grazing

Spatially explicit grazing and vegetation structure

project details

• type of model: cellular automata
• features of model: spatial patterns and indicators, single vegetation type
• type of pressure: grazing, aridity
• developed by: Florian Schneider and Sonia Kéfi (CNRS / University of Montpellier 2)
• project status: in review (May 2015)
  • paper submitted
  • paper published
  • model code published (R and NETLOGO)
  • model code documentation available

project outline

Spatial models of vegetation cover so far have considered grazing mortality a rather constant pressure, affecting all plants equally, regardless of their position in space. In the known models it usually adds as a constant to the individual plant risk (Kéfi et al 2007 TPB). However, grazing has a strong spatial component: Many plants in rangelands invest in protective structures such as thorns or spines, or develop growth forms that reduce their vulnerability to grazing. Therefore, plants growing next to each other benefit from the protection of their neighbors.

Such associational resistance is widely acknowledged in vegetation ecology but hardly integrated in models as a cause for spatially heterogenous grazing pressure. It also renders the plant mortality density dependent, which has important impacts on the bistability of the system.

We investigate how the assumption of spatially heterogeneous pressure alters the bistability properties and the response of spatial indicators of catastrophic shifts.

Distribution of plants dying due to grazing (red) at high vegetation cover (left) and low cover (right). Traditional models assume equal risk for all plants.

Our models assumes that plants that grow associated protect each other from grazing, whereas plants growing isolated suffer most.

In which way does your model help us to improve our understanding of the mechanisms and processes that cause sudden shifts in dryland ecosystems?

Besides the plant-soil feed-back, which is a well-established mechanism leading to spatial vegetation patterns in drylands, our cellular automaton model integrates more realistic aspects of grazing by taking its spatial heterogeneity into account. Grazing is one of the main causes of dryland desertification and it is notably heterogeneous in space, especially in patchy vegetated landscapes. Thus, our model provides a framework where multiple positive feed-backs interact dynamically in space and time to shape vegetation pattern and to define tipping points for the occurrence of sudden shifts.

Has your model helped to identify early warning signals and indicators for sudden shifts in drylands? If so, which ones?

Our model builds on previous models that had identified potential early warning signals of sudden shifts in drylands. We tested the validity of these signals in the case where the pressure exerted on the ecosystem was spatially heterogeneous. These simulations demonstrate that spatial indicators, namely vegetation cover, patch sizes, clustering coefficients and cumulative patch size distributions, have to be considered with caution. If spatially-explicit grazing is high, catastrophic shifts can occur even in apparently 'healthy' ecosystems (i.e. systems without indication of risk: high cover, low fragmentation and straight power-law distributions of patch sizes) due to the self-enhancing character of the positive feed backs involved. We conclude that a more profound knowledge on the interactive mechanisms is a prerequisite to infer the level of degradation from spatial structure.

Can your model predict where, when and under which conditions catastrophic shifts are likely to occur in European drylands?

Based on aerial images providing the vegetation spatial structure and knowledge about the type of stress (grazing, droughts), the spatial indicators can be quantified and used to categorise ecosystems. Our model predicts a qualitative succession of spatial pattern along the coinciding gradients of increasing environmental pressure and grazing pressure: High cover landscapes with large spanning clusters of vegetation are followed by fragmented landscapes with power-law distributions of patch sizes; these are succeeded by down-bent power law distributions, due to over-proportional fragmentation of large patches before the system degrades into a desert. Looking at spatial indicators thereby provides a level of stress exerted on the ecosystems and, if the type of stress exerted on ecosystems is the same, may help compare ecosystems with each other. However, the indicators do not provide the probability of a catastrophic shift (although it may be used to identify which factors favour catastrophic shifts).

Can your model simulate what effect SLM measures would have? Can your models tell us which sustainable land management (SLM) measures should be taken to prevent shifts?

In principle, yes. However, a requirement for such an application would be the translation of the abstract 'grazing pressure', which is not including the density dependence of livestock feeding, into more meaningful terms for land use management (e.g. livestock units per time and resulting plant mortalities).We aim to resolve these issues in the current phase of the project. Provided this translation, the stochastic model can be used to simulate different density based management scenarios (e.g. supplementary feeding, temporal or numerical population management) to assess their effect on dryland resilience.

Does your model provide any information on possibilities for restoration if a tipping point has already occurred?

Yes. Along the simulated gradients of environmental pressure and grazing pressure the model provides a probability of recovery from a very low vegetation cover, showing that the reversibility of degradation is more difficult if grazing pressure is high. After resolving the caveats described for Question 4, the model can be used to assess different restoration strategies (planting patterns and quantity, irrigation strategies, grazing exclusion) Please explain your answers in a few sentences, as obviously we are also interested in how your model does these things, or why it is suited/not suited.

SUMMARY of results

• We found that resilience of ecosystem declines with grazing pressure.
• Spatial indicators are quantitatively and qualitatively affected by spatially explicit grazing. Under high grazing pressure, no truncated cumulative patch-size distributions precede the collapse. The degradation happens to ecosystems with perfectly straight power-law distributions.
• The individual level mechanisms, such as spatially-explicit grazing pressure and local facilitation, interact indirectly to define the patch level "life-cycle". The total plant mortality becomes a highly dynamic function of vegetation cover: At high cover, the pressure approximates zero, due to the almost full protection against grazing that plants provide each other. At low cover, the mostly isolated plants suffer from grazing and will likely be lost as well.