By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
We determined the historical change in soil organic carbon (SOC) stocks from long-term field trials that represent major soil types and climatic conditions of northern China. Soil carbon and general circulation models were validated using these field trial data sets. We then applied these models to predict future change in SOC stocks to 2100 using two net primary production (NPP) scenarios (i.e., current NPP or 1% year−1 NPP increase). The conversion rate of plant residues to SOC was higher in single-cropping sites than in double-cropping sites. The prediction of future SOC sequestration potential indicated that these soils will be a net source of carbon dioxide (CO2) under no fertilizer inputs. Even when inorganic nutrients were applied, the additional carbon input from increased plant residues could not meet the depletion of SOC in parts of northern China. Manure or straw application could however improve the SOC sequestration potential at all sites. The SOC sequestration potential in northern China was estimated to be −4.3 to 18.2 t C ha−1 by 2100. The effect of projected climate change on the annual rate of SOC change did not differ significantly between climate scenarios. The average annual rate of SOC change under current and increased NPP scenarios (at 850 ppm CO2) was approximately 0.136 t C ha−1 yr−1 in northern China. These findings highlight the need to maintain, and where possible increase, organic carbon inputs into these farming systems which are rapidly becoming inorganic fertilizer intensive.
Soil organic carbon (SOC) sequestration in agricultural soil is directly affected by anthropogenic activities and climate change; both can alter net primary production (NPP) and organic matter decomposition [Yan et al., 2010]. Carbon inputs to soil can be increased in arable farming systems where (i) crop production has not yet achieved maximum water use efficiency and/or where irrigation is available, (ii) nutrient limitations are overcome with fertilizers, and (iii) where additional organic sources are applied; potentially converting agricultural soil into a net carbon store. The capacity for further SOC sequestration in agricultural soils is estimated at 140 to 170 Pg C [Lal, 2004], which is more than 10% of the existing global terrestrial SOC store. As such the Intergovernmental Panel on Climate Change [IPCC,2007a] has identified SOC sequestration as a cost-effective and environmentally friendly option to mitigate increasing atmospheric carbon dioxide (CO2).
China has more than 20% of the world population and 8% of the total world arable land [Food and Agriculture Organization,2010]. Agriculture was responsible for 15 to 18% of the total greenhouse gas emissions in China during 2007; with contributions from agricultural land being 43 to 47% from methane (CH4), 33 to 34% from nitrous oxide (N2O) and 19 to 23% from CO2 [Guo and Zhou, 2007; Lin et al., 2012]. Crop production is the major land use occupying an area of 122 million ha [National Bureau of Statistics of China, 2012] and accounts for 7 to 12% of the SOC stock under arable production systems worldwide [Schlesinger, 1999]. Furthermore, additions of organic waste to agricultural soil have occurred for thousands of years (e.g., Loessial soil) in China, aiding the stabilization of SOC [Liang et al., 2012; Zhao et al., 2008]. However, China is currently the largest consumer of inorganic fertilizer in the world, accounting for 90% of the global increase in use [Liu and Diamond,2005]. Through increased inorganic fertilizer use, adoption of modern plant cultivars and increased areas of irrigation, crop grain yields in China have approximately doubled between 1980 (wheat 1.9 t ha−1, maize 3.1 t ha−1) and 2010 (wheat 4.7 t ha−1, maize 5.4 t ha−1) [National Bureau of Statistics of China, 2012]. However, during this 30 year period SOC stocks in agricultural systems employing common management practices (i.e., tillage, inorganic fertilizers, straw removal, and no animal manure application) have only changed slightly; with a general decrease in arid/semiarid regions and increase in humid/semihumid regions [Sun et al., 2010; Yan et al., 2010; Yu et al., 2009]. Reported SOC changes in agricultural soils vary (−2.0 to 0.6% yr−1) [Yan et al., 2010] with an average SOC sequestration rate of 21.9 Tg C yr−1 between 1980 and 2000 [Sun et al., 2010]; equivalent to 0.21% of the estimated 10,070 Tg C stored in upland soils in China [Xie et al., 2007].
A change in crop growth will alter the carbon input to soil from plant residues, which is typically the main source of new SOC in arable land (unless manure is applied). Net primary production (NPP) is affected by climatic variables such as temperature, precipitation, atmospheric CO2, and the length of crop growth period [Ye et al., 2013]. It is reported that the yields of wheat and maize have responded negatively to warming at the global scale, although the impact on other crops (e.g., rice) is less certain [Lobell and Field, 2008]. Wan et al.  modeled future changes in SOC stocks for upland soils in China based on historical plant carbon input rates without consideration of manure or straw application. They predicted that SOC would decrease in most upland soils, especially in northern China. No consideration was given in their future predictions to increases in NPP based upon improved plant breeding and/or adoption of “best practice” agronomic management. It is expected that the rate of straw retention in China could increase from 40% [Gao et al., 2002] to 90% [Sun et al., 2010], and that no-tillage practices could be extended to 50% of the nations cropland by 2050; with organic manure inputs likely to remain the same (110 Tg C yr−1) [Li et al., 2003]. Based on a historical crop NPP increase of approximately 12 Tg C yr−1 from 1960 to 1999 [Huang et al., 2007] and a future increase in NPP of approximately 6 Tg C yr−1, Sun et al.  calculated a further 55% increase in NPP by 2050 for China (equivalent to a 1% annual NPP increase from 2000 to 2050).
In our study, we measured the historical change in SOC stocks from long-term field trials for the major soil types and climatic conditions of northern China and then modeled future changes in SOC stocks using different climate and carbon input scenarios. We wanted to quantify the difference in current SOC sequestration rates when organic residues are added to soils compared to common management practice. We also wanted to determine how climate change would alter SOC stocks and if future SOC stocks could be increased with inorganic fertilizer alone or if organic residue/manure inputs will be required. Our specific aims were to (i) measure the historical change in SOC stocks from eight long-term fertilizer trials (15–28 years; treatments of inorganic fertilizers and/or manure/straw application) that represent the major soil types and climatic conditions of northern China, (ii) use the historical climate and soil data to validate global climate models and the RothC carbon model for northern China, and (iii) model future changes in SOC stocks to 2100 under two plant carbon input scenarios (no change to NPP or an annual 1% NPP increase).
2 Methods and Materials
2.1 Field Research Sites
Our study consisted of eight long-term (i.e., 15–28 years) experimental sites on upland soils in the northern regions of China (Figure 1). The climate at these sites ranged from arid to semihumid and from mild to warm temperate. Annual mean temperature ranged from 4.5°C in the northeast to 14.5°C in the western central region, annual precipitation ranged from 127 mm in the northwest to 832 mm in the central eastern region, and evaporation was 1 to 18 times greater than precipitation (Table 1). The annual cropping rotation was either single or double crops and consisted of various crop sequences of predominately wheat or maize (Table 1). The four single-cropping trial locations were significantly cooler (4.5–8.0°C) and drier (127–540 mm) compared to the four double-cropping sites (11.0–14.5°C and 575–832 mm) (P < 0.05). The field research sites covered the major arable land soil types (Table 2) in northern China: Haplic Calcisol (Grey Desert soil, Urumqi site), Luvic Phaeozems (Black soil, Gongzhuling site), Anthrosol (Irrigated Desert soil, Zhangye site), Calcic Kastanozem (Dark Lossial soil, Pingliang site), Haplic Luvisol (Brown Fluvo-aquic soil, Changping site), Calcaric Cambisol (Fluvo-aquic soil, Zhengzhou and Xuzhou sites), and Cumulic Anthrosol (Lossial soil, Yangling site).
Table 1. Location and Climate at Each of the Experimental Field Trials
MT = mild temperate; SH = semihumid; SA = semiarid; A = arid; WT = warm temperate.
SC-M = single-cropping, maize; SC-MWW = single-cropping, 1 year maize followed by 2 years wheat; SC-MMWWWW = single-cropping, 2 year maize followed by 4 years wheat; DC-MW = double-cropping, maize/wheat annually.
Table 2. Initial Soil Physical and Chemical Properties at the Long-Term Experimental Sites
Long-term field plots (n = 8 for field trial location, n = 1–3 for within trial plot replicates) varied in size: 33 m2 at Zhangye and Xuzhou; approximately 200 m2 at Changping, Yangling, and Pingliang and approximately 400 m2 at Urumqi, Gongzhuling, and Zhengzhou. Inorganic nitrogen (N), phosphorus (P), and potassium (K) fertilizers were applied as urea, calcium superphosphate, and potassium chloride, respectively. There were three fertilizer treatments common to each field trial: no fertilizer (Control), inorganic fertilizer only (NP or NPK), and inorganic fertilizer plus manure (M) addition (NP + M or NPK + M). In addition, one additional fertilizer treatment was sampled depending on the site: manure only (M) at Zhangye and Xuzhou, and inorganic fertilizer plus straw (S) return (NP + S or NPK + S) at the other six sites. The total N applied (inorganic + organic) was equal for the NPK and NPK + M treatments at five sites but was higher in the NPK + M treatment at Zhangye, Pingliang, and Xuzhou due to an additional manure N application (Table S1 in the supporting information). For the NP + M and NPK + M treatments 30% of the total N was from the inorganic fertilizer, while the remainder was organic manure N.
Organic carbon input into soil included plant residues (plant roots + stubble) plus any treatment addition of organic manure or crop straw return. The average annual carbon inputs from manure, straw, and plant residues at each site are reported in Table S2. All aboveground biomass (not including stubble) was removed from the plots at harvest; the straw was returned to plots in the NP + S and NPK + S treatments. The average C/N ratio of straw was 67/1 for wheat and 50/1 for maize. The carbon input from roots was estimated by the ratio of belowground biomass to aboveground biomass. Total plant biomass carbon was proportioned to roots as 30% for wheat and 26% for maize [Li et al., 1994], and we assumed that 75.3% and 85.1% of the total root biomass were in the surface 20 cm of soil for wheat [Fang et al., 2011; Lu and Xiong, 1991; Ma, 1987; Miao et al., 1989], and maize [Li et al., 1992; Liu and Song, 2007], respectively. The contribution of carbon input from stubble was estimated using the ratio of stubble biomass to straw biomass. For wheat we used the average for fertilized plots of 13.1% and for control plots 18.3%, while for maize we used 3.0% for all plots [Xu et al., 2006]. To convert plant dry matter into the equivalent amount of carbon, we used the national average carbon concentrations for wheat (399 g C kg−1) and maize (444 g C kg−1) residues on an oven-dried basis [NCATS, 1994].
The source of manure changed with local availability (pig, goat, horse, or cattle) and varied with a C/N ratio between 11/1 and 26/1 (Table S2). Annual carbon inputs from manure ranged from 0.43 to 8.69 t C ha−1 yr−1 depending on the site and application year [Fan et al., 2008; NCATS, 1994; Xu et al., 2006]
Soil tillage was by a mouldboard plow. For single-cropping or double-cropping field sites tillage occurred before seeding once or twice a year, respectively. The tillage depth was 25 cm at Gongzhuling, Urumqi, and Zhengzhou where prior measurements determined that 80% of the straw/manure inputs would remain within the surface 20 cm of soil, while tillage was to 20 cm at other field sites where 100% of the straw/manure inputs remained within this soil depth [Xu et al., 2006].
2.2 Soil Analysis
Composite soil samples (0–20 cm depth) were randomly collected from each plot at each field site (n = 5–10 cores per composite sample; 5 cm in diameter) after harvest but before tillage (i.e., September–October). Soil samples were air dried before being sieved (<2 mm) for soil pH analysis (soil:water = 1:1), and then further ground (<0.25 mm) for determination of SOC [Walkley and Black, 1934], total N [Black, 1965], total P [Murphy and Riley, 1962], and total K [Kundsen et al.,1982]. Available N and K were measured following the method of Lu  and available P (Olsen-P) by the method of Olsen et al. .
The clay content (16–34%) in single-cropping sites was higher compared to the double-cropping sites (6–17%; Table 2). The initial SOC was also higher in sites with single cropping (5.24–13.23 g C kg−1) compared to double cropping (6.26–7.10 g C kg−1). Initial soil pH was between 7.2 and 8.7, and there was no significant change in pH during the experimental periods (data not shown). Soil Olsen P ranged from 3.4 to 23.3 mg P kg−1 and available K from 16 to 288 mg K kg−1. Initial nutrient levels at field sites were low with positive response of plant growth to inorganic fertilizer application [Shen, 1982].
2.3 Carbon Model
The climatic data used in the RothC model (RothC-26.3) [Jenkinson and Coleman,1999] consisted of monthly mean air temperature (°C), precipitation (mm), and open pan evaporation (OPE; mm). Temperature and precipitation data for each site were collected from the nearest meteorological station of the China Meteorological Administration. Because the OPE data were unavailable, we calculated potential evapotranspiration (PET) according to the Food and Agriculture Organization (FAO) Penman-Monteith method [Allen et al., 1998] and converted the PET to OPE by OPE = PET/0.75 [Jenkinson and Coleman,1999]. Since the land was irrigated at Urumqi, Zhangye, Yangling, Changping, Zhengzhou, and Xuzhou, we added summed irrigation water with precipitation.
Soil input data for modeling were based upon the clay content (%) and the initial SOC content (t C ha−1) for each trial site. To determine the inert organic matter (IOM) pool for the RothC model we used the equation IOM = 0.049 × SOC1.139 [Falloon et al., 1998]. Management data on monthly soil cover (bare or vegetated) were obtained from Xu et al. . In the RothC model the added straw was treated as crop residue and animal manure as farm yard manure.
In modeling each set of field trial data, we set the initial SOC value in the RothC model to the measured SOC content from the initial value of each field trial treatment plot (Table 2) and then simulated the change in SOC during the trial period for each set of fertilizer treatments. To run the model, it is necessary to specify the initial amount of SOC in each of five defined organic matter pools: Decomposable Plant Material (DPM), Resistant Plant Material (RPM), Microbial Biomass (BIO), Humified Organic Matter (HUM), and Inert Organic Matter (IOM). The allocation of SOC among the different pools was unknown for these field sites. However, as described by Jenkinson and Coleman  if we assume that the SOC content has reached equilibrium, then RothC can be run inversely to calculate the amount of carbon that is needed to enter the soil annually to maintain a specific level of SOC; the allocation of SOC into each of the four organic matter pools is defined at the same time. This is a standard means by which to parameterize this model to equilibrium; at which point the relative size of the carbon pools can be defined [see Jenkinson and Coleman,1999; RRes, 2007]. For plant residue C inputs we used a DPM: RPM ratio of 1.44 as this is a typical value for most agricultural crops and grasses [Jenkinson and Coleman,1999]. The average weather data (monthly mean air temperature (°C), precipitation (mm), and open pan evaporation (OPE; mm)) for each trial site from the start year to the end of the simulation was used in this equilibrium model run.
Once the starting SOC content and its initial allocation among the organic matter pools had been established, the model was run using carbon inputs according to the different carbon inputs scenarios: (A) from the initial year to 2010, the carbon inputs were the measured data for each year (Figure 2); (B) after 2010, there are two scenarios (i) for the current NPP carbon inputs scenario, the carbon inputs were the average values for each site during the experimental period which are listed in Table S2; (ii) for the 1% annual increase in NPP carbon inputs scenario, the carbon inputs are based on the values and justification provided in Sun et al.  for a 1% annual increase by the current NPP carbon inputs (see carbon inputs scenarios in detail in paragraph 18).
2.4 General Circulation Models
Two general circulation models (GCMs) were selected: BCCR, the Bjerknes Centre for Climate Research, University of Bergen, Norway, http://www.ipcc-data.org/ar4/model-BCCR-BCM2.html and IPSL, the Institute Pierre Simon Laplace, France, http://www.ipcc-data.org/ar4/model-IPSL-CM4.html. These two global change models represent a range of model characteristics and thus their climates scenarios. The future climate using BCCR is cold and dry, while the IPSL is warm and wet when compared to historical observations (Table S3). Both GCMs have been validated for use in China [Li et al., 2011], and we also found good agreement between models and historical climatic data (1971–2000) when assessed for the trial sites used in this study (e.g., observed versus estimated total net radiation at Urumqi; Figure S3). Here we used extremes in CO2 concentration scenarios of 550 ppm (B1) and 850 ppm (A2) [IPCC,2007b].
2.5 Climate Change Scenarios and Plant Residue Carbon Input Scenarios
We set five climate scenarios until the year 2100 for the RothC modeling: no climate change, BCCR GCM under two CO2 emission scenarios (B1, A2) and the IPSL GCM under two CO2 emission scenarios (B1, A2). Since the RothC model does not include the crop submodel routine, we set two carbon input scenarios: (i) current NPP (the average of the field trial experimental period) and (ii) 1% annual increase in NPP based on the agricultural productivity improvements proposed by Sun et al. —we extrapolated this NPP scenario to the year 2100 on the basis that similar further gains in NPP would be made as agricultural practices and crop breeding continue to make improvements to plant productivity.
2.6 Evaluation of Model Performance and Statistical Analysis
We determined the coefficient of determination (R2) to represent the degree of association between the modeled and measured data and the root-mean-square error (RMSE) to represent the magnitude of differences between the modeled and observed values [Smith et al., 1997]. Analysis of variance (ANOVA) and the least significant difference methods (P < 0.05) were applied to compare treatment and climate effects on crop yield, organic carbon input, and SOC dynamics. The t test was employed to assess differences in basic site information, soil properties, crop yield, organic carbon input, and SOC dynamics between single-and double-cropping sites.
3.1 Grain Yield and Carbon Inputs Estimation
Average annual grain yield from the eight long-term field trials was 4.0 t ha−1 for wheat and 6.3 t ha−1 for maize. Grain yield in the control plots was lowest (1.4 t ha−1 yr−1 for wheat and 3.3 t ha−1 yr−1 for maize) with a decreasing trend during the 15 to 28 years of field trials. Under the plots with fertilization (i.e., NPK, NPK + M, and NPK + S), the grain yield increased during the experimental period in 5/8 of the field trial sites with the exceptions being at Pingliang, Changping, and Xuzhou (Figure S1). Generally, the grain yield under NP + M/NPK + M plots was highest, but there was no significant difference between NP/NPK, NP + M/NPK + M, and NP + S/NPK + S plots. The additional manure or straw return had no significant effect on grain yield compared to inorganic only fertilizer application (Figure S1). The annual crop yield at double-cropping sites was almost two times that at single-cropping sites (Table S2). Total annual grain yield and total carbon inputs in plots receiving fertilizer (i.e., NPK, NPK + M, and NPK + S) were significantly higher than those in the control treatment without fertilization for both single-cropping and double-cropping sites; there was no significant difference between fertilizer treatments (P < 0.05, Table S2).
As reflected in the grain yield trend, the lowest carbon inputs were also found in the control plots of both single-cropped (0.85 t C ha−1 yr−1) and double-cropped (1.63 t C ha−1 yr−1) sites (Table S2). Where inorganic fertilizer alone was used, the carbon inputs were double that of the control and threefold the control when manure or straw was also applied (Figure 2). At Changping the trial site contained an additional carbon inputs from weeds (i.e., 1.0 t C ha−1 season−1 in the control and 0.25 t C ha−1 season−1 for other treatments) which was accounted for in the carbon inputs budgets.
3.2 SOC Change in Northern China
The initial SOC content was higher for single-cropping sites (24 t C ha−1) compared to double-cropping sites (20 t C ha−1; Table 2). After 15 to 28 years, in single-cropping sites, the SOC content in plots with inorganic fertilizer was significantly higher than in the control plot without fertilization. The annual carbon inputs was higher in double-cropping sites (1.6 to 5.7 t C ha−1 yr−1) than single-cropping sites (0.9 to 3.9 t C ha−1 yr−1; Table S2), but this was not reflected in the annual SOC change, which was greater in single-cropping compared to double-cropping sites when considered for the same amount of carbon inputs (Figure 3).
The SOC sequestration potential was negative in the control plots, which were a net source of CO2 (Table 3). However, even when NPK was applied the future SOC sequestration potential (until the year 2100) was only 4.5 and 5.7 t C ha−1 in double-cropping and single-cropping sites, respectively (Table 3). There was significantly higher SOC sequestration potential where inorganic fertilizer was combined with manure (i.e., 20.2 t C ha−1 in single-cropping sites and 16.1 t C ha−1 in double-cropping sites) or straw (i.e., 16.8 t C ha−1 in single-cropping sites and 13.5 t C ha−1 in double-cropping sites) compared to NPK only plots. Averaged across all eight sites, the SOC sequestration potential ranged from −4.3 t C ha−1 in the control plots to 18.2 t C ha−1 in NPK + M plots.
Table 3. Average Soil Organic Carbon (SOC) Stocks at the Initial and Final of the Experimental Trial Period and the RothC Modeled Estimate of the Future Potential for SOC With Current NPP Carbon Input Under a No–Climate Change Scenarioa
Initial SOC at Trial Site (1979–1990)
Final SOC at Trial Site (2001–2008)
Modeled SOC in 2100 at Trial Site
Future SOC Potential (2008–2100)
Single-cropping sites: Urumqi, Gongzhuling, Zhangye, and Pingliang; Double-cropping sites, Changping, Zhengzhou, Yangling, and Xuzhou. Within a cropping practice grouping letters denote significant differences (P < 0.05).
(t C ha−1)
NPK + M
NPK + S
NPK + M
NPK + S
NPK + M
NPK + S
3.3 Validation of RothC Model
The RothC model was able to adequately simulate SOC dynamics in all treatment plots (Figure 4) as modeled SOC values fitted well with the observed values (Figure 5). Both the modeled and observed SOC content showed a declining trend in control plots at most sites. Modeled SOC values were at steady state, or increased slightly, in plots that received inorganic fertilizer only (NP/NPK) but increased in plots with organic manure (M, NP/NPK + M, NP/NPK + S). The coefficient of determination (R2) between observed and modeled SOC contents ranged from 59% to 94% (P < 0.001) with a root-mean-square error of between 1.65% and 3.64% (Figure 5).
3.4 Comparison of the Climate Condition Prediction by GCMs Until 2100
The two GCMs predicted two different climate conditions by the end of 21st century. Compared with the first 20 years of the experiment, the average annual mean temperature (AMT) during the years 2080 to 2100 was predicted to increase at all sites (Table S3). Annual mean temperature increments predicted by the IPSL model were significantly higher than those predicted by the BCCR model (P < 0.05). Under both BCCR and IPSL models, for the highest CO2 emission scenario (i.e., A2), the increment of AMT ranged from 3.8 to 5.5°C at Gongzhuling and from 2.0 to 3.8°C at Yangling, which indicated that the colder high-latitude region would experience a larger change compared to the warmer low-latitude region. There was no clear trend in annual precipitation (AP) or annual evapotranspiration (AE) between sites or through time.
3.5 The Effect of Climate Change on SOC Change Until 2050 and 2100
First, we predicted the future change in SOC under the five climate conditions (no climate change, BCCR-B1, IPSL-B1, BCCR-A2, IPSL-A2) using the existing level of carbon inputs (i.e., current NPP; Figure 6). There was no significant difference in either the total annual SOC change or the normalized (i.e., per ton plant carbon inputs) annual SOC change among the five climate conditions. Total annual SOC change was in the order: NPKM > NPKS, NPK > Control (P < 0.05). However, when total annual SOC change was normalized, there was no significant difference among the plots that received fertilizer (i.e., NPK, NPK + S, NPK + M, P < 0.05). The different carbon inputs scenarios showed that under the current NPP scenario the average annual carbon change from the initial year to 2100 was 0.062 t C ha−1 yr−1 and 0.087 t C ha−1 yr−1 with IPSL-A2 and BCCR-A2 climate conditions, respectively.
With the increased NPP scenario the total SOC sequestration potential from the initial year to 2100 was 23.76 t C ha−1 (ranged from 1.63 to 39.63 t C ha−1 under different fertilization plots) under the BCCR-A2 climate condition, and 20.45 t C ha−1 (ranged from −0.80 to 37.13 t C ha−1 under different fertilization plots) with the IPSL-A2 climate condition. The relevant total annual SOC change from the initial year to 2100 was 0.211 t C ha−1 yr−1 (ranged from 0.008 to 0.355 t C ha−1 yr−1) and 0.182 t C ha−1 yr−1 (ranged from −0.011 to 0.333 t C ha−1 yr−1) with BCCR-A2 and IPSL-A2 climate conditions, respectively (Figure 6). The average annual carbon change under current and increased NPP scenarios when using the IPSL-A2 and BCCR-A2 climate conditions was 0.136 t C ha−1 yr−1.
Since there was no significant difference in SOC changes between these climate scenarios, we chose the highest temperature scenario, IPSL-A2, to examine how the different carbon input scenarios would alter SOC stocks into the future. Three sites are illustrated to represent three regions in northern China: Urumqi for northwest China, Gongzhuling for northeast China, and Zhengzhou for north China. During the years 2010 to 2050 or 2050 to 2100 the SOC conversation rate was higher under the increased NPP scenario at all three sites (except at Gongzhuling during 2010 to 2050 when compared to the current NPP scenario; Figure 7). For the model time period 2010 to 2050 the SOC conversion rate was always higher at Gongzhuling (i.e., the SOC conversion was 8.9% under the current NPP scenario, and 8.6% under the increased NPP scenario), and almost double of that at Zhengzhou (i.e., the SOC conversion was 3.9% under current NPP scenario, and 5.9% under increased NPP scenario). At the Urumqi site the SOC conversion was 3.6% under the current NPP scenario and 8.7% under an increased NPP scenario. For the average of these sites, the SOC conversion rate was lower during 2050 to 2100 (where the SOC conversion was 1.3% under current NPP scenario and 3.1% under increased NPP scenario) than during 2010 to 2050 when the SOC conversion was 5.4% under current NPP scenario and 7.0% under the increased NPP scenario. This indicated that the SOC storage potential of these soils was close to being reached in the later time period at all sites.
Under the IPSL-A2 climate scenario, representing the highest temperature scenario, even with an increased NPP carbon input scenario, the SOC content decreased in the control plots at all sites except Changping (Figure S2). Under this climate scenario for the NPK and NPK + S plots, the SOC increased at all sites except Urumqi, where the SOC decreased gradually. However, for the NPK + M plot the SOC increased at all sites under this climate scenario.
The social and economic stability of China largely depends on agricultural development. Cropland in northern China accounts for 65.8% of the 122 million hectares of total cropland in China [National Bureau of Statistics of China, 2012]. While inorganic fertilizers have played an important role in feeding the rapidly growing world population, the application of organic amendments to agricultural fields has declined [Ju et al., 2005]. The decomposition rate of SOC was shown to be faster when inorganic fertilizer was applied alone compared to with manure in Loessial soil in northwest China [Liang et al., 2012]. A future consequence of inorganic fertilizer use without organic amendments in China will be declining SOC stores and increasing CO2 release under current tillage practices. If China continues to maintain self-sufficiency in food production [Solot, 2006], then arable lands will need to increase productivity without causing loss of soil fertility.
The RothC model was able to accurately predict the SOC dynamics in agricultural upland soil in northern China. Yang et al.  and Guo et al.  applied the RothC model to upland soils (Black and Fluvo-aquic soils) in northern China, and both reported that the SOC predicted agreed well with the experimental data observed in unfertilized plots, in plots with inorganic fertilizers and where inorganic fertilizers were applied in combination with manure. Our results were consistent with Yang et al.  and Guo et al. , and we also found that the RothC model was suitable for use in predicting SOC stocks with straw application (Figures 4 and 5).
Addition of animal manures and return of crop straw are well recognized as positive management options to improve SOC as illustrated in this study. Farmers have used organic food waste and animal manures to maintain crop production and soil fertility for thousands of years in China [Yang, 2006]. However, with inorganic fertilizers currently being widely available, the application on manure to arable land has declined from 99.9% in 1949 to 25% in 2003 [Huang et al., 2006; Yang et al., 2010]. Under the current NPP carbon input scenario, the annual SOC change for the NPK + M treatment was 0.287 t C ha−1 yr−1 with no climate change, 0.252 t C ha−1 yr−1 assuming BCCR-A2 and 0.219 t C ha−1 yr−1 assuming IPSL-A2 climate conditions until 2100 in northern China. This would mean an additional 17.5 to 23.0 Tg C yr−1 sequestered to the end of this century if agricultural management practices were to apply NPK + M (without improvement in straw retention or conservation tillage practices). Assuming the increased NPP carbon input scenario, the SOC sequestered by 2100 would be 26.7 to 28.5 Tg C yr−1 under BCCR-A2 and IPSL-A2 climate condition, respectively. However, organic manure is now more commonly applied to vegetable crops than to grain crops—data from 200 agrometeorological stations confirmed this practice [Wan et al., 2011] which is unlikely to change. It should also be recognized that a larger land area is required to “grow” manure than the input of soil carbon [Schlesinger, 1999]. Increasing SOC stocks through manure application at one site may result in depletion of SOC at other. Thus, manure is not likely to yield an environmentally sustainable net sink for carbon across these large areas of arable land.
Throughout northern China, crop straw was historically used as fuel, animal feed and bedding, or burnt directly within the field. Since the 1980s, straw return was popularized by government policy as a practice to improve soil fertility and decrease air pollution by not burning. The area of agricultural land where straw was returned varied from 7% to 71% depending on the province; with an average 36.6% of all straw in China used to improve soil fertility [Gao et al., 2002; Lu et al., 2009]. Under the current NPP carbon input scenario, the annual SOC change in northern China until the year 2100 within the NPK + S treatment was 0.162 t C ha−1 yr−1 assuming no climate change (12.6 Tg C yr−1), 0.096 t C ha−1 yr−1 assuming BCCR-A2 (7.7 Tg C yr−1), and 0.064 t C ha−1 yr−1 assuming IPSL-A2 climate conditions (5.1 Tg C yr−1). Compared to NPK only, the return of straw could sequester an additional 7.2, 3.6, and 2.2 Tg C yr−1 SOC under these three climate conditions (i.e., no climate change, BCCR-A2, and IPSL-A2, respectively) in northern China by 2100.
In agricultural systems, tillage can be a major cause of SOC change; losses up to 50% of the starting SOC in surface soils (20 cm) have been observed after cultivation for 30 to 50 years when natural vegetation is converted to cultivated crops [Post and Kwon, 2000]. Based on the global database of 67 long-term agricultural experiments, West and Post  found on average that a change from conventional tillage to no-tillage can sequester 0.57 ± 0.14 t C ha−1 yr−1. In China, Lu et al.  determined that under the current climate situation that no-tillage can sequester 0.800 Tg C yr−1 (0.039 t C ha−1 yr−1). In our study, all the long-term experimental sites were plowed after harvest; as such we could not measure the effect of no-tillage on carbon turnover. However, it is reported that the carbon conversion rate is 8% per year in plowed systems and 10% per year in no-tillage systems [Duiker and Lal, 1999]. We attribute the lower SOC levels in the double-cropping sites in our study to the additional tillage each year associated with the planting of the second crop along with the higher soil temperatures at these sites—both tillage and temperature are likely to have increased organic matter decomposition rates [Zhang et al., 2010]. Assuming that organic manure inputs remain the same [Li et al., 2003] but that straw retention in China does increase [Gao et al., 2002; Sun et al., 2010], and that no-tillage practices can be extended Sun et al.  calculated a further 1% annual NPP increase from 2000 to 2050. Based on these improved agricultural management practices, this rate of increase in NPP would result in an annual SOC changes of −0.002 under the control plot to 0.284 t C ha−1 yr−1 under the NPK + M plot (a SOC conversion rate of 7.0%) until 2050 and a further 0.024 t C ha−1 yr−1 under the control to 0.209 t C ha−1 yr−1 under the NPK + M treatment (a SOC conversion rate of 3.1%) until 2100, assuming the IPSL-A2 climate condition scenario. This is equivalent to an additional 0.33 Tg C yr−1 until 2050 and 0.98 Tg C yr−1 until 2100 of SOC sequestered compared to a no change in NPP scenario.
The prediction of future SOC sequestration potential demonstrated that under no fertilizer input, these soils would be a net source of CO2 in most parts of northern China. Even when inorganic nutrients were applied, the additional carbon input from increased plant residues could not meet the depletion of SOC in the northwest sites. Manure or straw application could improve the carbon sequestration at all sites, with straw being a more likely option into the future. The future SOC sequestration potential in northern China was −4.3 to 18.2 t C ha−1 by 2100 under current carbon input and existing climate conditions. The effect of climate change on the annual rate of SOC change did not differ significantly between the five climate scenarios; under the higher CO2 emission scenario (i.e., A2) 8.1 t C ha−1 (0.062 t C ha−1 yr−1) and 10.7 t C ha−1 (0.087 t C ha−1 yr−1) will be sequestered under IPSL-A2 and BCCR-A2, respectively, with the current NPP C input scenario. Under the increased NPP C input scenario, 20.5 t C ha−1 (0.182 t C ha−1 yr−1) and 23.8 t C ha−1 (0.211 t C ha−1 yr−1) would be sequestered in northern China. This doubling in the potential of future SOC sequestration under an increased NPP scenario highlights the need to introduce both straw retention and no-tillage practices across the areas of northern China where this is not commonly practiced.
We acknowledge our colleagues for their unremitting efforts to the long-term experiments, and we are also very grateful to Wendy Wang, University of Maryland, USA, and Daniel Richter, Duke University, USA, for their constructive comments and suggestions. This research was financially supported by the National Science Foundation of China (41171239), the National Basic Research Program (2011CB100501), the Australian Research Council Future Fellowship Scheme (FT110100246), and a Chinese High-End Foreign Experts Visiting Professorship.