institutional knowledge-making. Laborde (2015) uses Foucault’s notion of heterotopia (Foucault, 1986) for this purpose (Ophir and Shapin, 1991; Smith and Agar, 1998). Positioning the modelling laboratory as a heterotopic space, Laborde discusses the complex relationships between the physical, segregated spaces where modelling work is conducted and the physical spaces (in Laborde’s case an Italian lake) which are the target of the modelling. Heterotopias are spaces set aside from the rest of society, albeit with multiple, overlapping relationships to other spaces and times (Johnson, 2013) – in the case of modelling labs, these relationships are the representational and cognitive links between model and target systems. Laborde’s use of heterotopia encourages reflection on the practices that make these heterogeneous spatial assemblages ‘hang together’ to produce authoritative knowledge (cf. Turnbull, 2000). Importantly, she encourages ‘a conceptual shift away from cognitive and epistemological differences between communities of practice and instead towards differences between the physical and social environments that support these practices’ (Laborde, 2015: 278). Spaces like the modelling lab, or modelled lakes and Antarctic ice sheets (O’Reilly, 2016), are polysemic, ‘with overlapping layers of different spatial formations’ (Livingstone, 2002: 16), requiring a spatial vocabulary that goes beyond locality and socio-spatial constitution to embrace a more fluid (but not frictionless – cf. Edwards et al., 2011; Rose, 2015) topology of circulating, mutable representations and virtual mediation between different sites and forms of expertise (Law and Mol, 2001; Blok, 2010; cf. Kinsley, 2014; Leszczynski, 2015; Ash et al., 2016).

The ‘universalizing instinct’ of climate science (Hulme, 2010b) is increasingly being supplemented, if not yet superseded, by a new focus on local variations (Stott and Thorne, 2010) and the development of new, locally-focused ‘climate services’ targeted at both public and private sector users (Krauss and Von Storch, 2012; Vaughan and Dessai, 2014; Webber and Donner, 2016). Following Rose (1993), we might say that climate science, as an assemblage of expertise, has moved from the identification and advocacy of new (global) problems to being institutionalized (and localized) as a form of government which acts through the regulated choices of citizens and ‘stakeholders’, and through increasingly marketized forms of substantive expertise (European Commission, 2014; see also Randalls, 2010b). We can therefore further our understanding of the co-production of modelling and modelled spaces by considering how climate models have participated in transforming the episteme of government at multiple scales (Dean, 1999; Whitehead, 2011: 229; Knox, 2014).

For example, PRECIS, a highly mobile regional climate model produced by the Hadley Centre, has since the early 2000s been disseminated or marketed to research institutes around the world, particularly the Global South (Jones et al., 2004). The making-mobile of this complex scientific tool through the production of a distinctive physical and social assemblage (its packaging in an easy-to-use software interface and institutional relationships with bodies like DFID and UNDP) can be seen as an important instance of the spatialization of the arcane practices and technologies of climate simulation (Schaffer, 1991), leading to new forms of national, predictive knowledge being pursued in government science institutes (Mahony and Hulme, 2012; Mahony, 2014). Regional models promise greater spatial realism through the ‘dynamical downscaling’ of global simulations, and we might identity a concurrent downscaling of political concentration. Although other mobile regional models exist (e.g. Pal et al., 2007), the Hadley Centre model is unique in its construction of an extensive network of national, governmental research sites. The spatialization of this model may thus be said to be not just reflective of particular ‘national’ cultures of knowledge-making (cf. Shackley, 2001), but to be part of a broader epistemic transformation of modes of prediction and scenario planning (e.g. Hulme and Dessai, 2008; Rickards et al., 2014) and of a broader regime of calculative and imaginative practices of anticipation (De Goede and Randalls, 2009; Anderson, 2010; Matthews and Barnes, 2016; McQuillan, 2016).

Practices of anticipation act through, and upon, present geographies (Anderson, 2010). The circulation of regional models, for example, has brought new geographies of cross-national scientific cooperation into being, with organizations like the Caribbean Community Climate Change Centre in Belize acting as regional hubs for predictive information and policy advice (see also the geography outlined in Giorgi et al., 2006). As Barnes (2015) has suggested, the act of defining the ‘region’ to be modelled can be deeply political – something recognized by IPCC authors in guidelines for regional modelling: ‘An important geo-political issue may be the importance of national representation in climate models in the context of international negotiations (i.e. it may matter if a country is or is not on the map)’ (Mearns et al., 2003: 21).

Existing studies of the role of models in politics have focused on how information and assumptions flow across social boundaries, or more simply on what kinds of information different decision-makers feel that they need. What is needed now, however, is further work on the actual political effects of new modes of simulation. By this we mean two things: the processes of constitutive and interactional co-production by which different spaces and scales (e.g. the nation, the region, the ‘local’) are jointly constituted as epistemic and political objects (Whitehead et al., 2007; Moore, 2008; Rangan and Kull, 2009); and the effects of this co-production on shifting societal engagements with climate change (Jónsdóttir, 2013; also Ahlborg and Nightingale, 2012; Beck et al., 2014). One debate to which geographers can contribute is about the role of simulation models in the ontological work of defining certain local weather events as anthropogenic or ‘natural’, and the implications of such distinctions for the social distribution of risk and responsibility (Hulme et al., 2011). Geographers can reflect on the extent to which simulation is a distinct practice of representation and of world-making, and how simulation becomes a site of heterogeneous engineering (Law, 1987) of equivalences and differences between objects, artefacts, practices and places in the production of new climatic knowledge spaces (Turnbull, 2000).