Large sensitivity in land carbon storage due to geographical and temporal variation in the thermal response of photosynthetic capacity

Summary Plant temperature responses vary geographically, reflecting thermally contrasting habitats and long‐term species adaptations to their climate of origin. Plants also can acclimate to fast temporal changes in temperature regime to mitigate stress. Although plant photosynthetic responses are known to acclimate to temperature, many global models used to predict future vegetation and climate–carbon interactions do not include this process. We quantify the global and regional impacts of biogeographical variability and thermal acclimation of temperature response of photosynthetic capacity on the terrestrial carbon (C) cycle between 1860 and 2100 within a coupled climate–carbon cycle model, that emulates 22 global climate models. Results indicate that inclusion of biogeographical variation in photosynthetic temperature response is most important for present‐day and future C uptake, with increasing importance of thermal acclimation under future warming. Accounting for both effects narrows the range of predictions of the simulated global land C storage in 2100 across climate projections (29% and 43% globally and in the tropics, respectively). Contrary to earlier studies, our results suggest that thermal acclimation of photosynthetic capacity makes tropical and temperate C less vulnerable to warming, but reduces the warming‐induced C uptake in the boreal region under elevated CO2.

used for ecosystem level model evaluation. Table S3 Global mean and standard deviation (µ ±s ) fields at the end of 1860 and 2100 (including variance s 2 ) and change in global land carbon.   Notes S1 Details of JULES model plant physiology: Photosynthesis model equations, leaf to canopy to grid level scaling, dynamic vegetation, stomatal conductance, leaf and plant respiration in JULES Following Kattge & Knorr (2007), we used the Farquhar et al., (1980) leaf C3 photosynthesis model as described in Medlyn et al., (2002). Accordingly, the net photosynthetic uptake (An) in [mol m -2 s -1 ] was calculated as the minimum of two limiting rates: Rubisco limited (Av) or electron transport limited (AJ) photosynthetic uptake both in [mol m -2 s -1 ] following Eqn S1 with Rd being the rate of leaf respiration.
In this study, scaling from leaf to canopy level (photosynthesis, respiration and stomatal conductance) was done using the big-leaf approach option within the JULES model. Canopy level flux was estimated the integral of the leaf level individual processes over the entire canopy leaf area. Remaining plant respiration components were estimated as a function of canopy respiration and individual tissue N:C ratios. Therefore, all plant respiration components retain the temperature response function of Vcmax. PFT specific parameters for biochemistry, photosynthesis and stomatal conductance are given in Table 2 of Clark et al. (2011). Scaling to ecosystem level or grid box level was done by adding the individual contributions (fluxes and stocks) of each PFT weighted by their gridcell fractional coverage. Competition between PFTs (broadleaf trees, needle-leaf trees, C3 and C4 grasses, and shrubs), i.e. dynamic vegetation, is included in all simulations. Competition is based on a prescribed dominance hierarchy -treeshrub-grass -where dominant PFTs limit expansion of subdominant PFTs.
Furthermore, JULES simulates a surface energy balance and includes a calculation of skin (leaf) temperature. Finally, calculations of photosynthesis, respiration, and full energy balance were done at hourly time scale and the vegetation dynamics module was updated every ten days.

Notes S2 Details of JULES-IMOGEN framework
Mean warming was calculated as a function of radiative forcing (via a parameterised energy balance model, EBM), which in turn is dependent on any altered atmospheric gas composition (Huntingford et al., 2010). Therefore, changes in the terrestrial carbon storage feedback on climate via atmospheric carbon dioxide (CO2) concentration. This is in addition to the changes in CO2 concentration directly due to fossil fuel burning, or draw-down into the oceans.
The oceanic uptake of atmospheric CO2 was calculated for each year, based on atmospheric CO2 concentration and temperature changes since pre-industrial, and up to that year, using an impulse response function. This was calibrated against the Princeton 3-D biogeochemical ocean model following Joos et al. (1996) as documented in Huntingford et al. (2004).

Notes S3 Calculation of leaf level photosynthetic temperature responses (Fig. 1)
We calculated the leaf-level temperature response of gross photosynthesis using the Farquhar et al. (1980)   . Tgrowth for the Geog configuration was based on the monthly pre-industrial air temperatures extracted from the CRU dataset at the three locations, as explained in the methods. By definition of the Geog configuration, Tgrowth is the same in 1860 and 2100. Tgrowth in Geog+Acclim was calculated for years 1860 and 2100 as the mean monthly air temperature of the months covered in the respective seasons for the two highlatitude gridcells and as the mean over the whole year for the tropical gridcell.
Monthly air temperatures at pre-industrial (1860) were taken from the CRU data set (New et al., 2000) and for 2100 extracted from the global JULES-IMOGEN simulation with the Geog+Acclim configuration, for each of three locations. Also, for computational simplicity, and only for these leaf-level photosynthetic temperature responses, internal CO2 concentration (Ci) was prescribed as 70% of the atmospheric CO2. We note that the optimal temperatures were likely overestimated as a result, since increasing vapour pressure deficit (VPD) with temperature is likely to drive stomatal closure (Lin et al., 2012); changing VPD was however accounted for in all the full global JULES-IMOGEN simulations. We used this information to produce temperature response curves over a range of relevant temperatures at each location.
To show the implication of the leaf-level results at the regional scale we extracted  Where 'x' can be DS or JV ratio, and 'i' denotes Jmax, Vcmax and JV.

Figure S1
Optimum temperatures (Topt) for Vcmax and Jmax and Jmax to Vcmax ratio at 25 o C with Tgrowth estimated from the CRU data set (New et al., 2000) for the growth periods of the years 1901-1910.