Search

Background: Understanding the reaction of ecosystems to climate change and anthropogenic pressure is a central question in ecology and environmental sciences. In the terrestrial tropics, theoretical and empirical works suggest that once external disturbances have reached a given threshold, forest-savanna systems can switch from one state to another. Considering the multiplicity of the tropical forest systems, we make the assumption that numerous shifts may actually occur within the forest itself, without changes in forest cover but with risks of critical modifications in forest functioning. Methods: To test this hypothesis, we used a finite mixture of regression models aiming at simultaneously predicting and grouping forest functional profiles at the stand level with respect to anthropogenic pressure, climate and soil. The model is built on a dataset of more than 140 000 plots of 0.5-ha each gathered from Central African forest companies. Forest stand functions are analyzed through two key functional traits: the successional status - pioneer vs. non-pioneer trees- and the leaf phenology - evergreen vs. deciduous trees. Results: Our model captured a significant part of variation in the functional composition over the study area and revealed how anthropogenic pressure, climate change, soils or their combination lead to profound modifications within the forests. In particular, we showed that shifts from evergreen to deciduous stands can be mediated both by anthropogenic pressure or climate change. Discussion: This work shows for the first time how external forcing may jointly lead to multiple shifts in the functional composition of tropical forests. Our model allowed to predict directional changes in forest functioning according to anthropogenic pressure and climate thus opening new perspectives in theoretical ecology, global vegetation modelling and in the understanding of the vulnerability of tropical forests to global changes. (Texte intégral)

International audience; Les forêts d’Afrique centrale vont subir de profondes modifications liées au changement climatique ces prochaines décennies. Pour que les sociétés humaines puissent anticiper ou s’adapter à ces changements, il est nécessaire de bien comprendre les mécanismes de fonctionnement de ces forêts et d’être à même de spatialiser leur variabilité. A partir d’une série temporelle de quatorze années d’images satellites MODIS à 250m de résolution par pixel, nous présentons ici une nouvelle carte des forêts d’Afrique centrale. La production de cette carte, basée sur le comportement phénologique de la canopée, reflète l’activité de la végétation en fonction des saisons

Les forêts d'Afrique centrale vont subir de profondes modifications liées au changement climatique ces prochaines décennies. Pour que les sociétés humaines puissent anticiper ou s'adapter à ces changements, il est nécessaire de bien comprendre les mécanismes de fonctionnement de ces forêts et d'être à même de spatialiser leur variabilité. A partir d'une série temporelle de quatorze années d'images satellites MODIS à 250 m de résolution par pixel, nous présentons ici une nouvelle carte des forêts d'Afrique centrale. La production de cette carte, basée sur le comportement phénologique de la canopée, reflète l'activité de la végétation en fonction des saisons.

International audience

International audience

Global carbon, water and energy cycles are substantially driven by vegetation phenology. In particular tropical rainforests have been shown to be a key component of the climate system as they act as major water vapor sources and carbon dioxide sink. For these reasons their evolution in response to both human pressure and climate change is critical. As compared to the Amazonian and Asian rainforests, the rainforest of Central Africa experiences slower deforestation rates, so that its main threat for the next decades might come from climate change. So far, the response and sensitivity of the Central Africa rainforest to the mean seasonal evolution and inter-annual variability of climate has attracted little interest. Indeed, most of the studies focus on its Amazonian counterpart and suggest that solar irradiation is the main driver of the annual and inter-annual variations of rainforest photosynthetic activity, and the Central Africa climate itself is not well documented. As a first step towards a better understanding of the Central Africa rainforest sensitivity to present-day climate variability and response to climate change, this study performs for a target region located between 0-5°N/12- 19°E (thus documenting forest areas from 5 countries) and using space borne observations, a detailed analysis of the rainforest photosynthetic activity mean seasonal cycle comparing it with those of climate variables considered as potential drivers, i.e. rainfall, cloudiness and solar irradiation. Several key points emerge from our study. First, the seasonal cycles of photosynthetic activity (EVI MODIS) and rainfall over our target region are both bimodal. However, the highest peak of EVI (March-May) coincides with the driest of the two rainy seasons while the lowest peak of EVI (September-October) coincides with the wettest of the two rainy seasons. Second, the two rainy seasons are not associated with two distinct lows in total solar irradiation and two distinct peaks in total cloudiness: the first rainy season (March-May) which is less rainy as compared to the second one (September-October), is also less cloudy and receives more total solar irradiation. This might explain the higher EVI values recorded. Third, the high total cloudiness recorded throughout the seasonal cycle actually hides marked seasonal variations in the frequency of the 5 main types of clouds analyzed. These cloud types have specific diurnal cycles which control those of solar irradiation (thus the daily light and energy available for photosynthesis), but also influence the remote sensed photosynthetic activity data (or index). Our results clearly show that (1) nor the two dry seasons, nor the two rainy seasons do compare in terms of mean rainfall, cloudiness, solar irradiation and temperature, and (2) water and light availability have a respective weight in the Central Africa rainforest photosynthetic activity which evolves throughout the seasonal cycle. They also suggest that any evolution, due to climate change, of the complex diurnal cycles of rainfall, nebulosity and solar irradiation which characterize the equatorial climate regimes might perturb the rainforest phenology and enhance these ecosystems vulnerability. (Texte intégral)

Understanding how environmental factors could impact population dynamics is of primary importance for species conservation. Matrix population models are widely used to predict population dynamics. However, in species-rich ecosystems with many rare species, the small population sizes hinder a good fit of species-specific models. In addition, classical matrix models do not take into account environmental variability. We propose a mixture of regression models with variable selection allowing the simultaneous clustering of species into groups according to vital rate information (recruitment, growth and mortality) and the identification of group-specific explicative environmental variables. We develop an inference method coupling the R packages flexmix and glmnet. We first highlight the effectiveness of the method on simulated datasets. Next, we apply it to data from a tropical rain forest in the Central African Republic. We demonstrate the accuracy of the inhomogeneous mixture matrix model in successfully reproducing stand dynamics and classifying tree species into well-differentiated groups with clear ecological interpretations. Copyright (c) 2014 John Wiley & Sons, Ltd.

Mapping the vegetation Carbon stocks is crucial to understand the global climate change. The Carbon stock maps have direct implications in economy and environmental policy. This is especially true in tropical forests where most of the uncertainties on carbon fluxes and stocks are concentrated. Substantial efforts have been done recently to map forest carbon in tropical areas, especially by using remote sensing-based approaches. However, there is no way to bypass a calibration step where biomass is locally measured through forest inventories. The great importance of this learning step and its possible issues has been documented, highlighting the importance of terrestrial datasets. In our work, we have gathered a very large dataset of forest inventories covering the Congo Basin. It consists of 73 000 0.5ha plots of commercial inventories covering 4 million hectares in Cameroon, Republic of Congo, Gabon, Central African Republic, and the Democratic Republic of the Congo. These terrestrial data are of great value to understand and model the spatial distribution of various forest properties, among which the Carbon stock. They can also make a great tool to control and improve the performance of the remote sensing methods. In our study, we rely on these plots to test the validity of previously published pantropical Carbon maps. After gathering the data with extra care due to the heterogeneous inventory methods, we used bioclimatic models, topography, and remote sensing observation to extrapolate the forest carbon estimates at the Congo basin scale. (Texte intégral)

Predicting the current and future natural distributions of species is challenging, especially in the tropics where large remote areas remain poorly known. Such challenge can only be met with an in-depth understanding of the drivers of species distribution, a well-designed and extensive survey and appropriate statistical models. In this study, we use a large dataset of forest inventories from logging companies, which provides information on the abundance of 215 tree genera, in more than 115,000 plots spread over four Central African countries. In order to predict the current and future distribution of these tree genera, we use a set of bioclimatic, geological and anthropogenic variables. We rely on a recently published methodology, called Supervised Component Generalized Linear Regression (SCGLR), which identifies the most predictive dimensions among a large set of predictors. Using a calibration and validation scheme, we show that the distribution of most tree genera can be well predicted over the whole study area. At the community level, the floristic and functional composition of tree genera is inferred with a high accuracy. Finally, using climatic and anthropogenic scenarios we predict the expected change in functional and phylogenetic structure of tree communities over the western part of the Congo Basin. Overall, our study provides useful ecological insights and shows that tropical tree distributions can be predicted with good accuracy, offering new perspectives to manage tropical forests at large spatial scales. (Texte intégral)

National audience; Les forêts d'Afrique centrale vont subir de profondes modifications liées au changement climatique ces prochaines décennies. Pour que les sociétés humaines puissent anticiper ou s'adapter à ces changements, il est nécessaire de bien comprendre les mécanismes de fonctionnement de ces forêts et d'être à même de spatialiser leur variabilité. A partir d'une série temporelle de quatorze années d'images satellites MODIS à 250 m de résolution par pixel, nous présentons ici une nouvelle carte des forêts d'Afrique centrale. La production de cette carte, basée sur le comportement phénologique de la canopée, reflète l'activité de la végétation en fonction des saisons.

Tropical forests, species distributions, generalized linear models, regression, partial least square, Fisher scoring algorithm Abstract: Understanding how environmental factors spatially impact species coexistence is of primary importance in ecology. Species distribution models are statistical or mathematical models combining observations of species occurrence or abundance with environmental factors. They are used to gain insights into ecological and evolutionary processes and to predict distributions across landscapes, sometimes requiring extrapolation in space and time. Generalized linear models (GLM) are frequently used for modeling species distributions. Nevertheless GLM can suffer from different limitations: (i) the classical approaches do not allow the simultaneous modeling of multiple species (ii) the level of model realism and the prediction robustness can be poor if relevant predictors are omitted. We propose a regularizing component-based technique, supervised component generalized linear regression (SCGLR), which extends the Fisher scoring algorithm by combining partial least square (PLS) regression with GLM estimation in the multivariate context. The proposed method can handle different types of dependent variables (occurrence, abundance, growth) and explanatory variables mixing different types, and account for additional covariates. We illustrate SCGLR and the associated R-package using a dataset built from the abundance of 27 common tree genera in the tropical moist forest of the Congo-Basin and 40 geo-referenced environmental variables collected in the CoForChange project.

International audience

International audience

International audience; The CoForChange project is focused on predicting the effects of global change on forest biodiversity in the Congo Basin region The objective of the study presented here is to characterize the spatial patterns of tropical forests in central Africa in relation with climatic variables (rainfall and light intensity) Hypothesis: Temporal remote sensing acquisition allow us to identify forest structure and functioning (evergreen/deciduous) as a response to the climatic environment

Human activities in tropical rainforests cause disturbances and degradation by opening up the forest canopy. Capacities for measuring the extent of the damage are essential to the calculation of carbon emissions under REDD+ programmes (Reducing Emissions from Deforestation and Forest Degradation). Remote sensing is a powerful tool to provide this type of information (e.g. on logging, mining or infrastructure projects). Various techniques have been used to identify and quantify canopy opening. A further step is to cross-reference past and present canopy opening in order to document the recovery of ecosystems after logging. The method proposed in this article involves a semi-automatic processing chain adapted for LANDSAT imagery. During post-processing, canopy opening information is extracted by means of specific algorithms. A spatial index, calibrated to a low-resolution radiometer, indicates present and past degrees of opening. This process produces estimations of forest degradation that can describe lowresolution remotely-sensed data (such as MODIS) used in continental mapping. These estimations are then cross-referenced with continental land cover maps to gain insights into present-day forest categories. This tool was developed under the CoForChange project, whose overall objective is to predict changes in forests and in tree species distribution in the Congo Basin as a result of global change, and to develop decision-making tools. The article focuses on a full-scale example in real time, in the rainforest located at the borders of the Central African Republic, the Republic of the Congo and Cameroon, which was analysed on a year-by-year basis with thirty years of LANDSAT image archives.

Human activities within tropical rain forests provoke perturbations and degradation. The capacity to measure the extent of these damages is essential for calculating carbon emissions under the REDD+ process. Remote sensing is a powerful tool to provide such information (logging, mining, infrastructure building). Various techniques to identify and quantify forest degradation have been used so far. One step further is to link in-time degradation and past-time degradation to document the recovery of ecosystems after logging. For this, we propose a processing chain adapted to Landsat and Spot imageries. Post-processing is devoted to extract degradation information using specific algorithms. A spatial index, sized to low resolution radiometer, is implemented to render the actual and past degree of degradation. It is then related to continental land cover maps to get insight in the actual state of particular forest categories. This tool is developed in the framework of the CoForChange program (http://www.coforchange.eu/project) which general objective is to predict changes in forest and tree species distribution in the Congo Basin due to global change and to elaborate decision-making tools. The presentation is focused on an example in real time and in real size. It covers all the moist forest of Central African Republic at a yearly time-scale during ten years of Landsat images archive.

Outil privilégié de la gestion durable, le plan d'aménagement cherche le plus souvent à concilier la production de bois et la conservation des multiples services rendus par la forêt. Le Cirad a développé une expertise unique en la matière. Il propose des méthodes et des outils d'aide à l'aménagement, fondés sur des scénarios comparatifs et des itinéraires techniques adaptés au cadre législatif et réglementaire de chaque pays. Ces outils offrent aux acteurs concernés, publics ou privés, les moyens de faire les choix de gestion les plus pertinents face aux enjeux actuels.

La gestion forestière vise des objectifs multiples: fonctions économiques, écologiques, sociales; maintien de la diversité, du rendement; pérennité des formations; gestion et conservation des ressources génétiques. Sur la base de connaissances et de modèles disponibles en matière de flux de gènes et de dynamique forestière, cette étude a permis: (1) l'élaboration d'un modèle démogénétique général prenant en compte simultanément les processus de reproduction (production et dispersion des graines et du pollen en lien avec la phénologie), de régénération ou recrutement, de croissance, compétition et mortalité; (2) l'implémentation informatique d'un tel modèle et l'intégration logicielle des relations disponibles pour cinq situations forestières diverses au sein d'une plate-forme de simulation qui mutualise des fonctionnalités génériques (création d'une bibliothèque "Genetics") tout en étant suffisamment modulaire, ouverte, et accessible aux chercheurs et aux utilisateurs; (3) la réalisation de simulations destinées, pour chaque situation étudiée, d'une part à identifier les éléments les plus sensibles du modèle démogénétique et les processus dont l'étude doit être approfondie et d'autre part à comparer divers modes de gestion forestière (sylviculture, exploitation forestière) pour ébaucher des indications quant à leurs effets sur les ressources génétiques.

Forest scientists build models to study, understand and represent stand growth and dynamics. Ad hoc computer tools are often developed to implement these models, to test and to explore the consequences of the underlying hypotheses. Some teams have also developed more integrated tools, often based on a parameterizable growth model. The aim of this chapter is to present CAPSIS, a more generic forest simulator integrating various models, and to discuss the advantages for modellers and foresters of using such a shared forest-modelling platform for research, management and education. CAPSIS is free portable software, with a kernel providing the overall structure for simulations and generic data descriptions for stands, trees, soil, etc. These descriptions can be completed in modules - one for each model - which implement a proper data structure and specific evolution functions. A plug-in architecture allows the user to add tools for management, data exportation and representation, or connections with other software. At the present time, CAPSIS hosts six modules and two specific libraries. This shared forest-modelling platform has already shown many advantages: allowing modellers to re-use submodels and share specific tools, and providing to foresters and students a single easy-to-use software platform to compare different management strategies.

International audience; Forest scientists build models to study, understand and represent stand growth and dynamics. They are particularly interested in the evolution of ecosystems, at the tree and compartment levels and in the consequences of forest management on volume, shape, wood quality, structural evolution of the stand or sensitivity to meteorological and sanitary problems. Specific computer tools are often developed to implement these models, to test and to explore the consequences of the underlying hypotheses on real or virtual stands. Sometimes, a team may invest in the development of a more integrated tool, based on a parameterisable growth model. The Capsis project aims at integrating several types of forest growth and dynamics models - stand models, distance independent or spatially explicit tree models,... - and providing forest management tools to establish and compare different silvicultural scenarios. The objective is to build a perennial, open and dynamic software platform (1) to contribute to the development of odels and test their sensitivity to some parameters by simulating the manager's actions, (2) to share tools and methods, (3) to compare results of different models, (4) to transfer models to the managers and (5) to serve as teaching material. Most models implemented in Capsis are described in a web database hosted by the European Institute for Cultivated Forest (www.iefc.net). Capsis is a portable software, designed around a kernel which provides an organizational data structure - session, project, scenario step. The kernel also proposes generic data descriptions - stand, compartment, tree,... These descriptions can be completed in modules - one for each model - which implement a proper data structure and a specific evolution function (growth, mortality, regeneration, dissemination,...) with a chosen simulation step. A plugin architecture provides the possibility to build tools for management, construction of graphics, data exportation, tree group construction, stand representation or connections with other software. At the present time, Capsis hosts six modules, several other integration projects are planned and model integration training sessions are periodically proposed.

The main challenge for forest managers is to define the optimal rotation period able to ensure resource sustainability on a long-term basis. For this purpose, Cirad-Forêt developed three types of forest dynamics models: a density-dependent matrix model, a hybrid model (matrix / single-tree distance-independent) and a single-tree spatial model. These models were built based on records from the experimental site of Paracou in French Guiana, where Cirad-Forêt has monitored the growth, survival and mortality of trees since 1984, None of those models has proved efficient enough to properly assess the requirements of forest managers as far as the future of particular valuable species are concerned. Simple matrix models, as those used in several forest concessions of central Africa, could bring quick answers to the managers. In this paper, we calibrate such models for the three most harvested species in French Guiana; assess the impact of the uncertainty of the input parameters on the predictions and give some advice on the proper use of this method. Finally, we show the relevance of examining separately the fate of different species, as species groups can hide endangered individual situations.

Capsis is a multi-platform simulation framework, being used by an increasing number of french modellers to build forest dynamics models (see the presentation entitled "CAPSIS: Computer-Aided Projections of Strategies In Silviculture", same workshop). Formerly written in SmaiTalk, Selva, a model dedicated to the study and simulation of tropical rainforest dynamics on a spatialized per tree basis, has been ported on this new framework. This paper address the process initiated by this porting, which lead to a completely new approach. A wide spectrum of requests emerged from interactions with current and future users of the model. In order to satisfy most of them and provide as global solutions as possible, generic mechanisms were prioritized. Depending on their needs and available data, users can configure "à la carte" models to handle an ecological context and then focus their attention on the specific points they are interested in.

Forest scientists build models to study, understand and represent stand growth and dynamics. They are particularly interested in the evolution of ecosystems, at the tree and compartment levels and in the consequences of forest management on volume, shape, wood quality, structural evolution of the stand or sensitivity to meteorological and sanitary problems. Specific computer tools are often developed to implement these models, to test and to explore the consequences of the underlying hypotheses on real or virtual stands. Sometimes, a team may invest in the development of a more integrated tool, based on a parameterisable growth model. The Capsis project aims at integrating several types of forest growth and dynamics models - stand models, distance independent or spatially explicit tree models... - and providing forest management tools to establish and compare different silvicultural scenarios. The objective is to build a perennial, open and dynamic software platform (1) to contribute to the development of models and test their sensitivity to some parameters by simulating the manager's actions, (2) to share tools and methods, (3) to compare results of different models, (4) to transfer models to the managers and (5) to serve as teaching material. Capsis is a portable software, designed around a kernel which provides an organizational data structure - session, project, scenario step. The kernel also proposes generic data descriptions - stand, compartment, tree... These descriptions can be completed in modules - one for each model - which implement a proper data structure and a specific evolution function (growth, mortality, regeneration, dissemination...) with a chosen simulation step. A plug-in architecture provides the possibility to build tools for management, construction of graphics, data exportation, tree group construction, stand representation or connections with other software. At the present time, Capsis hosts six modules, several other integration projects are planned and model integration training sessions are periodically proposed.

National audience; Présentation succinte du logiciel CAPSIS qui permet de simuler et de comparer différents scénarios sylvicoles en peuplements forestiers.

Cette sortie de terrain a comporté plusieurs arrêts dans plusieurs types de formations végétales : futaie équienne de hêtres avec jeune régénération de hêtres, hêtres isolés dans une prairie, futaie de pins à crochets avec jeune régénération d'épicéas, futaie équienne fermée de hêtres avec régénération agée de hêtres. Ces observations ont servi de base à l'exposé par E. Nicolini des bases de l'architecture. Ces exemples contrastés ont permis de discuter le protocole de mesure dendrométrique qui doit être mis en place en Guyane et/ou au Gabon