Roles of Climate Change and Increasing CO₂ in Driving Changes of Net Primary Productivity in China Simulated Using a Dynamic Global Vegetation Model

Abstract

Net primary productivity (NPP) is the key component of the terrestrial carbon cycle, and terrestrial NPP trends under increasing CO₂ and climate change in the past and future are of great significance in the study of the global carbon budget. Here, the LPJ-DGVM was employed to simulate the magnitude and pattern of China’s terrestrial NPP using long-term series data to understand the response of terrestrial NPP to increasing CO₂ concentration and climate change. The results showed that total NPP of China’s terrestrial ecosystem increased from 2.8 to 3.6 Pg C yr⁻¹ over the period of 1961–2016, with an annual average of 3.1 Pg C yr⁻¹. The average NPP showed a gradient decrease from the southeast to northwest. Southwest China and Northwest China, comprising mostly arid and semi-arid regions, exhibited the largest increase rate in total NPP among the six geographical regions of China. Additionally, large interannual variability around the NPP trends was presented, and NPP anomalies in China’s terrestrial ecosystem are strongly associated with the El Niño-Southern Oscillation (ENSO). Southwest China made the largest contribution to the interannual variability of national total NPP. The total NPP of China’s terrestrial ecosystem continuously increased with the concurrent increase in the CO₂ concentration and climate change under different scenarios in the future. During the period from 2091 to 2100, the average total NPP under the A2 and RCP85 scenarios would reach 4.9 and 5.1 Pg C yr⁻¹ respectively, higher than 4.2 and 3.9 Pg C yr⁻¹ under the B1 and RCP45 scenarios. Forests, especially temperate forests, make the largest contribution to the future increase in NPP. The increase in CO₂ concentration would play a dominant role in driving further NPP increase in China’s terrestrial ecosystems, and climate change may slightly attenuate the fertilization effect of CO₂ on NPP.

1. Introduction

The increase rate of CO₂ in the 20th century was faster than any stages during the past 22000 years [1]. According to observed data from Mauna Loa, Hawaii, USA, the mean CO₂ concentration reached 408.3 ppm in 2018, which was an increase of 29% compared to the observed value (316.0 ppm) in 1959 and an increase of 45% compared to the value (277 ppm) in the pre-industrial period. Global CO₂ emissions are still increasing sharply. The cumulative CO₂ emissions from anthropogenic sources reached 615 ± 80 Pg C between 1870 and 2017; of these emissions, 425 ± 20 Pg C was from fossil fuel and industrial processes and 190 ± 75 Pg C was from deforestation and other land use changes [2]. Approximately half of these CO₂ emissions remained in the atmosphere over the long term (250 ± 5 Pg C, 41%); others were sequestered by terrestrial ecosystems (190 ± 50 Pg C, 31%) and ocean ecosystems (150 ± 20 Pg C, 4%), while the fate of 4% (25 Pg C) of CO₂ emissions is unknown. The retention of CO₂ in the atmosphere causes an increase in the CO₂ concentration and has become one of the main drivers for global warming.

Net primary productivity (NPP) is the net carbon fixed by plants through photosynthesis. It quantifies the conversion of atmospheric CO₂ into plant biomass, and is identified as a key variable for sustainability [3,4], biodiversity [5] and environmental degradation (e.g., desertification, deforestation) [6,7]. In recent years, carbon sequestration in terrestrial ecosystem has an uptrend and great interannual variability [8,9,10]. However, there were large uncertainties in the current estimates of terrestrial carbon sinks due to the complexity and diversity of terrestrial ecosystems. The terrestrial carbon sink is the residual of NPP minus heterotrophic respiration. Therefore, accurate estimation of NPP is the prerequisite for constraining the uncertainties in estimated terrestrial carbon sinks.

China covers a vast land area, i.e., 9.6 million km², and the regional climate is dominated by Asian monsoon, with diverse climate types ranging from tropical to cold-temperate and from humid to extremely dry [11]. China has experienced explosive economic growth in recent decades; this growth has been detrimental to the environment through land-use change, consumption of resources, and pollution. The extensive economy led to the largest annual CO₂ emissions in the world, and a series of policies on climate change and CO₂ emissions, such as adjustments in the energy structure, improvements in the energy efficiency, and reforestation [12], have been implemented to achieve global emission-reduction targets. Therefore, China is one of the most critical and sensitive regions in the global climate system, and studies of NPP and its dynamics in China are helpful for improving our understanding of the carbon balance at both the regional and global scales.

A number of studies have been conducted to examine the magnitude and pattern of NPP in China with statistical model [13], light use efficiency models [14,15,16,17,18], and process-based biogeochemical models [19,20,21,22,23,24,25]. These estimates of total NPP in China range from 1.44 to 4.64 Pg C yr⁻¹, owing to differences in model structure, parameters, and input data. Large uncertainties were still present in the estimation of the total NPP in China. Total NPP in China presented increasing trend in recent years [11,19,24,26], where the total NPP changed obviously, and the reasons for this are still unclear. In addition, a few studies have been conducted to investigate the response of terrestrial NPP to different future scenarios of climate change and increasing CO₂ in China. Therefore, the assessment of long-term trends in the magnitude and patterns of terrestrial NPP in China was one of the objectives of this study.

Climatic extremes, such as droughts and heat waves, have significant impacts on NPP. In parallel with global warming, climatic extremes occurred much more frequently in recent years in China, for example, the low temperature freezing in early 2008 [27], severe national droughts in 2011[28] and six consecutive years of spring droughts in Yunnan province since 2009 and so on. These extremes might be related to EI Niño-Southern Oscillation (ENSO) [29]. The ENSO is a coupled ocean-atmosphere phenomenon and characterized by irregular fluctuations between warming and cooling of sea surface temperature (SST) in tropical eastern Pacific (EP) or central pacific (CP) every 2−7 years. The warming and cooling anomalies are known as EI Niño and La Niña events, respectively. ENSO is a major driver for interannual and seasonal anomalies of regional and even global climates [29,30]. A number of studies have indicated that precipitation and temperature anomalies caused by the ENSO have large impacts on NPP of terrestrial ecosystems [9,31,32,33,34]. For instance, the transition from EI Niño to La Niña during 1997–1998 caused a reduction of terrestrial NPP in eastern Africa related to decreased precipitation [31]. The prevalent La Niña since 2010 caused up to six consecutive seasons of increased precipitation in Australia, which resulted in a remarkable increase in terrestrial ecosystem carbon sequestration in 2011 [34]. The global gross primary productivity (GPP) changed concurrently with ENSO, and GPP anomalies were more strongly associated with the ENSO than La Niña [9]. However, the influence of ENSO on terrestrial NPP in China has not been well investigated. Analyzing the linkage between total NPP of terrestrial ecosystems in China and ENSO was also one of the objectives of this study.

CO₂ fertilization induces global warming and concurrently enhances vegetation growth. For example, 12 free air CO₂ enrichment (FACE) experiments in different global regions indicated that CO₂ increases of 50% would cause NPP increases of 12% [35]. However, CO₂ fertilization effect may be limited by various factors, such as water availability [36]. Positive effects of CO₂ fertilization in arid and temperate grassland ecosystems appeared in sufficient water conditions [37]. Furthermore, effects of CO₂ fertilization was influenced by the precipitation at different times of year. Increasing spring precipitation would promote the effects of CO₂ fertilization, while increasing non-spring precipitation would inhibit this effect [38]. At the same time, warmer climatic conditions resulting from increasing CO₂ may increase vegetation moisture stress [9] and attenuate the direct positive effects of CO₂ increase on NPP, especially in semi-arid ecosystem [37]. Compared with the averaged total NPP from 2000–2012, climate change alone simulated by the LPJ model would reduce global terrestrial NPP by 12.5% and 7.5% under A2 and RCP85 scenarios during the period of 2091–2100, respectively [39]. The extent to which CO₂ fertilization and climate change affect terrestrial NPP in China is unclear; examining this question is another objective of this study.

2. Materials and Methods

2.1. LPJ Model

As a tool used in this study, the Lund-Potsdam-Jena dynamic global vegetation model (LPJ) is a process-based, coupled, non-equilibrium biogeography-biogeochemistry model, which has adopted many features from the BIOME family of models [40,41]. This model defines 10 different plant functional types (PFTs) to describe the vegetation composition in each grid cell. Each PFT has different forms of vegetation phenology (evergreen, summergreen and raingreen), leaf shape (broad-leaved and needle-leaved), climate zone (tropical, temperate and boreal), and photosynthetic pathway (C3 and C4). Different bioclimatic and biochemical parameters are set to account for the variety of structure and functioning among vegetation types. Key ecosystem processes such as vegetation growth, carbon allocation, resource competition, mortality, population establishment, fire disturbance, and litter decomposition are represented in the model to simulate the terrestrial vegetation dynamics and exchanges of carbon and water between the land and atmosphere. Details about the LPJ model can be found in [42,43]. Now, the LPJ model has been widely used to calculate the global carbon budget [2], to assess the impact of plant water balance on agriculture [44,45], to quantify the impact of climate change on agriculture productivity [46], and to estimate the bioenergy production potential [47]. It has acted as a tool to identify the contribution of permafrost soil to the global carbon budget [48] and to investigate the influence of fire on the terrestrial carbon cycle [49] and the regulation of the nitrogen cycle on the dynamic global vegetation, hydrology, and crop growth [50].

2.2. Input Data

The LPJ model was driven by meteorological data, annual CO₂ data, and soil texture data. Monthly fields of mean meteorological data, including monthly mean air temperature, precipitation, cloud cover, and the number of wet days at a spatial resolution of 0.5° × 0.5° from 1901 to 2016, were taken from the Climate Research Unit (CRU), University of East Anglia, called CRU TS3.25 [51,52]. The annual historical global atmospheric CO₂ data from 1901 to 1958 were obtained from the Carbon Cycle Model Linkage Project [53,54], which was extended to 2016 using the atmospheric CO₂ data from the Mauna Loa observatory. The soil texture data were retrieved from the FAO soil data set [55,56,57].

This study also used outputs from two climate models (National Center for Atmospheric Research-Community Climate System Model (NCAR-CCSM) and the National Centre for Meteorological Research-Coupled Model (CNRM-CN)) under four IPCC SRES emission scenarios (AR4 A2 & B1 scenarios and AR5 RCP45 & RCP85 scenarios) to investigate the impacts of future climate and CO₂ changes on NPP of terrestrial ecosystems in China. Eight future climate scenarios (

Table 1. Details of eight future climate scenarios from CMIP3 and CMIP5.

NO	Scenarios	Projects	GCMs	SRES
1	CNRM-CM3-A2	CMIP3	CNRM-CM3	AR4-A2
2	CCSM3-A2	CMIP3	NCAR-CCSM3	AR4-A2
3	CNRM-CM3-B1	CMIP3	CNRM-CM3	AR4-B1
4	CCSM3-B1	CMIP3	NCAR-CCSM3	AR4-B1
5	CNRM-CM5-RCP45	CMIP5	CNRM-CM5	AR5-RCP45
6	CCSM4-RCP45	CMIP5	NCAR-CCSM4	AR5-RCP45
7	CNRM-CM5-RCP85	CMIP5	CNRM-CM5	AR5-RCP85
8	CCSM3-RCP85	CMIP5	NCAR-CCSM4	AR5-RCP85

) from CMIP3 and CMIP5 (Coupled Model Intercomparison Project Phase 5) were used to generate the future climates of 2017–2100.

The monthly fields of mean temperature, precipitation, and cloud cover in CMIP3 and CMIP5 during 2001–2100 were downloaded from the World Data Center for Climate (https://cera-www.dkrz.de/WDCC/ui/cerasearch/) and the LASG (http://www.lasg.ac.cn/), respectively. All the data were interpolated into 0.5° × 0.5° grids. The monthly wet days from 2001 to 2100 were calculated with projected precipitation and historical averages of monthly precipitation and wet days [51] as follows:

$$wet={\left(\frac{pre*{\overline{wet}}^{\frac{1}{0.45}}}{\overline{pre}}\right)}^{0.45}$$

(1)

where pre is the monthly precipitation in the future, $\overline{pre}$ is the historical mean monthly precipitation, and $\overline{wet}$ is the historical mean number of monthly wet days.

To adjust for climate model bias, the additive departures of projected temperature and multiplicative departures of projected precipitation were calculated to produce time series of the future climates for each grid during 2017–2100 [22]. Departures of projected temperature and precipitation were estimated according to anomalies of projected climates in CMIP3 during 2001–2016 and CMIP5 during 2006–2016 relative to historical climates from CRU TS3.25 during 2001–2016. The time series of the mean annual temperature (MAT) and mean annual precipitation (MAP) over China are shown in Figure 1. Overall, the MAT in China obviously increases, while the MAP has no clear tendency. The MAT in China changed from 6.36 °C to 8.24 °C during 1961–2016, with an average of 7.21 ± 0.48 °C and an increase rate of 0.024 °C∙yr⁻¹. The increasing rates (average of NCAR and CNRM outputs) of the MAT under the A2 and RCP85 scenarios were 0.053 °C∙yr⁻¹ and 0.054 °C∙yr⁻¹ during 2017–2100, respectively, which were significantly larger than those under the B1 (0.013 °C yr⁻¹) and RCP45 (0.022 °C yr⁻¹) scenarios. The MAP in China changed from 563 mm to 721 mm during 1961–2016, with an average of 631 ± 32mm. There were large interannual fluctuations in MAP during 2017–2100 in the different models, especially under the AR4 scenario.

The data set of the global CO₂ concentration under the IPCC AR4 and AR5 scenarios during 2017–2100 were downloaded from the IPCC DDC (http://www.ipcc-data.org/) and RCP Database (http://www.iiasa.ac.at/web-apps/tnt/RcpDb), respectively. The average of the references estimated from the ISAM and BERN-CC models was used in AR4. The CO₂ concentrations under the AR4 and AR5 scenarios over 10 years were interpolated into yearly CO₂ concentration values using the Fourier-fitting method. The results are shown in Figure 2. The observed CO₂ concentration increased from 317.64 ppm in 1961 to 404.21 ppm in 2016, with an annual rate of 1.52 ppm. Under the AR5 RCP85 scenario (AR5_RCP85), the CO₂ concentration increased to 935.87 ppm in 2100, which was the highest for the scenario without the inclusion of climate policy [58]. This scenario was followed by the AR4 A2 scenario (AR4_A2), with a value of 846 ppm in 2100. B1 (AR4_B1) and RCP45 (AR5_RCP45) are the medium stabilization scenarios, and the CO₂ concentration increased to 544.50 ppm and 538.36 ppm in 2100, respectively.

2.3. Modelling Protocol

The LPJ simulation starts from ‘bare ground’ and ‘spins up’ for 1000 model years until an approximate equilibrium has been reached with respect to the carbon pools and vegetation cover [42]. In this study, the LPJ model was driven with CRU data and CO₂ from 1901 to 1930 to spin up for 1000 model years; then, the standard simulation was run from 1901 to 2016. By comparing the simulated spatial distribution of the PFTs with the land cover maps of China (http://www.resdc.cn) in 2000, we revised the model parameters and repeated the above process until the simulated PFT spatial distribution was in general agreement with the land cover of China [23].

The calibrated model was applied to simulate the NPP of China over the period from 1901 to 2100 at a spatial resolution of 0.5° × 0.5°. The historical (1901–2016) climate and 8 future climate scenarios (2017–2100) were used together with the corresponding CO₂ to drive the model. Furthermore, we conducted two control experiments to analyze the influences of only climate change and only increasing CO₂ on NPP in the future. In one control experiment, only climate would change in the future, while the CO₂ concentration was set to at the observed value in 2016. In another experiment, only CO₂ would change in the future while the climate was set to a random climate in 2007–2016. Configurations of historical and future climate and CO₂ in different simulations are listed in

Table 2. Configurations of historical and future climate and CO₂ in different simulations.

Simulations	Scenario	History	Climate Change	CO2 Change
1	A2_CO2	CRU	No	A2
2	B1_CO2	CRU	No	B1
3	RCP45_CO2	CRU	No	RCP45
4	RCP85_CO2	CRU	No	RCP85
5	A2_C	CRU	A2	No
6	B1_C	CRU	B1	No
7	RCP45_C	CRU	RCP45	No
8	RCP85_C	CRU	RCP85	No
9	A2_Couple	CRU	A2	A2
10	B1_Couple	CRU	B1	B1
11	RCP45_Couple	CRU	RCP45	RCP45
12	RCP85_Couple	CRU	RCP85	RCP85

.

3. Results and Discussion

3.1. Temporal Trends of Terrestrial NPP in China and Linkage with ENSO

The national total simulated NPP in China changed with fluctuations, from 2.8 Pg C yr⁻¹ to 3.6 Pg C yr⁻¹, over the period of 1961–2016 (Figure 3), with an average of 3.1 ± 0.2 Pg C yr⁻¹. Total NPP increased at a rate of 0.0076 Pg C yr⁻¹ since 1961. Based on the mean of 251 published global terrestrial NPP estimates from 1862 to 2011 by Ito [7], the total NPP of China accounted for approximately 5.52% of the global total NPP (mean of 251 estimates: 56.2 Pg C yr⁻¹).

The annual total NPP demonstrated distinct trends in different stages during 1961–2016. The average NPP over the periods of 1961–1980, 1981–2000, and 2001–2016 were 2.98 ± 0.11 Pg C yr⁻¹, 3.11 ± 0.14 Pg C yr⁻¹, and 3.27 ± 0.18 Pg C yr⁻¹, respectively, and increased at rates of 0.0049 Pg C yr⁻¹, 0.0006Pg C yr⁻¹, and 0.0306 Pg C yr⁻¹ in these periods. The increasing rate of national total NPP was higher in the period of 2001–2016 than those in the periods of 1961–1980 and 1981–2000. Meanwhile, larger interannual fluctuations were observed during 1981–2000. In the period of 1991–2000, the total NPP declined with fluctuations.

The analyzing linkage of total NPP with ENSO was made to explain the interannual fluctuations of terrestrial NPP in China. The magnitude and pattern of precipitation in eastern China during summer and winter were greatly affected by the ENSO [59,60,61]. For example, during the development stage of the ENSO, deficient precipitation occurred over most of eastern China, with as much as 30%–50% decreases in precipitation in some areas. In contrast, abundant precipitation was observed in the Yangtze-Huai Plain. During the attenuation stage of the ENSO, precipitation mostly increases in the area of the middle-lower Yangtze plains and declines in the Yellow River basin and in northern, southern, and southwestern China [59]. Such ENSO-induced anomalies of precipitation would definitely affect NPP. Tao [62] found that NPP in China declined remarkably in some EI Niño years.

Since 1951, 14 global EI Niño events have been recorded, and the EI Niño events in 1982/1983, 1997/1998, and 2015/2016 were “super level”. In addition, the EI Niño events in 1965/1966, 2004/2005, 2009/2010, and 2015/2016 were CP EI Niño, while others were EP EI Niño [29]. We used MEI provided by NOAA-CPC to define the ENSO events [63], and associated these events with NPP in China during 1961–2016. As shown in Figure 3, the terrestrial NPP of China declined remarkably in the EP EI Niño years and increased in the CP EI Niño years. Before 2000, all ENSO events were EP EI Niño except for the events in 1965/1966, and the total NPP simulated by the LPJ model had relatively low values in these ESNO years. After 2000, the ENSO events were mostly CP EI Niño, and the total NPP had relatively high values in these ESNO years. Additionally, the NPP often increased in the La Niña year. However, continuous La Niña events (1973–1975, 1998–2000) might induce NPP to decrease.

3.2. Spatial Pattern of Terrestrial NPP in China

National average NPP in China was 322.05 g C m⁻² yr⁻¹ over the period of 1961–2016. As shown in Figure 4, NPP in China showed a gradient decreasing from the southeast to the northwest. The annual NPP was mostly below 50 g C m⁻² yr⁻¹ in extensive areas of Northwest China and western areas of Inner Mongolia covered with desert or the Gobi. In these areas, the climate is characterized by low temperatures and scarce precipitation. In the Greater and Lesser Xing’an and Changbai mountain areas of Northeast China, temperate needle forests or temperate deciduous broad-leaved forests are extensively distributed, and the annual NPP was in the range of 500 to 600 g C m⁻² yr⁻¹. Southwest, South central and East China lie in the subtropical monsoon climate zone, where abundant precipitation and warm temperatures are suitable for vegetation growth. The annual NPP in these areas was above 500 g C m⁻² yr⁻¹. Among these areas, Hainan province had the highest NPP, i.e., above 800 g C m⁻² yr⁻¹, in which tropical rain forests and monsoon forests were widely distributed. In the Yunnan-Guizhou Plateau, the Hengduan mountains areas, and the southeastern coastal areas, the vegetation types were mixed with tropical rain forests, subtropical evergreen broad-leaved forests and temperate transitional forests, and the NPP ranged from 600 to 700 g C m⁻² yr⁻¹.

In addition, the northeastern Plain and North China Plain are major grain crop and economic crop production areas of China. Temperature and precipitation are colimiting factors of vegetation growth, and the annual NPP here ranged from 200 to 400 g C m⁻² yr⁻¹. The Sichuan Basin is also a well-developed agricultural area with suitable climate conditions for plant growth, and the mean NPP was in the range of 400–500 g Cm⁻² yr⁻¹. The Qinghai-Tibetan Plateau and Inner Mongolian Plateau are major pasture areas of China, where precipitation is the major limiting factor affecting vegetation growth. The mean NPP was in the range of 200–400 gCm⁻² yr⁻¹.

The total NPP varied greatly in different regions of China. Totals and averages of NPP were calculated for six geographical regions for the periods of 1961–1980, 1981–2000, 2001–2016, and 1961–2016 (Figure 5). Northwest China had the largest area (32.37% of the national total) but the lowest total NPP among the six geographical regions. The total NPP and average NPP in Northwest China were 0.31Pg C yr⁻¹ (9.68% of the national total) and 101.43 g C m⁻² yr⁻¹, respectively, and both were much lower than those in other regions. Southwest China had the greatest portion of the national total NPP (0.90 Pg C yr⁻¹, 29.05% of the national total) due to the second largest area (24.41% of the national total) and high average NPP (392.84 g C m⁻² yr⁻¹). South-central China had the highest average NPP (566.98 g C m⁻² yr⁻¹). Total NPP here was 0.58 Pg C yr⁻¹. These regions were followed by East China, Northeast China and North China in the contribution to national total NPP. The total NPP in these three regions was 0.47, 0.43, and 0.41 Pg C yr⁻¹, respectively, and their average NPP was 561.45, 537.38 and 263.69 g C m⁻² yr⁻¹.

The transient trends of annual NPP in six geographical regions of China averaged over the periods of 1961–1980, 1981–2000, and 2001–2016 are shown in Figure 5. The total NPP showed gradually increasing trends in Northwest China, Southwest China, South-central China, and East China. Especially in Northwest China and Southwest China, the total NPP over the period of 2001–2016 increased by 20.00% and 16.05%, respectively, compared with that during 1961–1980. However, the total NPP in Northeast China and North China had no significant trend.

As shown in Figure 6, the temperatures increased in six regions of China over the periods of 1961–1980, 1981–2000, and 2001–2016, while precipitation showed distinguishable differences. Generally, an increase (decrease) in precipitation caused an increase (decrease) in total NPP in Northwest China, East China, and North China, but a decrease in precipitation in South-central China, which located in humid areas, didn’t cause a decrease in total NPP. Temperature played an important role in the increasing NPP in high altitude regions such as Southwest China.

Distinguishable difference in total NPP trends in six geographical regions led to the interannual variability (IAV) of terrestrial NPP in China from 1961 to 2016. An index [9] was used to assess the contribution of individual regions to the national total NPP IAV.

$$f_i=\frac{{\displaystyle \sum _t\frac{NP{P}_{i,t}\left|NP{P}_t\right|}{NP{P}_t}}}{{\displaystyle \sum _t\left|NP{P}_t\right|}}$$

(2)

where NPP_i,t is the total NPP anomaly (departure from the average NPP from 1961 to 2016 in this study) for region i at time t (in years in this study), NPP_t is the national total NPP anomaly, and |NPP_t| is the absolute national total NPP anomaly. f_i values are in the range of 0–1; a higher f_i indicates a larger contribution to the national total NPP anomaly.

NPP anomalies of six geographical regions are shown in Figure 7. Southwest China was found to make the largest contribution (30.32%) to the national total NPP IAV, followed by South-central China (18.81%); the two regions contributed nearly 50% to the national total NPP IAV. Northwest China (13.98%), East China (13.50%), Northeast China (11.96%), and North China (11.43%) played a similar role in regulating national total NPP IAV.

3.3. Comparison with Previous Studies

Table 3. National total NPP in China in previous and current studies.

Model	Period	NPP (Pg C yr−1)	Reference
TEM4	modern	3.65	[66]
LUE model	1992–1993	2.65	[17]
CASA	1997	1.95	[15]
CASA	1982–1999	1.80	[16]
CASA	1982–1999	1.44	[14]
CEVSA, GLO-PEN	1981–1998	3.05	[19]
CEVSA	1981–1998	3.09	[23]
LPJ	1961–2000	4.64	[20]
Remote Sensing model	1989–1993	3.12	[65]
BEPS	2001	2.24	[21]
CEVSA,CASA et al.	1980–2005	2.86	[67]
AVIM2	1981–2000	2.94	[68]
LPJ	1981–1998	3.11	[69]
BEPS	2000–2010	2.74	[24]
CASA	1980–2010	2.54	[25]
Statistical model	2000–2010	3.89	[13]
IBIS	1960–2016	2.46	[26]
CASA	2015	2.93	[18]
LPJ	1961–2016	3.10	This Study

shows the total NPP of China reported previously and in this study. Reported total NPP ranged from 1.44 to 4.64 Pg C yr⁻¹. The total NPP simulated by the CASA model ranged from 1.44 to 2.93 Pg C yr⁻¹; several estimates were lower, ranging from 1.44 to 1.95 Pg C yr⁻¹ [14,15,16]. The low total NPP output from the CASA model was possibly caused by the undervalue of maximum light use efficiency for some types of vegetation in China [64]. NPP simulated by the CASA model linearly changes with this parameter. Gao [65] analyzed the estimates of total NPP in China’s terrestrial ecosystems based on process-based models and remote sensing driven models in a meta-analysis, and the total NPP of these estimates averaged 2.83 ± 0.83 Pg C yr⁻¹. Additionally, large interannual variability of NPP exhibited in these estimates. Most of the estimates showed that total NPP peaked in 1990, and that valleys of NPP appeared in 1992 and 1997, similar to our results. The average total NPP of China in this study was 3.10 ± 0.2 Pg C yr⁻¹ during 1961–2016, higher than estimates of most previous studies. It should be kept in mind that the NPP estimated by the LPJ model is potential one, which is higher than actual NPP.

The Eddy Covariance (EC) technique has been widely used to measure the carbon and water fluxes at the ecosystem scale [70,71,72], and EC observations have also been used to assess the ecosystem models. The simulated annual GPP was validated with the flux data during the period of 2003–2010 summarized by Yu et al. [71]. As shown in

Table 4. Comparison of annual GPP simulated by LPJ with observations.

Code	Types	Latitude	Longitude	Period	GPPobs	LPJsim	RPE%
DHS	Forest	23.17	112.53	2003–2008	1367.26 ± 78.42	1953.98	42.91%
QYZ	Forest	26.73	115.05	2003–2008	1798.74 ± 100.44	1926.07	7.08%
CBS	Forest	42.40	128.10	2003–2008	1338.84 ± 108.86	1437.48	7.37%
HZ	Forest	51.78	123.02	2010	962.75	1133.02	17.69%
XSBN	Forest	21.95	101.20	2003–2008	2342.67 ± 174.10	2020.26	−13.76%
ALS	Forest	24.53	101.02	2009–2010	1848.33	1791.98	−3.05%
LS	Forest	45.33	127.67	2004	1351	1472.64	9.00%
YY	Forest	29.53	112.86	2006–2007	1974.8	1948.72	−1.32%
AQ	Forest	30.47	116.99	2006–2007	1859.2	1948.23	4.79%
XP	Forest	33.35	113.91	2010	1288.1	1425.93	10.70%
YC	Cropland	36.83	116.57	2003–2008	1746.62 ± 139.35	1250.08	−28.43%
TYC	Cropland	44.57	122.92	2004–2006	444.33 ± 56.36	356.00	−19.88%
WS	Cropland	36.65	116.05	2007–2008	1838	1208.18	−34.27%
NM	Grassland	44.53	116.67	2004–2008	231.66 ± 111.13	329.27	42.13%
XLD	Grassland	43.55	116.67	2006	294.44	392.83	33.42%
SJY	Grassland	34.35	100.55	2006	480.27	701.13	45.99%
DTZ	Wetland	31.58	121.90	2005	1512.63	1704.31	12.67%
PJ	Wetland	41.13	121.90	2005	1298.16	1371.52	5.65%

Note: RPE means Relative Predictive Error and is calculated as (LPJ_sim-GPP_obs)/ GPP_obs ×100%.

and Figure 8, the LPJ model performed well in simulating forest GPP, and the average error was 8.14% across 10 forest sites. At the DHS site located in Guangdong province, the vegetation type is subtropical evergreen broad-leaved forest, and the GPP here was overestimated by 42.91% compared with the observed value. In contrast, in the XSBN site of south Yunan, the vegetation is tropical rain forest, and the simulated GPP here was underestimated by 13.76%. The simulated GPP was obviously lower tower measurements at all cropland sites since cropland is not explicitly defined as a PFT, and was treated as grassland in the LPJ model. In China, intensive management, such as fertilizer application, the introduction of new varieties, and irrigation, enhance GPP. These mechanisms were not represented in the LPJ model. In grasslands and wetlands, the simulated GPP was overestimated. The mismatch of model spatial resolution and footprints of flux towers was one of the major reasons for the discrepancy between simulated and observed GPP. The spatial resolution in our simulations was 0.5° × 0.5°, while the footprints of flux towers were just a few square kilometers. The human-induced land cover change and spatial heterogeneity of climate and soil within 0.5° × 0.5° grids were not considered in the current study, which would at least partially explain the departure of simulated GPP from observations. Overall, the LPJ model was able to capture GPP variations among different sites (Figure 6).

3.4. Changes of NPP in the Future

The total simulated NPP under different climate scenarios from 2017 to 2100 are shown in Table 3. Overall, the total NPP of China will continue to increase in the future under different climate change scenarios with increasing CO₂ concentration and temperatures. The total NPP under the A2 and RCP85 scenarios would be significantly higher than those under the B1 and RCP45 scenarios, and the magnitude of deviation between them would be more remarkable after 2050. The average CO₂ concentrations and temperatures under the A2 and RCP85 scenarios would be approximately 18.4 ppm and 20.2 ppm and 0.18 °C and 0.31 °C, respectively, higher than those under the B1 and RCP45 scenarios during 2011–2050. On average, total NPP would be just 0.01 Pg C yr⁻¹ and 0.04 Pg C yr⁻¹ higher under A2 and RCP85 scenarios than that under B1 and RCP45 scenarios during 2011–2050. When the gaps of CO₂ concentrations and temperatures would increase to 175.0 ppm and 236.6 ppm and to 1.79 °C and 1.69 °C during 2051–2100, the gaps of the total NPP would widen to 0.48 Pg C yr⁻¹ and 0.66 Pg C yr⁻¹. Especially in the final period (2091–2100), the average total NPP in the A2 and RCP85 scenarios would reach 4.90 Pg C yr⁻¹ and 5.10 PgC yr⁻¹, which would be 0.70 Pg C yr⁻¹ and 1.20 PgC yr⁻¹ larger than that under the B1 and RCP45 scenarios.

As shown in

Table 5. Averages of total NPP simulated using future CO₂ concentration and projected climates by two climate models under different scenarios.

Periods	A2		RCP85		B1		RCP45
	NCAR	CNRM	NCAR	CNRM	NCAR	CNRM	NCAR	CNRM
2010s	3.4	3.4	3.3	3.5	3.3	3.3	3.4	3.4
2020s	3.7	3.4	3.6	3.7	3.4	3.4	3.4	3.5
2030s	3.8	3.7	3.6	3.8	3.7	3.5	3.7	3.6
2040s	3.9	3.8	3.7	4.0	4.0	3.8	3.8	3.9
2050s	4.1	4.0	3.9	4.0	4.2	4.0	3.9	3.8
2060s	4.5	4.2	3.9	3.9	4.5	4.1	4.0	4.1
2070s	4.6	4.3	4.0	4.0	4.7	4.6	4.0	3.8
2080s	4.9	4.6	3.9	4.2	4.9	4.9	4.0	4.1
2090s	5.0	4.8	4.2	4.2	5.1	5.1	3.9	3.9

and Figure 9, there are differences in annual total simulated NPP using the climates projected by different climate models (NCAR and CNRM). The average NPP simulated using the climates projected by two climate models during the periods of 2011–2010, 2046–2055, and 2091–2100 was used to indicate the NPP changes in the early, middle, and late stages, respectively. The results showed that forests would make significant contributions to increases in NPP in the future. In most temperate forests, the NPP would experience a sustained increase. The NPP in tropical and subtropical forests would also increase, but to a lesser extent.

As shown in Figure 9, there are small differences in the spatial pattern of NPP simulated using projected climates under different scenarios during 2011–2020. During 2046–2055, the NPP pattern would vary more significantly under AR5 scenarios than under AR4 scenarios. NPP in temperate forests would increase obviously overall, NPP in the Greater and Lesser Xing’an, Changbaishan, and Hengduan mountain areas would increase from 500–600 g C m⁻² yr⁻¹ during 2011–2020 to 600–800 g C m⁻² yr⁻¹ during 2046–2055. NPP in subtropical forests in the southeastern coastal areas of China would also increase from 600–700 g C m⁻² yr⁻¹ during 2011–2020 to 700–900 g C m⁻² yr⁻¹ during 2046–2055. However, the NPP increases in the Northeast Plain, North China Plain, and middle-lower Yangtze River plain would be small from the period of 2011–2020 to the period of 2046–2055. The patterns of simulated NPP under four different scenarios would show distinguishable differences during 2091–2100. The NPP in most forests would continue to increase under the A2 and RCP85 scenarios, but would have no increase or even decline under the B1 and RCP45 scenarios.

3.5. Influence of Future CO₂ Increase and Climate Change on NPP

Our simulations indicated that continuous increase in CO₂ concentrations would significantly enhance terrestrial NPP in China (Figure 10). The average CO₂ concentration was 393.24 ppm during 2007–2016, and would increase to 817.15 ppm and 894.37 ppm under the A2 and RCP85 scenarios and to 547.67 ppm and 535.83 ppm under the B1 and RCP45 scenarios during 2091–2100. In the context of the above CO₂ concentrations without corresponding climate change, the total of simulated NPP in China would reach 5.08 Pg C yr⁻¹ and 5.23 Pg C yr⁻¹ under the A2 and RCP85 scenarios during 2091–2100, respectively. These values are 1.68 Pg C yr⁻¹ (49.49%) and 1.83 PgC yr⁻¹ (53.86%) higher than the average NPP during 2007–2016 (3.40 Pg C yr⁻¹). Under the B1 and RCP45 scenarios, CO₂ concentration would increase at smaller magnitudes. As a consequence, the increases of NPP driven by sole CO₂ increase would be 0.82 PgC yr⁻¹ (24.02%) and 0.67 PgC yr⁻¹ (19.57%) by the end of the 21st century, respectively, much smaller than the increases under A2 and RCP85 scenarios.

The NPP simulated with only increase in CO₂ concentration was compared with the estimates that combined the effects of climate change and increasing CO₂. The total NPP (only CO₂ change) under the A2 and RCP85 scenarios would be 3.03% (0.15 Pg C yr⁻¹) and 2.54% (0.13 Pg C yr⁻¹) higher than couple scenarios during 2091–2100, respectively. The corresponding values would approximately 0.60% (0.03 Pg C yr⁻¹) and 3.58% (0.14 Pg C yr⁻¹) under B1 and RCP45 scenarios. The warmer climate in A2 and RCP85 scenarios induced slightly larger reductions of total NPP than B1 and RCP45 scenarios. As a whole, the warm climate in the future had a slightly negative influence on the NPP in China.

There are no obvious trends in the total NPP of China simulated with projected future climate under the A2, RCP85 and B1, RCP45 scenarios and CO₂ concentration value in 2016 (Figure 10). Compared with the average NPP (3.4 PgC yr⁻¹) in the past ten years (2007–2016), the solely future climate change would drive slight changes in total NPP with an increase of 0.78% (0.026 Pg C yr⁻¹) and decreasing by 1.64% (0.056 Pg C yr⁻¹) under the A2 and RCP85 scenarios respectively, and an increase of only 1.81% (0.062 Pg C yr⁻¹) and decrease of 0.12% (−0.004 Pg C yr⁻¹) under the B1 and RCP45 scenarios during 2051–2100, respectively. Previous studies have indicated that precipitation was the main limiting factor of increasing NPP, and that temperature was also an important assisting factor. In years of abundant (deficient) precipitation, the rise in temperature caused an increase (decrease) in NPP [19]. Interannual and spatial variations in precipitation in different regions of China led to trends and variability of NPP. Precipitation in Northwest China has increased with fluctuations since 1961, and the climate in Qinghai-Tibet Plateau area has become warmer and wetter [73]. The positive effect on NPP caused by increasing precipitation in the semi-arid ecosystems of China might offset the negative effect from warmer climates in other areas.

Compared with the estimates that combined the effect of climate change and concurrent increasing CO_2, the simulated total NPP with only climate change (CO₂ kept at the value in 2016) under the A2 and RCP85 scenarios would decreased by 1.44 Pg C yr⁻¹ (29.28%) and 1.68 Pg C yr⁻¹ (32.97%) during 2091–2100 (Figure 10a,c). The corresponding values would be 0.64 Pg C yr⁻¹ (15.31%) and 0.60 Pg C yr⁻¹ (15.40%) under the B1 and RCP45 scenarios (Figure 10b,d). Therefore, the future increase of NPP in China would be dominantly contributed to by the fertilization effect of CO₂ enrichment.

4. Conclusions

The LPJ model was employed to analyze the historical and future spatial and temporal patterns of NPP and assess the contributions of climate change and CO₂ enrichment to future NPP trend in China. The main findings of this study are:

(1)

Total NPP of China’s terrestrial ecosystem increased from 2.8 to 3.6 Pg C yr⁻¹ over the period of 1961–2016, with an annual average of 3.1 Pg C yr⁻¹. Southwest China acted as the largest contributor to national total NPP among the six geographical regions of China, while Northwest China played the smallest role. Total NPP exhibited gradually increasing trends in Northwest China, Southwest China, South-central China, and East China.

(2)

National total NPP exhibited large interannual variability. NPP anomalies in China’s terrestrial ecosystems were strongly associated with the El Niño-Southern Oscillation (ENSO). Overall, the national total of NPP declined remarkably in the EP EI Niño year while increasing in the CP EI Niño year. The national total NPP often increased in the La Niña year. However, continuous La Niña events might induce NPP to decrease. Southwest China made the largest contribution to the interannual variability of national total NPP, the next is South-central China.

(3)

The total NPP in the A2 and RCP85 scenarios would be obviously larger than those in the B1 and RCP45 scenarios. During the period from 2091 to 2100, the average total NPP under the A2 and RCP85 scenarios would reach 4.9 and 5.1 Pg C yr⁻¹ respectively, higher than 4.2 and 3.9 Pg C yr⁻¹ under the B1 and RCP45 scenarios. The future increase of NPP in China would be dominantly contributed to by the fertilization effect of CO₂, while the climate change would have slightly negative effects on NPP.