Climate change induced socio-economic tipping points: review and stakeholder consultation for policy relevant research

Tipping points have become a key concept in research on climate change, indicating points of abrupt transition in biophysical systems as well as transformative changes in adaptation and mitigation strategies. However, the potential existence of tipping points in socio-economic systems has remained underexplored, whereas they might be highly policy relevant. This paper describes characteristics of climate change induced socio-economic tipping points (SETPs) to guide future research on SETPS to inform climate policy. We review existing literature to create a tipping point typology and to derive the following SETP definition: a climate change induced, abrupt change of a socio-economic system, into a new, fundamentally different state. Through stakeholder consultation, we identify 22 candidate SETP examples with policy relevance for Europe. Three of these are described in higher detail to identify their tipping point characteristics (stable states, mechanisms and abrupt change): the collapse of winter sports tourism, farmland abandonment and sea-level rise-induced migration. We find that stakeholder perceptions play an important role in describing SETPs. The role of climate drivers is difficult to isolate from other drivers because of complex interplays with socio-economic factors. In some cases, the rate of change rather than the magnitude of change causes a tipping point. The clearest SETPs are found on small system scales. On a national to continental scale, SETPs are less obvious because they are difficult to separate from their associated economic substitution effects and policy response. Some proposed adaptation measures are so transformative that their implementations can be considered an SETP in terms of ‘response to climate change’. Future research can focus on identification and impact analysis of tipping points using stylized models, on the exceedance of stakeholder-defined critical thresholds in the RCP/SSP space and on the macro-economic impacts of new system states.


Introduction
In climate change research and communication, the concept of tipping points has received much attention. In the natural sciences, it articulates conditions at which the state of complex systems can abruptly alter as a result of small perturbations (Russill and Nyssa 2009). Concerns have been expressed about passing critical thresholds of large-scale elements of the climate system (Lenton et al 2008, Steffen et al 2018, although there is high uncertainty about the temperature thresholds at which these events occur (Kriegler et al 2009, Levermann et al 2012. Even small changes in climatic conditions can tip ecological systems to alternative states which can be hard to reverse (Scheffer and Carpenter 2003). Beyond the natural sciences, the notion of adaptation tipping points is used to assess under what conditions policies are no longer able to achieve their objectives, requiring additional action (Kwadijk et al 2010, Werners et al 2013. In the transformation and transition literature, tipping points indicate transformative changes in adaptation or mitigation strategies (Wise et al 2014, David Tàbara et al 2018, Farmer et al 2019. Besides their scientific use, tipping points have become a popular metaphor to communicate climate change impacts and their far-reaching consequences to a wider audience (Russill andNyssa 2009, van der Hel et al 2018).
Various studies have given an overview of tipping points in environmental systems, such as the climate system (Schellnhuber et al 2006, Lenton et al 2008, Levermann et al 2012, ecological systems (Scheffer and Carpenter 2003) and the universal warning signals these systems exhibit (Scheffer et al 2009, Ratajczak et al 2018. These studies have a strong focus on the biophysical sphere, and only incidentally touch upon the socio-economic. There is an emerging field on tipping points in coupled socio-ecological systems (Milkoreit et al 2018, Reyers et al 2018). Here, a tipping point is defined as: 'a threshold at which small quantitative changes in the system trigger a nonlinear change process that is driven by system-internal feedback mechanisms and inevitably leads to a qualitatively different state of the system, which is often irreversible' (Milkoreit et al 2018, p 11). However, in nearly all studies on coupled socio-ecological systems, the regime shift is located in the ecological rather than the socio-economic domain (Filatova et al 2016). In line with this observation, several authors have identified a lack of studies on tipping points in the socio-economic domain, whereas they might be highly policy relevant in the context of climate change (Filatova et al 2016, Kopp et al 2016, Kabir et al 2018, Milkoreit et al 2018. The objective of the paper is therefore to help to fill this gap and describe the characteristics of climateinduced socio-economic tipping points (SETPs), in order to guide future research on assessments of these SETPs for informing adaptation and mitigation policy. We do so in three steps. First, we conduct a literature review to create a typology of different branches of the tipping point literature and to define and characterize the SETPs in the perspective of this literature (section 3). Second, we report on a stakeholder consultation that was undertaken to identify examples of SETPs with policy relevance on a European scale and to explore their characteristics (section 4). Third, based on the literature review and stakeholder consultation, we provide suggestions for future research on the assessment of societal impacts of SETPs in order to inform adaptation and mitigation policies (sections 5 and 6).

Methods
This paper combines two methods: literature review and stakeholder consultation.

Literature review
The literature review was structured along two criteria emerging from existing review articles (table SI 1.1 available online at stacks.iop.org/ERL/15/023001/ mmedia): (1) The type of systems on which the tipping point concept is applied: e.g. physical, biological and social systems, which is a major source of discussion on tipping point definitions (Russill andNyssa 2009, Russill 2015).
We searched 'Tipping Point * ' AND 'Climat * Chang * ' in the Web-of-Science core collection database (n=508) (table SI 1.2), enriched by forward and backward citation tracking. To address criterion 1, this literature was positioned in a diagrammatic representation of systems exhibiting tipping points (figures 1, 3), after Barker (2003). To address criterion 2, after Milkoreit et al (2018), three qualitative questions were answered for each branch of literature: (1) How are system states distinguished? (2) What is the role of mechanisms underlying state stabilisation and shifts?
(3) How is abruptness or nonlinearity defined? In addition, the purpose and policy relevance of each branch were addressed. The review process methodologically followed Berrang-Ford et al (2015), see the supplementary information, SI 1. This supplement also contains an overview of model approaches used in studies describing climate-induced tipping points in socioeconomic and coupled socio-ecological systems (table SI 1.3).

Stakeholder consultation
To derive an overview of policy relevant tipping points in Europe, we organized a stakeholder workshop and carried out 28 additional semi-structured interviews resulting in a list of 22 SETP candidate examples (table SI 2.1). The whole process followed a co-design and co-delivery protocol and used a set of research coproduction success factors identified from the literature (Watkiss et al 2018). 24 stakeholders attended the workshop, representing a variety of sectors in Europe, including health, finance, public governance, tourism, agriculture, transport, insurance and nature conservation. Workshop attendants were divided in five thematic groups (table SI 2.2). Each group was asked to draw an inventory of SETPs in their sector, followed by a discussion on causal mechanisms. Similarly, the 28 semi-structured interviews (

Literature review results: types of tipping points
We identified four branches of tipping point literature, that are further explained below (sections 3.1-3.4) and summarized in table 1. In section 3.5, we create a synthesis and typology of these, to define SETPs in section 3.6. The literature on biophysical tipping points is most substantial in size, the other branches are smaller (figure 1).

Climate tipping points
Climate tipping points describe critical thresholds at which large-scale components of the Earth's climate system, at least sub-continental in scale, switch to a qualitatively different state due to a small perturbation (Lenton et al 2008). The classic example is the potential collapse of the North Atlantic deep water formation and the associated Thermohaline Circulation (THC): an ocean current mechanism which is a key determinant of the global climate (Steflen et al 2004). Other examples are: the self-amplifying melting of the West-Antarctic and Greenland ice-sheet, sea-ice melt in the Arctic, and state changes in the El-Nino Southern Oscillation (Levermann et al 2012, Lenton 2013.
Concerning stable states (characteristic 1 in table 1), early models of the THC described two clearly distinguished stable flow regimes and critical transitions (or bifurcations) between them (Stommel 1961, as cited by Lenton 2013, although there is still a debate on whether unambiguous stable states can be distinguished (Caesar et al 2018, Thornalley et al 2018, Nature Editorial 2018. The mechanisms (characteristic 2 in table 1) causing different stable states, as well as abrupt transitions between states, are reinforcing and dampening feedback processes in the climate system leading to nonlinear system behaviour. Abruptness (characteristic 3 in table 1) can be understood in two ways: change is rapid compared to a geological timescale or rapid compared to a policy-relevant time horizon, such as 100 years (Lenton et al 2008, Kopp et al 2016. The policy relevance of climate tipping points is mainly to inform about safe levels of climate change with respect to temperature stabilisation. Climate tipping points seem to have played a crucial role in the history of the Earth, and therefore, they also might have large impacts on current societies (Lenton 2013). This is for example reflected in the IPCC Fifth Assessment Report (IPCC 2014b)-with 'large singular events' as one of the 5 'reasons for concern' and in the IPCC 1.5°C-warming report (IPCC 2018). They also have been used as major justification for ambitious mitigation policies (Knutti et al 2016, Lemoine andTraeger 2016), such as the European Union Long-Term Strategy (EC 2018). Framing of climate tipping points is predominantly negative. They are often seen as being beyond the limits of adaptation (Watkiss et al 2015), so they can only be addressed by insurance-like mitigation (Weitzman 2014).

Ecological tipping points
The second branch of literature is ecological tipping points, better known as 'critical transitions' causing 'regime shifts' in biological or ecological systems (Scheffer andCarpenter 2003, Biggs et al 2018). This branch of literature is much wider than climate change discourse alone, but here, we focus on climate related examples. For example, increasing temperature and nutrient concentrations can tip lakes into eutrophic states (Scheffer et al 2015). Climate change could potentially cause drought-induced vegetation shifts, such as desertification in the Sahara (Martínez-Vilalta and Lloret 2016). Increasing CO 2 concentrations may shift coral reefs to entirely different ecosystems Concerning stable states, the literature emphasises that the term 'dynamic regimes' would be more appropriate, because ecological systems are hardly ever completely stable in the sense that no fluctuations occur (Scheffer et al 2001). Concerning mechanisms, Van Nes et al (2016) distinguish two ways in which an ecosystem can be tipped to an alternative state. The first mechanism is erosion of system resilience. Normally, a small perturbation would have been within the coping capacity due to system resilience, but with the resilience eroded, it can cause a state shift. This highlights that many causal factors may have contributed before a perturbation in one variable causes the actual tipping to occur and these could have additive or even synergistic interactions (Scheffer andCarpenter 2003, Ratajczak et al 2018). The second mechanism is an unusual large perturbation in one driver which on its own pushes the system over a critical point, from which it cannot return (Van Nes et al 2016). This literature emphasises that internal systemfeedbacks play a determining role in both the stabilisation and the rapid change between different dynamic regimes (Scheffer et al 2001). The clearest examples of tipping points are found in small-scale, heterogeneous and strongly interconnected systems . Abruptness is defined as rapid change compared to typical rates of change in the ecosystem (Ratajczak et al 2018).
The policy relevance of this literature is that it gives insights as to how a desired state of an ecosystem can be maintained or achieved by describing a 'safe operating space' (Scheffer et al 2015). For instance, preventing tipping points in iconic ecosystems such as coral reefs received major attention in the recent 1.5°Cwarming report (IPCC 2018). Monitoring and early warning systems can be useful to indicate when ecosystems approach a tipping point, noting these are also relevant for climate tipping points (Scheffer et al 2009, Wang et al 2012, Camarero et al 2015, Clements and Ozgul 2018.

Transformation or transition tipping points
Whereas the above literature mainly draws on complex dynamic systems theory, the literature about transformation tipping points is more qualitative and descriptive, i.e. closer to Gladwell's (2000) popular understanding of tipping points. The roots of this literature are in innovation and change theory (Rogers 1962) which studies why and how ideas and trends spread. It seeks to build on these mechanisms to deliver transitions in society that meet the adaptation and mitigation challenges posed by climate change (Loorbach and Rotmans 2006, Moser and Dilling 2007, van der Brugge 2009).
Distinction of clear stable states is emphasised less in this body of literature. The tipping point is an approximate indicator of the point separating a state A, in which a new strategy, behaviour, idea or technology is only adopted by a minority of early adopters, from a state B, in which it is adopted by a large majority (Loorbach and Rotmans 2006, Moser and Dilling 2007, Patenaude 2011, Sperling 2018. Others define states as institutional settings with different actors and altered actor networks (Westley et al 2011, Fuchs andThaler 2017). In contrast to the biophysical literature, this literature mainly talks about so-called 'positive tipping points' (David Tàbara et al 2018): the envisioned state B is more desirable than the current state A. One difficulty here is that only in retrospect, it is possible to properly describe shifts as being radical (Fuchs and Thaler 2017). Another complication is that these descriptions are subjective; whereas transformative change may pose opportunities for some stakeholders, it may be destructive for others, in particular those that depend on the established system (Young 2012).
The mechanisms that could cause a shift towards a 'fundamentally improved state' (Burch et al 2017) are diverse (Farmer et al 2019). In the mitigation domain, these could include policy instruments or market forces (EC 2018, Mercure et al 2018) as well as financial thresholds for competing technologies (IRENA 2018). However, it can also arise via other means. Westley et al (2011) highlight how institutional entrepreneurs can lower the threshold between the current and the envisioned state, causing tipping to occur. Geels and Schot (2007) articulate that slowly changing drivers may put pressure on an existing regime, so that at a certain point a window of opportunity may cause a breakthrough of a niche innovation. The policy relevance of this literature is that it gives insights into how governments could formulate policies and incentives in order to achieve successful change towards societies that embrace adaptation and mitigation strategies. It provides checklists for the design of 'transformative-

Adaptation tipping points
The fourth branch is the climate adaptation literature. Here, adaptation tipping points indicate thresholds where the magnitude of climate change causes an adaptation action to fail in meeting key performance indicators, policy or management objectives (Kwadijk et al 2010). The origins of this branch are in the adaptation decision support literature and in particular the focus on decision making under deep uncertainty , Ray and Brown 2015, Watkiss et al 2015. This field studies exceedance of threshold conditions under different futures using exploratory modelling approaches such as robust decision making (Hall et al 2012) and scenario discovery (Bryant and Lempert 2010). One of these adaptation decision support approaches is 'dynamic adaptive policy pathways', developed as a scenarioneutral approach to policy making . In order to illustrate these pathways using adaptation route-maps, the approach uses tipping point terminology to indicate the 'best-before' dates of policy actions for meeting a plan's objectives. It is best known from coastal and water management applications such as the Dutch Delta Program (Bloemen et al 2018) and the London Thames Estuary project (Ranger et al 2013), but has also been applied to other sectors including agriculture and forestry (Petr et al 2015, Prober et al 2017.
State descriptions of performance of actions across the pathway are binary: they either meet objectives (state A) or fail (state B). Descriptions of these thresholds can be formal, such as standards or laws, or informal, based on acceptability thresholds as assessed by stakeholders. In climate adaptation studies, the mechanism causing an 'adaptation tipping point', i.e. the exceedance of the threshold, is a change in climate conditions (e.g. sea-level rise (SLR), river discharge), causing a certain action or portfolio of actions to fail. Although this is a sudden and therefore an obvious nonlinear and abrupt switch, it does not necessarily correspond to a nonlinearity in the physical or socioeconomic domain, nor does it have to be the result of internal feedback mechanisms for state stabilisation and change (Werners et al 2013, Kopp et al 2016. However, some shifts in adaptation actions are so transformative in nature that they could qualify as tipping points consistent with the other branches of literature. This is, for instance, explicitly recognised in the IPCC AR5 with the explicit definition of 'transformational' versus 'incremental' adaptation (Kates et al 2012, IPCC 2014a.

Synthesis and typology
Based on the above analysis of the tipping point literature, we propose the following typology of tipping points for climate change (figure 2). The state shifts have different positions in the cause-effect chain from climate change to adaptation and/or mitigation, as shown in figure 3. Climate tipping points highlight that increasing greenhouse-gas concentrations may trigger strong nonlinear responses of large elements of the climate system. Ecological tipping points may have climate change as one of their drivers, and describe cases that lead to state shifts in ecological systems. For both, the state shifts occur in the biophysical system. Transformation tipping points reflect radical changes in human response to the impacts or challenges of climate change, and can lead to different mitigation or adaptation strategies, while also some of the adaptation tipping points reflect climate-induced exceedance of criteria or policy thresholds requiring transformative adaptation. Note that the literature explicitly mentions adaptation tipping points, but not mitigation tipping points. This notwithstanding the ample research addressing the huge technological and behavioural changes implied by decarbonisation pathways leading to a 2 or 1.5°C warming world. Transformational response, encompassing adaptation and mitigation, describes shifts in terms of climate action (by policy makers, private companies or individuals). We thus summarize adaptation and mitigation tipping points under the umbrella of transformational response to climate change, resting in the socioeconomic domain (figure 2).
Note that we contrast these tipping points in terms of transformational response (figure 3, top-right) with tipping points in terms of socio-economic impact (figure 3, bottom-right). Although both occur in the socio-economic system, the state shifts result from different mechanisms. Impact refers to cases where a state shift occurs in the socio-economic structure due to autonomous climate change and inaction or insufficient action from the human side. Response refers to cases where the socio-economic structure is deliberately and fundamentally altered, or planned to be altered, by human action, in anticipation or response to (potential) impacts (see Kates et al 2012). Also, an abrupt shift in response (such as a global carbon tax) may take a long time before it materializes in socioeconomic benefits, whereas the consequences of an impact tipping point may directly hit the economy. In summary, SETPs encompass socio-economic impact tipping points, adaptation tipping points and mitigation tipping points.

Defining climate change induced SETPs
From the above exploration, we are now ready to define a climate change induced SETP. It is 'a climate change induced, abrupt change of a socio-economic system, into a new, fundamentally different state (beyond a certain threshold that stakeholders perceive as critical).' These have the following characteristics: (1) Stable states: substantially different, more-or-less stable states or dynamic regimes at either side of some critical threshold. This distinction can be made qualitatively: showing by narrative how state A is fundamentally different from state B, or quantitatively: showing an S-shaped nonlinearity between climate drivers and socio-economic indicators.
(2) Mechanism explaining nonlinear behaviour. One should be able to provide a rationale-e.g. a causal pathway-on why these states are more or less stable and why there is a sudden transition between the two.
(3) Rapid, abrupt change. The state change should be rapid compared to the change usually observed in the system. Additionally, for the scope of this paper we look for policy relevant examples where climate change is a significant driver of the tipping point. Also, we focus on impact SETPs (rather than transformational response SETPs), because these seem to have least coverage in the existing literature.

Stakeholder consultation results: SETP examples in Europe
Stakeholder consultation identified 22 potential SETPs with policy relevance for Europe (table SI 2.1). Most frequently mentioned during the workshop (5 of 5 groups) was migration towards Europe, due to adverse climate change impacts in other continents. Most groups (3 of 5) also mentioned the reconfiguration of the energy network due to the demand for renewable energy. Other EU-wide examples were: the introduction of new vectors causing the outbreak of diseases; extreme heat mortality and labour productivity impacts due to increasing temperatures; and cascading impacts of road network and supply chain disruptions due to floods. Most of the examples had adverse consequences (20 of 22); only the business group came up with opportunities arising from climate-induced market reconfigurations. In general, the workshop attendees thought about tipping points as large-impact events such as economic shocks and were less concerned with the underlying system dynamics (nonlinear behaviour, feedbacks, threshold effects etc). Intentionally, the setting was one where stakeholders should not be restricted by any concise SETP definition and resulted in a comprehensive collection of concerns. Consequently, not all the introduced examples exhibit all tipping point    (Falk 2013). This has caused societal concerns because of the economic significance of the sector and its contribution to cultural identity (Interview #22, see table SI 2.3).
The state (Characteristic 1, table 2) of an individual resort can be described in terms of financial profitability. The mechanism (Characteristic 2, table 2) causing a state shift can be understood as follows. When a resort does not succeed in making a profit for a number of successive years, it will reduce its reserves and eventually make a decision to withdraw from the sector, or in certain conditions, it may fall into bankruptcy. As a rule of thumb, a resort needs 100 days of good snow conditions a year to be financially viable. Based on this rule, the number of viable existing resorts in the Alpine region could reduce from the baseline 91% (in 2007) to 61% and 30% under warming of 0°, 2°and 4°C, respectively (Abegg et al 2007, as cited in Damm et al 2014. Bankruptcy could occur abruptly (Characteristic 3, table 2), because there is large intra-annual variability in snow conditions on top of the slowly changing trend (Marke et al 2015). In the immediate term, bankruptcy is not necessarily irreversible, and some resorts have been successfully reopened (Falk 2013), but in the face of climate change, the viability of low-lying resorts for snow tourism in the medium to long-term is fundamentally irreversible, irrespective of supporting policies-e.g. the government privatizing unprofitable ski lifts and train lines (Interview #23).
From this case we take the following insights. First, the question whether collapse of a ski resort can be seen as a clear tipping point strongly depends on the scale of analysis. On a village scale, the disappearance of a resort may have large impacts, because a major share of local income is derived from it. On a larger spatial scale, the impacts are usually smaller because of economic diversification and substitution effects when tourism and employment shift to other areas. This is strongly region dependent (Abegg et al 2007), as can be seen by comparing two Austrian regions: Styria and Tyrol. Styria has many low-lying resorts but a diverse economy, so that individual resorts are likely to tip, but their overall economic impact may be limited. In contrast, Tyrol is very much dependent on income from winter tourism, but the resorts are situated at much higher altitudes, making them less likely to tip (Interview #22). Second, adaptation by artificial snow-making may significantly reduce the likelihood of tipping by prolonging the skiing season in the short or even medium term. However, this is very costly and has large environmental impact. Thirdly, there is a potential adaptation strategy to diversify. Several regions are considering a more fundamental shift towards gaining a large share of income from summer tourism. For example, they seek to revive the 'Sommerfrische', a historic fashion where rich families retreated from cities to spend the hot summer at higher altitudes (Interview #23). Whereas higher temperatures threaten winter tourism, it may be opportune in summer when people wish to escape the hot cities in Southern Europe.

Farmland abandonment in Southern Europe
In Southern Europe, climate change is projected to have disproportionally large impacts from the combination of higher temperatures and most likely lower precipitation ( The state (Characteristic 1, table 2) of a piece of land can be described in financial terms. Whether the land provides a viable financial rate of return is determined by land, input and production costs; productivity and quality of outputs; market prices; and ancillary income from subsidies or other farm activities. Furthermore, the most profitable farming activity at any location is dependent on the local climate and biophysical context. Climate change may alter the relative productivity of crops in certain regions, making the crop more favourable (unfavourable) if, on average, climate moves closer to (further away from) the optimum conditions for cultivating that crop (Prishchepov et al 2013). If land eventually is abandoned, this also impacts wider rural communities. At some point, what once was a lively agricultural area may have become what one of the interviewees described as the 'Spanish Laponia': abandoned farmlands, villages with little economic activity inhabited by mainly elderly people and the schools, shops and churches closed (Interview #25). The mechanism causing farmland abandonment (Characteristic 2, table 2) is the choiceprocess of individual farmers, influenced by several feedback mechanisms. With migrants leaving the area, the rural area gets less attractive to live in-especially for younger people, and the infrastructure remains underdeveloped (Interview #25). This process is very hard to reverse, which is reflected by many failed governmental attempts to repopulate rural areas (Interview #6). Concerning abruptness (Characteristic 3, table 2), abandonment may happen either sudden or gradually, depending on the interaction between environmental and socio-economic conditions (Estel et al 2015, Levers et al 2018. From this case we take the following insights. First, the clearest cases of tipping points are found on the community scale, where reinforcing mechanisms can rapidly change the socio-economic structure to a state of abandonment. Second, in many places, adaptation by irrigation has been effective in countering further farmland abandonment and rural exodus (Iglesias et al 2018, Interview #25). Southeast Spain and the Ebro Basin has taken-up a large amount of irrigation, and with that, a change of crops towards water-intensive cultivation such as maize (García-Ruiz and Lana-Renault 2011). This, however led to severe overexploitation of water resources, aggravated by inappropriate crop choices due to subsidy schemes (González-Gómez et al 2012). Considering that the main social contribution of irrigation is rural employment (Gómez-Limón and Picazo-Tadeo 2012), decreasing water availability may also tip some of the currently irrigated areas to abandonment (Interview #26).

SLR-induced migration
SLR constitutes a major threat for the densely populated coastal zones in Europe (Hinkel et al 2010, IPCC 2014b, Vousdoukas et al 2018 and worldwide (Hinkel et al 2014, IPCC 2019. Although flood defences can be upgraded to reduce damage, this is economically efficient during the 21st century for only 13% of the global coastline (comprising 90% of today's global floodplain population and 96% of today's global floodplain assets) . For the protected areas, however, rising sea-levels increase the potential of larger and larger flood disasters in the case of defence failure, because the floodplains, and flood depth in case of a disaster, continue to grow with SLR irrespective of protection. Furthermore, coastal protection often attracts further development in the floodplain, which further increases potential flood damages (Brown et al 2014. After the occurrence of a major flood disaster, affected regions can either decide to rebuild in the same site or to retreat from the coast, which is the tipping point considered here.
The state (Characteristic 1, table 2) of a coastal settlement can be described in terms of its population exposure: high and increasing exposure before and low and decreasing exposure after the tipping point. Three social mechanisms (Characteristic 2, table 2) can cause a state shift from high to low population exposure. The first process is migration, which refers to permanent or semi-permanent voluntary human mobility (Adger et al 2015). Migration is generally a slow process, driven by complex interrelations between many push and pull factors, and there is generally limited evidence that SLR (to date) has caused coastal migration (Adger et al 2015, McLeman 2018. The second process, displacement, refers to the involuntary and unforeseen movement of people due to armed conflicts or disasters (McLeman 2018, Mortreux et al 2018). Displacement is a fast process taking place during, or in the direct aftermath, of a disaster. The third process is relocation or managed retreat, initiated, supervised and implemented by governments (Hino et al 2017, Mortreux et al 2018. It is a slower process than displacement, but it is often initiated after people have been displaced by disaster. With rising sea levels, it is expected that governments will be confronted with this decision more frequently. The change in population exposure induced by both forced displacement and planned relocation following a disaster is abrupt (Characteristic 3, table 2) as compared to normal changes in population exposure due to two reasons. First, displacement always is abrupt as this is caused by a fast flood disaster. Second, relocation decisions by governments generally provide incentives such as buy-outs for households to move out of the hazard zone. For example, in the aftermath of the coastal flooding brought by Xynthia, the French Government offered to fully compensate all homeowners, based on the value of the real estate prior to the storm and most homeowners accepted within a year (Lumbroso and Vinet 2011). These kinds of incentives are in addition to existing individual incentives for migration and thus lead to faster population change.
From this case we take the following insights. First, the SETP of retreat from a coastal zone may be more likely to be triggered by an extreme event (Interview #10). Second, however, it could also be that adaptation to the increasing flood risk in itself causes a reconfiguration of the system (Interview #27, 28) which can be so transformative that it is a 'response to climate change' tipping point. This is for example seen in The Netherlands, where new studies on extreme SLR (DeConto and Pollard 2016, le Bars et al 2017) have triggered the debate on a long-term water management policy (Haasnoot et al 2018). Some radically different strategies are mentioned: the construction of a large dike in front of the entire coast combined with a permanent closure of all estuaries; elevating new buildings above sea level (Aerts et al 2008); or strategic retreat from areas with economic stagnation and population decline (Olsthoorn et al 2008, van der Meulen 2018. Third, in this context it is not only the magnitude, but also the rate of SLR that could cause a policy shift. It typically takes decades to prepare for large infrastructural adaptation measures such as flood defences (Hallegatte et al 2012). The rate of change may exceed the capacity of society to adapt in the traditional way and trigger a shift towards more transformative policies. Table 3 lists the characteristics that SETPs may share, drawing from the stakeholder interviews on other SETP candidates (table SI 2 .3, 2.4).

Generalisation of results
Concerning SETP descriptions, there are two observations. First, it seems that the clearest examples of SETPs are found on small system scales (table 3). This not only holds for ski resorts, but also for agriculture, where interviewees (#2, 5, 6, 24, 25, 26, numbers refer to table SI 2.3, 2.4) could provide several examples of local, abrupt transformations, but hesitated to claim that this would lead to significant country-scale economic effects. Vice versa, even the interviewees that opposed the whole idea of the existence of tipping points in socio-economic systems, agreed that some of the dynamics could be seen on a small system scale, which they however did not consider of enough concern to be called 'real' tipping points (#27, 28). Second, it appears that descriptions of SETPs are prone to a large degree of subjectivity. This can relate to the fundamental question whether there are critical thresholds at all, at which strong nonlinear behaviour occurs. For example, interviewee #20 says: 'In the health sector, we never speak of thresholds, it is not that when a temperature surpasses a certain degree that suddenly something happens.', whereas interviewee #18 gave several examples of outbreaks of infectious diseases following environmental conditions gradually moving beyond threshold conditions. Similarly, Interviewee #3 expects that in case of a catastrophic dike-breach in the Netherlands, 'they will think (K): there is no way of recovery', and the delta will be irreversibly changed. In contrast, interviewee #28 argues that 'even if there would be a large flood event, we probably will rebuild. We will not move to higher grounds. That is so contrary to Dutch nature.' On the same casus, Interviewee #27 highlights another dimension of subjectivity. Coastal retreat could be a tipping point from a social, but not an economic point of view-from which it simply is a continuation of the rational cost-benefit analysis which also underpins the current policy. 'People don't like change, but whether it is a dramatic problem from an economic point of view, remains a question.' Concerning mechanisms causing SETPs, interviewees distinguished several mechanisms that can push the socio-economic system over a tipping point. Most frequently mentioned, was a series of unfavourable conditions, that eventually tips the system into a new state. Above, we illustrated this for ski resorts and interviewees highlighted similar dynamics for farmland abandonment, collapse of insurance markets and willingness to rebuild after devastating hurricanes (table 3). In many cases, these extremes are perturbations on top of a gradual trend moving towards threshold conditions. Many of these thresholds are an interplay between climatic and financial factors: for profitable operation, agriculture needs a minimum amount of water and ski resorts a minimum amount of days with sufficient snow cover. In some cases, this implies that favourable conditions (for example for a certain crop) may shift from one region to another, which on the local level causes relatively abrupt shifts (for example in agricultural practices). Sometimes, governments need to radically reform their policies to maintain certain levels of performance. As we illustrated for the case of SLR, it can be the rate rather than the magnitude of change which exceeds society's capacity to respond and causes a system shift. As interviewee 10 puts it: 'One of the biggest threats to society is that the changes in the climate and the weather happen so rapidly, that we do not, or inadequately, have abilities to adapt to these changes.' In almost every case, interviewees highlighted a combination of climatic and co-determining socio-

Definition and examples of SETPs, description of critical thresholds and description of stable states
The tipping point occurs on a small system scale Agricultural shifts and farmland abandonment happen in certain regions of Spain, but these contribute little to total national agricultural income (6, 24, 25, 26, see 2, 5). Many low-altitude ski resorts may collapse, but the most well-known ski resorts are located at higher altitudes (8,22,23 (8). Sea-level rise induced retreat is more likely in Island states and parts of Bangladesh (10). Heat stress is a larger concern in India (10). Droughts may cause migration out of the fertile crescent (10) Whether the new state is (un)favourable is a subjective issue Disappearing of rivers with large iconic/religious value is a tipping point for certain groups (10). Collapse of certain agricultural systems may be favourable for the environment (24). Collapse of ski resorts could be favoured over adaptation with adverse impacts on nature (23).
Coastal retreat can be very unfavourable for some groups: a social tipping point, but still rational from a cost-benefit perspective: not an economic tipping point (27) Mechanisms leading to the occurrence of SETPs Co-determining drivers contribute to eroding system resilience Depletion of fossil fuels (1), consumer behaviour steering demand for cheap and unsustainable agricultural production (2), K, a main drivers for farmland abandonment are economic and demographic (25, 26), affordability of flood protection is determined by GDP-growth (27, 28) SETP is caused by a gradual increase of risk or economic unfavourability, causing a market/policy reconfiguration Insurance premiums rise above affordability threshold (3, 15). Collapse of a local ski resort because it is unprofitable (8,22,23). Nobody wants to take over a small, unprofitable farm (25). It is no longer cost-effective to protect a coastal community (4, 10, 27, 28). Traffic disruptions rise above formal or acceptability threshold ( economic conditions, which together explain a state shift: the resilience of a system against biophysical change already has been depleted by economic, demographic or social mechanisms (table SI 2.4). Nonlinear positive (self-enforcing) feedbacks in the socio-economic system may accelerate a transition. For example, increasing risk raises insurance premiums, which reduces their affordability, which in turn decreases the insurance uptake, which raises the premiums even further (Interview #3, 15). Similarly, concerns about the profitability of ski resorts may hinder further investments in the winter tourism sector, and rumours about accelerating SLR may worsen the investment climate of a delta which reduces the funds available for flood protection.

Discussion
Reflecting on the results of the stakeholder consultation against what is present in the tipping point literature to date, we find the following. First, change in the socio-economic system is the outcome of complex environmental and socio-economic drivers, as well as policy interventions. Consequently, the role of the climate driver can be difficult to attribute for each SETP example, particularly for impact SETPs. One should be careful not to overemphasize the role of biophysical change while neglecting the co-determining social, political and economic variables. Such a 'naïve' approach has received criticism in the literature on migration, the collapse of ancient societies and the impacts of natural disasters (Butzer 2012, Bettini 2017, Freeman 2017, Soens 2018. A relatively 'slow' process such as climate change will most likely only cause abrupt change of a socio-economic system when the system was already on the edge of its resilience due to co-determining factors, or when there is a very large disaster-like perturbation, or recurring smaller disasters, on top of the gradual trend (van Nes et al 2016). This also implies that impact SETPs are more likely to be found outside Europe, in developing countries with less socio-economic resilience against climate shocks (table 3, see Hinkel et al 2018 for an example for SLR).
Second, there are limitations to the proposed distinction between SETPs in terms of impacts and response. In practice, socio-economic impact and response are closely interrelated. Adaptation and mitigation measures can help to avoid some SETPs, although there are physical, economic and social limits to these measures, so that tipping points cannot always be avoided but possibly delayed. In some cases, the time required to implement adaptation measures becomes a limiting factor. This highlights that not only the magnitude, but also the rate of climate change may cause a tipping point. SETPs can thus occur because the rate of change exceeds society's capacity to adapt. Furthermore, some proposed adaptation measures are so transformative, that their mere implementation can be considered a response SETP.
Third, SETPs are more likely to be found on small system scales. On larger scales, economies are more diverse so that collapsing sectors may be substituted. At the same time, there are cases where small scale tipping points may aggregate to larger-scale socio-economic impacts, for example in a region with many low-lying ski resorts or a region with many vulnerable farms. This is in line with complex system theory, which indicates that homogeneous and interconnected systems are more likely to exhibit wide scale tipping . This also confirms the finding that although societal-wide collapse is seldom observed, natural disasters may have a large local impact on individuals and communities (Soens 2018).
Fourth, in this context, the stakeholder workshop illuminated tensions between different ways of framing tipping points. The concept had large policy resonance because of its popular meaning as 'large impact event'. This, however, caused a disregard of the systemdynamic properties of tipping points among workshop participants. In contrast, strict system-dynamic definitions of tipping points are found on small system scales in which policy makers showed less interest from a macro-economic perspective. Consequently, stakeholder consultation does not guarantee that all collected examples align with a (stricter) tipping point definition and should therefore be complemented with a concise evaluation of the characteristics of the examples supplied, to identify those that classify as tipping points. Nevertheless, these academia-stakeholder conversations were perceived as helpful. For academia, because it helped to describe indicators for system states in a policy-relevant way. For policy makers, because dialogue on system dynamics enhanced system understanding and helped to find better management strategies to avoid impact SETPs.
Fifth, including socio-economic dynamics adds an extra layer of complexity compared to the studies which only include biophysical dynamics. Prediction of occurrence and likelihood of SETPs is complicated by (a) the complexity and dynamic character of the socio-economic system and (b) the fact that humans proactively and autonomously may alter the system. In response to changing conditions, the socio-economic system might deliberately have been changed before the tipping was predicted. There is an increasing body of regime shifts literature in which these biophysical and socio-economic dynamics are studied in an integrated way under headings like CHANS (coupled-human and natural systems); SES (socio-ecological systems) and CHES (coupled human-environment systems). The model approaches in this literature (such as system-dynamic and agent-based models, see table SI 1.2) are suitable to study the dynamics of stylized cases to increase system understanding. So far, however, this literature has little coverage from the perspective of climate change . Also, because of its often stylized character, it gives little concrete suggestions for adaptation and mitigation strategies for policy makers.
To illustrate the last point, consider the potential impacts of accounting for SETPs in cost-benefit-analysis (CBA), a common tool for guiding climate policy (Tol 2003). Accounting for climate tipping points may significantly alter CBA because the extra economic damage justifies more radical mitigation policies, such as a higher carbon tax reflecting the higher social cost of carbon (Lenton and Ciscar 2013, Cai et al 2015, Van Der Ploeg 2016. Similarly, any additional damage from passing impact SETPs increases the costs of inaction, making mitigation measures more cost-effective. On a local scale, the inclusion of SETPs will mainly increase the benefits of local adaptation measures. Furthermore, when evaluating a management strategy that could lead to an impact SETP, one should account for additional transfer costs to an alternative strategy at some point in the future (Haasnoot et al 2019).

Conclusion
This paper set out to define and characterize SETPs, to identify policy-relevant examples and to give recommendations for further research. We showed how SETPs can be defined along the same characteristics as biophysical tipping points and suggested a distinction between: (a) SETPs in terms of transformational response to climate change and (b) SETPs in terms of socio-economic impacts. Impact SETPs, where slowonset changes in the biophysical system cause strong nonlinear behaviour in the socio-economic system, have least coverage in the literature. The stakeholder process delivered 22 candidate SETPs, which mostly have some general coverage in the literature, but not specifically from a tipping point perspective.
How could further SETP research support decision making on climate adaptation and mitigation? We propose the following research agenda: • For small-scale SETPs, develop models (e.g. agentbased or system-dynamic) that can represent the nonlinear dynamics causing regime shifts in socioeconomic systems upon gradual change in the biophysical system.
Use these models to assess the effectiveness of different adaptation and monitoring strategies.
Investigate the different mechanisms identified in the above stakeholder consultation, notably under what conditions the rate-of-change (rather than the magnitude of change) causes an SETP.
Investigate how, and under what conditions, small scale SETPs may scale up to wider socio-economic impacts.
• Study the wider (macro)economic impacts of SETPs by representing the new states in partial equilibrium models of specific sectors, or in general equilibrium models to study economy-wide consequences. Investigate how the likelihood and impact of SETPs can be reduced by mitigation policy.
• When SETPs cannot explicitly be modelled or quantified, use stakeholder consultation to identify potential thresholds indicating SETPs and to investigate their causal mechanisms.
• Explore the exceedance of these thresholds in climate and socio-economic scenarios.
• Use stakeholder-thresholds to define maximum levels of change (e.g. a temperature limit, or several consecutive unprofitable years) at which SETPs most likely will not occur.
Finally, it is clear that impact SETPs could have major policy consequences but have been omitted from many European and national impact studies to date. Including them in the academic literature and in policy discussion is therefore a useful addition for informing sound mitigation and adaptation strategies.

Data availability statement
Any data that support the findings of this study are included within the article or the supplementary information.

Funding sources
This paper has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 776479 for the project CO-designing the Assessment of Climate CHange costs. https://coacch.eu/