• Energy Research

Abstract Modelling phenology is crucial to assess the impact of climate change on the length of the canopy duration and the productivity of terrestrial ecosystems. Focusing on six dominant European tree species, the aims of this study were (i) to examine the accuracy of different leaf phenology models to simulate the onset and ending of the leafy season, with particular emphasis on the putative role of chilling to release winter bud dormancy and (ii) to predict seasonal shifts for the 21st century in response to climate warming. Models testing and validation were done for each species considering 2 or 3 years of phenological observations acquired over a large elevational gradient (1500 m range, 57 populations). Flushing models were either based solely on forcing temperatures (1-phase models) or both on chilling and forcing temperatures (2-phases models). Leaf senescence models were based on both temperature and photoperiod. We show that most flushing models are able to predict accurately the observed flushing dates. The 1-phase models are as efficient as 2-phases models for most species suggesting that chilling temperatures are currently sufficient to fully release bud dormancy. However, our predictions for the 21st century highlight that chilling temperature could be insufficient for some species at low elevation. Overall, flushing is expected to advance in the next decades but this trend substantially differed between species (from 0 to 2.4 days per decade). The prediction of leaf senescence appears more challenging, as the proposed models work properly for only two out of four deciduous species, for which senescence is expected to be delayed in the future (from 1.4 to 2.3 days per decade). These trends to earlier spring leafing and later autumn senescence are likely to affect the competitive balance between species. For instance, simulations over the 21st century predict a stronger lengthening of the canopy duration for Quercus petraea than for Fagus sylvatica , suggesting that shifts in the elevational distributions of these species might occur.

Essential to the French forestry and wood industry, the Sessile and Pedunculate oaks are threatened by the intensification of extreme climatic events, while the Pubescent oak, which is more resistant to drought, offers a promising alternative. Modelling is a key tool for exploring this alternative, but its complexity presents challenges in terms of transferability. Against this backdrop, two mechanistic models (CASTANEA and ForCEEPS) were used to assess the potential of Pubescent Oak in terms of maintenance and production in France by 2100. The projections show the species is declining in the warmer parts of its range, and that it may persist in the northern half of France and in mountainous areas. The volumes of wood harvested with the most dynamic management are lower, but nevertheless comparable to current removals of Sessile and Pedunculate oak. In the long term, the diameters of the individuals harvested (DBH > 25 cm) could allow them to be used as timber. The convergence of the two models highlights the importance of Pubescent Oak for the future of the forestry and timber industry, particularly outside its current area of distribution. The divergences between their projections for stand survival and productivity give a possible order of magnitude of the associated uncertainties, and are discussed in order to provide a critical analysis of forest modelling as a decision making tool. This analysis illustrates the importance of using different models to determine whether similar conclusions emerge despite different approaches, i.e. quantifying to a minimum and reducing uncertainties in order to produce more robust results. Essentiels à la filière forêt-bois française, les Chênes sessile et pédonculé sont menacés par l'intensification des événements climatiques extrêmes tandis que le Chêne pubescent, plus résistant à la sécheresse, offre une alternative prometteuse. La modélisation est un divergences de leurs projections concernant la survie et la productivité des peuplements donnent un ordre de grandeur possible des incertitudes associées, et sont discutées afin de fournir une analyse critique de la modélisation forestière en tant qu'outil d'aide à la décision. Cette analyse illustre l'importance d'utiliser des modèles différents pour déterminer si des conclusions similaires émergent malgré des approches variées, c'est-à-dire quantifier à minima et réduire les incertitudes afin de produire des résultats plus robustes.

Ecology Letters (2012)AbstractModel‐based projections of shifts in tree species range due to climate change are becoming an important decision support tool for forest management. However, poorly evaluated sources of uncertainty require more scrutiny before relying heavily on models for decision‐making. We evaluated uncertainty arising from differences in model formulations of tree response to climate change based on a rigorous intercomparison of projections of tree distributions in France. We compared eight models ranging from niche‐based to process‐based models. On average, models project large range contractions of temperate tree species in lowlands due to climate change. There was substantial disagreement between models for temperate broadleaf deciduous tree species, but differences in the capacity of models to account for rising CO2 impacts explained much of the disagreement. There was good quantitative agreement among models concerning the range contractions for Scots pine. For the dominant Mediterranean tree species, Holm oak, all models foresee substantial range expansion.

The structure of a forest stand, i.e. the distribution of tree size features, has strong effects on its functioning. The management of the structure is therefore an important tool in mitigating the impact of predicted changes in climate on forests, especially with respect to drought. Here, a new functional-structural model is presented and is used to assess the effects of management on forest functioning at a national scale.The stand process-based model (PBM) CASTANEA was coupled to a stand structure module (SSM) based on empirical tree-to-tree competition rules. The calibration of the SSM was based on a thorough analysis of intersite and interannual variability of competition asymmetry. The coupled CASTANEA-SSM model was evaluated across France using forest inventory data, and used to compare the effect of contrasted silvicultural practices on simulated stand carbon fluxes and growth.The asymmetry of competition varied consistently with stand productivity at both spatial and temporal scales. The modelling of the competition rules enabled efficient prediction of changes in stand structure within the CASTANEA PBM. The coupled model predicted an increase in net primary productivity (NPP) with management intensity, resulting in higher growth. This positive effect of management was found to vary at a national scale across France: the highest increases in NPP were attained in forests facing moderate to high water stress; however, the absolute effect of management on simulated stand growth remained moderate to low because stand thinning involved changes in carbon allocation at the tree scale.This modelling approach helps to identify the areas where management efforts should be concentrated in order to mitigate near-future drought impact on national forest productivity. Around a quarter of the French temperate oak and beech forests are currently in zones of high vulnerability, where management could thus mitigate the influence of climate change on forest yield.

Climate change is imposing drier atmospheric and edaphic conditions on temperate forests. Here, we investigated how deep soil (down to 300 cm) water extraction contributed to the provision of water in the Fontainebleau-Barbeau temperate oak forest over two years, including the 2018 record drought. Deep water provision was key to sustain canopy transpiration during drought, with layers below 150 cm contributing up to 60% of the transpired water in August 2018, despite their very low density of fine roots. We further showed that soil databases used to parameterize ecosystem models largely underestimated the amount of water extractable from the soil by trees, due to a considerable underestimation of the tree rooting depth. The consensus database established for France gave an estimate of 207 mm for the soil water holding capacity (SWHC) at Fontainebleau-Barbeau, when our estimate based on the analysis of soil water content measurements was 1.9 times as high, reaching 390 ± 17 mm. Running the CASTANEA forest model with the database-derived SWHC yielded a 185 gC m-2 y-1 average underestimation of annual gross primary productivity under current climate, reaching up to 687 ± 117 gC m-2 y-1 under climate change scenario RCP8.5. It is likely that the strong underestimation of SWHC that we show at our site is not a special case, and concerns a large number of forest sites. Thus, we argue for a generalisation of deep soil water content measurements in forests, in order to improve the estimation of SWHC and the simulation of the forest carbon cycle in the current context of climate change.

The effects of climate changes on carbon and water fluxes are quantified using a physiologically multi-layer, process-based model containing a carbon allocation model and coupled with a soil model (CASTANEA). The model is first evaluated on four EUROFLUX sites using eddy covariance data, which provide estimates of carbon and water fluxes at the ecosystem scale. It correctly reproduces the diurnal fluxes and the seasonal pattern. Thereafter simulations were conducted on six French forest ecosystems representative of three climatic areas (oceanic, continental and Mediterranean areas) dominated by deciduous species (Fagus sylvatica, Quercus robur), coniferous species (Pinus pinaster, Pinus sylvestris) or sclerophyllous evergreen species (Quercus ilex). The model is driven by the results of a meteorological model (ARPEGE) following the B2 scenario of IPCC. From 1960 to 2100, the average temperature increases by 3.1 °C (30%) and the rainfall during summer decreases by 68 mm (-27%). For all the sites, between the two periods, the simulations predict on average a gross primary production (GPP) increase of 513 g(C) m-2 (+38%). This increase is relatively steep until 2020, followed by a slowing down of the GPP rise due to an increase of the effect of water stress. Contrary to GPP, the ecosystem respiration (Reco) raises at a constant rate (350 g(C) m-2 i.e. 31% from 1960 to 2100). The dynamics of the net ecosystem productivity (GPP minus Reco) is the consequence of the effect on both GPP and Reco and differs per site. The ecosystems always remain carbon sinks; however the sink strength globally decreases for coniferous (-8%), increases for sclerophyllous evergreen (+34%) and strongly increases for deciduous forest (+67%) that largely benefits by the lengthening of the foliated period. The separately quantified effects of the main variables (temperature, length of foliated season, CO2 fertilization, drought effect), show that the magnitude of these effects depends on the species and the climatic zone.