Rethinking Tipping Points in Spatial Ecosystems.

The higher flammability of tropical savanna, compared with forest, plays a critical role in mediating vegetation‐environment feedbacks, alternate stable states, and ultimately, the distribution of these two biomes. Multiple factors contribute to this difference in flammability, including microclimate, fuel amount and fuel type. To understand this transition in flammability, we studied fuel characteristics and microclimate across eight savanna–forest boundaries in south‐central Brazil. At each boundary, the environment was monitored for one week with automated measurements of near‐surface wind speed, air temperature, relative humidity and presence of dew. Manual measurements were performed to quantify fuel amounts and fuel moisture. These data were used to parameterize the fire behaviour model BehavePlus5 in order to simulate fire behaviour over the savanna–forest boundary. There were strong gradients across the boundary in all variables with the exception of total fuel load. During the day, savannas had higher wind speed and air temperature, and lower relative humidity and fuel moisture than forests. Although fuel loads were similar in savanna and forest, savanna was characterized by lower fuel bulk density, largely because of the presence of grasses. Based on these measurements, the fire behaviour model predicted savanna fires to be faster, more intense, and with greater flame lengths, relative to forest. A sensitivity analysis indicated that the primary cause of these differences was the low fuel bulk density characteristic of grassy fuels, with lesser contributions from wind speed, fuel moisture and total fuel load. These results indicate that the dominance of grassy fuels is the primary cause of the high flammability of savanna.

The direct effects of climate change are increasing the frequency of high-intensity fire events in many ecosystems across the globe, including wet Eucalypt forests of south-eastern (SE) Australia. Recurrent high-intensity fire can alter vegetation structure and composition, and the resultant alternative vegetation states may be more likely (positive feedback) or less likely (negative feedback) to burn again than the vegetation community replaced. These indirect effects of climate change have been reported for a range of different ecosystems across the globe. However, a common limitation to many empirical studies is the narrow temporal range of observations, often limited to a single fire season. In turn, this limits our understanding of the potential for vegetation-mediated indirect effects of climate change to generate positive or negative fire feedbacks across the range of climate conditions common to the region. In wet Eucalypt forests of SE Australia, dead fuel moisture content (FMC) is a key determinant of fire activity, and therefore a useful metric on which to quantify the potential for feedbacks across alternative forest states. To quantify potential for indirect effects of climate change to alter future fire activity, FMC was modelled in the open and at seven alternative forest states to wet Eucalypt forest using a process-based FMC model. The model was run using a long-term climate dataset spanning 1973 – 2020, which were transferred from macro- to microclimate values using forest structural properties derived from lidar. Hourly FMC outputs were summarised to fuel availability (FMC < 16% for one hour each day) to understand the potential for positive, negative of no feedbacks on potential fire activity. Mean annual FMC was significantly different between each alternative forest state across each of the 48 years of climate data – which act as a replicate for different climate conditions in our experimental design. By quantifying these differences using the metric of fuel availability, we have demonstrated that statistically significant differences in FMC translate into meaningful differences in the context of potential fire activity. Overall, the results show strong positive and negative feedbacks across the alternative forest states compared to the mature wet Eucalypt forest that they replaced, which were greater than age-related differences within the wet Eucalypt forest sites. Overall, our results support the hypothesis that indirect effects of climate change, acting through vegetation conversion to alternative forest states, have a substantial impact on the potential for future fire activity, with important implications for land and fire managers in this region.

A self-reinforcing positive feedback is regarded as a critical process for maintaining alternative stable states (ASS); however, identification of ASS and quantification of positive feedbacks remain elusive in natural ecosystems. Here, we used large-scale field surveys to search for ASS and a positive feedback mechanism under a wide range of habitats on the Tibetan Plateau. Using multiple methods, we proved that three stable states exist that accompany alpine marsh degradation. Positive feedbacks between changing soil moisture and plant community composition forced the ecosystem into another stable state, and the alteration of water use efficiency (WUE) of the component species contributed to this shift. This study provides the first empirical evidence that positive feedback loops maintain ASS in the alpine marsh ecosystem on the Tibetan Plateau. Our research revealed the powerful driving role of plants in transitions between states, which may support the conservation and restoration of global alpine marsh ecosystems.

We congratulate Drake and colleagues (Drake et al. 1993) for their creative and elegant experimental study of some critical issues in community ecology; we count ourselves among a minority of ecologists who believe that such 'bottle experiments' (Kareiva 1989) can tell us much about the real world. However, we wish to raise some caveats, involving both technical aspects of Drake et al.'s experimental design, and the interpretations of the results proffered by these authors. It is useful to define some terms. 'Community convergence' occurs when one or more communities reach the same 'state', in terms of identities, absolute and relative abundances of constituent species. Historically, this has usually meant convergence to the same point equilibrium, but more recent thinking on higher order attractors (e.g. limit cycles) and stochastic boundedness (Chesson & Ellner 1989) allows a possibility that species abundances might only converge in terms of some dynamic pattern, e.g. to oscillations with the same frequency spectra, or to random variables with the same probability distributions. Communities could fail to converge if external, abiotic conditions differed. More interestingly, communities with the same external conditions would fail to converge if the state space of the governing dynamical system had 'alternative stable states'-more than one locally stable, point equilibrium, or higher-order attractor. It is useful to distinguish cases where one or more of the alternative stable states is on the boundary of the state space, because such boundary attractors correspond to community states in which one or more species are extinct. Such cases we call 'priority effects': the classical example is provided by the Lotka-Volterra competition equations, when each species suffers more from interspecific than intraspecific competition, and the species with a sufficiently large initial relative abundance excludes the other. If alternative stable states occur only in the interior of the state space, communities have the same species composition, but differ in absolute and relative abundances of species. When priority effects occur, communities fail to converge in the identity of constituent species. These are important issues. In the real world, communities are frequently disturbed in ways that alter species abundances. If a community has alternative stable states, such a disturbance can potentially move the community from one state to another. In such cases, good knowledge of local processes and interactions would at best enable one to catalogue a set of possible community states. More exact prediction of community state would require historical knowledge of disturbances. Despite a number of plausible theoretical models in which alternative'stable states occur (Polis, Myers & Holt 1989; Sinclair 1989; Case 1990; Drake 1990; Jones, Hassell & Pacala 1993), and their reported occurrence, including priority effects, in natural communities (Cole 1983; Kneidel 1983; Barkai & McQuaid 1988; Moss 1989; Sinclair 1989; Dublin, Sinclair & McGlade 1990; Knowlton, Lang & Keller 1990; Pech et al. 1992), the existence of, and mechanisms generating these phenomena in the field remain poorly researched. Occam's razor would assume a single equilibrium until proved otherwise. Parsimony has no logical primacy, however. Invasion experiments in laboratory microcosms are ideal tools to look for alternative stable states in communities, because replicated communities under the same external conditions can be created, and invasions provide radically differing initial relative abundances, from which divergence to alternative states might be revealed. Priority effects would be detected if such experimental communities diverged in species identities, and more generally, alternative stable states would be detected if absolute and relative abundances (or their dynamic patterns) diverged. Of course, a sufficiently long period of undisturbed observation of each community is needed to allow convergence to occur, if it is going to. Several generations of the longest-lived resident organisms, or the time required for achievement of the apparent carrying capacity for total biomass, or of a few turnovers of total biomass, are probably minimum estimates of the time required. Where several invasions are imposed on local communities (as in Drake et al. 1993), the time allowed for potential convergence must be counted from the date of the last invasion, since invasions are a form of disturbance, even if unsuccessful, and may slow convergence. Moreover, if convergence to higherorder attractors, rather than point equilibria, is possible, very long runs of data may be needed to assess whether dynamic patterns of species abundances have converged or not. 484

Mohammed Armani.

Nonlinear ecological responses to anthropogenic forcing are common, and in some cases, the ecosystem responds by assuming a new stable state. This article is an overview and serves as the introduction to several articles in this BRIDGES cluster that are directed toward managers interested in dealing with nonlinear responses in freshwaters, particularly streams. A threshold or breakpoint occurs where the system responds rapidly to a relatively small change in a driver. The existence of a threshold can signal a change in system configuration to an alternative stable state, although such a change does not occur with all thresholds. In general, a mechanistic understanding of ecological dynamics is required to predict thresholds, where they will occur, and if they are associated with the occurrence of alternative stable states. Thresholds are difficult to predict, although a variety of univariate methods has been used to indicate thresholds in ecological data. When we applied several methods to one type of response variable, the resulting threshold values varied 3-fold, indicating that more research on detection methods is necessary. Numerous case studies suggest that the threshold concept is important in all ecosystems. Managers should be aware that human actions might result in undesirable rapid changes and potentially an unwanted alternative stable state, and that recovery from that state might require far more resources and time than avoiding entering the state in the first place would have required. Given the difficulties in predicting thresholds and alternative states, the precautionary approach to ecosystem management is probably the most prudent.

The role of fire in expanding the global distribution of savanna is well recognized. Empirical observations and modeling suggest that fire spread has a threshold response to fuel-layer continuity, which sets up a positive feedback that maintains savanna-forest bistability. However, modeling has so far failed to examine fire spread as a spatial process that interacts with vegetation. Here, we use simple, well-supported assumptions about fire spread as an infection process and its effects on trees to ask whether spatial dynamics qualitatively change the potential for savanna-forest bistability. We show that the spatial effects of fire spread are the fundamental reason that bistability is possible: because fire spread is an infection process, it exhibits a threshold response to fuel continuity followed by a rapid increase in fire size. Other ecological processes affecting fire spread may also contribute including temporal variability in demography or fire spread. Finally, including the potential for spatial aggregation increases the potential both for savanna-forest bistability and for savanna and forest to coexist in a landscape mosaic.

Detecting vulnerability of humid tropical forests to multiple stressors

Cellular protrusions are typically considered as distinct structures associated with specific regulators. However, we found that these regulators coordinately localize as propagating cortical waves, suggesting a common underlying mechanism. These molecular events fell into two excitable networks, the signal transduction network STEN and the cytoskeletal network CEN with different wave substructures. Computational studies using a coupled‐network model reproduced these features and showed that the morphology and kinetics of the waves depended on strengths of feedback loops. Chemically induced dimerization at multiple nodes produced distinct, coordinated alterations in patterns of other network components. Taken together, these studies indicate: STEN positive feedback is mediated by mutual inhibition between Ras/Rap and PIP2, while negative feedback depends on delayed PKB activation; PKBs link STEN to CEN; CEN includes positive feedback between Rac and F‐actin, and exerts fast positive and slow negative feedbacks to STEN. The alterations produced protrusions resembling filopodia, ruffles, pseudopodia, or lamellipodia, suggesting that these structures arise from a common regulatory mechanism and that the overall state of the STEN‐CEN system determines cellular morphology.

An application of the Spatial Ecosystem And POpulation DYnamics Model (SEAPODYM) is developed for the North Atlantic albacore (Thunnus alalunga) population. We investigate the spatiotemporal dynamics of this species, distinguishing the influences of environment and international fishing. Incorporating historical data (1960–2008), a maximum likelihood approach allows the estimation of biological parameters (thermal and oxygen tolerance) and stock spatial distribution varying over time. Juvenile albacore are predicted in warm surface waters, whereas adults inhabit cooler and deeper waters. Positive correlations between juveniles and tropical large-scale climate indices highlight the importance of environmental drivers when estimating stock recruitment biology and spatiotemporal distribution. A methodology is proposed to use SEAPODYM outputs to estimate stock abundance and maximum sustainable yield (MSY). MSY is computed taking into account the spatial dynamics of the species and the environmental variability and is based on a mechanistic modelling of larval recruitment. MSY estimates converge towards an asymptotic value (15 997 t) of the same magnitude than standard stock assessment estimates conducted for the international tuna commission. In agreement with all assessment studies, the stock status is estimated from overfished in the 1990s to recovered in the 2000s. Our results show that the stock recovery results both from fishing actions, including total allowable catches established in the 2000s, and from the beginning of a North Atlantic Oscillation warm phase, leading to more favourable recruitment conditions. Following a parsimonious ecosystemic approach, SEAPODYM offers a faithful and spatially dynamic modelling framework that now includes direct tools for spatialized management advice and for distinction between environmental and fishing effects.

Increasingly frequent and intense wildfires threaten to push ecosystems beyond their ecological thresholds. Fire-adapted forests can be vulnerable to ecosystem conversion under shortening fire-return intervals. To evaluate the impacts of fire regime change, we contrasted Eucalyptus forests that retained canopy cover under tolerable regimes (reference forests) with alternative states of markedly reduced canopy. These alternative states established following short-interval wildfires and have persisted through subsequent fires. Alternative states were dominated by more uniform distributions of small trees (< 20 cm diameter) than reference states, with 97% fewer large trees (> 20 cm diameter). These changes in structure translated into declines in several key ecosystem functions. The loss of large, hollow-forming trees and limited Eucalyptus regeneration indicate long-term constraints on habitat availability for hollow-dependent fauna. Coarse woody debris mass was 38% lower in alternative states, indicating decreased resources for detritivores, fungi, and ground-dwelling mammals. Aboveground carbon stocks were reduced by 59%, with a much higher proportion (98%) stored in small trees than reference states (20%). Lower fire tolerance of small trees suggests a greater vulnerability of carbon stocks to subsequent fires in alternative states compared to reference states. These findings highlight the long-term ecological consequences of altered fire regimes. In addition, emphasising the management and restoration of forest structural integrity is essential to maintain ecosystem functions in fire-prone landscapes that are increasingly vulnerable under changing climates.

Theories to explain the diversity of species have required that individual species occupy unique niches and/or vary in their response to environmental factors. Positive interactions within a species, although common in communities, have not been thought to maintain species diversity because in non-spatial models the more abundant species always outcompetes the rarer species. Here, we show, using a stochastic spatial model, that positive intraspecific interactions such as those caused by positive frequency dependence and/or priority effects, can maintain species diversity if interactions between individuals are primarily local and the habitat contains areas that cannot be colonized by any species, such as boulders or other physical obstructions. When intraspecific interactions are primarily neutral, species diversity will eventually erode to a single species. When the landscape is homogeneous (i.e. does not contain areas that cannot be colonized by any species), the presence of strong intraspecific interactions will not maintain diversity.

The possibility that some ecosystems can exist in alternative stable states has profound implications for ecosystem conservation and restoration. Current ecological theory on multistability mostly relies on few-species dynamical models, in which alternative states are intrinsically related to specific non-linear dynamics. Recent theoretical advances, however, have shown that multiple stable ‘cliques’—small subsets of coexisting species—can be present in species-rich models even under linear interactions. Yet, the mechanisms governing the appearence and characteristics of these cliques remain largely unexplored. In the present work, we investigate cliques in the generalized Lotka–Volterra model with mathematical and computational techniques. Our findings reveal that simple probabilistic and dynamical constraints can explain the appearence, properties and stability of cliques. Our work contributes to the understanding of alternative stable states in complex ecological communities.

Ecosystems can undergo abrupt, often irreversible transitions between alternative states —phenomena termed critical transitions or regime shifts— with profound consequences for biodiversity, ecosystem services, and human well-being. Early warning signals (EWS) derived from time series analysis offer the prospect of anticipating such transitions before they occur, potentially enabling preventive management intervention. This review provides a comprehensive synthesis of EWS methods for ecological systems, encompassing theoretical foundations, statistical indicators, empirical applications, and emerging methodological frontiers. We examine the dynamical basis of EWS in critical slowing down theory, wherein systems approaching bifurcation points exhibit characteristic statistical signatures including rising autocorrelation, increasing variance, and spectral reddening. We present a systematic overview of proposed indicators (Table 1), discuss moving-window frameworks for their computation, and critically evaluate preprocessing requirements and sensitivity to analytical choices. Empirical applications across major ecosystem types---including lakes, coral reefs, grasslands, forests, and marine fisheries---reveal both successes and limitations, with EWS performance depending critically on data quality, transition mechanism, and system-specific dynamics (Table 2). We address recent advances including machine learning approaches, non-equilibrium thermodynamic indicators, multivariate extensions, and the important distinction between bifurcation-induced, noise-induced, and rate-induced tipping. We conclude with recommendations for specialists, emphasizing the integration of EWS within broader monitoring frameworks, systematic sensitivity analysis, and the interpretation of indicators as probabilistic assessments of changing resilience rather than deterministic predictions of imminent collapse.

The models investigated in this lab demonstrate how an ecological system can exist in alternative stable states and how the dynamics of alternative stable states change when moving from a site to an entire landscape. In this exam-ple, a site could persist as either a young, frequently burned site (dominated by longleaf pine) or an old, infrequently burned site (dominated by oaks). In-terestingly, however, the dynamics of a particular site changed when the site was placed within the context of its landscape, and the dynamics of a site greatly depended on the specific landscape context in which it was placed. Also, the dynamics of any single site were quite different from those of the entire landscape. When sites were grouped together to create a landscape, a site tended to push its neighboring sites toward its own particular state, re-ducing the range of system parameters for which alternative stable states could persist. In the forests of northern Florida, the ability of fire to spread, and conse-quently the rate at which patches of hardwood or pine either grow or shrink, is influenced by the pattern of hardwoods and pine across the landscape. The combination of spatial heterogeneity and positive feedbacks makes the dy-namics of forests difficult to predict from the study of a local site, since the processes that control a site can be affected by the properties of its neighbors. However, the use of models allows one to examine how the configuration of habitats across a landscape can affect the resilience of landscapes in the face of disturbance. Furthermore, while the models used in this lab are based on the forest dynamics of northern Florida, many ecosystems exhibit alternative stable states (Dublin et al., 1990; Knowlton, 1992; Scheffer et al., 1993). The dynamics of these ecosystems can be analyzed using similar models and techniques.