" Idol with feet of clay ": reliable predictions of forest ecosystem functioning require flawless climate forcings.
Climate change affects various aspects of the functioning of ecosystem, especially photosynthesis, respiration and carbon storage. We need accurate modelling approaches (impact models) to simulate the functioning, vitality and provision of ecosystem services of forests in a warmer world. These impact models require climate data as forcings, which are often produced by climate models comparing more or less well with observational climate data. The bias percentage of the climate forcings propagates throughout the modeling chain from the climate model to the impact model. In this study, we aimed to quantify these bias percentage, addressing three questions: (1) Do the impact model predictions vary when forcing it with different climate models, and how do the predictions under climate model vs. observational climate forcing differ? (2) Does the variability in the impact climate simulations caused by climate forcings fade out at large spatial scale? (3) How the fact of using simulated climatic data affects the process-based model predictions in the case of stressful events? To answer these questions, we present results obtained over the historical period (e.g. 1970-2010) with the CASTANEA ecophysiological forest model and use the data from three climate models. Our analysis focuses on French forests, studying European beech (Fagus sylvatica), temperate deciduous oaks (Quercus robur and Q. petraea), Scots pine (Pinus sylvestris) and spruce (Picea abies) monospecific stands. We show that prediction of photosynthesis, respiration and wood growth highly depends on the climate model used, whether debiased or not, and also on species and region considered. Overall, we observed an improvement of prediction after a monthly mean bias or monthly quantile mapping correction for three model considered, but not with the same success. Then we highlighted a large variability in the processes simulated by the impact model under different climate forcings when considering the plot (i.e. scale of a few hectares) scale. This variability fades out at larger scale (e.g. the scale of an ecological region, i.e. 100 km2), owing to an aggregation effect. Moreover, process predictions obtained under different climate forcings are more variable during driest years. These results highlight the necessity to quantify bias and uncertainties in climate forcings before predicting fluxes dynamics with process-based model.