• Reduce text

    Reduce text
  • Restore text size

    Restore text size
  • Increase the text

    Increase the text
  • Print

Tree in winter. © INRA, Hervé Cochard

How light and wind shape tree architecture

To understand why trees look the way they do, an interdisciplinary team of researchers from the École Centrale de Marseille, INRA, AgroParisTech, CNRS, and the University of Aix-Marseille1 developed an innovative model that simulated the evolution of a forest ecosystem over a 200,000-year period. In this virtual ecosystem, trees compete for light, change their growth patterns in response to wind, and experience storms that can snap their branches. The researchers found that light competition and wind selected for fractal forms—morphological patterns that repeat themselves (are "self-similar") across different scales. Such self-similarity* has been observed in real trees by ecologists and forestry scientists. Consequently, it is possible that light and wind acted in tandem over evolutionary time to sculpt the current architecture of trees. These findings were published on October 18, 2017, in Nature Communications.

Updated on 01/31/2018
Published on 10/18/2017
Keywords: forest - tree - GROWTH

Recently, researchers in biomechanics and tree ecophysiology from INRA and AgroParisTech joined with physicists studying complex systems from the École Centrale de Marseille and CNRS to tackle an interdisciplinary project. Their goal? To model a forest ecosystem in which trees grow, reproduce, and die over evolutionary time. They used this model to test a novel hypothesis: that the allometric relationships seen in tree structural traits are driven by selective pressures exerted by light competition and wind.

A model incorporating the best available knowledge on trees

In the model's virtual forest ecosystem, trees capture light and photosynthesize, allocate the resulting energy among their organs, grow branches, and produce seeds that fall to the ground and germinate. To model these processes, the scientists incorporated up-to-date knowledge about how plants respond to light and wind. More specifically, the model includes two recent discoveries: first, that the position of new branches depends on light levels in the surrounding foliage and, second, that wind affects branch thickness. The latter case is an example of plants responding to mechanical stimulation by changing their growth patterns, a phenomenon known as thigmomorphogenesis. Wind-induced thigmomorphogenesis has a major influence on the production of woody biomass in temperate ecosystems. The model also uses findings in meteorology and biomechanics to simulate how branches snap during storms. All this knowledge was integrated into an innovative model of great computational efficiency.

Of seeds, selection, and thousands of computational hours

MechaTree Programme. © Centrale Marseille, C. Eloy
MechaTree Programme © Centrale Marseille, C. Eloy
Once the model was ready, the researchers planted the first generation of virtual seeds and allowed natural selection to run its course. However, natural selection requires fodder, namely genetic variation. The physiological processes described above—responsiveness to light, sensitivity to wind, and energy allocation—were all quantitatively described by various parameters. These parameters thus functionally represented the tree's genotype. Random genetic mutations caused the parameter values to vary across time, from generation to generation. Competing for light and buffeted by the wind, the trees were subject to selective pressures and evolution occurred. The researchers decided to model forests as islands, isolated systems in which no seeds or pollen arrived from the exterior. This choice has practical benefits because research in evolutionary ecology has shown that natural selection occurs more rapidly in island ecosystems.

The modeling program, called MECHATREE, was used to generate hundreds of forest islands. The first generation of seeds had randomly determined physiological parameter values. The seeds germinated and grew into trees, resulting in dense forests. Certain individuals were at a genetic disadvantage and disappeared because of self-thinning or wind-induced mortality. Others persisted but varied in their reproductive success, providing the raw material for natural selection. This process continued, and, generation after generation, seeds were produced. Certain species came to dominate the forests, while others vanished.  
Thousands of computational hours later, representing nearly 200,000 years in the life of the forest ecosystem, the researchers examined the traits of the surviving tree species. They made a gratifying discovery: unlike trees in previous models, their virtual trees displayed not just some but all the patterns of allometric scaling* that have been seen in nature. For example, the trees followed the self-thinning rule, displayed a fractal dimension consistent with available empirical data, and respected the observed allometric relationship between branch length and diameter. Their morphology was even consistent with the predictions of Leonardo da Vinci's Rule of the Trees.

Recognition of the selective roles played by light competition and wind

A broader-scale question remains: what can the model tell us about the evolution of tree architecture? As mentioned above, the aim was to examine the evolutionary roles played by light and wind. The model’s results suggest that foliage transparency and light competition are the primary drivers of the fractal dimension of trees. In contrast, wind-induced thigmomorphogenesis shapes trunk and branch thickness along the tree as well as branch emergence patterns.

The researchers do not exclude the possibility that other factors, such as sap transport hydraulics, are evolutionarily relevant. Indeed, the relative importance of sap transport and wind resistance as selective forces may depend on the environmental context in which species evolve. These issues aside, this research is the first to underscore that light competition and wind play a crucial role in shaping tree architecture. This discovery will significantly change the field of forest ecology. It will also force us to rethink how we represent trees and revisit our assumptions about what goes into making a tree.

* Tree self-similarity

Trees belong to different phylogenetic groups, and the tree growth form seems to have appeared several times over the course of plant evolution. What are the architectural similarities currently shared by trees? How did these similarities arise? These are not new questions.

Indeed, Leonardo da Vinci (1478–1518) remarked that the sum of the cross-sectional areas of all the branches above a branching node was equal to the cross-sectional area of the trunk immediately below the node. This observation became known as his Rule of the Trees. Later, when scientists started to study fractals, they saw that natural trees were characterized by self-similarity, displaying a fractal dimension of 2.5. In tandem, forestry scientists and ecologists noted allometric scaling in tree traits. For example, there is a relationship between mean tree biomass and tree density for a given forest stand (the self-thinning rule) as well as between tree height and trunk diameter. Allometric scaling is said to exist when two traits, signified by x and y, follow a power law relationship (y = kxa). Finally, at the individual level, rules describing “taper” (how thickness changes with height) have been observed.
However, the researchers that developed MECHATREE were interested in a broader mechanistic question: what is driving these recurrent empirical patterns? Until recently, a two-faceted explanation was commonly evoked. Tree architecture was considered to be driven by the optimization of sap transport hydraulics (from the roots to the leaves). It was also thought to reflect optimal mechanical design—maximizing the span of the trunk and branches while still supporting the load exerted by body mass (without buckling or drooping). However, this explanation has some weaknesses. First, it is temporally short sighted: it does not adequately address the dynamics of allometric scaling in trees as they grow. Second, no mention is made of evolution; the selective forces that acted to optimize these functions remain undiscussed. Finally, it entirely ignores two other factors that are just as important as sap transport and buckling risk, namely light competition and wind resistance.  It is these issues that the model sought to explore.  


1The scientists that participated in this interdisciplinary project came from the following research units: French Research Institute for Non-Equilibrium Phenomena (IRPHE), University of Aix-Marseille—CNRS—École Centrale de Marseille; Joint Research Unit for Forest and Wood Resource Studies (LEFRoB), INRA—AgroParisTech; and Joint Research Unit for the Integrative Physics and Physiology of Trees in Fluctuating Environments (PIAF), INRA—University of Clermont-Auvergne.

Scientific contact(s):

  • Bruno Moulia (33 (0)4 43 76 14 23) Joint Research Unit for the Integrative Physics and Physiology of Trees in Fluctuating Environments (PIAF), INRA—University of Clermont-Auvergne
  • Christophe Eloy / Professor of Fluid Mechanics, French Research Institute for Non-Equilibrium Phenomena (IRPHE), University of Aix-Marseille—CNRS—École Centrale de Marseille
Press Relations:
INRA News Office (33 (0)1 42 75 91 86)
Associated Division(s):
Environment and Agronomy, Forest, Grassland and Freshwater Ecology
Associated Centre(s):


Christophe Eloy, Meriem Fournier, André Lacointe & Bruno Moulia. Wind loads and competition for light sculpt trees into self-similar structures, Nature Communications doi:10.1038/s41467-017-00995-6