Increased forest coverage will induce more carbon fixation in vegetation than in soil during 2015–2060 in China based on CMIP6

2.1.1. Carbon density and carbon flux data

2.1.2. Environment variables

2.2.1. Calculation of carbon stocks

The carbon stocks of vegetation (cVeg) and soil (cSoil) in China were calculated according to the weighted sum of their carbon density and area and subsequently, converted to a unit of 'PgC' (1 PgC = 10¹⁵ gC). Additionally, equation (1) was also used to calculate the carbon stocks in four climatic regions (He et al 2019; supplementary figure 1): temperate continental, temperate monsoonal, high-cold Tibetan Plateau, and subtropical-tropical monsoonal:

$$\begin{equation}{\operatorname{C} _{\operatorname{stock} }} = \mathop \sum \limits_{i = 1}^N \operatorname{C} _{\operatorname{density} }^i{\,}\times{\,}{A_i} \times {10^{-15}}\end{equation} \tag{ 1 }$$

where ${\operatorname{C} _{\text{stock}}}$ is the carbon stock (cVeg or cSoil; PgC) in China or the four climatic regions, $\operatorname{C} _{\operatorname{density} }^i$ and ${A_i}$ are the carbon density (gC m⁻²) and area (m²) of pixel i, respectively

2.2.2. Carbon stock or sink ratios of vegetation to soil

Vegetation to soil carbon stock ratios (${\operatorname{Ratio} _{\operatorname{stock} }}$; equation (2)) were calculated for China and each climate region to compare variations. For further comparisons of the carbon sequestration capacities of vegetation and soil during the study period, carbon sink ratios (${\operatorname{Ratio} _{\operatorname{sink} }}$, equation (3)) were estimated:

$$\begin{equation}{\operatorname{Ratio} _{\text{stock}}} = \frac{\text{cVeg}}{\text{cSoil}}\end{equation} \tag{ 2 }$$

$$\begin{equation}{\operatorname{Ratio} _{\text{sink}}} = \frac{\text{cVeg }\,{\text{slope}}}{\text{cSoil }\,{\text{slope}}}\end{equation} \tag{ 3 }$$

where ${\operatorname{Ratio} _{\text{stock}}}$ and ${\operatorname{Ratio} _{\text{sink}}}$ are the ratios of vegetation to soil carbon stocks and carbon sinks, respectively, and $\operatorname{cVeg} \,{\text{slope}}$ and $\operatorname{cSoil} \,{\text{slope}}$ are the temporal trends of vegetation and soil carbon stocks from 2015 to 2060.

2.2.3. Effect of environmental factors on carbon stock modifications

Before conducting attribution analysis, the LUH2 dataset was reclassified into the land use types (forest, grassland, cropland, city, and bare land) analyzed in this paper according to the classification scheme of supplementary table 1. The average coverage (%) of LUH2 land use has relatively small differences with MODIS observations, and forests increased and grasslands declined from 2015 to 2020, which is consistent with MODIS (supplementary table 2). The annual changes in the average coverage (%) of each land use type in China and the four climatic regions were observed (supplementary figure 2). The results showed that the main land use change characteristics in China and the four climatic regions were: conversion of grasslands to forests and bare lands (occurring in China and the high-cold Tibetan Plateau climatic region), conversion of grasslands to forests (occurring in the subtropical-tropical monsoonal climatic region), conversion of grasslands to bare lands (occurring in the temperate continental climatic region), and conversion of grasslands and croplands to forests and bare lands (occurring in the temperate monsoonal climatic region) under the SSP126 scenario (supplementary figure 2). Consequently, to avoid collinearity, the percentage coverages of target land use type (i.e. forests or bare lands) (supplementary figure 2(f)) were used to represent land use change factors (${X_{{{\operatorname{LC} }_k}}},\,k = 1\,\operatorname{or} \,\;2$) in the attribution analysis. A map of their changes from 2015 to 2060 was shown in supplementary figure 3.

Major contributors to carbon stock changes in China under the SSP126 scenario were identified using equation (4) (He et al 2019) to calculate the contribution (${C_i}$) of precipitation (${X_{\operatorname{PRE} }}$), temperature (${X_{\operatorname{TEM} }}$), CO₂ (${X_{{{\operatorname{CO} }_2}}}$), and land use change factors (${X_{{{\operatorname{LC} }_k}}}$) on the trend of carbon stocks (Y), expressed as the rate of the product of the standardized environmental factors (${X_i}$) and its regression coefficient (${b_i}$) in the multiple linear regression equation (5):

$$\begin{align}{C_i} & = \frac{{\operatorname{d} \left( {{b_i} \times {X_i}} \right)}}{{\operatorname{d}t}},\,i = \operatorname{PRE} ,{\text{TEM}},\,{\text{CO}_2},\,{\text{LC}_k},\,\nonumber\\ k & = 1\,\operatorname{or} \;\,2\end{align} \tag{ 4 }$$

where t is the study period, i.e. 2015–2060, d/dt is the time derivative, ${X_i}$ are the annual standardized values of environmental factors, i.e. precipitation (${X_{\operatorname{PRE} }}$), air temperature (${X_{\operatorname{TEM} }}$), atmospheric CO₂ (${X_{{{\operatorname{CO} }_2}}})$, and land use change factors (${X_{{{\operatorname{LC} }_k}}}$) in China or the four climatic regions during 2015–2060 on average. The land use change factors represent the percentage coverages (%) of different land use types in different regions: forest and bare land in China and the temperate monsoonal and high-cold Tibetan Plateau climatic regions, forest in the subtropical-tropical monsoonal climatic region, and bare land in the temperate continental climatic region (supplementary figure 2(f)). Other environmental factors have historical simulations consistent with observations (supplementary figure 4), and their annual changes during 2015–2060 were shown in supplementary figure 5. b_i represents the response coefficient of carbon stocks to each factor, which was calculated using equation (5):

$$\begin{align}Y & = {b_0} + \sum {b_i}{X_i} + \varepsilon ,{ }i = \operatorname{PRE} ,{\text{TEM}},{\text{CO}_2},{\text{LC}_k},\nonumber\\ \,k & = 1\,\operatorname{or} \,\;2\end{align} \tag{ 5 }$$

where b₀ is a constant and is the residual. Y represents the carbon stocks. ${b_{\operatorname{PRE} }}$, $\,{b_{\operatorname{TEM} }}$, $\,{b_{\operatorname{CO} 2}}$, and ${b_{{{\operatorname{LC} }_k}}}$ are the responses of carbon stocks to standardized annual precipitation, annual average temperature, atmospheric CO₂, and land use change factors, respectively, during the study period.

2.2.4. Uncertainty between simulations of ESMs

All calculations were performed not only in the ensemble average of multiple ESMs simulations, but also in each ESM simulation. Furthermore, we used the standard error (SE; equation (6); McHugh 2008) of the statistical results of the nine ESMs to characterize the uncertainty between multi-model simulations:

$$\begin{equation}{\text{SE}}\frac{{{\text{SD}}}}{{\sqrt n }}\end{equation} \tag{ 6 }$$

where SD and n are the standard deviation and count of a set of numbers, respectively.

The vegetation carbon stock was lower than soil carbon stock in China for all pixels (Ratio_stock < 1; figures 1(a)–(c)), and the average vegetation carbon stock was 29 ± 1 PgC from 2015 to 2060, accounting for approximately 26 ± 6% of the soil carbon stock (113 ± 23 PgC; figure 1(f)). Initially, the Ratio_stock decreased rapidly and then increased slightly with increasing latitude, whereas it first increased and then decreased with increasing longitude (figures 1(d) and (e)). The Ratio_stock values above 0.5 were mainly observed at 21–29° N (figure 1(d)).

Zoom In Zoom Out Reset image size

The highest Ratio_stock were observed in the subtropical-tropical monsoonal (0.55 ± 0.12) and temperate monsoonal (0.21 ± 0.06) climatic regions, with high vegetation carbon stocks of 17 ± 1 PgC and 8 ± 1 PgC, respectively (figure 1(f)). However, in the high-cold Tibetan Plateau and temperate monsoonal climatic regions, carbon was mostly stored in the soil, and the ratio of vegetation to soil carbon was 0.11 ± 0.02 and 0.08 ± 0.02, respectively (figure 1(f)).

Vegetation and soil carbon in China showed increasing trends from 2015 to 2060, with a total carbon sequestration of 10 ± 1 PgC (figure 2(a)). Specifically, the carbon sequestration rate of vegetation (0.19 ± 0.01 PgC yr⁻¹) was approximately 4.31 times greater than that of soil (0.04 ± 0.01 PgC yr⁻¹; figure 2(e)). The temperate monsoon climatic region's Ratio_sink (9 ± 1) was the highest (figures 2(d) and (e)) among the climatic regions because the growth rate of vegetation carbon was relatively high (0.067 ± 0.006 PgC yr⁻¹), while that of soil carbon was low (0.007 ± 0.003 PgC yr⁻¹) (figures 2(b), (c) and (e)). The greatest increase in vegetation carbon stocks was observed in the subtropical-tropical monsoon climatic region at 0.097 ± 0.008 PgC yr⁻¹, which was approximately 5.52 ± 2.05 times the soil carbon stock (figure 2(e)). Conversely, Ratio_sink was low in the high-cold Tibetan Plateau (1.66 ± 0.38) and temperate continental (1.04 ± 0.77) climatic regions, which presented relatively high soil carbon sequestration potential (figures 2(b), (c) and (e)).

Zoom In Zoom Out Reset image size

This study also explored the impacts of environmental factors on the dynamics of vegetation and soil carbon stocks in China. Vegetation and soil carbon responded positively to temperature and precipitation (figures 3(a) and (b)), while their contributions to stocks were small owing to the low response coefficients (figures 3(a)–(d)). Compared with temperature and precipitation, major contributions to the increased vegetation and soil carbon stocks of China were observed from increased forest coverage during the study period, with a contribution of up to 81 ± 0.4% and 82 ± 3% in vegetation and soil, respectively (figures 3(c) and (d)). Carbon stocks were highly correlated with the normalized forest coverage (figures 3(a) and (b)), and with increasing forest coverage, more carbon was stored in vegetation than in soil (figure 2(e)) due to a higher response coefficient of vegetation carbon (2.10 ± 0.28) than that of soil carbon (0.49 ± 0.15) (figures 3(a) and (b)). Moreover, the substantial increase in forest coverage (supplementary figure 2) promoted an expansion in vegetation carbon stocks in the subtropical-tropical monsoonal and temperate monsoonal climatic regions from 2015 to 2060 (figures 2(e) and 3(c), (d)).

Zoom In Zoom Out Reset image size

Rising CO₂ concentrations were a secondary factor in the increase in China's vegetation and soil carbon stocks, contributing 65 ± 1% and 60 ± 6%, respectively; presenting a similar response in the high-cold Tibetan Plateau climatic region, but playing a role as important as the forest in two monsoonal climatic regions (figures 3(c) and (d)).

The increase in China's vegetation and soil carbon was impacted by bare land expansion (response coefficients: −1.43 ± 0.53 and −0.30 ± 0.17; figures 3(a) and (b)) that encroached on grasslands in the temperate continental, high-cold Tibetan Plateau, and temperate monsoonal climatic regions, contributing −55 ± 4% and −50 ± 12% to the increase in China's vegetation and soil carbon, respectively (figures 3(c) and (d)).

The ensemble average of the CMIP6 simulations revealed that China's vegetation and soil carbon stocks will increase from 2015 to 2060 under the SSP126 scenario, and the magnitude of vegetation and soil carbon sinks were approximately consistent among most ESMs of CMIP6 (supplementary table 3). Furthermore, the ensemble average of the vegetation carbon sink of the nine models in the 1990s (0.049 ± 0.009; table 2) was similar to that of the process-based model of Cao et al (2003) but lower than that of the Ground survey during the same period (Fang et al 2007, 2018). The soil carbon sink (table 2) was close to that of the inventory-satellite-based estimation and process-based models of Piao et al (2009) in the same period. Moreover, the ratios (1.258 ± 0.585–1.443 ± 0.459) of vegetation to soil carbon sinks in China of CMIP6 during 1990–2010 were consistent with most studies (table 2, Piao et al 2009, Jiang et al 2016, Fang et al 2018). Furthermore, the CMIP6 simulations revealed that the Ratio_sink will increase to 2.344 ± 2.227 in 2015–2020 and 4.776 ± 2.300 in 2055–2060, indicating an enhancement in vegetation carbon sinks, with a magnitude of approximately 0.209 ± 0.031 PgC yr⁻¹ by 2060 (table 2). Additionally, the CMIP6-based vegetation and soil carbon sinks in China during 2015–2060 were lower than those reported by Huang et al (2022) and Cai et al (2022), which may be caused by the following reasons. On the one hand, several parameter-sparse empirical models do not take into account nutrient limitations and carbon losses caused by climate change, especially extreme climates, compared with ESMs of CMIP6. On the other hand, carbon sinks are generally lower under the SSP126 scenario with low forcing used in this study than under the SSP245 and SSP585 scenarios with high forcing (Cai et al 2022, Huang et al 2022).

However, China's vegetation carbon stock in CMIP6 was higher than that in previous studies (table 3), especially in the ACCESS-ESM1-5 model (supplementary table 3), with magnitudes of 25 ± 2 PgC during 1980s, 24 ± 2 PgC during 1990s, and 25 ± 2 PgC during 2000s (table 3), whereas that in the ground survey and process-based model was only 13–15 PgC in previous decades (table 3, Li et al 2004, Xu et al 2018). Compared with vegetation, the soil carbon estimation from CMIP6 (around 111–112 PgC) during 1980–2010 was relatively reasonable, although slightly higher than that in previous studies (75–90 PgC, Xie et al 2007, Tang et al 2018), because the soil depth of ESMs in CMIP6 was greater than the 1 m depth commonly used in these studies (table 1). However, the uncertainty of soil carbon size among ESMs was large, for example, the China's soil carbon in CMCC-CM2-SR5 or CMCC-ESM2 model was seven times higher than that in the IPSL-CM6A-LR model (supplementary table 3). Conclusively, vegetation carbon sinks of the CMIP6 ESMs were higher than soil carbon sinks, and the ratio of vegetation to soil carbon sinks will continue to increase in the future, although estimates of vegetation carbon stocks were higher than previous studies over the same period. Therefore, the prior calibration of the carbon pool size using the uniformly standardized vegetation and soil carbon data from ground surveys is necessary before biogeochemical model simulation in China. Initial calibration can also improve the representation of the processes related to vegetation and soil carbon changes, such as turnover loss of carbon pool (Koven et al 2013, Wu et al 2019, Boucher et al 2020, Ziehn et al 2020).

Increases in China's vegetation and soil carbon are mainly affected by the combined effects of increased forest coverage and CO₂ concentrations from 2015 to 2060 under the SSP126 scenario, with regional differences. Increased forest coverage will promote a greater increase in vegetation carbon stocks than that soil carbon stocks in the subtropical-tropical and temperate monsoonal climatic regions because of the difference in response coefficients to an increase in forest coverage. Similarly, a previous study reported that ecological restoration projects such as afforestation contributed more than half of the contribution to the enhancement of China's carbon stocks in project areas (Lu et al 2018). Moreover, the two monsoonal climatic regions are appropriate for potential afforestation (Zhang et al 2021). CO₂ had an effect equivalent to forest coverage in two monsoonal climatic regions but has a weaker contribution in the high-cold Tibetan Plateau with a high soil carbon sequestration potential as indicated by the relatively low Ratio_sink. However, grassland encroachment by bare land will weaken the increases in vegetation and soil carbon stocks in the high-cold Tibetan Plateau and temperate monsoonal climatic regions, which has experienced significant grassland degradation over the past few decades (Bardgett et al 2021). These results imply that afforestation in monsoon regions and grassland protection in the high-cold Tibetan Plateau and temperate monsoonal climatic regions need to be strengthened to increase carbon sinks in the future.

Compared with the SSP245 (23%) and SSP585 (6%) scenarios, the contribution of vegetation and soil to the absorption of anthropogenic CO₂ emissions (0.61 PgC yr⁻¹; supplementary figure 12) was the greatest in 2060 under the SSP126 scenario (53%; figure 4(a)), and the total carbon sequestration rate of vegetation and soil was 0.33 ± 0.07 PgC yr⁻¹.

Zoom In Zoom Out Reset image size

China's actual forest coverage was 23.2% in 2020 (Central People's Government of the People's Republic of China 2021a), and the Chinese government has set a phased target to increase forest coverage (24.1% in 2025; 25% in 2030; and 26% in 2035) (National Development and orm Commission of the People's Republic of China 2020, Central People's Government of the People's Republic of China 2021a, 2021b). Assuming that the response coefficients of vegetation and soil carbon stocks to individual factors remain unchanged (or changed within ±15%), the total carbon sink potential in China will reach 0.42 PgC yr⁻¹ (0.37–0.48 PgC yr⁻¹) in 2060 under the planned policies (figures 4(b) and (c)), which can absorb approximately 69% (60%–79%) of the projected anthropogenic CO₂ emissions from China in 2060. Note that we only provided a potential vegetation and soil carbon sink in the future using China's average forest coverage of the planned policies. However, the unknown spatial location of these afforestation sites may introduce uncertainty into the predicted carbon sinks. Based on the increased forest map of LUH2 (supplementary figure 3(a)), modeling carbon sinks in different spatial distribution schemes of afforestation is needed in future work, which may bring additional carbon sink increments over the SSP126 scenario.

Therefore, China's ecosystem will generate a major carbon sink under China's planned policies. This can contribute to the achievement of China's carbon neutrality goal by 2060. Additionally, it provides strong support and broad implications for carbon sink improvement and net-zero emission achievement globally.

There were large differences in China's soil carbon between the nine ESMs analyzed in this study, with a maximum (227 PgC) in the CMCC-ESM2 model and a minimum (32 PgC) in the IPSL-CM6A-LR models, respectively (supplementary table 3), both of which deviated greatly from the results of the Ground survey (table 3). Although the magnitudes of China's vegetation carbon were similar among most ESMs (BCC-CSM2-MR, CanESM5, CESM2-WACCM, CMCC-CM2-SR5, CMCC-ESM2, NorESM2-LM, MPI-ESM1-2-LR), which was much higher than the results of the ground survey (table 3). Fortunately, the uncertainty of vegetation carbon sinks among the nine ESMs was small, and soil carbon sinks were similar among most ESMs (supplementary table 3). Moreover, changes in vegetation and soil carbon stocks were positive for all ESMs (supplementary table 3), which improved the confidence of our results. Nonetheless, the effects of the high uncertainty in the magnitude of soil carbon between the models and the relatively high vegetation carbon on the carbon cycle are unclear and need to be assessed in future work.

In this study, we used multiple linear regression to analyze the responses of vegetation and soil carbon to precipitation, air temperature, and CO₂, as well as land use change, but did not consider the nonlinear effects of multiple factors. Actually, the interaction between these factors has an impact on carbon dynamics (Garten et al 2009, Wolf et al 2011, Sui et al 2013). Although we have modeled differential responses across four climatic regions, this rough consideration of interactions between climatic conditions and environmental factors may not be sufficient. An in-depth analysis of the impact of nonlinear relationships between different factors on the carbon cycle is required in the future.