Impact of changes in climate and CO2 on the carbon-

15 Documenting year-to-year variations in carbon-sequestration potential in terrestrial ecosystems is crucial 16 for the determination of carbon dioxide (CO2) emissions. However, the magnitude, pattern and inner 17 biomass partitioning of carbon-sequestration potential, and the effect of the changes in climate and CO2 18 on inner carbon stocks, remain poorly quantified. Herein, we use a spatially explicit individual based19 dynamic global vegetation model to investigate the influences of the changes in climate and CO2 on the 20 enhanced carbon-sequestration potential of vegetation. The modelling included a series of factorial 21 simulations using the CRU dataset from 1916 to 2015. The results show that CO2 predominantly leads 22 to a persistent and widespread increase in above-ground vegetation biomass carbon-stocks (AVBC) and 23 below-ground vegetation biomass carbon-stocks (BVBC). Climate change appears to play a secondary 24 role in carbon-sequestration potential. Importantly, with the mitigation of water stress, the magnitude of 25 the aboveand below-ground responses in vegetation carbon-stocks gradually increases, and the ratio 26 between AVBC and BVBC increases to capture CO2 and sunlight. Changes in the pattern of vegetation 27 carbon storage was linked to regional limitations in water, which directly weakens and indirectly 28 regulates the response of potential vegetation carbon-stocks to a changing environment. Our findings 29 differ from previous modelling evaluations of vegetation that ignored inner carbon dynamics and 30 https://doi.org/10.5194/gmd-2021-383 Preprint. Discussion started: 22 November 2021 c © Author(s) 2021. CC BY 4.0 License.

demonstrates that the long-term trend in increased vegetation biomass carbon-stocks is driven by CO2 31 fertilization and temperature effects that are controlled by water limitations.  (Harris et al., 2020). Because the CRU dataset is a monthly based dataset, the monthly meteorological 117 data were converted into daily climatic variables by supplementing daily climatic variability within each 118 month using the National Centre for Environmental Prediction (NCEP) daily climate dataset. The NCEP 119 data, displayed using the T62 Gaussian grid with 192 × 94 points, was interpolated into a 0.5° grid (which 120 corresponds to the CRU dataset) using a linearly interpolation method. By combining the CRU data, with 121 the interpolated NCEP dataset, we were able to directly obtain the most of driving meteorological data 122 (details in Sato et al. (2020)). Neither the CRU nor NCEP datasets included downward shortwave and 123 longwave radiation. Thus, daily cloudiness values in the NCEP were used to calculate radiation values 124 using empirical functions (Sato et al., 2007). These data were all aggregated to a daily timescale with 0.5° 125 resolution to run SEIB-DGVM. 126 individuals without human disturbances, leading to a more properly calculated and accurate 173 representation of the responses of potential vegetation biomass to external environmental change. 174 Therefore, SEIB-DGVM, in general, effectively represents plant competition and function dynamics 175 under environmental change (Sato et al., 2007). 176

Parameterization of daily allocation 178
Flexible allocation schemes about resources and biomass are set up in the framework of the SEIB-DGVM 179 biogeochemical model. Atmospheric CO2 is assimilated by the photosynthesis of both woody and grass 180 foliage, and then is added into the non-structural carbon of the plant. This non-structural carbon of 181 photosynthetic production is allocated to all the plant organs (foliage, trunk, root, and stock), supplying 182 what is needed for the maintenance and growth of each organ. When the non-structural carbon is greater 183 than 0 during the growth phase, the following dynamic carbon allocation is executed for each individual 184 plant at the daily time scale, such that: 185 (1) When the fine root biomass (massroot) of wood or grass does not satisfy minimum requirements for 186 fulfilling functional balance (massleaf/FRratio), the mass of non-structural carbon is allocated to the root 187 biomass to supplement the deficit. Here, massleaf is the leaf biomass, and FRratio is the ratio of massleaf to 188 massroot satisfying the functional balance. 189 (2) The stock biomass is supplemented until it is equal to leaf biomass. This scheme is active after the 190 first thirty days of the growing phase. 191 (3) Woody leaf biomass is constrained by three limitations of the maximum leaf biomass, which are 192 calculated as follows: 193 is maximum leaf area of PFTs per unit biomass (m 2 m −2 ), and 1 represents the area of 200 transport tissue per unit biomass, and is a constant (dimensionless). If the massleaf is less than the 201 minimum ( 1 , 2 , 3 ) , the mass of non-structural carbon is allocated into leaf biomass to 202 supplement the deficit. 203 Grass leaf biomass is supplemented until the leaf area index of grass equals the optimal leaf area index, 204 which are calculated as: 205 where is optimal leaf area index (m 2 m −2 ), is the grass photosynthetically active radiation 207 (μmol photon m −2 s −1 ), is the light-saturated photosynthetic rate (μCO2 m −2 s −1 ), is the light-use 208 efficiency of photosynthesis (mol CO2 mol photon −1 ), is the cost of maintaining leaves per unit leaf 209 mass per day (g DM g DM −1 day −1 ), is day length (hour), and is light attenuation coefficient at 210 midday. 211 (4) When non-structural carbon is less than 10 g dry mass (DM) PFT −1 or annual NPP is less than 10 g 212 DM PFT −1 in the previous year, the following daily simulation processes (5~6) will be skipped. 213 (5) When total woody biomass is more than 10 kg DM, which defines the minimum tree size for 214 reproduction, 10% of non-structural carbon is transformed into litter. 215 (6) During the simulation of trunk growth, the remaining structural carbon is allocated to sapwood 216 biomass. There is no direct allocation to heartwood, which is transformed slowly from sapwood biomass. 217 For grass PFTs biomass, the densities of all organs comprising the biomass never decline below 0.1 g 218 DM m −2 even if the environment is deteriorated for grass survival. A more detailed description of SEIB-219 DGVM is given by Sato et al. (2007).  According to the flexible allocation scheme, SEIB-DGVM allocates and stores the biomass carbon in 232 four pools of woody PFT (foliage, trunk, root, and stock) and three pools of grass PFT (foliage, root, and 233 stock). To investigate the fractional variability of carbon-sequestration potential between the pools, we 234 partitioned potential vegetation carbon-stocks based on the physiological function of the plant (Figure  235 A1). The root-shoot ratio (R/S) has been widely used to investigate the relationship between aboveground 236 vegetation biomass to belowground vegetation biomass and is considered an important variable in the 237 terrestrial ecosystem carbon cycle (Zhang et al., 2016). In this study, we adjusted the method of 238 calculating the R/S ratio by distinguishing between the aboveground vegetation biomass carbon-stock 239 (AVBC) and the belowground vegetation biomass carbon-stock (BVBC). AVBC includes biomass carbon 240 from woody foliage, woody trunk, and grass foliage, while BVBC includes biomass carbon from woody 241 fine roots and grass fine roots, excluding the stock pool. Thus, 242 where AVBC is aboveground vegetation biomass carbon-stock (kg C m −2 ), BVBC is belowground 244 vegetation biomass carbon-stock (kg C m −2 ), is the leaf biomass carbon-stock of wood (kg 245 C m −2 ), and is the trunk biomass carbon-stock of wood (kg C m −2 ), including both branch 246 and structural roots. This biomass is simplistically attributed to aboveground organs and is used primarily 247 to support the plant.
is the leaf biomass carbon-stock of grass (kg C m −2 ), whereas 248 and are functional root (fine roots) biomass carbon-stocks of wood and grass, 249 separately (kg C m −2 ), which absorb water and nutrition from soil.
To further quantify the relative contributions of varying atmospheric CO2 concentrations, precipitation, 261 temperature, and radiation, we performed six factorial simulations after the spin-up procedure using 262 different input variables between 1916 and 2015 (Table 1). Other drivers included wind velocity and 263 relative humidity. Consistent with previous studies (Zhu et al., 2016;Piao et al., 2006), the contribution 264 of CO2 to the trend in carbon-stocks trend was defined as the ratio of the carbon-stock increase from 265 simulation S2 to that of simulation S1. The contributions of precipitation, temperature, radiation, and 266 other factors were calculated by subtracting simulation S2 from each corresponding simulation (S3, S4, 267 S5, S6, respectively), then dividing by simulation S1. 268

Non-parametric test methods 269
Each driving factor (atmosphere CO2, precipitation, temperature, and radiation) has a different influence 270 on the carbon-stock, so it is difficult to make a simple pre-assumption about the population distribution 271 pattern for factorial simulations. We used the non-parametric Mann-Kendall and Sen's slope estimator 272 statistical tests (Gocic and Trajkovic, 2013)  Locally available water strongly regulates and limits the response of carbon-stocks to changes in climate 281 and CO2. We defined an aridity index (AI) to distinguish between the global hydrological regions for 282 comparing the long-term trend in carbon-stocks over different hydrological environments, and for 283 quantifying the influences of each hydrological environment on the variations in the trends. The AI was 284 defined as: 285 where ̅ is the multiyear mean precipitation (mm year −1 ), and ̅̅̅̅̅ is the multiyear mean potential 287 evapotranspiration (mm year −1 ), which was calculated by the Penman-Monteith model (Monteith and 288 Unsworth, 1990). As in a previous study (Chen et      116.18 ± 2.34 Pg C (or ~15.60%) and dominates the positive global carbon-stock trend; BVBC also 343 increases 3.08 ± 0.14 Pg C (or ~18.03%) over the past century. 344

Spatial variability in estimated AVBC and BVBC trends 354
Based on the carbon-stock partitioning method, we found that the integrated carbon-stock as well as the 355 above-and belowground carbon-stocks over the period of 1916-2015 exhibited a remarkable spatial 356 heterogeneity. Figure 6a shows that an increase in vegetation carbon-stocks occurred over regions and 357 global aggregate levels during the entire study period. About 57.39% of the terrestrial grids exhibited an 358 increase with a noticeable trend (p<0.05) in biomass carbon-stock; 53.82% of global grids possessed 359 increases that were statistically significant at the p=0.01 level. To determine the contributions of each 360 fraction (AVBC, BVBC) to the integral change in the potential vegetation carbon-stock, we partitioned 361 and present the historical spatial and temporal patterns for each fraction separately (Figure 6b, 6c). AVBC 362 contributes 97.33% to the total incremental change (116.18 ± 2.34 Pg C), with about 51.32% of the grids 363 possessing a noticeable positive trend (p=0.01). Generally, spatial patterns of AVBC and the integral 364 carbon-stock are consistent (Figure 6a, 6b), which further supports the argument that AVBC dominates 365 the trend in carbon-stocks in most regions. Although the proportion of the total change in carbon-stocks 366 is small (3.08 ± 0.14 Pg C), about 61.00% of the land surface shows an increase in BVBC; of these 367 terrestrial grids, 55.81% was characterized by a significant p=0.01 increase.  Eurasia. Negative trends in AVBC/BVBC ratios are found in northern America, southern Europe, and 376 tropical Africa. 377

Responses of AVBC and BVBC to environmental drivers 378
The responses of AVBC and BVBC to changes in climate and CO2 are both positive at the global level 379 (Figure 7a, 7c), although regionally, they exhibit both negative and positive responses (Figure 7b, 7d).    Climate change induced by the greenhouse effect explains part of the increase in carbon-stocks, but 404 unlike CO2 fertilization, climate has dramatic negative effects on some vegetated regions. Figure 7  405 illustrates that temperature is the largest climatic contributor to the change in AVBC (13.83%, 2.572 g 406 m −2 yr −1 ), followed by precipitation (8.51%, 1.572 g m −2 yr −1 ) and radiation (-3.19%, -0.649 g m −2 yr −1 ). 407 The spatial distribution shows that temperature predominantly influences the change in AVBC, 408 influencing over 27.56% of the global vegetated regions, followed by precipitation (21.88%) and 409 radiation (20.67%). Modelled BVBC trends based on the factorial simulations have similar 410 spatiotemporal patterns to AVBC. The effects of temperature on BVBC are stronger than AVBC, because 411 fine root tightly correlates with temperature (Gill, 2000). Meanwhile, there is a difference in the negative 412 contribution of precipitation to the change in BVBC at the global level (-2.76%, -0.013 g m −2 yr −1 ). It 413 should be noted that trends in the global carbon-stock can be largely attributed to the influences of CO2, 414 precipitation, temperature, and radiation (Figures 8, 9). Nonetheless, at the regional scale, the 415 contributions of other factors should be considered, such as humidity and wind speed. The effects of 416 these other factors dominate trends in AVBC in over 16.05% of the regions that increased and 6. humid region during the past hundred years. Drivers attributed to increase BVBC density changed from 444 0.011 ± 0.001 kg C m −2 in the hyper-arid regions to 0.044 ± 0.005 kg C m −2 in the humid regions during 445 the same period ( Figures A3, A4). With a lessening of water stress (from hyper-arid to humid area), the 446 response of the carbon-stock to changes in climate and CO2 gradually became more noticeable. The 447 robust pattern in the regional average density of the carbon-stock shows that terrestrial water limitations 448 strongly limit the enhanced magnitude of the carbon-stock. 449 rate of change in the AVBC/BVBC ratio gradually decreased. Vegetation utilizes a tolerance strategy to 457 allocate biomass, storing more biomass carbon in roots to resist enhanced water stress (Chen et al., 2013). 458 In humid regions (AI>0.65), the proportion of AVBC increases more than that of BVBC to obtain more 459 resources like CO2 and radiation energy, leading to an increase in the AVBC/BVBC ratio. Conforming to 460 the optimal partitioning hypothesis, plants store more carbon in shoots and leaves in environments where 461 water is more available and shift more carbon to roots when water is more limited (Yang et al., 2010;462 Mcconnaughay and Coleman, 1999). Terrestrial water availability has a strong regulating effect on the 463 spatial pattern of growth in the carbon-stock, demonstrating that the effects of the changes in climate and 464 CO2 on the dynamics of the vegetation carbon-stock are controlled by the terrestrial water gradient. 465  (Figures 10, 11), which suggests that limited water availability constrains the response 500 magnitude of the changes in carbon-stocks to changes in CO2 and climate. In contrast, we found that 501 indirect factors constrain the impact of increasing water stress on the response of carbon-stocks. Although 502 vegetation carbon-stocks dramatically increase under the effects of climate and CO2 changes, vegetation 503 in humid regions stores more biomass (and carbon) in aboveground plant organs (trunk and foliage) to 504 obtain nutrients and light. Dryland vegetation lowers the AVBC/BVBC ratios and stores more biomass 505 below ground to enhance the capture of water resources. Terrestrial ecosystems utilize sensitive strategies 506 to allocate and store biomass to adjust to local hydrological conditions, which is consistent with optimal 507 partitioning theory (Mcconnaughay and Coleman, 1999). A significant conclusion is that water 508 constraints not only confine the responses of vegetation carbon-stocks to drivers of variability, but also 509 constrain the proportion of biomass carbon-stocks in above-and belowground fractions.

Code and data availability statement 540
The code of SEIB-DGVM version 3.02 can be download from http://seib-dgvm.com/. Climatic Research 541 Unit data can be downloaded from https://crudata.uea.ac.uk/cru/data/hrg/. The soil physical parameters 542 can be downloaded from www.iges.org/gswp. The reconstructed CO2 concentration dataset and SEIB 543 code can be downloaded from http://seib-dgvm.com/. In model validation, Ecosystem Model-Data 544