Solar radiation variation weakened the boost of gross primary production by vegetation restoration in China’s most forestry engineering areas during 2001–2020

Over the past decades, ecological restoration initiatives in China have made great progress in restoring degraded forests and increasing vegetation coverage, yet the carbon sequestration effects of these initiatives in the context of climate change are not clear. In this study, we assessed the effects of vegetation restoration on gross primary production (GPP) in China’s forestry engineering areas, where large-scale vegetation restoration programmes were launched, during 2001–2020 by disentangling the respective roles of land cover change (LCC), CO2 fertilization, and climate changes using a two-leaf light use efficiency model. We found that LCC attributed by the vegetation restoration dominantly accelerated the increase of GPP in seven out of the eight areas, and CO2 fertilization played a near-equivalent role in all areas. By contrast, the changes in different climate factors contributed to GPP variations diversely. The solar radiation variation greatly inhibited the vegetation GPP over time in seven out of these areas, and the changes in air temperature and vapor pressure deficit regulated GPP inter-annual variations without clear trends in all areas. This study advances our understanding of the contribution of China’s afforestation on its forest GPP in a changing climate, which may help to better manage forests to tackle the challenge of the climate crisis in the future.


Introduction
Terrestrial gross primary production (GPP), as the largest component of the global terrestrial carbon budget, plays an important role in the global carbon cycle (Ichii et al 2005, Zhao et al 2005, Piao et al 2009, Ahlström et al 2015).GPP is vulnerable to changes in climate and atmospheric CO 2 concentration and disturbances from human activities (Friedlingstein et al 2010, Li et al 2015).Thus, it is of great importance to investigate the fate of regional GPP in a changing climate with human interferences.
The changing climate and the rising atmospheric CO 2 concentration affect vegetation growth and carbon cycle strongly.They alter the physiological constraints on plant photosynthetic rates (Nemani et al 2003, Zhao and Running 2010, Piao et al 2013, Chen et al 2019).The increase in CO 2 concentration increases the intercellular CO 2 content and thus promotes leaf photosynthesis (Piao et al 2013, Jiang and Ryu 2016, Chen et al 2019).Solar radiation (Rad) is considered to be the main driver of the negative effect of global GPP from 1982 to 2015 (Sun et al 2018).Warming further reduces the pressure of air temperature (Ta) to increase GPP in the northern high latitude cold regions (Nemani et al 2003), while increasing Ta can increase vapor pressure deficit (VPD) and thus exacerbate drought stress (Liu et al 2019), leading to a decrease in GPP (Ciais et al 2005, Zhao and Running 2010, Anderegg et al 2015).
In the late 20th century, the Chinese government launched large-scale vegetation restoration programmes, including the Global Canopy Program and natural forest conservation projects, to mitigate land degradation, desertification and soil erosions (Yang et al 2014, Li 2021).Through these projects, vegetation cover on land surface has increased significantly, effectively limited soil erosion (Xin et al 2011, Liu et al 2012a, Sun et al 2015) and contributed significantly to regional and global carbon sinks (Guo et al 2013, Fang et al 2014, Li et al 2015, Xiao et al 2015).The Three-North Shelter Forest Program, initiated in 1978 (Zhang et al 2016), has contributed to an increase of 8.05% in forest cover and a slight increase in GPP (Xie et al 2020).The Pearl River Shelter Forest Program has brought significant changes in forest cover since 1996 (Hasan et al 2019, Xiao et al 2019) and land cover change has become the main contributor to forest GPP increase since 2011 (Zhang et al 2022).The forestry engineering areas in China have different climate change characteristics (Yang et al 2014, Li 2021), GPP could be affected differently by LCC, CO 2 and climate change in individual ecological areas, but existing studies are often incomplete in the consideration of impacting factors with empirical attribution frameworks (e.g. by linear regressions), hampering our understanding on the capacity of vegetation to sequester carbon.Thus, a systematic understanding of the effects of China's vegetation restoration initiatives over time in the context of changing climate is urgently needed.
In this study, we assess the effects of the vegetation restoration on GPP in China's forestry engineering areas during 2001-2020 by disentangling the respective roles LCC, CO 2 fertilization, and climate changes using a two-leaf light use efficiency (TL-LUE) model with scenario simulations.The objectives of this study are: (1) to characterize the dynamic of the GPP in eight forestry engineering areas with different vegetation types since 2001; (2) to disentangle and quantify the individual and combined effects of changes in LCC, CO 2 and climate factors (Rad, Ta and VPD) on the GPP of eight forestry areas in China.

Meteorological and CO 2 concentration data
The daily meteorological data were interpolated from 753 meteorological stations across the country for the period of 2001-2020 with a spatial resolution of 500 m × 500 m.The climate variables include solar radiation (Rad), air temperature (Ta) and relative humidity.Both daily Rad and daily Ta are in good agreement with the tower observations, and Ta showed high agreement with the 0.5 • monthly air temperature data from the China Meteorological Administration (http://cdc.cma.gov.cn)(Liu et al 2016).
Monthly CO 2 concentration was used in the simulation which was obtained from direct measurements at the Mauna Loa Observatory in Hawaii (https://gml.noaa.gov/ccgg/trends)calculated from hourly observations.

Study area
The spatial distribution and specific information of the eight areas in China (www.resdc.cn/data.aspx?DATAID=138) are shown in figure 1 and table 1.

Flux data for model validation
To test the performance of TL-LUE model in simulating GPP (GPP_TL), the model was validated at both site and regional scales.At site scale, monthly GPP data from 12 sites (91 site years) of ChinaFlux were selected, including 4 forest sites, 2 wetland sites, 5 grassland sites, and 1 cropland site (supplementary table 1).

TL-LUE model
The latest version of the TL-LUE model (Bi et al 2022) simulates GPP as follows: where ε msu and ε msh denote the maximum LUE of sunlit and shaded leaves in the vegetation canopy, respectively; f (VPD), g (Ta ) and c(CO 2 ) were regulation scalars of VPD, Ta and atmospheric CO 2 concentration, respectively, which were described in  Bi et al (2022) and Zhou et al (2016); APAR su and APAR sh are the PAR absorbed by sunlit and shaded leaves, respectively, which were calculated as: where α denotes albedo, which varies among vegetation types; β is the leaf inclination angle (60  (2022).As for scaling from hourly scale to daily scale, we followed the method by Chen et al (1999).Firstly daily averaged solar declination was calculated, then daily averaged solar zenith angle was calculated, and finally daily averaged PARdir and PARdif were calculated according to equations ( 2) and (3).

Simulation setup
Six scenarios were simulated to quantify the impact of changes in LCC, CO 2 and meteorological factors on GPP over the period of 2001-2020 (table 2).
In Scenario I, the TL-LUE model was driven by the observed data of land cover, LAI, CO 2 and meteorological factors (Rad, VPD and Ta) that followed the historical changes or the period of 2001-2020 to simulate GPP.
In Scenario II, the TL-LUE model was driven by the same datasets of CO 2 and meteorological factors (Rad, VPD and Ta) as in Scenario I, except the land cover and LAI dataset.The land cover and LAI data are kept at 2001 levels.
In Scenario III, the same datasets of land cover, LAI and meteorological factors (Rad, VPD and Ta) as in Scenario I, and CO 2 in 2001 was used to drive the TL-LUE model to explore the effects of CO 2 changes on vegetation GPP after 2001.
In Scenario IV, the same land cover, LAI, CO 2 , Ta and VPD datasets as in Scenario I, except Rad, which is kept at 2001 levels to analyze the effect of Rad changes on vegetation GPP over two decades.
In Scenario V, the same land cover, LAI, CO 2 , Rad and VPD data sets were used as in Simulation I, except Ta, which is kept in 2001 to analyze the effect of Ta changes on vegetation GPP over the last 20 years.
In Scenario VI, the same land cover, LAI, CO 2 , Rad and Ta data sets were used as in Simulation I, except VPD, which is kept in 2001 to analyze the effect of VPD changes on vegetation GPP over the last 20 years.

Calculating the effects and cumulative effects of impact factors on GPP
The differences between dynamic and static simulations on vegetation GPP was used according to the equation: where GPP_ Dynamic is the GPP in Scenario I, GPP_ Static is the GPP results from Scenarios II-VI, and ∆GPP is the dynamic effect of changes in LCC, CO 2 , Rad, Ta and VPD on vegetation GPP, respectively.The cumulative impact of each factor on GPP is expressed as: where j is the year ranging from 2001 to 2020; i was the the ith pixel.

Model validation and evaluation
The site-scale validation showed that TL-LUE model was able to track the seasonal and inter-annual variations of GPP (figure 2).The R 2 ranged 0.67-0.96for the 12 sites.We also evaluated our simulation with two other GPP products (GPP_Fluxsat and GPP_GOSIF) at the regional scale (supplementary figure S1).98.57% of the vegetation areas had a significant positive correlation between monthly GPP_TL and GPP_Fluxsat (p < 0.05) with an average R 2 of 0.78.And monthly GPP_TL was significantly positively correlated with GPP_GOSIF (p < 0.05) with an average R 2 of 0.80 in 99.40% of the vegetation areas.

Long-term trends of annual GPP and contributions of LCC and CO 2 fertilizations over different engineering areas
Firstly, the annual total GPP and long-term trends in the eight areas from 2001 to 2020 were analyzed (figures 4(a) and (b)).Among these areas, the largest total annual GPP locates at Area VI, which has an average of 1.49 Pg C yr −1 .This is mainly because, Area VI is located in the south of China, where has abundant forest and grassland with vegetation-friendly climatic environment.This area also has the largest increasing trend, with a value of 0.0106 Pg C yr −2 .In terms of total annual GPP, it was followed by Area VII (0.53 Pg C yr −1 ) and Area I (0.50 Pg C yr −1 ).The mean total annual GPP in Area I was slightly lower than that in Area VII, but the growth rate was higher (0.0065 and 0.0078 Pg C yr −2 , respectively), resulting in a decreased difference in total annual GPP over time.In comparison, areas VIII, V, IV, III and II had relatively low annual GPP and trends.We further analyzed the contributions of total annual GPP by different vegetation types (supplementary figure S2(a)).The average annual GPP for forest, cropland, and grasslands for each area were ranged from 1.97 × 10 6 to 4.96 × 10 8 , from 5.09 × 10 7 to 3.23 × 10 8 , and from 3.35 × 10 7 to 1.38 × 10 9 T C yr −1 , respectively.Then, the impacts of LCC on GPP in the eight areas during 2001-2020 were investigated (figures 4(c) and (d)).Overall, the areas VI, I and VII were subjected to higher impacts of LCC on GPP, with average values of 22.50, 18.46, 17.04 and 18.82 Tg C yr −1 , respectively.Their impacts were also significantly increased, with 3.26, 2.59 and 3.06 Tg C yr −2 , respectively.This was followed by the areas IV and II (7.77 and 6.76 Tg C yr −1 , respectively), whose impacts on GPP were moderately increased (1.13 and 1.02 Tg C yr −2 , respectively).It implied that vegetation GPP was less affected by LCC in areas IV and II, albeit that GPP was affected by higher LCC per unit area in areas IV and II.In comparison, the GPP in Area III, VIII and V were less affected by LCC, with averages of 1.98, 0.84 and −1.67 Tg C yr −1 , respectively, and trends in total annual impact were 0.34, 0.28 and −0.22 Tg C yr −2 , respectively.The impact of LCC on average annual GPP for forest, cropland, and grasslands for each area were ranged from 1.05 × 10 5 to 1.18 × 10 7 , from −1.77 × 10 6 to 4.02 × 10 7 , and from 7.42 × 10 5 to 1.29 × 10 7 T C yr −1 , respectively (supplementary figure S2(b)).
Lastly, the impacts of CO 2 changes on GPP in the eight areas during 2001-2020 were investigated (figures 4(e) and (f)).The GPP in Area VI affected by CO 2 changes during 2001-2020 was relatively higher, with an average value of 20.10 Tg C yr −1 .Its increasing trend of the impact was also high, with 2.34 Tg C yr −2 .It was followed the areas VII and I, with means of 8.29 and 6.19 Tg C yr −1 , respectively.However, the increasing trends were moderate, at 0.93 and 0.71 Tg C yr −2 , respectively, which showed that the GPP in areas VII and I was positively affected by lower CO 2 changes.In comparison, the GPP in areas VIII, V, IV, III and II were minimally affected by CO 2 changes, with averages of 3.13, 2.75, 1.55, 1.48 and 1.42 Tg C yr −1 , respectively, and the increasing trends of annual sums of impacts were low, with 0. 35, 0.30, 0.19, 0.17 and 0.17 Tg C yr −2 , respectively.The impact of CO 2 changes on average annual GPP for forest, cropland, and grasslands for each area were ranged from 2.80 × 10 4-6.00 × 10 6 , from 4.92 × 10 5 -1.88 × 10 7 , and from 7.19 × 10 5-4.91 × 10 6 T C yr −1 (supplementary figure S2(c)).and −0.17 Tg C yr −2 , respectively.Area VII was positively influenced by Rad changes on GPP, and the 20 year mean value of the positive influence was 7.98 Tg C yr −1 , with a trend of −0.79 Tg C yr −2 .The impact of Rad variations on average annual GPP for forest, cropland, and grasslands for each area were ranged from −6.90 × 10 5 -1.89 × 10 6 , from −1.61 × 10 7 -9.51 × 10 6 , and from −1.18 × 10 7 -1.05 × 10 6 T C yr −1 (supplementary figure S2(d)).

Impact of the change of climate factors on long-term GPP over different engineering areas
Then, we analyzed the impact of Ta on GPP (figures 5(c) and (d)).It shows that Ta changes had a positive effect on GPP in Area VI from 2001 to 2020, with 20 year mean values of 17.60 and 9.87 Tg C yr −1 , respectively, and the increasing trends of the impact were slightly higher, with 1.04 and 1.10 Tg C yr −2 , respectively.The GPP in areas I, IV and II were affected by Ta changes with relatively low total annual values of 1.36, 1.13 and 0.12 Tg C yr −1 , respectively, and the increasing trends of the impact were relatively low at 0.11, 0.08 and 0.01 Tg C yr −2 , respectively, showing that the vegetation GPP in areas I, IV and III were less affected by Ta changes.Areas VIII, V, VII and III were negatively affected by Ta changes on GPP, with 20 year means of −1.95, −0.60, −0.49 and −0.33 Tg C yr −1 , respectively, but the trends of the impact were not prominent, at −0.08, −0.07 and −0.07 Tg C yr −2 respectively.The impact of Rad variations on average annual GPP for forest, cropland, and grasslands for each area were ranged from −2.38 × 10 5 -7.35 × 10 6 , from −1.71 × 10 6 -1.95 × 10 7 , and from −9.55 × 10 5 -7.61 × 10 5 T C yr −1 (supplementary figure S2(e)).
Lastly, we analyzed the impact of VPD on GPP (figures 5(e) and (f)).It shows that the GPP in areas VI and VII were negatively affected by VPD changes during 2001-2020, and the total annual value of negative impact was obvious, with 20 year mean values of −11.80 and −5.41 Tg C yr −1 , respectively.The trend of total annual impact was also more pronounced, with −0.32, and −0.21 Tg C yr −2 , suggesting that vegetation GPP was more negatively affected by changes in VPD.In areas VIII and V, GPP were negatively impacted by changes in VPD, with the 20 year annul mean values of −1.05 and −0.83 Tg C yr −1 , respectively.In areas I, IV, II and III, GPP were positively affected by changes in VPD, with 20 year means of positive impacts of 3.93, 2.08, 0.94 and 0.44 Tg C yr −1 , respectively.The impact of VPD variations on average annual GPP for forest, cropland, and grasslands for each area were ranged from −2.47 × 10 6 -3.72 × 10 5 , from −1.37 × 10 7 -3.44 × 10 6 , and from −1.50 × 10 6 -1.04 × 10 6 T C yr −1 (supplementary figure S2(f)).

Cumulative impact of impact factors on vegetation GPP during 2000-2020
The cumulative impact of changes in LCC, CO 2 , Rad, Ta and VPD on GPP during 2001-2020 were analyzed for the eight areas (figure 6 and table 3).In Area I, LCC mainly dominated the increase in GPP, with a cumulative effect of up to 369.16 Tg C by 2020.In contrast, CO 2 , VPD and T a have a comparatively lower positive cumulative impact on GPP.However, the positive effects of the LCC, CO 2 , Ta and VPD changes on GPP were offset by 2012 as a result of the Rad changes, which had a cumulative effect of −386.21Tg C on GPP.Similarly, in Area II, LCC mainly dominated the increase in GPP, with a cumulative effect of up to 135.20 Tg C by 2020.CO 2 , VPD and Ta had the lowest cumulative impact on GPP with 28.47, 18.77 and 2.47 Tg C respectively, contributing  S3).In Area III and IV, the land cover was mainly transformed from grasslands to croplands and forests (figure 4 and supplementary figure S3).In Area V, cropland and forests transformed into grasslands and urban, which resulted to the weak negative effects on GPP.In Area VI and VII, the land cover was mainly transformed from grasslands and cropland to forests.In Area VIII, the land cover was mainly transformed from grasslands and cropland to forests and urban, which is located in the eastern part of China, and the rapid urbanization in the past 20 years has offset partially the positive impacts of LCC.
The important role of LCC on GPP changes was also found in the three northern regions over 1982-2017 (Xie et al 2020), with the fallowing program (Feng et al 2016 ) and the grazing program (Gang et al 2018) preventing grassland degradation and contributing more than 1 g C m −2 yr −1 to GPP trends.Zhang et al (2022) found that in a study of Area VII for nearly 20 years, the positive impact of LCC on GPP also increased rapidly after 2010 due to the increase in the number of plantation forests and the rapid growth of new plantation forests (Zhang et al 2014, Tong et al 2018).Li et al (2017) found that human activities in Area IV were the main influencing factor of vegetation change, with a contribution of 55% and over 60% in some areas.The impact of LCC on GPP has shown a highly remarkable increasing trend over the past 20 years, indicating that ecological conservation policies in China have made great progress in restoring degraded forests and increasing vegetation productivity despite the adverse effects of urbanization (Chen et al 2021, Yue et al 2021).

Radiation variation decreased GPP in most areas
Rad variation weakened GPP in the eight engineering areas, except Area VII, with a decreasing trend, in line with the finding by Chen et al (2021) for the study area of China.In Area I, which locates in northern China, the impact of Rad showed a decreasing trend since 2001 (supplementary figure S4), and vegetation GPP was positively correlated with PAR (Piao et al 2006), so this area was increasingly negatively affected by Rad variation.Similarly, the northeastern part of Area VI is the area with strong Rad reduction (figure 4), and the negative effect of Rad change on GPP per unit area exceeds −100 g C m −2 yr −1 (figure 4), making Area VI more negatively affected by Rad change.Area VII is located in southern China with a warm and humid climate and limited Rad due to cloud cover (Li et al 2017, Feng andWang 2019).Enhanced Rad after 2001 led to a longer growing season, which increased vegetation GPP (Sun et al 2018).In addition, cloud cover increased the scattered Rad ratio and thus enhances LUE (Gu et al 2003, Rap et al 2018), which is more conducive to improving vegetation productivity (Zhou et al 2020), making Rad changes promote vegetation GPP in Area VII (Zhang et al 2022).

Regulations of Ta and VPD on GPP inter-annual fluctuation
The response of vegetation GPP to Ta is complex, since warming extends the vegetation growth period (Keenan et al 2014), but increases water stress (Reich et al 2018) that inhibiting GPP as well.The annual total impact of Ta changes on GPP and its trend varied widely across different regions.Compared with other factors, GPP was more influenced by Ta changes in Area VI, which was in line with the findings by Qu et al (2018) and Ye et al (2020).Ta has generally increased in the Yangtze River basin since 2001, which accelerated land carbon uptake (Farquhar et al 1980, Piao et al 2013), and the positive effect of Ta change on GPP is high.Although Area I has the largest area among the eight areas, it is located in northwestern China and is more restricted by water, which still increases slightly in the context of increasing desertification (Wang et al 2010), benefiting from human intervention and management (Xie et al 2020).Area I, IV, II and III are located in severe water shortage areas, where decreased VPD had a positive effect on vegetation productivity (Tuo et al 2018), coinciding with the finding by Xie et al (2020).The increased VPD in Area VI lead to the negative impact on GPP, which was consistent with the finding by Ye et al (2020).

Uncertainties of this study
Some uncertainties exist in this study.Firstly, the CO 2 concentration data used in this study for 2001-2020 are monthly averaged CO 2 data measured at the Mauna Loa Observatory in Hawaii, with the same value of CO 2 concentration for each month globally.However, the CO 2 concentration actually varies in time and space, and such treatment could lead to uncertainty when disentangling the effect of CO 2 on GPP.Secondly, the TL-LUE model does not take into account the effects of nitrogen deposition and tree age, which would lead to the inability to quantitatively explore the effects of these factors on GPP.Stand age has a non-linear relationship with forest carbon sink (Zhou et al 2015).Although some studies reported that the increasing hydraulic limitations would reduce GPP in aging forest (Drake et al 2011), and some suggested that triose phosphate utilization limitation would lead to decrease of GPP (Barnard and Ryan 2003), till now it lacks of sufficient observations to quantifying the effects of tree age in GPP.Therefore, in future studies, the perturbation functions of other influencing factors should be incorporated in the model to facilitate such attribution studies.Thirdly, leaf angle should change with vegetation types, but there was no dataset for depicting the spatial distributions of leaf inclination angles, so the mean leaf inclination angle as 60 degrees was used in this study, which could lead to some uncertainties of GPP estimation.Fourthly, the approach used to scale up input variables from instantaneous to daily scales by simply averaging them may lead to some uncertainties in GPP simulation.In addition, topography affects the surface solar radiation, as terrain shading can reduce direct radiation, and diffuse radiation can be amplified due to reflected flux from surrounding terrain (Wang et al 2018, Zhang et al 2019), which further affect the GPP dynamics.In this study, the effects of topography on GPP variations were not considered, which should be paid attention in future studies, especially for mountainous areas.

Conclusions
In this study, we disentangled the roles of LCC, CO 2 fertilization, and climate changes on GPP in the eight forestry engineering areas during 2001-2020 using the TL-LUE model with scenario simulations.The main findings are: (1) Among the eight areas, LCC attributed from the forestry engineering initiatives greatly accelerated the increase of GPP in seven areas except Area V, in which cropland and forests transformed into grasslands and urban, resulting to the weak negative effects on GPP.In addition, CO 2 fertilization played a near-equivalent role as LCC in all areas.
(2) Rad changes decreased the GPP in seven areas due to the decrease of radiation during 20 years, except Area VII.In Area VII, the increased radiation led to the positive effect on GPP.
(3) The increased VPD in the areas V, VI, VII and VIII affected the GPP negatively and the decreased VPD in the other four areas impacted the GPP positively.Compared with other areas, GPP in Area VI was more influenced by Ta changes.
These findings could improve our understanding of the contribution of China's afforestation on its forest productivity in a changing climate, which may help to better manage forests to tackle the challenge of the climate crisis in the future.
1. Remote sensing data Land cover data were obtained from the MODIS Land Cover product MCD12Q1 v006 dataset (https:// lpdaac.usgs.gov/products/mcd12q1v006/)with an annual temporal resolution and a spatial resolution of 500 m × 500 m (Sulla-Menashe et al 2019) from 2001-2020 with the International Geosphere-Biosphere Programme classification system.Leaf area index (LAI) data from GlobMap LAI version V3 for 2001-2020 were obtained by inversion of MODIS surface reflectance data (Deng et al 2006, Liu et al 2012b) with a temporal resolution of 8 d and a spatial resolution of 500 m × 500 m.The long-term LAI was compared with field measurements, showing an error of 0.81 on average (Liu et al 2012b).

Figure 1 .
Figure 1.Spatial distribution of eight forestry engineering areas (a).I is the Three-north shelterbelt program (TNSP), II is the afforestation program for Taihang mountain (THSP), III is shelterbelt program for Liaohe river (LRSP), IV is the shelterbelt program for middle reaches of Yellow river (YRSP), V is shelterbelt program for Huaihe river and Taihu lake (HRSP), VI is shelterbelt program for upper and middle reaches of Yangtze river (YRSP), VII is shelterbelt program for Pearl river (PRSP), VIII is coastal shelterbelt program (CSP).Land cover data of from MCD12Q1 v006 in eight areas in 2020 (b).ENF, EBF, DNF, MF, CS, OS, WS, SAV, GRA, WET, CRO and NOV means evergreen needle leaf forest, evergreen broadleaf forest, deciduous needle leaf forest, mixed forest, open shrubland, woody savannas, savannas, wet, cropland, grassland, non-vegetation, respectively.
Figure 3(a)  shows the spatial distribution of the annual mean GPP from 2001 to 2020 in the eight engineering areas.Overall, GPP showed a clear gradient of low to high form the north to the south.Among these areas, the highest annual GPP was in Area VII (1927.19g C m −2 yr −1 ), followed by Area VIII (1502.15g C m −2 yr −1 ) and in Area VI (1288.57g C m −2 yr −1 ).The lowest annual GPP was in Area IV (629.81 g C m −2 yr −1 ) and Area I (479.03g C m −2 yr −1 ).In general, LCC and CO 2 fertilizations have positively impacted the GPP over the eight areas (figures 3(b) and (c)).The largest impacts of LCC on GPP were mainly in northeast and southern China (figure3(b)).The impact of CO 2 fertilization on GPP varies significantly in spatial distribution, decreasing from the southeast coast to the northwest inland.The largest impacts were mainly in southern China, with the impact on GPP up to 30 g C m −2 yr −1 , while the least impacts were mainly in southwest, northwest and north China, with less than 3 g C m −2 yr −1 (figure3(c)).Differently, the climate factors have both negative and positive effects over different engineering areas, mostly playing negative roles (figures 3(d)-(f)).Most vegetation GPP was negatively affected by Rad variations, with the largest impacts exceeding −100 g C m −2 yr −1 , mainly in Northeast and Central China.Only a small proportion of vegetation GPP was positively affected by Rad variations, mainly in the south-west China (figure 3(d)).In addition, the positive effect of Ta change on GPP locates in the high GPP areas which exceed 15 g C m −2 yr −1 except the Sichuan Basin (figure 3(e)).Notably, the GPP in areas VI and VII were weakened by VPD changes most strongly (figure 3(f)).

Figure 3 .
Figure 3. Spatial distribution of mean GPP in eight areas in China from 2001 to 2020 (a), and spatial distribution of annual mean value of impact of LCC (b), CO2 (c), Rad (d), Ta (e) and VPD (f) on GPP from 2001 to 2020.

Figure 4 .
Figure 4. Long-term trend of the annual GPP (a) and that of the contributions of LCC (c) and CO2 (e) on GPP, and average value of vegetation GPP per unit area (b) and the impact of LCC (d) and CO2 (f) change on GPP per unit area in eight areas from 2001 to 2020.T-test was used to analyze the significance of the different factors on GPP variations.The symbols ' * * ' and ' * ' represent significance levels at p < 0.01 and 0.05 < p < 0.01, respectively.

Figure 5 .
Figure 5.Long-term trend of the annual impact of Rad (a), Ta (c) and VPD (e) on GPP, and the impact of change of Rad (b), Ta (d) and VPD (f) on GPP per unit area in eight areas from 2001 to 2020.T-test was used to analyze the significance of the different factors on GPP variations.The symbols ' * * ' and ' * ' represent significance levels at p < 0.01 and 0.05 < p < 0.01, respectively.

Figure 6 .
Figure 6.Cumulative effects of five impact factors (LCC, CO2, Rad, Ta and VPD) on GPP in eight areas from 2001 to 2020.

Table 1 .
Information of eight forestry engineering areas.

Table 3 .
Cumulative impact of five impact factors on GPP in eight areas from 2001 to 2020.Tg C. The positive effects of LCC, CO 2 , Ta and VPD changes on GPP were completely offset by 2013.In Area III, the Rad variation mainly dominated the decrease in GPP, with a negative cumulative effect of up to −109.15Tg C by 2020.The cumulative effect of Ta on GPP was also negative.In Area IV, LCC mainly dominated the increase in GPP, with a cumulative impact of up to 155.50 Tg C by 2020.VPD, CO 2 and Ta have the lower cumulative impact on GPP.Conversely, GPP was negatively affected by changes in Rad with a cumulative impact of −52.09Tg C. In Area V, Rad variation dominated the GPP reduction, with a negative cumulative effect of up to −186.12 Tg C by 2020.Meanwhile, the cumulative effects of LCC, VPD and Ta on GPP are also negative, at −33.47 Tg C,