-
Net ecosystem production (NEP) was used to represent C uptake and was calculated as follows:
where NPP is the net primary production and HR is the heterotrophic respiration, both in Pg C yr−1, at the regional and global scales.
The C accumulation after 1901 was calculated as:
where Ct is the sum of NEP in each year i (i = 1, 2…t) and the unit of NEP is Pg C yr−1.
In this study, we estimated the P limitation using two different biogeochemical cycles: the C–N cycle and C–N–P cycle. In the C–N–P experiment, we assumed that increases in C uptake and storage were limited by P. As we investigated how a new biochemical cycle changed the C uptake and C storage, the changes of each variable (Δx) due to the inclusion of P were calculated as the difference between the C–N–P and C–N experiments:
where
$ {x}_{t}^{\mathrm{C}\mathrm{N}} $ represents the change in each variable when using the C–N cycle (i.e., without the P cycle) t times relative to the baseline value in 1901, and$ {x}_{t}^{\mathrm{C}\mathrm{N}\mathrm{P}} $ is the same but when using the C–N–P cycle (i.e., considering the effect of P), relative to the initial values (units: Pg C yr−1).The magnitude of P limitation was calculated as:
where
$ {\mathrm{f}}_{t}^{\mathrm{P}} $ represents the magnitude of P limitation at the tth time and the unit is %.The above formulae are based on different biogeochemical cycles. We assumed that the N:P ratios were constant for all pools under the C–N cycle but that the C:N:P ratios differed among different pools under the C–N–P cycle. When higher and negative scores (
$ {\mathrm{f}}_{t}^{\mathrm{P}} $ ) were obtained, it implied that there were greater P restrictions on C uptake. Conversely, smaller or positive scores were characterized as low or no nutritional restriction. -
In this study, we used version 2.0 of CABLE, including a global biogeochemical model (CASA-CNP) (Wang et al., 2010). The biogeochemical model predicts both canopy leaf area index and maximal leaf carboxylation rate (vcmax), which are not specified in advance. The effects of nutrient limitation on the terrestrial C cycle under different representative concentration pathways have previously been evaluated using this version of CABLE (Zhang et al., 2013, 2014; Peng et al., 2020).
We employed each of the two biochemical cycles in CABLE (version 2.0), including interactive C and N cycles (C–N cycle), reported previously by Peng et al. (2020), and fully cooperative C, N, and P cycles (C–N–P cycle), for simulating C accumulation in China. The model includes three plant, three litter, and three soil pools for the C, N, and P cycles. The method by which CABLE calculates nutrient limitations has been reported previously (Zhang et al., 2013). There are five submodels in this model: radiation, canopy micrometeorology, surface flux, soil and snow, and biogeochemical cycles (Zhang et al., 2013). Good performance has been shown in estimating vegetation productivity and sizes of C pools, as compared with observations from eddy flux measurements or other process-based land surface models (Wang et al., 2007, 2012; Zhang et al., 2016; Peng et al., 2020).
There were eight types of meteorological inputs from 1901 to 2100: temperature, specific humidity, air pressure, downward solar radiation, downward longwave radiation, rainfall, snowfall, and wind speed. The meteorological variables from 1901 to 2005 were generated from the meteorological variables of GCP-TRENDY data. Using the method as reported by Qian et al. (2006), the GCP-TRENDY meteorological datasets were interpolated from six-hourly to hourly at a 1.9° (latitude) by 2.5° (longitude) spatial resolution. From 2006 to 2100, using the same spatial resolution, hourly meteorological variables were generated from the daily corresponding variables using CESM version 1.0 under the most severe Representative Concentration Pathway (RCP) scenario (RCP8.5; Riahi et al., 2011).
For each simulation, we spun up the model by recycling the meteorological forcing for 1901 to 1910 until all C and N pools reached equilibrium. If the difference in any pool size between two successive cycles was <0.01%, it was judged to be in equilibrium for the model. These equilibrium values of all state variables were used as the initial conditions to perform the simulations for the period 1901–2100. In addition, for all simulations, the vegetation cover for the 1990s was used based on the land cover classification of the International Geosphere Biosphere Program (IGBP) data at a 0.5° by 0.5° spatial resolution (Loveland et al., 2000). Subsequently, it was interpolated into CABLE plant functional types by Kowalczyk et al. (2016). Thus, in this study, we were unable to account for the impacts due to changes in land use (Peng et al., 2020).
To assess the impacts of climate change, atmospheric CO2, and N deposition on modeling terrestrial C uptake in China, we performed four different simulations using CABLE, based on each of the two different biochemical cycles (Table 1). The simulations were as follows: simulations CN1 and CNP1 included atmospheric CO2, climate, and N deposition, which varied over time; simulations CN2 and CNP2 fixed atmospheric CO2 at the 1901 level of about 295.4 ppm but allowed climate and N deposition to vary over time; simulations CN3 and CNP3 fixed the climate at the 1901 level but allowed atmospheric CO2 and N deposition to vary over time; and simulations CN4 and CNP4 used the 1901 N deposition of 20 Tg N yr−1 (Lamarque et al., 2010, 2013) but allowed atmospheric CO2 and climate to vary over time for all model years. The experiments for each cycle (C–N and C–N–P) are named CN and CNP, respectively. The terrestrial C uptake differences between the CNP and CN simulations were used to examine the effect of P limitation.
Simulation CO2 Climate N deposition CN1/CNP1 Time-varying Time-varying Time-varying CN2/CNP2 Fixed at 1901 Time-varying Time-varying CN3/CNP3 Time-varying Fixed at 1901 Time-varying CN4/CNP4 Time-varying Time-varying Fixed at 1901 Table 1. CABLE simulations used in this study under the C–N–P cycle and C–N cycle.
Under historical and future conditions from 1901 to 2100, relative to initial conditions in 1901, the differences in NPP were estimated for simulations CN1 and CNP1. We also compared our results with results from CMIP6, globally and for China. The CMIP6 models used were ACCESS-ESM, BCC-CSM2-MR, CanESM5, CLASS-CTEM, CESM2, CNRM-ESM2-1, IPSL-CM6A-LR, and UKESM1-0- LL. Further details are provided in Table 2.
Model name Spatial resolution Land component Full N cycle Full P cycle Fire Reference ACCESS-ESM1.5 1.875° × 1.25° CABLE Yes Yes No Ziehn et al. (2020) BCC-CSM2-MR 1.125° × 1.125° AVIM2 No No No Wu et al. (2019) CanESM5 2.81° × 2.81° CLASS-CTEM No No No Arora and Scinocca (2016) CESM2 0.9° × 1.25° CLM5 Yes No Yes Danabasoglu et al. (2020) CNRM-ESM2-1 1.4° × 1.4° ISBA-CTRIP No No Yes Séférian et al. (2016) IPSL-CM6A-LR 2.5° × 1.3° ORCHIDEE No No No Hourdin et al. (2020) UKESM1-0- LL 1.875° × 1.25° JULES-ES1.0 Yes No No Sellar et al. (2019) Table 2. Details of the models from CMIP6 used in this study (Peng et al., 2021).
New emission scenarios in CMIP6 driven by different socioeconomic models, i.e., the SSPs, have replaced the RCPs used previously in CMIP5. Among the scenarios, SSP5-8.5 and RCP8.5 are the most severe in terms of emissions. We chose CMIP6 under SSP5-8.5 instead of CMIP5 under RCP8.5, and this choice did not influence our conclusions. The reason was because of the focus in this study on the impacts of a new biogeochemical cycle on C uptake and C storage in China, which is an aspect better represented in CMIP6 models. Correspondingly, we used CMIP6 outputs to check whether our results fell within a reasonable range.
Simulation | CO2 | Climate | N deposition |
CN1/CNP1 | Time-varying | Time-varying | Time-varying |
CN2/CNP2 | Fixed at 1901 | Time-varying | Time-varying |
CN3/CNP3 | Time-varying | Fixed at 1901 | Time-varying |
CN4/CNP4 | Time-varying | Time-varying | Fixed at 1901 |