Introduction

The global carbon cycle is greatly affected by climate warming, with varying effects observed across climate zones and ecosystems1. For alpine ecosystems, climate warming not only promotes vegetation growth to take up more carbon dioxide (CO2) but also stimulates permafrost thawing to emit more CO2, which probably weakens the regional carbon sink and converts it into a carbon source2,3. Ecological restoration projects (ERPs) defined as the process of assisting the recovery of an ecosystem that has been degraded, damaged or destroyed4. ERPs are the most important component of nature-based climate solutions and have been proven to be an efficient and cost-effective means to enhance the carbon sinks of terrestrial ecosystems5,6,7,8. However, whether the implementation of ERPs can offset the negative impact of permafrost thawing on carbon sinks remains unknown.

Both the vegetation and soil of alpine ecosystems are affected by climate change. Temperature and precipitation are the main cause that constrain vegetation growth to uptake carbon. Rising temperature can prolong the vegetation growth time and increase the peak photosynthesis rate to uptake more CO2 in the atmosphere9,10. Precipitation is the main cause affecting the rate at which vegetation sequesters carbon through photosynthesis in drylands11,12,13. Climate warming probably aggravates the water shortages in drylands, water shortages would shorten vegetation growth periods and reduce carbon uptake14,15. Previous analyses have found that recent climate change has accelerated vegetation growth16,17 to obtain greater carbon sink in Alaska, Canada and the Tibetan Plateau18,19. Moreover, recent advances concluded that future climate change would stimulate vegetation growth to further promote carbon assimilation in these areas20,21. In addition, rising surface temperature may accelerate permafrost thawing to increase carbon emissions22. It has been reported that the amount of carbon stored in permafrost regions is twice that in the atmosphere23,24. Even a small portion of this frozen carbon pool is released into the atmosphere in the form of CO2 and CH4, which would likely significantly increase the magnitude of future climate change3,25. It is estimated that approximately 120 ± 85 Gt of carbon emissions will be generated from thawing permafrost by 2100, which is equivalent to 5.7 ± 4.0% of the total anthropogenic emissions for the Intergovernmental Panel on Climate Change (IPCC) representative concentration pathway (RCP) 8.5 scenario and would increase global temperatures by 0.29 ± 0.21 °C or 7.8 ± 5.7%26.

ERPs play a crucial role in mitigating global warming as part of nature-based climate solutions27. Prior studies pointed out that global forest restoration may sequester more than 200 GtC, which is approximately the total amount of CO2 that has been emitted into the atmosphere globally in the past 20 years7. Furthermore, ERPs are more cost-effective than other options28,29 (e.g., carbon capture, utilization and storage technologies) and can enhance ecosystem services6. Meanwhile, the enhancement of ERPs in increasing carbon sinks is also affected by climate change. Researches have shown that the increasing temperature and rising CO2 have amplified China’s carbon sink contributed by forest restoration and grassland restoration by 28.94% and 54.75%, respectively30. Additionally, drought aggravated by climate change also weakens the effectiveness of ERPs in increasing carbon sinks6,31. Furthermore, climate change will also cause the areas suitable for restoration to increase or decrease7.

Exploring the effectiveness of EPRs in facing the negative impacts of climate change on alpine ecosystems is essential to enhancing the understanding of nature-based climate solutions in achieving carbon neutrality. The Qinghai‒Tibet Plateau (QTP) is a crucial component of the global alpine ecosystem, characterized by widely distributed permafrost32,33, and situated in southwestern China (Fig. 1a). Many ERPs (including several key national ERPs, such as Natural Forest Protection Project, Grain for Green Program, Returning Grazing Land to Grassland Project, Yangtze River Shelter Forest Project) have been implemented on the QTP34,35. Forest restoration and grassland restoration are the two major approaches included in ERPs implemented on the QTP36,37. These ERPs have shown marked effectiveness in expanding forest and grassland area (Fig. 1b) and enhancing carbon sinks35,38,39,40. Previous studies already concluded that a warmer (Fig. 1c) and wetter (Fig. 1d) climate leads to an increase in vegetation growth and carbon absorption on the QTP41,42, and this trend will continue in the future43,44. Climate change will also accelerate permafrost thawing, leading to a significant increase in carbon emissions33,45. It was suggested in many previous studies35,46 that implementation of ERPs was effective in maintaining and enhancing regional carbon sink, but the effect of future climate change was seldom taken into consideration. Therefore, it remains unclear whether ERPs are still efficient in increasing regional carbon sink under future climate change, especially when facing carbon loss from climate change-induced permafrost thawing. Thus, we chose the QTP and utilized the modified Biome-BGCMuSo, which included the effect of permafrost thawing on the carbon cycle, to address this issue. The goals of this study were to (a) clarify the influence of ERPs on the carbon budget in the QTP and how the ERPs interacted with future climate change during 2020–2060; (b) examine the effectiveness of ERPs in combating permafrost thawing-induced carbon loss.

Fig. 1: Spatial pattern of land use/cover and ecological restoration projects, temporal changes in climate.
figure 1

a Spatial pattern of land use/cover (LUC) in 2020 on the Qinghai-Tibetan Plateau (QTP); b spatial pattern of implemented ecological restoration projects (ERPs) on the QTP; c spatial pattern of permafrost and seasonally frozen ground (obtained from Wang et al.33); d spatial pattern of DEM; e mean annual temperature change and f total annual precipitation change from 2020 to 2060 under the SSP245 and SSP585 scenarios; shading represent the standard error associated with those estimates.

Full size image

Results

The effect of climate change on carbon budget of QTP ecosystems

Compared with the reference scenario, future climate change (based on Shared Socioeconomic Pathway (SSP) scenarios) will bring down the regional cumulative net ecosystem productivity (NEP) by 38.25% (495.25 ± 124.60 Tg C) and 12.05% (156.25 ± 253.40 TgC) in 2020–2060 under SSP-245 and SSP-585, respectively (Fig. 2a, Supplementary Fig. 1). Although net primary productivity (NPP) is expected to increase (Supplementary Fig. 2), heterotrophic respiration (RH) will increase faster in 2020–2060 (Supplementary Fig. 3), which, in turn, leads to a decline in cumulative NEP. The effect of climate change on NEP shows obvious spatial heterogeneity, i.e., it causes a decline in NEP in western QTP but an increase in NEP in eastern QTP (Fig. 3). The decrease of NEP will occur in approximately 46.11% and 42.30% of the total area, while the increase of NEP will occur in approximately 23.96% and 28.16% of the total area under SSP-245 and SSP-585, respectively (Fig. 3).

Fig. 2: The effect of different factors on carbon sink.
figure 2

a The impact of CNEP changes from 2020 to 2060 owing to different factors compared with the Reference scenario. TP Temperature + Precipitation, CO2 Rising CO2; b The impact of different factors on carbon budget. Veg_Soil: the effect temperature and precipitation change without considering permafrost thawing; Permafrost: the effect of permafrost thawing; ERPs_TP: the interaction between ERPs and TP; ERPs_CO2: the interaction between ERPs and rising CO2; ERPs-E_TP: the interaction between expanding ERPs and TP; ERPs-E_CO2: the interaction between expanding ERPs and rising CO2; both shading and error bars represent the standard error associated with those estimates.

Full size image
Fig. 3: Spatial pattern of different factors on carbon sink.
figure 3

Spatial pattern of (a) ERPs induced CNEP change (ΔCNEP) compared with Reference scenario; spatial pattern of future climate change causing ΔCNEP compared with Reference scenario under (b) SSP-245 and (c) SSP-585; spatial pattern of ERPs with future climate change causing ΔCNEP compared with Reference scenario under (d) SSP-245 and (e) SSP-585; spatial pattern of ERPs-E with future climate change causing ΔCNEP compared with Reference scenario under (f) SSP-245 and (g) SSP-585; Area ratio of (h) ERPs, future climate change under (i) SSP-245 and (j) SSP-585, ERPs with future climate change under (k) SSP-245 and (l) SSP-585, ERPs-E with future climate change under (m) SSP-245 and (n) SSP-585 induced ΔCNEP compared with Reference scenario.

Full size image

Future temperature + precipitation (TP) changes will affect the carbon cycle in vegetation and soil, causing the NEP to decrease by 369.36 ± 63.33 Tg C and 393.99 ± 89.34 Tg C in 2020–2060 under SSP-245 and SSP-585, respectively (Fig. 2b). Moreover, TP changes will also stimulate permafrost thawing, warming-induced permafrost thawing will cause the NEP to decrease by 530.22 ± 173.76 Tg C and 582.94 ± 168.40 Tg C in 2020–2060 under SSP-245 and SSP-585, respectively (Fig. 2b). Overall, TP changes will cause NEP to decrease by 69.38% (899.58 ± 179.92 Tg C) and 75.35% (976.93 ± 195.39 TgC) in 2020–2060 under SSP-245 and SSP-585, respectively (Fig. 2b). The influence of TP changes and rising CO2 on NEP change is opposite. Rising CO2 will lead NEP to increase by 31.13% (403.63 ± 80.73 Tg C, SSP-245) and 63.30% (820.67 ± 164.13 TgC, SSP-585) in 2020–2060, respectively (Fig. 2b). The effects of rising CO2 on increasing NEP almost totally offset by TP changes (Fig. 2b). Moreover, the effects of TP changes and rising CO2 also vary in different areas (Supplementary Fig. 4). The decrease in NEP caused by TP changes mostly occurred in the southwestern region (Supplementary Fig. 4, Supplementary Fig. 5), while the increasing NEP contributed by rising CO2 mostly occurred in the eastern region (Supplementary Fig. 4).

The effect of ERPs on the carbon budget

ERPs will enhance carbon sink in the QTP during 2020–2060. Future climate change will increase regional peak photosynthesis rate, extend growth period, and subsequently further amplify the sink enhancement of ERPs (Fig. 2). In addition, future climate change can also change the areas suitable for forest restoration and grassland restoration (Supplementary Fig. 6), which will also affect the contribution of ERPs in increasing carbon sink (Fig. 2).

It was predicted that ERPs could make a contribution of about 372.96 ± 81.41 Tg C (approximately 94.31% from forest restoration) during 2020–2060 without considering the amplifying effect of future climate change (Fig. 2b, Supplementary Fig. 1). If the amplifying effect of future climate change on ERPs was considered, the carbon sink contributed by ERPs is expected to increase by 703.47 ± 128.16 Tg C (for SSP-245, reaching 1076.44 ± 215.29 Tg C) and 772.40 ± 113.19 Tg C (for SSP-585, reaching 1145.37 ± 229.07 Tg C), respectively. The effort of ERPs totally offset the permafrost thawing-caused carbon sink decline and led the regional carbon sink to increase by approximately 44.77% (580.49 ± 315.48 Tg C) and 76.29% (989.11 ± 260.59 Tg C) during 2020–2060 under SSP-245 and SSP-585, respectively (Fig. 2a, Supplementary Fig. 1). The influence of different climate change factors on NEP were varied. The warmer and wetter climate increased the ERPs-induced cumulative NEP change by 66.11 ± 13.22 Tg C and 281.37 ± 56.27 Tg C during 2020–2060 under SSP-245 and SSP-585, respectively (ERPs_TP) (Fig. 2b), while rising CO2 will enhance ERPs-induced cumulative NEP change by 637.37 ± 127.47 Tg C and 491.03 ± 98.21 Tg C in 2020–2060, respectively (ERPs_CO2) (Fig. 2b).

A warmer and wetter climate will enlarge the suitable area for forest restoration by 7.45 Mha (SSP-245) and 11.78 Mha (SSP-585), while the suitable area for grassland restoration will decrease by 1.75 Mha (SSP-245) and 2.87 Mha (SSP-585) in the 2060 s (Supplementary Fig. 6-7). The decrease in the suitable area for grassland restoration was mostly due to the increase in the areas used for forest restoration as a priority. If these changes are included in future restoration plans, the carbon sinks contributed by ERPs will further increase by 1187.66 ± 183.97 Tg C (for SSP-245, reaching 1560.62 ± 237.53 Tg C) and 1596.21 ± 231.41 Tg C (for SSP-585, reaching 1969.17 ± 319.24 Tg C) (Fig. 2b). With the implementation of the ERPs, the regional total carbon sink will further increase by approximately 82.11% (SSP-245) to 139.82% (SSP-585) (Fig. 2a, Supplementary Fig. 1). Among them, the warmer and wetter climate would increase the ERPs-induced carbon sink increment by 328.43 ± 65.69 Tg C (SPP-245) and 618.01 ± 123.60 Tg C (SPP-585) in 2020-2060, respectively (ERPs_E_TP) (Fig. 2b). While rising CO2 would increase the ERPs-induced carbon sink increment by 859.23 ± 171.85 Tg C (SPP-245) and 978.20 ± 195.64 Tg C (SPP-585) in 2020–2060, respectively (ERPs_E_CO2) (Fig. 2b).

Discussion

The effect of climate change on the ecosystem carbon budget on the Qinghai‒Tibet Plateau

When the permafrost thawing is not considered (TPC2), future climate change will promote carbon sinks to increase in the QTP (Fig. 2). However, spatial heterogeneity of climate change will lead to various effects on carbon sink in different subregions (Fig. 4, Supplementary Figs. 4, 5, and 8). Climate change will cause the eastern QTP to become warmer and wetter to enlarge the ecosystem carbon sink, while make the western QTP to get warmer and drier to diminish the CO2 uptake (Supplementary Figs. 4, 5, and 8). Recent advance has indicated that the effect of increase in precipitation is greater than the rise in temperature in determining the vegetation growth to sequester CO2 in the QTP. The increase in precipitation can promote soil moisture to significantly advance spring green-up date and delay their autumn senescence dates47.

Fig. 4: Carbon sink change caused by different factors along the aridity index gradients.
figure 4

CNEP change caused by (a) TP change and (b) TP + CO2 change along the aridity index (AI) gradients; ERPs with (c) TP change and (d) TP + CO2 change along the AI gradients; ERPs-E with (e) TP change and (f) TP + CO2 change along the AI gradients. error bars represent the standard error associated with those estimates.

Full size image

Fortunately, positive effects from the warming and humidification climate increased carbon sink in the eastern QTP succeed offset the warming and drying climate caused carbon loss in the western QTP, which leads to the increase of net carbon sink finally (Fig. 2b). While the effect of permafrost thawing was considered, positive effect from climate change on vegetation was totally offset (Fig. 2b). Conversely, some studies claimed that the increase in vegetation carbon uptake will surpass permafrost thawing, thus inducing an increase in carbon sink48,49. It is worth noting that these studies did not consider the effect of deep permafrost thawing on carbon emissions. Otherwise, the carbon sink will weaken and may be converted into a carbon source in the future45,50.

Although uncertainties exist in these studies, it is certain that permafrost thawing will not only accelerate the emission of CO2 but also stimulate the emission of CH4 and N2O, which holds greater climate warming effects51,52. Furthermore, prior studies also reported that future climate change will accelerate greenhouse gas emissions in thermokarst lakes53 and permafrost collapse54,55,56. In addition, the priming effect that exists in permafrost thawing will further accelerate carbon emissions57. Thus, the carbon emissions caused by permafrost thawing are probably greatly underestimated. Recent advancements have also suggested that the effect of permafrost thawing in stimulating carbon emissions deeply exceeds the accelerated vegetation growth to increase carbon sinks, which challenges the realization of global climate mitigation25,58,59.

The effect of ecological restoration on the carbon budget modulated by climate change

The implementation of ERPs has great potential in strengthening carbon sinks in the future (Fig. 2a), and forest restoration can obtain a greater carbon sink than grassland restoration (Supplementary Fig. 9). The effect of ERPs in carbon sink is also varied in different subregions, carbon sink contributed by ERPs mainly occurred in the eastern part of the QTP (Fig. 3a). The western QTP is much colder and drier than the eastern QTP, which would constrain the implementation of vegetation restoration actions largely60,61. Meanwhile, future climate change can amplify the carbon sink contributed by ERPs, and made it strong enough to totally offset permafrost thawing-induced carbon loss (Fig. 2). Climate change will extend the period for photosynthesis by altering phenology in newly planted vegetation62. It has already been proven that the start of the growing season (SOS) advanced, on average, by 0.28 d/y, while the end of the growing season (EOS) was delayed by an estimated 0.33 d/y during 1982–2014 on the QTP63. Furthermore, rising temperatures will increase the carbon sink by promoting the peak carbon uptake rate because the current temperature in the QTP is slower than the optimal temperature for photosynthesis10,64. The QTP is expected to be warmer and wetter in the future (Fig. 1c, d), so the carbon sink contributed by ERPs will be amplified under this climate change background (Fig. 4). Rising CO2 will also stimulate newly planted vegetation to absorb more carbon30,65. Compared with TP changes, rapidly rising CO2 (Supplementary Fig. 10) has much greater amplifying effect on ERPs-induced carbon sink increment (Fig. 5).

Fig. 5: Approaches to quantify the effect of climate change and ERPs on carbon budget in the QTP.
figure 5

The light green, red, blue and dark green solid lines represent the effect of ERPs, TP change, rising CO2 and climate-enlarged ERPs on carbon sink; the dark green dashed line represents the effect of TP change on suitable areas for forest restoration and grassland restoration. The light green, red, blue and dark green color numbers represent the effect size of ERPs, TP change, rising CO2 and climate-enlarged ERPs on carbon sink under SSP-245 and SSP-585, respectively. The black number represent the area for forest restoration or grassland restoration. * represents the areas that suitable for grassland restoration change to forest restoration owing to higher priority; # represents the carbon sink contributed by increased suitable area for restoration bring by the TP change.

Full size image

The impact of climate change on ERPs-induced carbon sink will also vary in different regions (Fig. 3). It is expected that the carbon sink contributed by ERPs will increase in the eastern region but decrease in the western region under climate change (Fig. 3), probably because the climate is warmer and wetter in the eastern part, but warmer and drier in the western part (Supplementary Fig. 8). Warmer and wetter climates will stimulate the growth of these newly planted trees and grass to enlarge the contribution of ERPs increasing carbon sink, while warmer and drier climates will constrain the growth of these newly planted trees and grass reduce the contribution of carbon sink contributed by ERPs66,67. Although rising temperature and rising CO2 would enhance photosynthesis, the limitation of water availability may constrain the promotion of photosynthetic capacity68,69. Recent study also concluded that moisture was the more important factor constraining the vegetation’s carbon sink than temperature in the QTP70. Furthermore, rising CO2 plays a greater role in enhancing carbon sinks in the areas with warmer and wetter climate backgrounds than in the areas with colder and drier climate backgrounds (Supplementary Figs. 4 and 11).

Temperature and moisture are the major factors that constrain the distribution and growth of trees71,72. It has been reported that warmer and wetter climate-induced tree lines have continued to climb during the past decades in the QTP73,74. According to our results, with a warmer and wetter climate, the areas suitable for restoration will be enlarged (Supplementary Fig. 12), and the carbon sink from ERPs will be greatly improved when the restoration is implemented in these areas (Fig. 2b). Recent advances have predicted that future climate change will decrease the global areas suitable for forest restoration, the area suitable for forest restoration will still increase at high latitudes7. Although the QTP is in a low latitude area, the average altitude is greater than 4000 m (Fig. 1f). The increase in altitude, similar to the increase in latitude, will cause the decrease in temperature. Climate warming will increase the area suitable for vegetation expansion occurred not only at high latitudes but also at high altitudes74,75,76. Thus, we should take this opportunity to gain a greater carbon sink through ERPs. However, uncertainties still exist in the area of vegetation restoration and the corresponding rate on the QTP. For instance, there are large differences among future climate change prediction data from multiple Earth System models used in CMIP6, the reliability of this data was widely discussed in prior studies77,78. Meanwhile, owing to the restriction of terrain, climate and soil, the survival rate and growth rate of restored vegetation in the QTP was much lower than that in the eastern plain79,80,81. Thus, how reliable the increase in areas suitable for restoration that we predict owing to warmer and wetter climates in the future should be treated with caution. Additionally, it should be noted that although the carbon sink contributed by ERPs will be able to reverse the permafrost thawing-caused carbon loss in the whole QTP (Fig. 5), the carbon sink contributed by grassland restoration will offset by permafrost thawing (Supplementary Fig. 9). The main reason for this situation is that the areas where grassland restoration was implemented fall in the main regions with permafrost thawing.

Implications

The QTP has already reached carbon neutrality and made a carbon profit of approximately 15.20 ± 5.09 Tg C/yr48,82. However, future climate change-caused carbon loss will probably largely offset this carbon profit and threaten the maintenance of carbon neutrality. It was reported that most of the carbon released from permafrost will be in the form of CO2, with only approximately 2.7% in the form of CH483. CH4 has a higher global warming potential, and almost half of the effect of future global permafrost-zone carbon emissions on climate forcing is likely to be from CH483. Likewise, if the CH4 emission-caused warming potential is the same as that of the CO2 caused by permafrost thawing in the QTP, permafrost thawing-caused greenhouse gas emissions will be largely increased compared to our results. In this case, the implementation of ERPs will not be sufficient to offset the permafrost thawing-caused carbon loss. Thus, consistent implementation of ERPs is indispensable to reduce or even avoid this negative effect as much as possible. Considering the great pressure to achieve carbon neutrality in China by 2060, the implementation of ERPs to promote carbon sinks in the QTP is not only helpful for maintaining carbon neutrality in the QTP but also beneficial for the realization of China’s carbon neutrality. In addition, the fact that carbon sinks contributed by ERPs will be amplified by climate change also suggests that we should take this chance to gain a greater carbon sink (Fig. 5). Although the implementation of ERPs probably increases carbon emissions by digging soil, some researchers have found that the carbon leakage caused by this disturbance is limited compared to the carbon sink contributed by ERPs themselves84,85. Thus, it is necessary to implement ERPs to increase carbon sinks to offset the negative impact of permafrost thawing in the future.

Prior studies demonstrated that restoration in alpine ecosystems may reduce albedo to increase surface temperature, which offsets its efforts by increasing ecosystem carbon sinks to mitigate climate change86. It should be noted that vegetation in alpine ecosystems reduce albedo was mainly occurred in winter. Increases in vegetation cover and height generally mediate the effect of increasing summer air temperatures on soil temperatures to reduce permafrost thaw87. Given that soil respiration caused by the thawing of frozen soil is controlled by temperature, reducing the temperature in summer can reduce carbon emissions more than increasing carbon emissions by raising the temperature in winter88. Additionally, recent research also suggested that accelerated vegetation growth induces the increase in evapotranspiration to cool temperatures and offset the warming effect caused by reduced albedo in the QTP89,90. Thus, increase vegetation cover through natural based climate solution in the alpine areas is still efficient.

In addition to the QTP, permafrost is also widely distributed in eastern Russia, Alaska, northern Canada and Europe. Future climate change will also stimulate permafrost thawing in these areas to emit by approximately 100–500 Gt CO2-eq (including CO2 and CH4) until 210022,91. In contrast, recent advance found that future forest restoration potential of the permafrost region reaches 411.59 Mha, accounting for 43% of the global forest restoration potential7. If these areas are reforested, approximately 98.45 Pg C can be sequestrated92. Meanwhile, if other natural based climate solutions locally and amplification effect of climate change on vegetation growth are also considered, permafrost thawing induced carbon emissions are likely to be offset totally. Our findings provide new insights for these areas to counter future climate change through nature-based solutions. Moreover, it is necessary to strengthen the protection of the existing ecosystems to avoid greater carbon emissions from disturbance and damage caused by future climate change30. ERPs can also increase multiple ecosystem services (including water retention, soil retention and sandstorm prevention)6. However, ERPs may lead to the invasion of woody plants into grasslands, thereby affecting local biodiversity93. Thus, it is necessary to be cautious in controlling the scale of forest restoration. Furthermore, large economic costs are also the main factor limiting the scale of ERPs94,95.

Uncertainty analysis

Our estimated NEP is consistent with the inventory and Multi-Scale Synthesis and Terrestrial Model Intercomparison Project (MsTMIP)-based method but lower than the atmospheric inversion and eddy correlation-based methods (Supplementary Fig. 13). The NEP derived from atmospheric inversion is limited by the number and distribution of atmospheric CO2 observation sites and cannot accurately partition the carbon fluxes of different types of ecosystems, especially in the QTP96,97. Previous studies found that eddy correlation flux towers are mainly distributed in areas with minor human disturbances, making it difficult to take forest age differences and ecosystem heterogeneity into account, which further leads to biases when measured fluxes are scaled up to the regional scale96,98. Moreover, the potential source of biases at the regional scale also includes disturbances such as logging, fire and land cover change, the neglection of which could also lead to overestimates in the regional ecosystem carbon sink98,99. It has been proven that the global NEP estimated from the eddy covariance method is 8 times that of the global land carbon sink99. Additionally, although we did not consider the effect of grazing on carbon sinks in grasslands in this study, a previous study pointed out that the effect of grazing on carbon sinks in grasslands is very limited48. Thus, our results are still reliable. Additionally, other factors including terrain and soil can affect vegetation growth and ecosystem carbon sink. Prior study revealed that the vegetation growth was ruled by the orientation of the slopes on the QTP, vegetation growth better uptakes more CO2 on polar-facing slopes than equatorial-facing slopes within warm and dry areas, but uptakes less CO2 on polar-facing slopes than equatorial-facing slopes within cold and wet areas100. Similarly, in colder regions, earlier soil thaw onset generally facilitated spring phenology, and longer soil thaw durations tended to increase the growing season soil moisture content, which could in turn enhance vegetation growth to absorb more CO2101,102. But in relatively warm regions, earlier thaw onset and longer thaw duration could possibly exacerbate the growing season water stress and limit vegetation growth103. These factors probably affect the efforts of ERPs to increase carbon sinks and interact with future climate change. Thus, the influence of these factors needs to be discussed in future studies.

Methods

Methodology for carbon budget analysis

Estimation of the carbon budget through Biome-BGCMuSo (BBMS)

The BBMS is an ecosystem process-based model developed from the Biome-BGC model104,105. The BBMS simulates the storage and fluxes of water, carbon, and nitrogen between ecosystems and the atmosphere106. This model can simulate photosynthesis, respiration, allocation of organic matter, litter and decomposition of plant tissues, and circulation and migration of nutrients in different ecosystems107. Improvements in the BBMS compared to the Biome-BGC included the addition of new modules, such as multiple soil layers, optimized processes related to soil moisture and senescence, and management practices104. The net ecosystem productivity (NEP) was calculated based on Eq. (1):

$${NEP}={GPP}-{MR}-{GR}-{HR}$$
(1)

where \({GPP}\) denotes gross primary productivity, which is estimated independently for the sunlit and shaded canopy fractions by Farquhar’s photosynthesis algorithm108; \({MR}\) denotes the maintenance respiration from the leaves, stems and roots109; \({GR}\) denotes the growth respiration, which is estimated as the fraction of \({GPP}\) minus \({MR}\); and \({HR}\) denotes the heterotrophic respiration, which is deemed the sum respiration of different litter and soil carbon pools.

To describe the effect of soil freeze–thaw processes on the carbon cycle and reconcile the difference in projected active soil organic carbon (SOC) change derived from observed data (Supplementary Fig. 14) and the logic of soil carbon change in the BBMS, we established Eq. (1) to address this issue by referring to prior studies45,110.

$${{SOC}}_{{BBMS},i}={{SOC}}_{{BBMS},i-1}\times \left(1+\frac{{{SOC}}_{{RF},i}-{{SOC}}_{{RF},i-1}}{{{SOC}}_{{RF},i-1}}\right)$$
(2)

where \({{SOC}}_{{BBMS},i}\) and \({{SOC}}_{{BBMS},i-1}\) denote the SOC in the BBMS of year i and i-1, respectively; \({{SOC}}_{{RF},i}\) and \({{SOC}}_{{RF},i-1}\) denote the SOC of year i and i-1 based on the random forest method developed in a previous study110. The effect of future climate change-caused permafrost thawing on active layer thickness (Supplementary Fig. 15) and active soil organic carbon was based on the method derived from Wang et al.33.

Estimation of the effect of land use/cover change on NEP

To further improve the capacity of the BBMS to simulate the impact of land use/cover changes (LUCCs) on NEP, we constructed a simplified mechanism by summing the area fraction change of PFTs (plant functional types) in each 20 × 20 km2 grid cell to quantify the impact of LUCCs on NEP (Eq. (3)) with reference to previous studies6,30.

$${{NEP}}_{i}={f}_{F,i}\times {{NEP}}_{F,i}+{f}_{C,i}\times {{NEP}}_{C,i}+\cdots +{f}_{G,i}\times {{NEP}}_{G,i}$$
(3)

where \({{NEP}}_{i}\) denotes the NEP in grid cell i; \({f}_{F,i}\), \({f}_{C,i}\), and \({f}_{G,i}\) denote the area fractions of forest, cropland and grassland in grid cell i, respectively; and \({{NEP}}_{F,i}\), \({{NEP}}_{C,i}\), and \({{NEP}}_{G,i}\) denote the NEP of forest, cropland and grassland in grid cell i, respectively.

Estimation of the effect of ERPs and future climate change on the carbon budget

Simulation protocol

To separate the effects of climate change and ERPs under conditions with or without considering climate change on NEP, we constructed 4 scenario groups to address this issue. Explicit scenario setting rules and the meaning of these scenarios are described in Table 1.

Table 1 Summary of experimental simulation protocols
Full size table

Estimation of the ecological restoration potential

Future ERP practices in the QTP include forest restoration and grassland restoration, and they are affected by climate change. Explicit steps used to explore this issue are described in the following paragraphs:

To explore the suitable areas for forest restoration, first, we created squares with a radius of 2 km in nature reserves (regions with limited human activity) through random sampling (Supplementary Fig. 16). Then, we used ArcGIS 10.2 to extract tree cover and the relevant environmental factors (comprising soil, topographic and climate) (Supplementary Table 1) information in all squares to establish a random forest machine learning regression (RFMLR) model to predict the potential tree cover developed by Bastin et al. 7. Consistency in both training and validation was good, indicating the feasibility of our established RFMLR model (Supplementary Fig. 17). Next, we used the trained model to project the potential tree cover under current climate conditions (Supplementary Fig. 18). Finally, we converted the potential tree cover map into a forestland map (defined as >20% tree cover) and deducted areas with forest, cropland, wetland and impervious surfaces to identify the potential forest restoration extent. Considering that future climate change will affect areas that are suitable for forest restoration, we used predicted future climate data derived from CMIP6 (Supplementary Table 1) as climate input data to run the constructed RFMLR model to predict potential tree cover (Supplementary Fig. 18) and the extent of forest restoration (Supplementary Fig. 12). For scenarios relevant to ERPs, we tentatively assumed that the forest restoration rate was consistent with the historical forest restoration speed in China46. Considering that expanding ERPs (ERPs-E) aims to accelerate achieving carbon neutrality through nature-based solutions, we assumed that the forest restoration rate will further improve to finish the forest restoration tasks in all potential forest restoration areas before 2060 for scenarios related to ERPs-E.

For grassland restoration, considering that temperature and precipitation are the most important factors that influence the distribution of grassland111, we analyzed the relationship between the area fraction of grassland cover and temperature/precipitation in rural areas (areas with population density less than 5 people/km2, Supplementary Fig. 19) of the QTP. Then, we identified the key thresholds of temperature and precipitation that constrain the growth of grass and used them to predict the potential grass cover (Supplementary Fig. 20). Next, we converted the potential grass cover map into the extent of the potential grassland restoration map by deducing areas with forest, cropland, grassland, wetland and impervious surfaces (Supplementary Fig. 12). We also predicted the effect of future climate change (using the climate data provided by CMIP6) on the extent of potential grassland restoration (Supplementary Fig. 12) to meet the needs of the simulation rules in Table 1. Notably, for areas that are suitable for forest restoration and grassland restoration at the same time, we prioritized forest restoration. For the speed of grassland restoration, we tentatively assumed that the rate of future grassland restoration was consistent with the rate of grassland restoration in 2000–2020 by referring to the practice used in prior studies46,112.

The potential areas for forest restoration and grassland restoration under current, SSP-245 and SSP-585 climate conditions are shown in Supplementary Fig. 12.

Scenario comparison to disentangle different factors induced NEP change

Equation (4) was used to calculate the cumulative NEP (CNEP) difference between different scenarios and the Ref scenario to quantify the effect of different factors induced NEP change.

$${\Delta {CNEP}}_{k}={\sum }_{i={strat}}^{{end}}{{NEP}}_{k,i}-{\sum }_{i={start}}^{{end}}{{NEP}}_{{Ref},i}$$
(4)

where \({\Delta {CNEP}}_{k}\) is the difference in CNEP between the k scenario (including all scenarios in Group Climate change, Ecological restoration and Expanding ecological restoration) and the Ref scenario from the start year to the end year; \({{NEP}}_{k,i}\) is the NEP under scenario k of year i; and \({{NEP}}_{{Ref},i}\) is the NEP under the Ref scenario of year i.

To quantify the effect of permafrost thawing on carbon cycle, we established two scenarios named TPa and TPb Scenario TPa indicated the condition that using the predicted temperature and precipitation data to drive the updated BBMS which considered the effect of permafrost thawing, while scenario TPb indicated the condition that using the predicted temperature and precipitation data to drive the original BBMS which did not consider the effect of permafrost thawing (Table 1). Then Eq. (4) was used to calculate the difference in CNEP under the condition that considering permafrost thawing or not due to climate change compared with Ref scenario. The difference between them was the effect of permafrost thawing on CNEP (equation described in Table 2). Likewise, equations to calculate other factors induced CNEP change also through similar routines, all these equations were summarized in Table 2.

Table 2 The contribution of cumulative NEP change is ascribed to various factors
Full size table

To further track the effect of different ecological restoration practices on the NEP, we established a simplified scheme to calculate the NEP change caused by different ecological restoration practices in each 20 × 20 km2 grid. First, we overlapped land use/cover (LUC) maps of the start and end years to identify different ecological restoration practices in each grid according to the rules defined in Supplementary Fig. 21. Second, we extracted the area fractions of PFTs involved in ecological restoration practices (Supplementary Fig. 22). Next, we tracked the ecological restoration practices relevant to LUC trajectories (Supplementary Fig. 23) and calculated the NEP in different ecological restoration practices of these scenarios using the equations established in our prior studies6,30.

Future climate data correction

Explicit information for future temperature and precipitation projections from CMIP6 models we used is described in Supplementary Table 2. Since climate simulations from CMIP6 models are generally biased compared to observations, Eqs. (5) and (6) were used to bias-correct future temperature and precipitation projections derived from CMIP6 models113.

$${T}_{{future},{bias}-{corrected}}={T}_{{current},{obs}}+({T}_{{future},{CMIP}6}-{T}_{{current},{CMIP}6})$$
(5)
$${P}_{{future},{bias}-{corrected}}={P}_{{current},{obs}}+({P}_{{future},{CMIP}6}-{P}_{{current},{CMIP}6})$$
(6)

where \({T}_{{future},{bias}-{corrected}}\) and \({P}_{{future},{bias}-{corrected}}\) are the bias-corrected temperature and precipitation during the future period; \({T}_{{current},{obs}}\) and \({P}_{{current},{obs}}\) are the observed temperature and precipitation during the overlapping period (2015–2018) derived from CMFD114; \({T}_{{future},{CMIP}6}\) and \({P}_{{future},{CMIP}6}\) are the projected temperature and precipitation during the future period based on CMIP6 model outputs; and \({T}_{{current},{CMIP}6}\) and \({P}_{{current},{CMIP}6}\) are the projected temperature and precipitation during the current period based on CMIP6 model outputs.

Model validation

We compared our simulated ALT and SOC stock in permafrost with results obtained from Wang et al.33. The high consistency between them along the altitudinal gradient implied the confidence of our estimated SOC (Supplementary Fig. 24). This result was consistent with the magnitude measured in Wang et al.33. Moreover, we compared our simulated GPP/NEP with flux site observation records in the QTP obtained from the China Flux Observation and Research Network (http://www.chinaflux.org). The spatial distribution of these flux sites is marked in Supplementary Fig. 14. Explicit information on these flux sites is described in Supplementary Table 3. Notably, a high degree of consistency between these comparisons indicated the reliability of our results (Supplementary Fig. 25). We also compared our simulated total NEP in the QTP with estimated values obtained from previously published studies through varied approaches. The consistency in magnitude and tendency suggested that our results were also feasible at the regional scale (Supplementary Fig. 13).