Tree mortality risks under climate change in Europe: assessment of silviculture practices and genetic conservation networks

General context: Climate change can positively or negatively affect abiotic and biotic drivers of tree mortality. Process-based models integrating these climatic effects are only seldom used at species distribution scale. Objective: The main objective of this study was to investigate the multi-causal mortality risk of five major European forest tree species across their distribution range from an ecophysiological perspective, to quantify the impact of forest management practices on this risk and to identify threats on the genetic conservation network. Methods: We used the process-based ecophysiological model CASTANEA to simulate the mortality risk of \textit{Fagus sylvatica}, \textit{Quercus petraea}, \textit{Pinus sylvestris}, \textit{Pinus pinaster} and \textit{Picea abies} under current and future climate conditions, while considering local silviculture practices. The mortality risk was assessed by a composite risk index \textit{(CRIM)} integrating the risks of carbon starvation, hydraulic failure and frost damage. We took into account extreme climatic events with the \textit{CRIM$_{max}$}, computed as the maximum annual value of the \textit{CRIM}. Results: The physiological processes' contributions to \textit{CRIM} differed among species: it was mainly driven by hydraulic failure for \textit{P. sylvestris} and \textit{Q. petraea}, by frost damage for \textit{P. abies}, by carbon starvation for \textit{P. pinaster}, and by a combination of hydraulic failure and frost damage for \textit{F. sylvatica}. Under future climate, projection showed an increase of \textit{CRIM} for \textit{P. pinaster} but a decrease for \textit{P. abies}, \textit{Q. petraea} and \textit{F. sylvatica}, and little variation for \textit{P. sylvestris}. Under the harshest future climatic scenario, forest management decreased the mean \textit{CRIM} for \textit{P. sylvestris}, increased it for \textit{P. abies} and \textit{P. pinaster} and had no major impact for the two broadleaved species. By the year 2100, 38\% to 90\% of the conservation units are at extinction threat (\textit{CRIM$_{max}$}=1), depending on the species. Conclusions: Using a process-based ecophysiological model allowed us to disentangle the multiple drivers of tree mortality under current and future climate. Taking into account the positive effect of increased CO$_2$ on fertilization and water use efficiency, the average risks may increase or decrease in the future depending on species and sites. However, considering extreme climatic events, future projections are as pessimistic than those obtained with bioclimatic niche models. Abbreviation for column: X Longitude Y Latitude LAImax Leaf area index max reach Nha Density per hectar Vha Volume per hectar NEE Net ecosystem exchange NPP net primary production Reco Respiration ecosystem GPP Gross primary production Etveg Evapotranspiration canopy Etsol Evapotranspiration sol TR tree transpiration ETP evapotranspiration potentiel BiomassOfReserves Biomass of reserve rw ring width dbh diameter at breast heast height height BBday Budburst date rFD risk of frost CRIM_max Maximum combined risk index of mortality reach rNSC risk of carbon starvation rPLC risk of embolism rPLC_max Maximum risk of embolism reach CRIM combined risk index of mortality Climate Climatic model rNSC_max maximum risk of carbon starvation reach rFD_max Maximum risk of frost reach Scenario_Sylvicol null means no silvulcture simulated species species Country Country alt_watch altitude of climate simulated grid_watch number of the pixel point of WATCH grid_eurocordex number of the pixel point of Eurocordex Pinus_sylvestris 0 abscence ; 1 presence Fagus_sylvatica 0 abscence ; 1 presence Quercus_petraea 0 abscence ; 1 presence Picea_abies 0 abscence ; 1 presence Pinus_pinaster 0 abscence ; 1 presence