+ de critères

How to better understand the forest carbon cycle

Eric Dufrêne (CNRS) takes the floor

How to associate different monitoring networks to better understand the carbon cycle in forest ecosystems? The CASTANEA model


In a context of climate change, the stakes are double for our forests: they must adapt to rapidly changing environmental conditions while they also play a role in attenuating climatic upheavals through their ability capture and store atmospheric carbon, most notably in wood.

We use models to mobilize our existing knowledge in order to simulate natural phenomena and to try to respond to questions about how forests will evolve in the future.

Basically, we can separate forest growth models into two main categories:

  • models based on dendrometry, which extrapolate growth from observed local growing conditions assuming a "stable" future climate;
  • models of bio-physical processes, which offer short-term extrapolations of how trees will respond to climate changes.

Recently, attempts have been made to combine these two types of models.

The global model called ORCHIDEE is based on broad functional types while the CANTANEA model explicitly takes into account forest species. In France, each of these models have been combined with forest stand structure models. These hybrid models include long-term stand structure dynamics and therefore allow us to extend the predictive capacity of process-based models to the whole forest cycle.
The hybrid model CASTANEA-SSM also makes it possible to evaluate the combined effects of climate changes and silvicultural regime on stand production and survival depending on the dominant tree species.

In both types of models, to calibrate the equations and test the accuracy of their predictions, on-site observations and measurements are required. Dendrometric models use a restricted number of equations and parameters but process-based models and hybrids require large numbers of equations and parameters in order to represent the wide variety of processes simulated.

The CASTANEA model was developed in the mid-1990s drawing on knowledge in plant organ ecophysiology and forest canopy bio-climatology. CASTANEA was first designed as a Soil-Vegetation-Atmosphere (SVAT) model to predict water vapour and CO2 fluxes between forest ecosystems and the atmosphere on an hourly basis throughout the year. The goal was to simulate the effects of an increase in atmospheric CO2 on forest ecosystems through experiments with open-top chambers in controlled environments (European ECOCRAFT Project). Within a decade, however, advances in flux measurements through eddy covariance in forest canopies made it possible to calibrate and test CASTANEA under real conditions in forest stands (the European CarboFlux and later Carbo-Europe projects).

Thus, after an important series of studies in the Hesse beech forest, CASTANEA was adapted and tested for several other tree species thanks to data from the European (for Holm oak, sessile oak, Scots pine, spruce, maritime pine) and American (for Douglas fir) flux tower networks. The main objective was to reproduce daily, seasonal and inter-annual variations in water vapour and CO2 fluxes.
Meanwhile, work began elsewhere to improve our knowledge of carbon reserves in order to simulate growth in the main components of the tree, in particular the trunk. The CASTANEA (SVAT) flux model was adopted as a carbon balance model and used to simulate growth for several species (sessile oak, beech, Holm oak, spruce).

Simulating carbon allocation to the different plant compartments is much more dependent on species than it is for gaseous fluxes. Simulating fluxes requires specific parametrisation but the rules (or processes translated into equations) are often rather "generic". Simulating carbon allocation not only requires specific parametrisation but, in addition, the rules can vary considerably from one species to another. It is also difficult to simulate net carbon absorption or release for a given forest rotation since tree age and size have considerable effects on parametrisation and sometimes on allocation rules.

To respond to the challenge set by this variability, different sets of data must be integrated from monitoring networks at the landscape scale (for example, the Fontainebleau forest), the country scale (France) and the European scale. Typically, the number of plots or sites in a monitoring network is inversely proportionate to the number of variables measured. On the other hand, there is no relation between the spatial scale of the network and the number of plots or variables concerned.

RENECOFOR data, either alone or in combination with data from other networks, have made it possible to parametrise the bud burst model, to parametrise and develop the leaf-yellowing model, to test the schema for carbon allocation connected with the management model (ceci n'est pas clair pour moi et je n'ai trouvé aucune reference internet pour m'aider - au secours!) , and finally, to develop and test a model for fruiting (in progress).

Pour en savoir plus