Indicators

Identification of the indicators

2017-05-17T10:56:12+00:00

Contributing Authors:	Sonia Kéfi, Florian Schneider, Alain Danet, Alexandre Génin, Angeles G. Mayor, Susana Bautista, Max Rietkerk, Koen Siteur
Editor:	Jane Brandt
Source document:	Kéfi, S., Schneider, F. Danet, A., Génin, A. Mayor, A. G., Bautista, S. Rietkerk, M. Siteur, K. 2017. Report on indicators for critical thresholds. CASCADE Project Deliverable 6.2, 26 pp

1. Generic early warning signals

Theoretical studies have suggested that a number of ‘generic’ indicators could be derived based on a phenomenon that appears to be universal prior to bifurcations (i.e. points at which the stability of a systems changes, such as a tipping point): critical slowing down [27]. Critical slowing down means that the time needed for a system to return to equilibrium following a small disturbance gets longer as the system approaches a bifurcation point (Figure 1b-c). In other words, closer to a bifurcation, the system has a harder time recovering from perturbations, and the capacity of the system to absorb perturbations without shifting to a different state decreases.

" data-placement="top" data-content=" Figure 1: Generic early warning signals" title="">

Far from the tipping point resilience is high (b): the ecosystem lies in a steep basin of attraction. Small disturbances are damped by high recovery rates back to equilibrium. Rate to recover from perturbations is high (b), the dynamics are characterized by low variance (d), and low correlation between subsequent states (d). Close to the tipping point resilience is low (c): the ecosystem lies in a less steep basin of attraction. Rate to recover from perturbations is low (c), the dynamics are characterized by high variance (e), and high correlation (e). Figure modified from [28].

Critical slowing down has direct statistical signatures that have led to the definitions of generic early-warning signals of ecosystem degradation [21]. First, critical slowing down can be assessed by measuring the recovery rate of the system upon a disturbance (which should decrease as a system approaches a bifurcation point) [29] (Figure 1b-c). Second, slowing down leads to an increase in variance prior to a tipping point: the state of the ecosystem should fluctuate more widely around its equilibrium [30] (Figure 1d-e). Third, there is an increase in autocorrelation: the state of the ecosystem resembles more its previous state when it is close to a bifurcation point [31] (Figure 1d-e).

Results highlights

Theoretical models predict that recovery rate, variance and autocorrelation are statistical properties of the system dynamics that change in predictable ways prior to bifurcation points in general, and tipping points more specifically.

The generic early-warning signals have been developed and tested in a number of models (e.g. [32]). In harsh environments such as drylands, recruitment of woody plants often depends on nurse plants that ameliorate stressful conditions and facilitate the establishment of seedlings under their canopy. For example, C. Xu, S. Kéfi and colleagues [33] used an individual-based model and demonstrated that these facilitative interactions may cause a treeless and a woodland state to be alternative stable states on a landscape scale if nurse plant effects are strong and if the environment is harsh enough to make facilitation necessary for seedling survival (Figure 2A). A corollary is that under such conditions, environmental change can bring drylands to tipping points for woody plant encroachment (path 4-3-1 in Figure 2A) or woodland collapse (path 1-2-4 in Figure 2A). We showed that the proximity of tipping points can be announced by the generic early-warning indicators, i.e. by slowness of recovery of woody vegetation cover from small perturbations (because of critical slowing down; Figure 2B, C) as well as by elevated temporal and spatial auto-correlation and variance (Figure 2D-G).

" data-placement="top" data-content=" Figure 2: A. Change in woody cover as a function of environmental harshness." title="">

As harshness increases, the woodland collapses catastrophically into a treeless state (path 1-2-4). After a collapse, a decrease in environmental harshness can lead to an abrupt recovery of the woody vegetation following the path 4-3-1. The ecosystem therefore exhibits two tipping points located around point 2 (woodland collapse tipping point) and point 3 (woody encroachment tipping point). B-C: The recovery rate of the ecosystem upon small perturbations slows down towards both tipping points (dashed lines). D-G: Temporal indicators of critical slowing down. Variance (standard deviation, D-E) and temporal correlation (lag-1 autoregressive coefficient, AR1, F-G) in simulated time series rise towards tipping points (dashed lines) for a shift from high to low woody cover (woodland collapse) as well as for a shift from low to high woody cover (woody encroachment). Figure modified from [33].

The genericity of early warning signals. Most studies on the generic early-warning signals had initially focused on models exhibiting tipping points and catastrophic shifts, and it was unclear how these indicators behaved in systems approaching other types of bifurcations. In particular, it was unclear whether the early-warning signals were specific to catastrophic shifts or whether they could also occur in cases of abrupt but reversible ecosystem responses. In CASCADE we tested the behavior of the generic early-warning signals as a model system approached different types of bifurcations [34].

We found that all indicators showed consistent patterns for a variety of bifurcations. In particular, we found that the generic early-warning signals were not specific to catastrophic bifurcations but also preceded non-catastrophic transitions [34]. The generic early-warning signals can generally be detected in situations where a system is slowing down, i.e. becoming increasingly sensitive to external perturbations, independently of whether the impeding change is catastrophic or not.

Results highlights

Slowing down and its statistical signatures can generally be used as indicators of degradation, also in systems where we have no reason to expect catastrophic transitions. Our results also imply that indicators specific to catastrophic shifts are still lacking.

2. Spatial generic early warning signals: indicators specifically based on spatial information

Studies have shown that slowing down in space takes place in an analogous way to slowing down in time [35,36]. Spatial variance and spatial correlation between near-neighbors are expected to rise as a system is approaching a bifurcation point. However, in models that show a strong spatial structure, like drylands, it has been shown that most of the generic early warning signals can fail [32]. In such cases, indicators specific to those systems need to be developed.

In drylands, vegetation is characterized by spatial patterns formed by the isolated vegetation patches interspersed with bare soil. In addition to the spatial generic early warning signals, studies have suggested that changes in the spatial vegetation patterns themselves could be indicative of environmental deterioration in semi-arid ecosystems [24,26]. In particular, the shape of the vegetation patches [1,37] and the distributions of vegetation patch sizes could indicate that an ecosystem is degrading [25,26].

In CASCADE we reviewed these spatial indicators of ecosystem degradation suggested by the theoretical literature [23] (early-warning signals and patch-based indicators), and we developed a methodological framework for the practical quantification and the interpretation of these indicators on real data (Figure 3).

" data-placement="top" data-content=" Figure 3: Flow chart of the analyses to perform on a spatial data set to quantify indicators of degradation along stress gradients. Figure from [23]." title="">

We developed a statistical toolbox in the free programming R environment whose code is freely available online as well as a webpage aiming at describing and explaining the various indicators (in time and space) and their theoretical foundation, giving some concrete examples of case studies and references from the literature (see »Early warning signals toolbox).

3. Refining the understanding and use of indicators

Patch-based indicators and spatial stressors. The theoretical foundations of early warning signs of catastrophic shifts had so far assumed that pressures on ecosystems distribute homogeneously in space. While this may be valid for some pressures, it is most certainly not true for others such as livestock grazing, which is not only a major human supply factor, but also a primary trigger of desertification. In CASCADE we developed a dryland vegetation model including grazing and its spatial component (see »Simulated pressures and ecosystem responses; [8]), and we investigated the behaviour of the spatial indicators of degradation in this model.

Our model analysis shows that spatially-explicit grazing disrupted patch growth and put even apparently 'healthy' drylands under high risk of catastrophic shifts (Figure 4). Our study highlights that the spatial indicators of degradation can fail in ecosystems where the pressure is spatially heterogeneous, such as grazed drylands. Our results may very well generalize to other ecosystems exhibiting self-organized spatial patterns where a spatially-explicit pressure disrupts pattern formation.

" data-placement="top" data-content=" Figure 4: Landscapes classified based on the cumulative patch size distribution of the vegetation" title="">

(i: full cover; ii: up-bent power law with spanning clusters; iii: pure power-law; iv: down-bent power law; v: desert) along gradients of environmental and grazing pressures. Note that at high grazing pressure, a vegetation collapse was not preceded by down-bent power laws. Figure adapted from [8].

Periodic patches and patch adaptation in response to environmental changes. CASCADE partners [38] studied a simplified vegetation reaction-diffusion-advection model producing periodic vegetation patterns. Model studies revealed that patterned ecosystems may respond in a non-linear way to environmental changes, meaning that gradual changes can lead to sudden desertification. In Siteur et al. [38] we studied this response through a novel stability analysis of patterned vegetation states. We found that, besides direct critical transitions through decreased rainfall, patterned vegetation states may also adapt, depending on the rate of environmental change and the amount of noise. Rapid environmental change and lack of noise resulted in a drastic critical transition (top right panel), while patterns could adapt in the case of slow environmental change (top left panel). In the model, the vegetation patterns adapted to environmental change in two ways: 1) by adapting biomass while the wave number of the periodic patches remained the same and 2) by adapting wave numbers. We were able to construct so-called ‘Busse balloons’, showing the surface in parameter planes for which stable patterned vegetation states can be found (grey area). These findings shed a more nuanced light on the earlier suggestions that regular patterns in those systems would indicate bistability and proximity to catastrophic shifts [24,37]. Indeed, these model results suggest that ecosystems may adapt and catastrophic shifts may be avoided, if environmental changes are sufficiently slow.

Including rainfall intensity. How annual and seasonal rainfall volumes in arid and semi-arid regions will change in the coming decades is subject to much uncertainty, according to global climate model projections. In contrast, projections of changes in rainfall intensity show strong trends. Rainfall intensity has an important impact on spatial infiltration patterns of water in patchy arid ecosystems [39,40], and it is unknown if and exactly how the projected changes in rainfall intensity are going to affect the productivity and functioning of patterned semiarid ecosystems.

CASCADE partners [41] performed a model analysis to address that question and concluded that projected increases in rainfall intensity could induce and enhance alternative stability of semi-arid ecosystems. We also found that under certain conditions both an increase and a decrease in mean rainfall intensity could push the system over a critical threshold, resulting in a regime shift to a bare desert state. This finding was attributed to the fact that water can be lost from the system in two ways. During high intensity rain events, a fraction of the water flows through the vegetation bands and is lost as runoff, while during low intensity events a large portion of the water infiltrates in the bare interbands, where it is less available to plants and can eventually be lost due to soil evaporation and percolation.

Results highlights

Considering rainfall intensity as a variable may help in assessing the proximity to regime shifts in patterned semiarid ecosystems, and monitoring losses of resources through runoff and bare soil infiltration could be used to determine ecosystem resilience.

4. Developing new, additional indicators

In addition to the early-warning signals and the patch-based indicators, new indicators were also suggested in CASCADE, which are hereafter presented (see Table 1 for an overview of all these indicators).

Table 1: Table summarizing the different indicators studied in WP6.

Indicator name	Description	Advantages	Drawbacks	References
Cover	Percentage of the ground covered by vegetation	Easy to understand, easy to measure	Fails in the case of ecosystem shift	[18–20]
Temporal generic indicators	Temporal variance, auto-correlation at lag 1 and temporal skewness calculated on time series of a variable of the ecosystem state (e.g. cover, abundance of a key species…)	Works independent of the type of changes expected in the ecosystem (i.e. both with continuous degradation and catastrophic shifts)	Requires to use more advanced statistical tools (but tools freely available); Requires detailed time series of the variable	[21,22]
Spatial generic indicators	Spatial variance, near-neighbour correlation and spatial skewness calculated on spatial data (e.g. aerial images on which vegetation abundance or presence/absence can be estimated in space)	Works independent of the type of changes expected in the ecosystem; Only a few spatial snapshots in time are required to get an idea of the trend that an ecosystem follows through time	Requires to use more advanced statistical tools (but tools freely available) Requires spatial data or maps with sufficient resolution	[21,23]
Patch-based indicators	Metrics quantifying the shapes, sizes, distribution of patch sizes present in the landscape and power law range (PLR, i.e. the proportion of the distribution that fits a power law)	More powerful than generic indicators for patchy landscapes (i.e. landscapes with a strong spatial structure); Works independent of the type of changes expected in the ecosystem	Only works on patchy landscapes; Requires to use more advanced statistical tools (but tools freely available) Requires spatial data or maps with sufficient resolution	[23,25,26,32]
Flowlength	Metric quantifying the connectivity of bare-soil areas in the landscape, which is a proxy for how much resource can be lost from the system.	More powerful than cover for patchy landscapes; Works independent of the type of changes expected in the ecosystem	Only works on patchy landscapes; Requires to use more advanced statistical tools	[16,42]
Network-based indicators	Metrics calculated on spatial data after transformation them into a network of interaction. In such network the mean and variance of the node degree are followed.	More powerful than generic indicators	Requires to use more advanced statistical tools	[43]

Hydrologically-based indicators: Flowlength (connectivity-based indicators). Vegetation cover and pattern, and therefore the bare-soil connectivity, largely determine runoff and thereby the potential of the ecosystem to conserve (or leak) resources such as water, soil and nutrients. An indicator of degradation based on bare-soil connectivity was developed, referred to as Flowlength [42]. Flowlength measures the connectivity of bare-soil areas in a given landscape (by calculating the average of the runoff pathway lengths from all the cells in the system), and thereby estimates the potential of the landscape to lose resources. Flowlength assumes that bare-soil areas behave as sources of runoff and sediments that are trapped by downslope vegetated areas, which behave as sinks of resources.

The effect of landscape resource loss (estimated with Flowlength) on plant establishment was simulated in a dryland vegetation model [16]. Our model analysis showed a non-linear inverse relationship between bare soil connectivity (here Flowlength) and vegetation cover. This means that if bare-soil connectivity increases above certain values (for example because of cover loss), a disproportional loss of resources would take place, greatly limiting plant establishment. This results in a positive feedback which accelerates the shift of the ecosystem into a degraded state (see »Simulated pressures and ecosystem responses; [16]). In other words, considering the effect of bare-soil connectivity on vegetation recruitment increases the probability of catastrophic shifts in dryland.

Our results further suggest a higher sensitivity of the bare-soil connectivity index (Flowlength index) to changes in the spatial organization of the vegetation during the transition to a degraded state, in comparison with bare-soil (or vegetation) cover, which shows a rather linear evolution during this transition. This means that bare-soil connectivity could be a better indicator of degradation that bare soil.

Results highlights

Changes in vegetation pattern and associated hydrological connectivity may be more informative early-warning indicators of dryland degradation than changes in vegetation cover. An acceleration of bare-soil connectivity observed in spatially-explicit time-series data may therefore provide an early warning of imminent shift.

Network-based indicators. A vegetation model that exhibits a catastrophic shift to desertification was investigated, and spatio-temporal data (i.e. a simulated field of vegetation biomass) was translated into a network of interactions [43]. A network is defined by two sets of objects, the so-called nodes, and the set of their mutual connections, namely their links.

The nodes were defined as the biomass grid cells of the discretized model. To define the links between the nodes, the zero-lag temporal correlations between the biomass time series at the different nodes were considered. More precisely, two nodes were linked, if the temporal cross-correlation of the time series of two nodes were statistically different. The most basic characteristic of a network is called its degree distribution, for which the degree of a node is defined as the number of links of the node. We followed the changes in network properties, here the mean and the variance of the node degrees, changed along the transition to desertification.

We found that the average and variance degree showed a markedly increase when decreasing rainfall in the model, before it collapsed at a certain rainfall rate. Our study suggests that basic network characteristics could offer novel indicators for identifying an upcoming desertification in semi-arid ecosystems [43].

Results highlights

Comparing the performance of these network-based indicators with the generic early-warning signals based on variance and autocorrelation, we found that network-based indicators were more sensitive to the presence of the transition point. The network based indicators hence offer a promising alternative to detect ecosystem degradation.

Note: For full references to papers quoted in this article see

» References

Validation of the indicators

2017-05-17T10:57:22+00:00

Contributing Authors:	Sonia Kéfi, Florian Schneider, Alain Danet, Alexandre Génin, Angeles G. Mayor, Susana Bautista, Max Rietkerk, Koen Siteur
Editor:	Jane Brandt
Source document:	Kéfi, S., Schneider, F. Danet, A., Génin, A. Mayor, A. G., Bautista, S. Rietkerk, M. Siteur, K. 2017. Report on indicators for critical thresholds. CASCADE Project Deliverable 6.2, 26 pp

The development of early warning signals to detect the onset of regime shifts in marine and terrestrial ecosystems has received increasing attention during the last decade. The theoretical interest for these indicators has created a novel and promising framework for studying tipping points in ecological systems. The challenge, however, is whether these indicators can be applied in reality.

To evaluate the indicators, and more precisely the patch-based indicators, as indicators of dryland degradation, we used a data set from another European project, BIOCOM, coordinated by Fernando Maestre (Madrid, Spain) [44]. This work is part of the Ph.D. thesis of Miguel Berdugo, co-supervised by Sonia Kéfi, Fernando Maestre and Santiago Soliveres. The database contains vegetation and soil data of 224 drylands from all around the world. For each site, the dataset contains the estimated plant cover, the frequency of positive plant-plant interactions, 16 soil variables (related to the carbon, nitrogen and plosphorous cycles) hereafter called ‘functions’, and the aridity index (AI, precipitation/potential evapotranspiration).

From these sites, we retained for this study those from which we could gather »»Google EarthTM or »»VirtualEarthTM images good enough for visually identifying vegetation patches. The resulting 115 sites used for the analyses are located in 13 countries and differ widely in their abiotic (elevation, temperature and precipitation) and biotic (vegetation type, cover and number of species) features.

We used the combination of remote sensing and field data to evaluate the links between vegetation cover, patch-size distribution and multifunctionality (the ability of ecosystem to provide several soil fertility related services at the same time; it was measured as the average Zscore of the 16 soil variables; see [44] for a description of the approach).

We found that the observed vegetation patch-size distributions always fitted heavy-tailed distributions with varying levels of curvature (see »Identification of the indicators). Distributions showing strong curvatures have a relatively low proportion of patch sizes that fit a power law (i.e. a low Power Law Range, hereafter referred to as PLR). These curvatures are caused by the lack of the largest and/or of the smallest vegetation patches compared to what would be expected in a pure power law: PL-like sites whose patch size distribution fits best a power law, and ii) and non PL-like sites which had more curved distributions.

Moreover, we found a bimodal distribution of multifunctionality values in our field sites, which suggests contrasting multifunctionality states in global drylands. This can be interpreted as the existence of two alternative states in multifunctionality in global drylands. More specifically, mapping the number and value of estimated alternative states along the aridity gradient studied reveals a range of aridity values (between 0.2 and 0.4, meaning that 1-AI is between 0.6 and 0.8) for which two multifunctionality levels coexist across our sites (Figure 7). The type of patch-size distribution was significantly associated with the two multifunctionality states observed (PL-like sites in the upper branch and non-PL-like sites in the bottom branch).

" data-placement="top" data-content=" Figure 7: Relationship between aridity (x-axis) and multifunctionality (y-axis)." title="">

Variation of the ‘stable’ states (i.e. local minima of the stability landscape (black line) along the aridity gradient studied for multifunctionality. AI = aridity index (annual precipitation / annual evapotranspiration). Contour lines represent the estimated potential energy from which the ‘stable’ states are derived as local minima, i.e. blue color represent more stable states and red color represent less stable states. Figure from [45].

Changes in patch-size distributions indicate a spatial reorganization of the existing cover, which is related to processes influencing the functioning of drylands, such as soil erosion. Modifications in spatial patterns can also reflect important variations in the structure of plant communities unrelated to changes in cover.

Results highlights

While plant cover is the best linear predictor of multifunctionality in global drylands, patch-size distributions are better reflecting non-linear changes in this variable. Our findings support the use of vegetation patterns as functional indicators in drylands, and pave the way for developing effective strategies to monitor desertification processes.

Note: For full references to papers quoted in this article see

» References

Results, implications and outlook

2017-05-17T10:57:45+00:00

Contributing Authors:	Sonia Kéfi, Florian Schneider, Alain Danet, Alexandre Génin, Angeles G. Mayor, Susana Bautista, Max Rietkerk, Koen Siteur
Editor:	Jane Brandt
Source document:	Kéfi, S., Schneider, F. Danet, A., Génin, A. Mayor, A. G., Bautista, S. Rietkerk, M. Siteur, K. 2017. Report on indicators for critical thresholds. CASCADE Project Deliverable 6.2, 26 pp

1. Results

A number of indicators of ecosystem degradation are currently available in the literature (see Table 1 in »Identification of the indicators for those studied in CASCADE). So-called generic early-warning signals are simple metrics (return rate after a perturbation, variance, correlation) based on the phenomenon of critical slowing down, which occurs when a system approaches a bifurcation point, i.e. a point at which the system stability is going to change drastically. These indicators can be quantified on both temporal and spatial data. In the case of spatially-structured ecosystems, such as drylands, these generic indicators have been shown to be very likely to fail [32], and additional indicators, based on the ecosystem spatial structure have been suggested: in particular the shape of the vegetation patches and the shape of the patch size distribution.

In CASCADE, we reviewed these indicators, and proposed a work flow of how to quantify them on real data (see »Identification of the indicators; [46]). The code to apply these indicators on ecological data (e.g. aerial images of the landscapes) has also been made available (see »Early warning signals toolbox).

These indicators were tested in a number of more particular cases, and we identified situations in which they are expected to fail. In particular, by comparing their behavior along different types of transitions, we showed that they are not specific to catastrophic shifts, but also occur along non-catastrophic transitions [34]: they are therefore indicators of ecosystem degradation because their detection points to the fact that the ecosystem is having a harder time recovering from perturbations, but they do not indicate what type of transition the system is approaching. A model study taking the spatial component of grazing into account showed that this mechanism, by affecting vegetation patch growth, affected ecosystem resilience (by increasing the probability of catastrophic shifts at high grazing pressures) and the efficiency of patch-based indicators at announcing upcoming ecosystem degradation [8]. This grazing model analysis warns about the blind use of the patch-based degradation indicator without knowing the characteristics of the stressor and their interactions with the intrinsic mechanisms of the ecosystem. Another model study focusing on rainfall intensity, one of the major changes expected in dryland climate in the coming decades, suggested that explicitly considering rainfall intensity may help in assessing the proximity to regime shifts in patterned semiarid ecosystems, and that monitoring losses of resources through runoff and bare soil infiltration could be used to determine ecosystem resilience [41].

A number of studies performed in CASCADE additionally proposed new indicators or approaches. Using a dryland vegetation model, including erosion feedbacks, Mayor et al. [16] suggested that changes in bare-soil connectivity along a degradation gradient (resulting from changes in both plant cover and spatial patterns) may be more informative than changes in plant cover as early-warning indicators of dryland degradation. This is in agreement with recent empirical evidence [48]. Moreover, we found that basic network characteristics could offer novel indicators for identifying an upcoming desertification in semi-arid ecosystems and that the performance of these network-based indicators could be superior to these of the generic early-warning signals based on variance and autocorrelation [43].

Finally, the last task of the CASCADE modelling work was to evaluate these indicators on real data in an attempt to validate their use and efficiency. To do this, we used a large-scale data set from another European project, BIOCOM, in which we could quantify patch-based indicators on 115 dryland sites located world-wide and compare them to field-based measurements reflecting ecosystem functioning (summarized in a metric called multifunctionality). We found that abrupt changes in multifunctionality along an aridity gradient could be reflected by the patch-size distribution of vegetation. By providing the first link between plant spatial patterns and multifunctionality in global drylands, our study provides strong empirical and mechanistic support to the use of these patterns as indicators of discontinuous changes in ecosystem functioning.

2. Implications for management

The results of this part of CASCADE have a number of practical implications in terms of predicting dryland degradation.

Our results provide support for the use of indicators based on the spatial structure of the vegetation cover (patch-size distribution, Flowlength) to assess the ecosystem degradation level.
Our results nonetheless warn about the need for well identifying the main stressors at play in the ecosystem considered (e.g. rainfall and grazing) since they can affect the type of indicator to follow and their reliability.
Our studies have put forward a number of new indicators (Flowlength and network-based indicators) that need further testing and validation in future studies.

Jointly, all those indicators, when simultaneously evaluated and if they all converge in their trends, can help identifying the critical point at which measures should be adopted to prevent drastic changes in ecological conditions before they happen. These spatial indicators can be evaluated on spatio-temporal ecosystem data that are becoming increasingly available through e.g. aerial images.

More globally, our results suggest that ecosystems with aridity indices between 0.2 and 0.4 are especially sensitive to further disturbances [45]. In areas where aridity is expected to reach such values in the future [49] or where grazing is rising due to a higher demand in livestock products, such increased pressures could force the sites in this sensitive climatic envelope into a low multifunctionality state (i.e. degradation).

Results highlights

A key result of our study is that these abrupt changes in multifunctionality can be reflected by the patch-size distribution of vegetation, which is related to critical changes in the way dryland ecosystems are organized.

3. Outlook

Our results also pave the way for more systematically testing these indicators, in various dryland sites (worldwide) and under various drivers, since our model analyses suggest that the nature of the driver and its characteristics can affect the efficiency and the reliability of the indicators. Steps in that direction have already being initiated in CASCADE (e.g. analyses of spatial images from CASCADE field site by Utrecht University Ph.D. student Myrna de Hoop).

Simultaneously, the statistical tools needed to evaluate these indicators needs to be developed, tested and made available so that they can be widely applied. As already mentioned, tools and information about them have already been made available by CASCADE and these tools will keep being updated.

A key element currently lacking from the validation of the indicators is a quantitative measure of the pressure at play. In the work of Berdugo and colleagues [45], the indicators of ecosystem degradation have been clearly correlated with metrics reflecting ecosystem functioning (so-called multifunctionality), but no measure or information about the pressures at play in the different field sites available were available. Again, a step in that direction will be taken by the upcoming study from Myrna de Hoop since dung counts have been measured in the field in that case and can constitute a proxy for the level of grazing pressure. Moreover, quantifying anthropogenic pressures is an explicit goal of a newly funded European project on desertification, BIODESERT (coordinated by Fernando Maestre).

Note: For full references to papers quoted in this article see

» References

Early warning signals toolbox

2017-05-17T10:58:49+00:00

Contributing Authors:	Sonia Kéfi, Florian Schneider, Alain Danet, Alexandre Génin, Angeles G. Mayor, Susana Bautista, Max Rietkerk, Koen Siteur
Editor:	Jane Brandt
Source document:	Kéfi, S., Schneider, F. Danet, A., Génin, A. Mayor, A. G., Bautista, S. Rietkerk, M. Siteur, K. 2017. Report on indicators for critical thresholds. CASCADE Project Deliverable 6.2, 26 pp

Accompanying our review of the spatial indicators of degradation currently available in the literature [23], we developed a statistical toolbox (“earlywarnings package”) in the free programming R environment whose code is freely available online at »»Github: Early warning toolbox.

This toolbox allows quantifying all spatial indicators reviewed in [23] on spatial data sets. The toolbox is constantly being updated. Alexandre Génin (Ph.D. student in Montpellier with Sonia Kéfi) is working on the next version of the code with Sonia Kéfi. This new version will be made available in 2017.

With our collaborator Vasilis Dakos, we also set up a webpage aiming at describing and explaining the various indicators (in time and space) and their theoretical foundation, giving some concrete examples of case studies and references from the literature (Figure 1). See »»Early Warning Signals Toolbox. This webpage will also keep being updated with the latest development regarding spatial indicators.

" data-placement="top" data-content=" Figure 1: Screenshot of the early warning signal webpage" title="">

Note: For full references to papers quoted in this article see

» References

References

2017-05-11T08:38:29+00:00

Contributing Authors:	Sonia Kéfi, Florian Schneider, Alain Danet, Alexandre Génin, Angeles G. Mayor, Susana Bautista, Max Rietkerk, Koen Siteur
Editor:	Jane Brandt
Source document:	Kéfi, S., Schneider, F. Danet, A., Génin, A. Mayor, A. G., Bautista, S. Rietkerk, M. Siteur, K. 2017. Report on indicators for critical thresholds. CASCADE Project Deliverable 6.2, 26 pp

References cited in articles in this section of CASCADiS

Rietkerk M, Dekker SC, de Ruiter PC, van de Koppel J. Self-Organized Patchiness and Catastrophic Shifts in Ecosystems. Science. 2004;305: 1926–1929. doi:10.1126/science.1101867
Suding KN, Gross KL, Houseman GR. Alternative states and positive feedbacks in restoration ecology. Trends Ecol Evol. 2004;19: 46–53. doi:10.1016/j.tree.2003.10.005
Reynolds JF, Smith DMS, Lambin EF, Turner BL, Mortimore M, Batterbury SPJ, et al. Global desertification: building a science for dryland development. Science. 2007;316: 847–851.
Millenium Ecosystem Assessment. Ecosystems and human well-being: desertification synthesis. World Resources Institute, Washington DC, USA; 2005.
Scheffer M, Carpenter S, Foley JA, Folke C, Walkerk B, Walker B. Catastrophic shifts in ecosystems. Nature. 2001;413: 591–596.
Solomon S, Qin D, Manning M, Chen Z, Marquis M, Averyt KB, et al. IPCC, 2007: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, Geneva. Cambridge University Press. Cambridge, UK; 2007.
Maestre FT, Eldridge DJ, Soliveres S, Kéfi S, Delgado-Baquerizo M, Bowker MA, et al. Structure and Functioning of Dryland Ecosystems in a Changing World. Annu Rev Ecol Evol Syst. 2016;47: In press.
Schneider FD, Kéfi S. Spatially heterogeneous pressure raises risk of catastrophic shifts. Theor Ecol. 2016;9: 207–217. doi:10.1007/s12080-015-0289-1
Realpe-Gomez J, Baudena M, Galla T, McKane AJ, Rietkerk MG. Demographic noise and resilience in a semi-arid ecosystem model. Ecol Complex. 2013;15: 97–108.
Baudena M, Baeza J, Bautista S, Eppinga MB, Hemerik L, Kéfi S, et al. Fire and succession in Mediterranean forests. In preparation;
Benateau S, Schneider FD, Rillo M, Kéfi S. Facilitation and the evolution of plant trait syndromes in Mediterranean vegetation. In preparation;
Vasques A, Baudena M, Kéfi S, Bautista S, Vallejo VR, Keizer JJ, et al. How do the interactions between fire regime and pre-fire pine structure determine potential shifts in plant community composition? In preparation;
Díaz-Sierra R, Baudena M, Verwijmeren M, Kéfi S, Rietkerk MG. The effect of positive plant interactions along stress gradients: a modelling approach. In preparation;
Verwijmeren M, Baudena M, Wassen MJ, Diaz-Sierra R, Smit C, Rietkerk M. Rainfall amount, rainfall intermittency and grazing determine the co-existence and interaction direction of two plant species in a semi-arid ecosystem. In preparation;
Danet A. Résistance à l’herbivorie et stabilité des milieux arides. Supervision: Sonia Kéfi and Florian Schneider. BEE master degree, Montpellier; 2014.
Mayor ÁG, Kéfi S, Bautista S, Rodríguez F, Cartení F, Rietkerk M. Feedbacks between vegetation pattern and resource loss dramatically decrease ecosystem resilience and restoration potential in a simple dryland model. Landsc Ecol. 2013;28: 931–942. doi:10.1007/s10980-013-9870-4
Kéfi S, Holmgren M, Scheffer M. When can positive interactions cause alternative stable states in ecosystems? Funct Ecol. 2016;30: 88–97. doi:10.1111/1365-2435.12601
Tongway DJ (David J, Hindley NL, CSIRO Sustainable Ecosystems. Landscape function analysis manual [electronic resource] : procedures for monitoring and assessing landscapes with special reference to minesites and rangelands / D.J. Tongway and N.L. Hindley. Canberra, A.C.T: CSIRO Sustainable Ecosystems; 2004.
Maestre FT, Escudero A. Is the patch size distribution of vegetation a suitable indicator of desertification processes? Ecology. 2009;90: 1729–1735. doi:10.1890/08-2096.1
Kéfi S, Alados CL, Chaves RCG, Pueyo Y, Rietkerk M. Is the patch size distribution of vegetation a suitable indicator of desertification processes? Comment. Ecology. 2010;91: 3739–3742. doi:10.1890/09-1915.1
Scheffer M, Bascompte J, Brock WA, Brovkin V, Carpenter SR, Dakos V, et al. Early-warning signals for critical transitions. Nature. Nature Publishing Group; 2009;461: 53–59. doi:10.1038/nature08227
Dakos V, Carpenter SR, Brock WA, Ellison AM, Guttal V, Ives AR, et al. Methods for Detecting Early Warnings of Critical Transitions in Time Series Illustrated Using Simulated Ecological Data. Yener B, editor. PLoS ONE. 2012;7: e41010. doi:10.1371/journal.pone.0041010
Kéfi S, Guttal V, Brock W a, Carpenter SR, Ellison AM, Livina VN, et al. Early warning signals of ecological transitions: methods for spatial patterns. PloS One. 2014;9: e92097. doi:10.1371/journal.pone.0092097
Rietkerk M, Dekker SC, de Ruiter PC, van de Koppel J, Ruiter PC De, Koppel J Van De. Self-Organized Patchiness and Catastrophic Shifts in Ecosystems. Science. 2004;305: 1926–1929. doi:10.1126/science.1101867
Kéfi S, Rietkerk M, Roy M, Franc A, de Ruiter PC, Pascual M. Robust scaling in ecosystems and the meltdown of patch size distributions before extinction. Ecol Lett. 2011;14: 29–35. doi:10.1111/j.1461-0248.2010.01553.x
Kéfi S, Rietkerk M, Alados CL, Pueyo Y, Papanastasis VP, ElAich A, et al. Spatial vegetation patterns and imminent desertification in Mediterranean arid ecosystems. Nature. 2007;449: 213–217. doi:10.1038/nature06111
Wissel C. A universal law of the characteristic return time near thresholds. Oecologia. 1984;65: 101–107.
Dakos V, Carpenter SR, Nes EH Van, Scheffer M. Resilience indicators : prospects and limitations for early warnings of regime shifts. Philos Trans R Soc B-Biol Sci. 2015;370: 20130263. doi:http://dx.doi.org/10.1098/rstb.2013.0263
van Nes EH, Scheffer M. Slow recovery from perturbations as a generic indicator of a nearby catastrophic shift. Am Nat. 2007;169: 738–747. doi:10.1086/516845
Carpenter SR, Brock WA. Rising variance: a leading indicator of ecological transition. Ecol Lett. 2006;9: 311–318. doi:10.1111/j.1461-0248.2005.00877.x
Held H, Kleinen T. Detection of climate system bifurcations by degenerate fingerprinting. Geophys Res Lett. 2004;31: L23207. doi:10.1029/2004GL020972
Dakos V, Kéfi S, Rietkerk MG, Van Nes EH, Scheffer M. Slowing Down in Spatially Patterned Ecosystems at the Brink of Collapse on JSTOR. Am Nat. 2011;177: E153–E166.
Xu C, Van Nes EH, Holmgren M, Kéfi S, Scheffer M. Local facilitation may cause tipping points on a landscape level preceded by early warning indicators. Am Nat. 2015;186: E81–E90. doi:10.1086/682674
Kéfi S, Dakos V, Scheffer M, Van Nes EH, Rietkerk M. Early warning signals also precede non-catastrophic transitions. Oikos. 2013;122: 641–648.
Dakos V, Nes EH van, Donangelo R, Fort H, Scheffer M. Spatial correlation as leading indicator of catastrophic shifts. Theor Ecol. 2009;3: 163–174. doi:10.1007/s12080-009-0060-6
Guttal V, Jayaprakash C. Spatial variance and spatial skewness: leading indicators of regime shifts in spatial ecological systems. Theor Ecol. 2009;2: 3–12. doi:10.1007/s12080-008-0033-1
Kéfi S, Eppinga MB, de Ruiter PC, Rietkerk M. Bistability and regular spatial patterns in arid ecosystems. Theor Ecol. 2010;3: 257–269. doi:10.1007/s12080-009-0067-z
Siteur K, Siero E, Eppinga MB, Rademacher JDM, Doelman A, Rietkerk M. Beyond Turing: The response of patterned ecosystems to environmental change. Ecol Complex. 2014;20: 81–96. doi:10.1016/j.ecocom.2014.09.002
Rodriguez-Iturbe I, Porporato A, Ridolfi L, Isham V, Coxi DR. Probabilistic modelling of water balance at a point: The role of climate, soil and vegetation. Proc R Soc A. 1999;455: 3789–3805.
Thompson S, Katul G, Konings A, Ridolfi L. Unsteady overland flow on flat surfaces induced by spatial permeability contrasts. Adv Water Resour. 34: 1049–1058.
Siteur K, Eppinga MB, Karssenberg D, Baudena M, Bierkens MFP, Rietkerk M. How will increases in rainfall intensity affect semiarid ecosystems? Water Resour Res. 2014;50: 5980–6001. doi:10.1002/2013WR014955
Mayor ÁG, Bautista S, Small EE, Dixon M, Bellot J. Measurement of the connectivity of runoff source areas as determined by vegetation pattern and topography: A tool for assessing potential water and soil losses in drylands - Mayor - 2008 - Water Resources Research - Wiley Online Library. Water Resour Res. 2008;44. Available: http://onlinelibrary.wiley.com/wol1/doi/10.1029/2007WR006367/abstract
Tirabassi G, Viebahn J, Dakos V, Dijkstra H a., Masoller C, Rietkerk M, et al. Interaction network based early-warning indicators of vegetation transitions. Ecol Complex. Elsevier B.V.; 2014;19: 148–157. doi:10.1016/j.ecocom.2014.06.004
Maestre FT, Quero JL, Gotelli NJ, Escudero A, Ochoa V, Delgado-Baquerizo M, et al. Plant species richness and ecosystem multifunctionality in global drylands. Science. American Association for the Advancement of Science; 2012;335: 214–218.
Berdugo M, Kéfi S, Soliveres S, Maestre FT. Plant spatial patterns identify alternative ecosystem multifunctionality states in global drylands. Nat Ecol Evol. In press;
Kéfi S, Guttal V, Brock WA, Carpenter SR, Ellison AM, Livina VN, et al. Early Warning Signals of Ecological Transitions: Methods for Spatial Patterns. Solé RV, editor. PLoS ONE. 2014;9: e92097. doi:10.1371/journal.pone.0092097
Petchey OL, Pontarp M, Massie TM, Kéfi S, Ozgul A, Weilenmann M, et al. The ecological forecast horizon, and examples of its uses and determinants. Ecol Lett. 2015;18: 597–611. doi:10.1111/ele.12443
Bautista S, Mayor ÁG, Bourakhouadar J, Bellot J. Plant Spatial Pattern Predicts Hillslope Runoff and Erosion in a Semiarid Mediterranean Landscape. Ecosystems. 2007;10: 987–998. doi:10.1007/s10021-007-9074-3 49. Feng S, Fu Q. Expansion of global drylands under a warming climate. Atmospheric Chem Phys. Copernicus GmbH; 2013;13: 10081–10094.