Abstract
Regime shifts are large, abrupt and persistent critical transitions in the function and structure of systems (1, 2). Yet it is largely unknown how these transitions will interact, whether the occurrence of one will increase the likelihood of another, or simply correlate at distant places. Here we explore two types of cascading effects: domino effects create one-way dependencies, while hidden feedbacks produce two-way interactions; and compare them with the control case of driver sharing which can induce correlations. Using 30 regime shifts described as networks, we show that 45% of the pair-wise combinations of regime shifts present at least one plausible structural interdependence. Driver sharing is more common in aquatic systems, while hidden feedbacks are more commonly found in terrestrial and Earth systems tipping points. The likelihood of cascading effects depends on cross-scale interactions, but differs for each cascading effect type. Regime shifts should not be studied in isolation: instead, methods and data collection should account for potential teleconnections.
Introduction
Regime shifts occur across a wide range of social-ecological systems, they are hard to predict and difficult to reverse (3, 4), while producing long term shifts in the availability of ecosystems services (5). A regime shift means that a system has moved from one stability domain to another, implying a shift from one set of dominating processes and structures to another set (2, 6–8). Changes in a key variable (e.g. temperature in coral reefs) often makes a system more susceptible to shifting regimes when exposed to shock events (e.g. hurricanes) or the action of external drivers (9). Over 30 different regime shifts in social-ecological systems have been documented (10), and similar non-linear dynamics are seen across societies, finance, language, neurological diseases, or climate (11, 12). As humans increase their pressure on the planet, regime shifts are likely to occur more often and severely (13).
A key challenge for science and practice is that regime shifts, in addition to being hard to predict, can potentially lead to subsequent regime shifts. A variety of cascading regime shifts have been reported (See Table S1). For example, eutrophication is often reported as a preceding regime shift to hypoxia or dead zones in coastal areas (14). Similarly, hypoxic events have been reported to affect the resilience of coral reefs to warming and other stressors in the tropics (15). If, why and how a regime shift somewhere in the world could affect the occurrence of another regime shift, however, remains largely an open question and a key frontier of research (16, 17) that we here aim to address.
Research on regime shifts is often confined to well-defined branches of science, reflecting empirical, theoretical (18) or predictive approaches (9, 19). These approaches require a deep knowledge of the causal structure of the of system or a high quality of spatio-temporal data. Hence research on regime shifts has generally focused on the analysis of individual types of regime shifts rather than potential interactions across systems. This paper takes another approach and instead explores potential interactions among a large set of regime shifts. We propose and investigate two types of interconnections: domino effects and hidden feedbacks. Domino effects occur when the feedback processes of one regime shifts affect the drivers of another, creating a one-way dependency (9, 17, 20). Hidden feedbacks rise when two regime shifts combined generate new (not previously identified) feedbacks (16, 20); and if strong enough, they could amplify or dampen the coupled dynamics. Furthermore, we contrast these cascading effects, where the occurrence of a regime shift give rise to subsequent regime shifts, with the potentially multiplying albeit different effect of two regime shifts being caused by common drivers. Sharing drivers is likely to increase correlation in time or space among regime shifts, but not necessarily interdependence (13, 17).
Hypotheses of cascading effects
Domino effects occur when a variable that belongs to a feedback mechanism in a regime shift acts as driver in another. A feedback mechanism is a self-amplifying or dampening process characterized by a pathway of causal processes that return to its origin creating a cycle. Theory often treats drivers as slow variables, which assumes that their change are relatively slower than changes in state variables (21, 22). Therefore, we expect that domino effects will be mostly dominated by connections from regime shifts that occur at larger spatial scales with slower temporal dynamics to regime shifts that operate at smaller spatial scales with faster temporal dynamics. Conversely, we expect that hidden feedbacks will occur when scales match (both in space and time), because for a new feedback to emerge they need to be somewhat aligned in the scale at which the process operate. Additionally, we expect that drivers sharing is more context-specific; in this case, regime shifts occurring in similar ecosystem types or land uses will be subject to relatively similar sets of drivers, increasing the likelihood of driver sharing.
Discovering how regime shifts might be interconnected requires a method for studying patterns in relational data. We analysed regime shifts as networks of drivers and feedback processes underlying their dynamics. These directed signed graphs allow us to explore driver co-occurrence, directional pathways and emergent feedback cycles of coupled regime shift networks (See SM Methods). The empirical basis for our investigation draws from the regime-shifts database(10), to our knowledge the largest online repository of regime shifts in social-ecological systems (Fig S2, SM Methods). It offers syntheses of over 30 types of regime shifts and > 300 case studies based on literature review of > 1000 scientific papers (10). The database consistently describes regime shifts in terms of their alternative regimes, drivers, feedback mechanisms, impacts on ecosystem services, and management options. It provides a set of 75 categorical variables (Table S2) about impacts, scales, and evidence types (10) that we used to test our hypotheses. Regime shifts are also described using consistently coded causal-loop diagrams as a summary of the drivers and underlying feedbacks of each regime shift (Fig S1). We converted the causal diagram for each regime shift into a network by creating the adjacency matrix A, where Ai,j is 1 if there is a connection or zero otherwise. A link between two nodes in these networks means that there is at least a scientific paper reviewed in the database providing some evidence for such causal relationship (13). Link sign Wi,j represents a link polarity taking –1 if the relationship is expected to be negative, or 1 when positive (SM Methods). Based on the signed graphs, we merged pairs of regime-shifts networks and created three response variables matrices that correspond to each of the cascading effects (Fig 1).
We tested our three hypotheses using as response variable for drivers sharing the number of drivers shared, for domino effects the number of directed pathways that connect two regime shifts, and for hidden feedbacks it is the number of k-cycles (where k denotes cycle length) that emerge on the joined network that do not exist on the separate causal networks (Fig. 1, SM Methods). Each of these response variables are matrices that can be represented as weighted networks in our statistical framework, in which nodes are regime shifts and the link depends on the cascading effect described (Fig. 1). Thus, our research question can be rephrased as ‘What is the likelihood of a link between regime shifts in the response variable network, and what features change this likelihood?’ As explanatory variables we use the regime-shift database categorical variables (N = 75, Table S2, see SM methods for processing of variables). We focused on how similar two regime shifts are based on the categorical variables they share, and how the similarity increases or not the likelihood of having a link. We analyzed patterns of connections using exponential random graph models (SM Method). All these networks contain weighted links of count data; therefore, the specification for the models have a Poisson reference distribution (23).
Driver sharing
Aquatic regime shifts tend to have and share more drivers, although the driver sharing is not exclusively with other aquatic regime shifts (Fig. 2a, Fig S3). The resulting matrix for driver sharing is a one-mode projection of the 2-mode network composed by 79 drivers and 30 regime shifts (Fig 1). The highest concentration of driver sharing is found between regime shifts in kelps, marine euthrophication, the collapse of fisheries and hypoxia. Terrestrial and polar regime shifts tend to have fewer and more idiosyncratic sets of drivers.
Large-scale regime shifts in polar and sub-continental areas (e.g. Monsoon weakening) have fewer drivers but are hotspots of sharing, typically including climate-related drivers. In support of previous results (13), we find that the most co-occurring drivers are related to food production, climate change and urbanization (Fig. 2b). Regime shifts are more likely to share drivers when they occur in similar ecosystem types and impact similar regulating services (p << 0.001, Fig 3, Table S1). The likelihood of sharing drivers is also affected by occurring on similar land use, impacting similar cultural services and aspects of human wellbeing, matching spatial scales (but not temporal ones), as well as having similar types of evidence and reversibility (p < 0.05, Fig 3, see Table S3 for comparison with null and alternative models fitted).
Intuitively, the sharing of drivers is proposed as a potential mechanism that can correlate regime shifts in space and time, but not necessarily make them interdependent (13, 17) unless they also share feedback mechanisms. Only 35% of all pair-wise combinations of regime shifts are coupled solely by driver sharing. Time correlation is debated given that spatial heterogeneity can break the synchrony induced by the sharing of drivers (17), meaning that contextual settings matter for such correlations to emerge. Spatial heterogeneity is also attributed as a mechanism that can smooth out critical transitions and soften their abruptness (24, 25). Yet, with or without masking mechanisms, identifying common drivers is useful for designing management strategies that target bundles of drivers instead of well-studied variables independently, increasing the chances that managers will avoid several regime shifts under the influence of the same sets of drivers (13, 26). For example, management options for drivers such as sedimentation, nutrient leakage, and fishing can reduce the likelihood of regime shifts such as eutrophication and hypoxia in coastal brackish lagoons, as well as coral transitions in adjacent coral reefs.
Domino effects
Domino effects were investigated by searching variables that belong to feedback mechanisms in one regime shift and at the same time are drivers of another (Fig 1, see SM Methods). Despite having relatively fewer drivers, regime shifts that account for most domino effects include the Greenland ice sheet collapse, the Monsoon weakening, primary production in the Arctic Ocean, river-channel changes, soil salinization, weakening of the thermohaline circulation and the shift from tundra to forests (Fig 4). Conversely, the regime shifts that receive most influence through domino effects are mangrove transitions, kelp transitions and transitions from salt marshes to tidal flats. The maximum number of pathways found was 4, and the variables that produce most domino effects relate to climate, nutrients and water transport (Fig 4). In line with our expectations, regime shifts that contain variables that will in turn be drivers of other regime shifts typically have large spatial scales and slow temporal scales: thermohaline circulation collapse, river channel change, monsoon weakening, and Greenland ice sheet collapse. On the other hand, regime shifts that receive the influence are often marine and their time and space dynamics contained more locally. The statistical models support this observation (Fig 3): regime shifts whose time scales are on the range of weeks to months are more likely to receive influence from regime shifts whose dynamics occur on the scale of years to decades (p < 0.1), but we did not find evidence for spatial scales (Table S4). Both, having a link and having a high number of pathways are significant (p < 0.01). The odds of having higher numbers of domino effects are increased when regime shifts impact similar regulating services and similar aspects of human well-being (p < 0.001).
Hidden feedbacks
Contrary to the results found for driver sharing, most hidden feedbacks occur in terrestrial and earth systems (Fig 5) and typically have higher feedback length. The regime shifts with higher numbers of connections (16 to 18 out of 30 possible) are forest to savanna, monsoon weakening, thermohaline circulation, desertification, primary productivity of the Arctic Ocean and the Greenland ice sheet collapse. Key variables that belong to many of these hidden feedbacks are related to climate, fires, erosion, the functional role of herbivores, agriculture and urbanisation (Fig 5). The statistical analysis (Fig 3) shows that there is fewer hidden feedbacks that one would expect by random, but when they do occur, the odds of having multiple feedbacks coupling two regime shifts is 7.33 times higher. The odds of two regime shifts been connected through hidden feedbacks is not affected by occurring on similar land uses or ecosystem types. Yet, the likelihood of hidden feedbacks increases if the pair of regime shifts impact similar ecosystem processes, impact similar regulating and cultural services, and especially if they match spatial scales but not necessarily temporal ones (Fig 3, p < 0.001, see table S5 for null and alternative models fitted), suggesting cross-scale interactions in time.
Out of the 870 pair-wise combinations of regime shifts analysed (when taking into account directionality), ~35% are solely coupled through sharing drivers, ~4% through domino effects, and ~2% through hidden feedbacks (Fig 6). Furthermore, 28% of these pair-wise combinations are coupled through two different types of connections, and 10% by all three of them. Only for 167 (19%) pair-wise combinations we can be certain with the current dataset that there is no cascading effects. However, the discovery of new drivers or feedback mechanisms underlying these dynamics can reduce this estimate.
Discussion
Our results show we can expect ecosystems to experience cascading regime shifts. Regime shifts that occur in similar ecosystem types and land uses are more likely to share drivers, confirming our expectation. Our analysis also suggests that sharing drivers increase when regime shifts operate at similar spatial scales, or when they impact similar provisioning or regulating services. We expected that domino effects will be more common between regime shifts whose dynamics occur at larger spatial scales and slower dynamics in time towards regime shifts more localised in space and faster in time. However, we only found support for the cross-scale dynamic in time but not in space. Additionally, aquatic regime shifts often received the domino effect supporting the idea that water acts as a potential transport mechanism linking regime shifts in one-way interactions (27). Last, we expected that hidden feedbacks would emerge when scales match both in space and time, as well as under similar ecosystem types. Our results only confirm matching in spatial scales and also suggest that hidden feedbacks are common when regime shifts impact similar ecosystem processes, regulating and cultural ecosystem services.
We found 45% of all pair-wise combinations of regime shifts are connected by a plausible domino effect or hidden feedbacks (Fig 6, Fig S3). Domino effects and hidden feedbacks are often disregarded because research on regime shifts is divided by disciplines and tipycally focus on one system at the time. Consequently, data collection and hypothesis testing for coupled systems has largely remained unexplored (16, 17). Our findings align with previous results on the type of variables and processes that might couple distant regime shifts (See Table S1), highlighting the role of climate and transport mechanisms for nutrients and water (Figs 2, 4, 5).
Recent literature (refs. in Table S1) reports potential linkages between euthrophication and hypoxia, hypoxia and coral transitions, shifts in coral reefs and mangroves transitions, or climate interactions. Other examples in the terrestrial realm report potential increase in Arctic warming from higher fire frequency in boreal forest or permafrost thawing. Regime shifts in the Arctic can impact any temperature driven regime shift in and outside the Arctic (28), including the weakening of the thermohaline circulation. Moisture recycling is a key underlying feedback on the shift from forest to savanna or the Indian monsoon; but also has the potential to couple ecosystems beyond the forest that depend on moisture recycling as an important water source. Changes in moisture recycling can affect mountain forest in the Andes, nutrient cycling in the ocean by affecting sea surface temperature and therefore regime shifts in marine food webs, or exacerbation of dry land related regime shifts. These connections between regime shifts represent an emergent literature on regime shifts interactions (Table S1 and references therein). We contribute to this endeavour with a network-based method that allow us to explore plausible cascading effects and distinguish potential correlations from true interdependencies. Distinguishing whether the coupling is expected to be correlational because of driver sharing, a one-way causation (the domino effect), or a two-way interaction (the hidden feedbacks), provides a useful set of hypotheses for future research.
While our method identifies plausible connections between regime shifts, identifying under what conditions plausible becomes probable requires more detailed understanding on regime shifts mechanisms. Empirical studies and modelling syntheses are required to translate our identification of possible mechanisms into context sensitive forecasts. Dynamic models of this type of dynamics require careful assumptions about parameter values as well as functional form of the system’s equations. Generalised modelling is a promising technique that does not require particular assumptions allowing the researcher to reach more general conclusions based on stability properties of the system (29, 30). Another potential avenue for future research is looking at how transport mechanisms couple far-apart ecosystems. One example already mentioned is the moisture recycling feedback (31). Another important social teleconnection (sensu (16)) could be with allocation of resources through international trade, investigating how demand of resources in certain countries can shape the state space of ecosystems from the providing countries. An important lesson from our study is that regime shifts can be interconnected, they should not be study in isolation assuming they are independent systems. On the contrary, methods and data collection that takes into account the possibility of cascading effects needs to be further developed.
The frequency and diversity of regime shifts interconnections suggests that current approaches to environmental management and governance are substantially underestimating the likelihood of cascading effects. More attention should be paid to how the Earth is social-ecologically connected (16), how those connections should be managed, and how to best prepare for regime shifts. Our research suggests that regional ecosystems can be transformed by ecosystem management far away, and conversely, can themselves drive the transformations of other distant ecosystems. Decisions made in one place can undermine the achievement of sustainable development goals in other places. For example, it has been shown that what happens in the Arctic does not stay in the Arctic (28, 32), many Arctic regime shifts have the potential to impact non-Arctic ecosystems far away and the provision of their ecosystem services. It implies that whoever does make decisions on management is not necessarily the one that has to deal with the impacts. This issue is evident in governance of water transport systems, whether run-off or atmospheric transport, but it is applicable to other dynamics connecting far-away ecosystems through other mechanisms such as climate change, fire, nutrient inputs or trade. Our results highlight variables that are key for domino effects and hidden feedbacks. They are also good observables for monitoring early warning indicators of the strengthening of regime-shifts coupling.
Conclusions
Regime shifts occur across a wide range of ecosystems. However, how a regime shift somewhere in the world could affect the occurrence of another regime shift remains an open question and a key frontier of research. We propose two types of cascading effects that can connect different regime shifts: domino effects, and hidden feedbacks; and compare them with the case of driver sharing. To assess these cascading effects among regime shifts we developed a network-based method that identified plausible cascading effects. Regime shifts that occur in similar ecosystems and land uses are more likely to share drivers, and consequently respond similarly to environmental change. One-way directional interactions or domino effects are common but not exclusive to aquatic regime shifts, it highlights the role of water transport and tend to emerge between regime shifts whose dynamics occur at similar temporal scales. Two-way interconnections or hidden feedbacks are more common in terrestrial systems, they occur often between regime shifts that match spatial scales and impact similar ecosystem processes. These results suggest that the Anthropocene can be expected to increase ecological surprise. How and when nonlinear change can be transmitted across space and time in the earth system should be considered in assessments of future environmental change and planning.
Acknowledgements
We are grateful to the contributors, reviewers and developers of the regime shifts database. This work was supported by FORMAS grant 942-2015-731 to JR. JR designed the research, JR and GP curated the data, JR wrote the code and run the analysis with guidance from GP, ÖB and SL; JR, GP, ÖB and SL wrote the paper. Authors declare no competing interests. Data from the regime shifts database is publicly available at www.regimeshifts.org. Upon publication, the version of the database used, curated causal networks, and code will be made available in both the regime shifts database and an online repository (e.g. figshare or dryad). The development version of the code is available at: https://github.com/juanrocha/Domino