Effects of climate warming on net primary productivity in China during 1961–2010

Abstract The response of ecosystems to different magnitudes of climate warming and corresponding precipitation changes during the last few decades may provide an important reference for predicting the magnitude and trajectory of net primary productivity (NPP) in the future. In this study, a process‐based ecosystem model, Carbon Exchange between Vegetation, Soil and Atmosphere (CEVSA), was used to investigate the response of NPP to warming at both national and subregional scales during 1961–2010. The results suggest that a 1.3°C increase in temperature stimulated the positive changing trend in NPP at national scale during the past 50 years. Regardless of the magnitude of temperature increase, warming enhanced the increase in NPP; however, the positive trend of NPP decreased when warming exceeded 2°C. The largest increase in NPP was found in regions where temperature increased by 1–2°C, and this rate of increase also contributed the most to the total increase in NPP in China's terrestrial ecosystems. Decreasing precipitation depressed the positive trend in NPP that was stimulated by warming. In northern China, warming depressed the increasing trend of NPP and warming that was accompanied by decreasing precipitation led to negative changing trends in NPP in large parts of northern China, especially when warming exceeded 2°C. However, warming stimulated the increase in NPP until warming was greater than 2°C, and decreased precipitation helped to increase the NPP in southern China.

in 1996 first accepted the maximum threshold of 2°C warming for the earth's climates (Schleussner et al., 2016); since then, the threshold of 2°C warming had become a common standard and has been accepted by parties of the United Nations Framework Convention on Climate Change (UNFCCC, 2012). Representative of many countries attending the 2015 climate conference in Paris signed an agreement to set a goal of limiting global warming to less than 2°C when compared with preindustrial era (UNFCCC, 2015). However, the threshold of 2°C warming is not a scientific prediction (Jaeger & Jaeger, 2011), and researchers need to conduct in-depth discussions related to how terrestrial ecosystems would respond to the climatic warming exceeding 2°C. At present, the average global temperature has increased almost 1°C since preindustrial times with the large regional difference in magnitude of warming (IPCC, 2014). The response of different ecosystems to different magnitudes of warming is far from clear, especially in China. This may hinder our future ability to manage the ecosystems in response to climatic warming in different regions.
Net primary productivity (NPP) represents the production of gross photosynthesis minus autotrophic respiration and is considered as a critical indicator for researchers who analyze the effects of climate change on terrestrial ecosystems (Ito, 2011). Quantifying the interannual variability in NPP would help us to understand the terrestrial carbon dynamics and underlying mechanisms in response to climate change (Twine & Kucharik, 2009). Numerous studies have demonstrated that warming can stimulate plant growth and carbon uptake (Delpierre et al., 2009;Oberbauer et al., 2007;Sullivan, Arens, Chinmner, & Welker, 2008;Wu, Dijkstra, Koch, Peñuelas, & Hungate, 2011). However, increased air temperature also stimulates autotrophic respiration in plants (Burton, Melillo, & Frey, 2008;Heimann & Reichstein, 2008;Knorr, Prentice, House, & Holland, 2005). Therefore, our knowledge of how NPP might respond to different magnitudes of warming is far from clear (Niu et al., 2008;Wu et al., 2011), because this response represents an integrated effect of changes in temperature and water status on photosynthesis and respiration (Angert et al., 2005;Ciais et al., 2005;Kang, Kimball, & Runing, 2006;Sullivan et al., 2008). The effects of warming on NPP will be either enhanced or weakened, depending on whether precipitation is decreasing or increasing correspondingly (Chen, van der Werf, de Jeu, Wang, & Dolman, 2013;Kang et al., 2006;Wu et al., 2011). In addition, site experiments have demonstrated that the effects of temperature increase and altered precipitation vary widely and are highly dependent on the ecosystem types and climate zones involved (Davi et al., 2006;Nemani et al., 2003;Niu et al., 2008;Wu et al., 2011). In the middle and high latitudes of the Northern Hemisphere, plant growth is mainly limited by temperature (Schwartz, Ahas, & Aasa, 2006;Wang et al., 2011), and recent climatic warming has enhanced ecosystem productivity and carbon uptake (Delpierre et al., 2009;Deng & Chen, 2011;Nemani et al., 2003;Potter et al., 2003). While in arid and semiarid areas, the ecosystem productivity in response to warming climate is mainly mediated by precipitation (Fang, Piao, Tang, Peng, & Ji, 2001;Knapp et al., 2002;Mitchell & Csillag, 2001;Niu et al., 2008).
The complex topography of China contributes to producing very diverse ecosystems, such as with the effect of the uplift of the Tibetan Plateau and the varied climate regimes from the East Asian monsoons to the western arid (Ni, 2011). These very special environmental characteristics make China a region that is particularly vulnerable to climate change (Ni, 2011), sparking concerns over the response of Chinese terrestrial ecosystems to climate change (Cao et al., 2003;Ju, Chen, Harvey, & Wang, 2007;Mu, Zhao, Running, Liu, & Tian, 2008;Peng & Apps, 1997;). Some studies have shown that China's terrestrial NPP has increased in response to increases in temperature, altered precipitation, and elevated CO 2 concentrations (Cao et al., 2003;Fang et al., 2003). However, responses of NPP to climate change could be remarkably diverse because of the high level of land surface heterogeneity in China and regional differences in climate change. Understanding how terrestrial NPP responds to historical warming trends and altered precipitation is very important for researchers tasked with predicting the effects of future climate change (Wang et al., 2011).
Here, we used the Carbon Exchange between Vegetation, Soil and Atmosphere (CEVSA), a process-based model, to quantify the effects of warming on NPP in China during 1961-2010. Our main objectives were to clarify (1) whether there is a temperature threshold above which NPP no longer increases with warming; (2) whether changes in precipitation modify the response of NPP to warming; (3) if a temperature threshold exists, whether it differs among ecoregions and the effects of altered precipitation.

| The CEVSA model
The CEVSA model simulates carbon synthesis as well as water and energy exchange among vegetation, soil, and atmosphere and also models the interactions between ecosystem and environmental conditions. Detailed information on model structure and algorithms can be found in our previous publications (Cao, Prince, & Shugart, 2002;Cao & Woodward, 1998a,b;Cao et al., 2003;Gu et al., 2007;Tao et al., 2007). In CEVSA, temperature influences the photosynthesis, respiration, and stomatal behaviors. Soil water content, which is determined by the difference between precipitation and evapotranspiration in the soil, influences the ecophysiological processes by affecting the stomatal conductance.

| Data sources
The meteorological datasets used to drive the CEVSA model include mean air temperature, precipitation, relative humidity, and cloud cover at a 10-day time-step. All meteorological datasets covering from 1954 to 2010 were provided by the National Meteorological Information Center of China, including 756 meteorological stations scattered across China. A 0.1° × 0.1° gridded meteorological dataset were obtained from the interpolation of the station observation data using ANUSPLINE (Hutchinson, 1989). Annual atmospheric CO 2 concentrations were downloaded from the CO 2 • earth website (https:// www.co2.earth/). The Chinese soil texture classification system was adopted, and the 1:14,000,000 soil texture map of China (The Institute of Soil Science, 1986) was digitized. The vegetation distribution map was derived from 1 km resolution Global Land Cover 2000 database (European Commission, Joint Research Centre, 2003). In order to be consistent with other input data, we aggregated the GLC2000 to obtain the 10 km land cover data by using ArcInfo 10.2. Each grid cell in the new vegetation map was assigned the value of land cover type that had the largest fraction in a 10 km grid cell.
Validation data came from three sources: (1) the recalculated National Forest Inventory (NFI) dataset (Luo, 1996;Wang, Zhou, Jiang, & Yang, 2001;Zhao & Zhou, 2004), (2) aboveground net primary productivity (ANPP) observed in Chinese grassland (Guo et al., 2012;Hu et al., 2010), and (3) cropland NPP recalculated based on agricultural statistical data. The recalculation of ANPP to NPP involves parameters such as ratio of aboveground and belowground biomass, turnover, etc. (Fan et al., 2008(Fan et al., , 2009. Errors within all parameters would be propagated into the estimation of NPP. Therefore, we compared the observed ANPP with modeled NPP directly. County agricultural statistical data include planting area and yield of main food crops. The yield can be converted into NPP based on the water content of grain and a harvest index (Lobell, Hicke, Asner, Field, & Los, 2002;Yan, Liu, & Cao, 2007), and these data can be used to validate the regional simulation of ecosystem models. Here, for the purpose of validation, the crop NPP was estimated using Equation (1)

| Model simulations
The CEVSA model was run at a 0.1° × 0.1° spatial resolution for the entire area of China with a 10-day time-step. First, we ran the model by using a 30-year averaged climatic data ) and a fixed CO 2 concentration level in 1954 until the model reached equilibrium status and the initial state parameters were obtained. Then, the simulation was conducted using time-variant climate and atmospheric CO 2 data for the period 1954-2010. In addition to including the all combined simulation data such as climate change and elevated CO 2 concentration, we also conducted three single-factor simulations to reveal the relative effects of temperature, precipitation, and CO 2 concentration on long trend of NPP. These three simulations were completed using (1) only air temperature change data; (2) only precipitation change data; and (3) only CO 2 concentration change data. The modeling results from 1961 to 2010 were used in our analyses. The run from 1954 to 1960 was designed to eliminate the effect of initial status on simulations for the period 1961-2010 (Cao et al., 2002).

| Trend and statistical analysis
We applied a least-squares linear regression model to determine trends in temperature, precipitation, and NPP during the period 1961-2010 using Equation (2): where Y is the mean annual temperature (MAT), mean annual precipitation (MAP), or NPP; X is the year; a is the intercept; b is the slope, which represents the trend of variables during last 50 years and indicates the absolute change per year; and e is the residual error.
Because the study period comprises 50 years, the total change (TC) of MAT, MAP, and NPP was calculated by multiplying the slope b by 50 years, and the percentage change (PC) as TC divided by the 50year average value of the parameters (temperature, precipitation, or NPP). All significant tests for trend analyses are t test.

| Model validation
We have validated the CEVSA model against multiscale observations/ measurements in previous studies Tao et al., 2007). In the present study, we compared the CEVSA-estimated average NPP with those from process-based as well as remote sensing models. Second, the CEVSA model does not consider the effects of all kinds of natural and human disturbance, such as nitrogen deposition, fire, harvest, management, on ecosystems. For example, the CEVSA model does not quantify the management in cropland, including irrigation, fertilization, and farming system, so the model may underestimate NPP in some areas, especially in irrigated cropland. Third, some errors were come from the recalculation of observation and statistical data.
All field NPP used as validation data were not measured directly. It was calculated based on variables measured directly, such as tree height, diameter at breast height, biomass, and statistical grain yield. All the calculations would bring errors into the validated data.

| The responses of total NPP to different magnitudes of climatic warming and the effects of changing precipitation
Warming resulted in an increase in NPP in China with or without the effects of CO 2 fertilization, while decreasing precipitation resulted in a decline in NPP (Figures 2 and 3 The percentage change of NPP increased with increasing magnitude of warming (Table 1). The largest increase in NPP was found in regions where temperature increased 1-2°C regardless of any increase or decrease in precipitation, which contributed more than The second assessment report of the IPCC (Houghton et al., 1996) integrated the research results of more than 1,000 scientists and pointed out that the risk of serious negative effects from climate change would increase significantly if the average global temperature increased over 2°C above preindustrial levels. In 1996, the European Union initially proposed 2°C as a red line for climatic warming that should not be exceeded (European Environment Agency, 1996), and climate change scientists of the European Union released "The 2°C target" evaluation report in 2008 (EU Climate Change Expert Group, 2008). This report pointed out that human beings might not suffer from the expected effects of climate change on the economy, society, and environment if the increase in the average global temperature was not held below 2°C (Jaeger & Jaeger, 2011). No evidence supports the concept that human society will be able to adapt to the climate change if the increase in average global temperature is above 3°C or 4°C (Knutti, Rogelj, Sedláček, & Fischer, 2016;Kypreos & Magné, 2013). However, in the present study, the total NPP is not expected to decrease even if the temperature increased by over Therefore, 2°C did not become the threshold of limiting ecosystem functions in whole China even though the rate of increase in NPP declined at higher temperature.

| The different responses of NPP to warming in different ecoregions and the effects of altered precipitation
The total NPP in China has exhibited an increasing trend with obvious spatial variations (Table 2, Figure 4). Over the past 50 years, North   (Dai, Zhang, Wang, Guo, & Wang, 2010;Dan, Ji, & Ma, 2007). All these indicate that ecosystems have different responses to changes in increased temperature and precipitation in different regions and climatic zones. Therefore, climate change research should not promote the use of a single threshold or a target for warming in a global context.
A better approach would be to establish regional thresholds for warming based on the different responses of various ecosystems to warming and changes in precipitation in different regions and climatic zones.
Therefore, models such as CEVSA could be a useful tool to determine reasonable regional warming thresholds. Falloon et al. (2007) suggested that while the pattern of warming is widespread in all areas, regional disparities between warming and NPP may be related to variations in precipitation, which shows strong variations regionally. Generally, a decrease in precipitation did not influence the response of an increase in NPP to warming (Table 1). However, the change in precipitation played different roles in responses of NPP to warming in different ecoregions (Table 2).
Temperature had been considered to be the principal limiting factor on vegetation growth in temperate and relatively cold areas (Nemani et al., 2003). Our results suggest that warming does not always result in an increase in NPP, even in areas where temperature was considered as the primary controlling factor for NPP. Recent research (Boisvenue & Running, 2006;Nemani et al., 2003;Running et al., 2004) and our results have shown that the vegetation growth across the entire northern China was limited by precipitation. Water availability plays a dominant role in plant growth and net ecosystem productivity in some regions (Niu et al., 2008;Potts, Huxman, Enquist, Weltzin, & Williams, 2006;Weltzin et al., 2003). In Northeast and North China as well as Inner Mongolia, NPP showed a decreasing trend in regions where temperature increased by more than 1°C while precipitation decreased (Table 2). Obviously, a combination of warming and decreased precipitation tend to be colimiting on trends of NPP in northern China, while warming enhanced the stress of water deficit induced by decreased precipitation on plant growth. In addition, the water shortage induced by warming could also reduce the increasing trend of NPP even when precipitation increased .
In southern China, which receives the greatest amount of precipitation (more than 1,000 mm), decreased precipitation does not necessarily result a decline in NPP and may even exacerbate the increase in NPP (Table 2 and Figure 4c). The stimulation of decreased precipitation on NPP in southern China may be originated from an increase in radiation induced by decreased precipitation; this occurred because the vegetation growth in southern China is limited by photosynthetically active radiation received by the canopy (Nemani et al., 2003). Photosynthetically active radiation was calculated using the CEVSA model, which was determined by cloud cover; therefore, it was influenced by the change of precipitation. In fact, the change of surface solar radiation was influenced by many factors, such as sunshine duration, cloud cover, the presence of aerosols. Various studies suggested a decrease in surface solar radiation has occurred in southern China during the past decades, with a partial recovery more recently (Wang, Huang, & Zhang, 2009;Zheng, Guan, Cai, Wu, & Liu, 2011); however, the level of diffuse photosynthetically active radiation increased in all of China, especially in southern China (Ren, He, Zhang, & Yu, 2014).
In Northwest China, the increase in precipitation has been associated with the largest percentage change in NPP. Some evidence from satellite data and other sources has also confirmed the positive trends in NPP during the past 30 years in the northwestern part of China (Dai et al., 2010;Dan et al., 2007). Previous studies also confirmed the climate evolved from warm-dry to warm-wet in northwest China (Liu, Feng, Ma, & Wei, 2009;Shi et al., 2003). Increased precipitation stimulated plant growth which was limited mainly by water scarcity in northwest China, so the vegetation index was obviously increased and the number of days experiencing sand-dust storm decreased (Shi et al., 2003).

| Uncertainty analysis and future research needs
Uncertainties are inevitable in modeling regional NPP and its response to global changes (Ito, 2011). The sources of uncertainty derive mainly from three aspects: (1) the input data, parameters, and the spatial resolution, including meteorological data, land use/land cover dataset, and soil data. Disturbances may also bring uncertainties into the simulated results.
An accurate estimation of the carbon budget of the terrestrial ecosystem is important during the evaluation of the effect and relative contributions of multiple environmental factors on productivity and carbon accumulation in terrestrial ecosystems (Ren et al., 2007;Tian et al., 2011;Wang et al., 2011). Incorporating the effects of all natural and anthropogenic factors into the model simulation will require greater effort related to the clarification and quantification of all these processes in the models. With the further development of the ecosystem model, it is necessary to analyze the effects of all these factors on ecosystem and their relative contributions to the carbon cycle. In the future, we may identify the uncertainty range of modeled carbon fluxes at a regional scale by using data-assimilation techniques (Cao, Yu, Liu, & Li, 2005;Ito, 2011;Rayner et al., 2005;;Zhang et al., 2012).
In addition, the multiple model intercomparison provides another method that can be used to evaluate the uncertainties in model simulations of NPP in response to climate change and variability.
In the present study, we only examined the responses of NPP to warming and altered precipitation in different ecoregions. We also need to understand the mechanisms that control the different responses of NPP to warming and changing precipitation by conducting simulation experiments as well as by making observations in different ecosystems and climate zones. Nevertheless, previous studies have demonstrated the stimulation of warming on productivity (Delpierre et al., 2009;Oberbauer et al., 2007;Sullivan et al., 2008;Wu et al., 2011) and the varied impact of altered precipitation (Chen et al., 2013;Kang et al., 2006;Wu et al., 2011

CONFLICTS OF INTEREST
The authors declare no conflict of interest.