Multiple Tree-Cover States in the Earth System

The boreal forest is an ecosystem of key importance in the Earth system, nonetheless, its dynamics regarding smooth changes and critical ecosystem transitions have not been systematically investigated. Henceforth, in the first part of my project, I study the relationship between the boreal tree-cover distribution and eight globally-observed environmental factors, chosen among those of known major importance for the boreal ecosystem. Additionally, I develop a methodology, based on the framework introduced by Staver, Archibald, and Levin in 2011, to detect the location of potential areas with alternative tree-cover states under the same environmental conditions. 


Satellite observations allow to draw a global picture of the possible multistability of the boreal forest and determine the environmental conditions shaping it. However, to investigate the causes of the existence of alternative tree-cover states, a bottom-up approach is necessary. Thus, in the second part of my project, I develop a conceptual model to better understand whether the competition between species with different evolutionary traits could explain the observed multistability. In fact, boreal trees have a high functional diversity, i.e., the diversity of species’ traits. By taking this in consideration, I create a simple, yet not simplistic, picture of the multiple stable states of the boreal forest. Moreover, I employ my model to simulate the sensitivity of tree cover to changes in environmental factors and to stochastic disturbances.

In the context of climate change, temperature changes could greatly affect forest resilience and cause an expansion of regions with alternative tree-cover states. The boreal ecosystem, in particular, is undergoing environmental changes more rapidly and intensely than other regions on Earth, and its surface temperature has been increasing approximately twice as fast as the global average. Hence, in the third part of my project, I investigate how the multimodality and multistability of the boreal forest could evolve at high latitudes under different scenarios of anthropogenic climate change. To this avail, I combine in a synergistic approach and further develop the previously introduced top-down and bottom-up frameworks to simulate scenarios with higher levels of CO2. Furthermore, I suggest implications for how to improve the representation of tree cover in dynamic global vegetation models.


Finally, I consider the limitations of current methods for climate projections, and how the combination of observations and modelling can help gain a better understanding of the dynamics between alternative tree-cover states. I also make the case that, to study forest multistability, it is necessary to analyse the components and the drivers playing a role, and consider a different level of complexity than the one currently allowed by global models.


Forest ecosystems represent one of the most important component of the Earth, covering almost a third of the land surface. Within their three trillion trees, they harbour a large proportion of global biodiversity, and provide countless ecological and socioeconomic services, to natural systems, and humankind as well. 

In the beautiful story L’Homme qui plantait des arbres [The man who planted trees] by Giono, a shepherd is able to single-handedly change the climate of a desolate valley in the foothills of the Alps by planting oak trees. This story, albeit admittedly fictional, contains two elements of truth. In fact, forests can influence the climate system through biophysical and biogeochemical processes. A dense forest has a low surface albedo and can mask the high albedo of snow, allowing its canopy to trap irradiation, and contributing to planetary warming. At the same time, forests support and help regulate the hydrological cycle through evapotranspiration, causing a cooling effect on climate. Moreover, forests constitute a large reservoir of global terrestrial carbon, and sequester large amounts of natural and anthropogenic carbon.

However, in recent years, forests around the world are undergoing several changes in structure, species composition, extent, and function. The roots of these changes are both natural and anthropogenic. In fact, they originate from a combination of environmental factors, such as rising CO2 concentrations, nitrogen deposition, extreme precipitation and temperature anomalies, and local drivers, for instance forest management, grazing, wildfires, and other disturbances.

With the projected increase in greenhouse gases concentrations due to human activity, such environmental and climate factors are likely to become more prominent in the future decades. Consequently, forest ecosystems will be progressively more affected, and their resilience, i.e., their ability to recover from disturbances maintaining similar structure and functioning, will decrease.

Therefore, it is of critical importance to foster our understanding of forest dynamics under global anthropogenic change. Unfortunately, though, forest ecosystems around the world are not all equal, and exhibit different responses to climate change. Furthermore, it is notoriously difficult to predict how a specific forest will evolve, due to their complexity, and their intrinsic feedbacks and nonlinearities. For these reasons, increasing attention has been given to the observation and interpretation of the response of forest ecosystems to past and present climate changes.

A question of particular interest and concern is whether tree cover, being one of the defining variables of forests, will show a smooth response to disturbances and climate change, or exhibit rapid shifts between alternative stable states. This brings us to the second element of truth in the story L’Homme qui plantait des arbres. At the beginning of Giono's story, the arid valley of Provence is a desolate and treeless land, covered only in wild lavender. By the end of the story, after four decades of efforts, however, the valley is covered by a oak forest, vibrant with life and water streams. The element of truth, here, is that certain regions of the world can exhibit alternative stable ecosystems, and that transitions between them can occur in a short time.

Occasionally, in fact, even if environmental conditions change gradually, instead of fluctuating around a smooth trend or stable state, ecosystems abruptly collapse or transition to a dramatically different regime. One of the most prominent examples of this phenomenon is, perhaps, the shift in vegetation cover that occurred in the Sahara between 5000 and 6000 years ago. After millennia of gradual changes, the vegetation in northern Africa suddenly shifted from the humid and verdant conditions known as the "green Sahara", to the world's largest warm desert. 

In recent years, many studies have touched upon the topic of ecosystem shifts and alternative states. Notably, with regards to vegetation cover, it has been hypothesised that tropical forests, savannas, and treeless areas represent three alternative stable states, which can be supported under the same climate conditions. Evidence for such tree-cover multistability has been inferred, locally, from field observations and fire exclusion experiments, and, more broadly, from mathematical models and remotely-sensed satellite observations. Through the use of satellite observations, in particular, it is possible to compare climate and tree-cover data from different regions and continents at the same time, allowing a more global perspective.

There are two key pieces of evidence to support the multistability hypothesis. The first one comes from remotely-sensed observations, and it consists of the fact that, in the tropics, the tree-cover distribution exhibits three distinct modes, corresponding to the forests, savannas, and treeless states. Multimodality of the frequency distribution of states is the spatial analogue of sudden jumps in a time series, and can be caused by the presence of alternative attractors which create more or less sharp boundaries between contrasting states. Importantly, multimodality does not necessarily imply the presence of alternative states, as it can be a result of the multimodality of one of the driving factors. Nevertheless, this is not the case for the tropics, where the distribution of the main variables determining tree cover, namely precipitation, rainfall seasonality, and soil properties, cannot differentiate between the three observed modes.

The second key piece of evidence comes from the presence of a positive feedback between fire and vegetation. A high tree-cover fraction suppresses the occurrence of fire, due to coarse fuels and a more humid microclimate. Hence, as tree cover increases, beyond some point flammability decreases, further promoting tree-cover densities in a positive feedback towards a closed canopy. On the contrary, at a lower tree-cover fraction, fire frequency is enhanced due to the higher grass cover which provides drier conditions and easily ignitable fuels. The increase in fire frequency, in turn, prohibits the establishment of trees, further promoting the presence of flammable grasses, in a runaway change towards open savanna. As shown with fire exclusion experiments, this positive feedback can maintain a savanna where climate and soils would otherwise support a closed-canopy forest. By taking in consideration the vegetation-fire feedback, it is possible to differentiate between the three alternative modes, both in observations and conceptual models. Furthermore, this has recently led to a first global assessment of multiple stable states of tree cover due to the fire-vegetation feedback in a dynamic vegetation model.

A similar tree-cover distribution has recently been detected in a completely different region, the boreal forest. In fact, an analysis of the vegetation cover from remote sensing revealed the existence of distinct alternative modes in the frequency distribution of boreal trees. These modes correspond to a sparsely vegetated treeless state, an open woodland "savanna"-like state, and a forest state, and are comparable to the ones found in the tropics. Specifically, it has been observed that, over a broad temperature range, these three vegetation modes coexist, whereas regions with intermediate tree cover are relatively rare. Moreover, the multimodality of the tree cover does not ensue from the distribution of two of the main environmental conditions, namely temperature and precipitation, driving the boreal forest dynamics. As for the tropics, these lines of evidence suggest that multiple stable tree-cover states might be present, acting as attractors. 

Contrary to the case of tropical savannas and forests, the boreal ecosystem doest not exhibit any known positive feedback capable of maintaining three alternative states, and multimodality alone is not proof of the existence of alternative states. Another important difference is that the distribution of tropical vegetation is, essentially, in large part determined by only two factors: fire interactions coupled with water availability from rainfall. On the other hand, despite a low diversity of tree species, the boreal forest's structure depends on interactions between several variables, including air temperature, precipitation, available solar radiation, presence of permafrost, depth of forest floor organic layer, forest fires, insect outbreaks and more.