Regime Shifts in the Anthropocene: Drivers, Risks, and Resilience

Many ecosystems can experience regime shifts: surprising, large and persistent changes in the function and structure of ecosystems. Assessing whether continued global change will lead to further regime shifts, or has the potential to trigger cascading regime shifts has been a central question in global change policy. Addressing this issue has, however, been hampered by the focus of regime shift research on specific cases and types of regime shifts. To systematically assess the global risk of regime shifts we conducted a comparative analysis of 25 generic types of regime shifts across marine, terrestrial and polar systems; identifying their drivers, and impacts on ecosystem services. Our results show that the drivers of regime shifts are diverse and co-occur strongly, which suggests that continued global change can be expected to synchronously increase the risk of multiple regime shifts. Furthermore, many regime shift drivers are related to climate change and food production, whose links to the continued expansion of human activities makes them difficult to limit. Because many regime shifts can amplify the drivers of other regime shifts, continued global change can also be expected to increase the risk of cascading regime shifts. Nevertheless, the variety of scales at which regime shift drivers operate provides opportunities for reducing the risk of many types of regime shifts by addressing local or regional drivers, even in the absence of rapid reduction of global drivers.

trophic levels, and especially sea urchins, the main herbivore, are kept in low numbers. In addition, urchins are prevented from grazing in the kelp forest through the effect of kelp foliage sweeping over the rocks due to its flexibility and the force of waves 5 . Urchin barrens are formed where the kelp canopy has been deforested by urchins. Trophic-level cascades are responsible for urchin overabundance. Overfishing of the apex predators can lead to shifts in the dominance of consumers at lower trophic levels 6 . This has been particularly observed in the case of overfishing of sea otter, cod and haddock, which have released urchins from their predatory controls. Large populations of urchins form grazing fronts that graze down strongly on kelp forest and can survive in adverse conditions by feeding on turf-forming algae 7 .
In some cases trophic-level cascades have allowed other species to reach sufficient population sizes to exert some control on urchins, for instance lobsters and crabs in the North Atlantic 1 . In addition, declines in predatory fish have created a market for urchins, especially in Japan, establishing a new human induced control on the herbivore. In some cases, harvesting of urchins for this market has led to the reestablishment of kelp forest, despite the urchin fishery being prohibited 1,8 .
Turf-forming algae may be considered another regime where opportunistic species with simple and less diverse elements dominate a seascape previously dominated by kelp, a perennial species with structurally complex community 9 . The feedback is given by the ability of turf-forming algae to persist under conditions of elevated nutrients, frequently attributed to coastal urban settlements, inhibiting the recruitment of kelp species 2 .
Once the system shifts from one regime to another, the main ecosystem impact is the loss of habitat complexity due to kelp defoliation. Kelp is a three dimensional structure that offers shelter and food for many species; urchin barrens and turfs do not have this characteristic. This loss is associated with reduction of food web complexity and loss of functional groups 6 , with varying effects on fisheries. Some valuable fish species may diminish since kelp forests provide nursery areas. However, invertebrate species such as lobster and crab have shown increases in population 1 . The abundance of such lower-level consumers reflects an exacerbation of the "fishing down food web" effect 10 . The ecosystem service impacts of algae turfs are likely to be similar to those related to coastal eutrophication. Such effects include abundance of richnutrient environment species as shellfish, bad odors and the associated consequences for recreational and aesthetic values. In addition, kelps support a multimillion dollar industry of canopy-cropping for alginates 1 . This product is commercially important in pharmaceutical and chemical industry.

Feedback mechanisms
The kelp regime is maintained by a healthy food web, usually with four trophic levels that keep urchin populations under control. The urchin regime is maintained by reduction of predators due to lack of habitat or fishing pressure. The turf-forming algae regime is maintained by an environment over-enriched by nutrients either from sediments from land or upwelling nutrients from the deep ocean. Note that canopyforming algae and turf-forming algae are functional groups competing for resources and space in the ecosystem. All feedbacks are local and well established. predator and prey in the food web can be viewed as a feedback. In a nutshell, the more abundant the prey is, the more resource the predator has and viceversa, producing a balancing feedback loop (yellow, brown and red in Fig 2). This feedback is represented in the CLD aggregated across functional groups: between apex predators and lobsters -meso-predators; between the latter and urchins; and between urchins and algae groups. Lobsters and other mesopredators help maintain kelp forests by regulating urchin populations at low densities. ▪ Structure feedback (Local, well established): Canopy forming macroalgae maintain more complex habitat structure that in turns favor the presence of high biodiversity (dark and light blue feedbacks in Fig 2). When diverse predatory species are present in the ecosystem, these predators regulate urchin populations, which maintains kelp forests.

Urchin barren ▪ Predation feedback (Local, well established):
The ecosystem is dominated by urchin barrens when the predation feedback amongst urchins and kelps is strong, while the predation feedback amongst lobster or meso-predators and urchins is weak. It may be related to changes in water temperature that favors urchin barren establishment, or fishing pressure that reduces meso-predator abundance. ▪ Structure feedback (Local, well established): As urchin barrens dominate, less structural habitat complexity is provided by kelp forest. Thus, mesopredators habitat requirements may be affected, reducing their abundance and the predation pressure on urchins.

Turf-forming algae • Competition feedback (Local, well established):
In turf-forming algae dominance regime, the competition feedback is favoring turfs, reducing space, nutrients and light availability for kelps to develop.

Drivers
Two key direct drivers are identified: overfishing functional groups 1,6 and input of nutrients [2][3][4] . The latter is related to deposition of wastewater from urban settlements and agriculture in adjacent catchments. Strong rain events and floods represent shock events for kelp ecosystems given the pulse input of nutrients. El Niño or global warming events may generate water stratification. As consequence, nitrogen concentration declines and kelps become nitrogen limited 1 . In addition pollution discharges and sedimentation may play a synergistic role as stressors. In Tasmania for example, global warming has favored the reproduction of urchins which in synergy with lobster fishing has reduced kelp resilience 3,4 .
Gorman and Connell (2009) also report that the loss of kelp dominated areas undermine kelp's ability to reestablish in disturbed areas. While deforested areas surrounded by kelp patches are more likely to return to the kelp regime, isolated kelp disturbed patches are more likely to stick in the turfed regime.
Important shocks (e.g. droughts, floods) that contribute to the regime shift include: ▪ Rain and floods (regional, well established): Strong rain events and floods represent shock events for kelp ecosystems given the pulse input of nutrients or by perturbing habitat structure. Nutrients in turn unbalance the competition between kelps and turfs favoring the development of the latter; which can use the excess of nutrients faster than kelps and also take advantage of the turbidity conditions generated by nutrients. ▪ ENSO (global, well established): El Niño events or global warming events may generate water stratification. As a consequence, nitrogen concentration declines and kelps become nitrogen limited 1 . In addition pollution discharges and sedimentation may play a synergetic role as stressors.
The main external direct drivers that contribute to the shift include: ▪ Overfishing (regional, well established): overfishing functional groups is one of the most important drivers of kelp transitions 1,3,6,11 . Fishing pressure reduces control of mid predators on urchins favoring the formation of turfs. When fishing is strong enough on urchins, it may favor the formation of turfs as well. ▪ Nutrients inputs (regional, well established): Input of nutrients is another key driver of the regime shift, both natural from deep ocean upwelling or anthropogenic runoff 2 . Nutrients inputs increase sedimentation and turbidity, favoring conditions for turf to outcompete kelps.
The main external indirect drivers that contribute to the shift are: ▪ Food demand (local-regional, speculative): Higher food demand usually stimulates agriculture, both as expansion of agricultural frontier or increase of fertilizers use to increase yield. It also increases fishing pressure on the food web. ▪ Agriculture (regional, well established): Agriculture often requires the use of fertilizers. When soils are eroded or washed, fertilizers run downstream increasing nutrient input to lakes and rivers. ▪ Urban growth (global, well established): Urban growth in coastal zones increases the production of sewage that is rich in nutrients. It also increases the water runoff on the urban landscape, which transports nutrients into coastal water. ▪ Deforestation (regional, well established): Deforestation and poor agricultural management can accelerate, in magnitude and frequency, the nutrient runoff from agricultural lands. Deforestation increases landscape fragmentation and facilitates landscape conversion to agriculture. Both reduce the capacity of the landscape to retain water in the soil, accelerating erosive processes and runoff of nutrients 12 . ▪ Global warming (regional, speculative): Global warming is expected to increase average water surface temperature. It is also expected to increase the gradient between land and ocean temperatures, strengthening winds parallel to the coast and as result increasing upwelling of deep ocean water 13 . This could increase nutrient inputs to coastal ecosystems at the regional scale. On the local scale, global warming has favored the reproduction of urchins which acting in synergy with lobster fishing has reduced kelp resilience 3,4 .

Comparison with other regime shifts and their CLDs
The previous example shows the structure of our data collection framework. Based on a review of the literature we collect information about how the regime shift works, the alternative regimes, important feedback processes underlying the shift, and reported drivers. All information is synthesized in both structured and unstructured text as well as a graphical representation of the system structure (CLD). Here we compare the CLD of Kelp transitions already introduced with another example from the database to exemplify how we deal with different direct and indirect drivers. Direct drivers are defined as those that directly affect the feedback mechanism underlying the shift while indirect drivers only affect other direct/indirect drivers. In the kelp example, fishing and nutrient inputs are direct drivers while water stratification is an indirect driver. However, the position of the driver changes if the system boundaries change to capture the dynamics of other regime shifts. Below we present the CLD for Greenland Ice Sheet collapse (Fig 3), where water stratification is part of the red reinforcing feedback. For Greenland, climate change plays a role of direct driver and greenhouse gases is an indirect one. Both of them, however, appear as indirect drivers in the kelp CLD.
For the purpose of our network analysis, we included all drivers (direct, indirect or within a feedback) and mark its qualitative position in the network as link type (Fig 1  on paper), and quantitatively measure its directedness as the shortest path to feedback loops included in the statistical analysis of exponential random graph models (S2 Table).