Introduction

Agriculture forms the foundation of China’s national economy. The large contribution of this industry to carbon emissions in the world’s most populous nation is therefore receiving growing attention, especially from the country’s federal government and research community. Further, agriculture is the world’s second-largest source of greenhouse gas (GHG) emissions (FAO 2009), accounting for about 20 % of global carbon emissions from human activities (Paustian and Vernon 1998). Analyses indicate that worldwide, 70 % of current atmospheric methane originates from agricultural production (Mosier et al. 1998), and that agricultural activities contribute roughly half of anthropogenic CH4 and N2O emissions, at 47 and 58 %, respectively (IPCC 2006). China produces the most GHG emissions of any developing country; its total agricultural emissions were 604 Mt CO2 eq in 1994 and 686 Mt CO2 eq in 2007, accounting for 16.57 and 9.2 % of the country’s total GHG emissions, respectively (NDRC 2004; Chen and Zhang 2010).

Agriculture is not only a major source of greenhouse gas emissions, but conversely also one of the economic sectors that are most vulnerable to climate change. Agricultural production is tightly coupled with the condition of the environment—efficient agricultural production will inevitably require high investment of material inputs, and the attendant high-energy consumption and high emissions have become a primary cause of the deterioration of natural resources and environmental conditions in China. Currently, population expansion and economic development pressure prevent the increase of agricultural production in developing countries via expansion of the area of arable land. Instead, high-yield crop varieties, fertilizers, pesticides, agricultural machinery and other modern agricultural investments are the only way to increase yield per unit area. The dependence of “carbon-intensive agriculture” on fossil fuels has a much more severe impact on natural resources, the environment, and climate, due to increasing agricultural carbon emissions. If agricultural production continues to be dependent on fossil fuels, the influence of agriculture will continue to exacerbate global climate change (Smith et al. 2007b; Verge et al. 2007).

The most commonly used approach to measure agricultural carbon emissions in China is the IPCC inventory and its methods. The total amount of agricultural carbon emissions (i.e., the sum of emissions from crop and livestock production) was assessed, to analyze the factors that drive it (Dong et al. 2013; Min and Hu 2012; Tian and Zhang 2013). Notably, for different sources of carbon emissions from the crop industry and from the livestock and poultry industry, the estimated inventory of emissions and their driving factors varied, especially for systems with a high production of livestock and poultry (i.e., in terms of output value and yield). For example, in the actual measurement of agricultural carbon emissions, the crop industry produces a large proportion of emissions from material inputs (such as chemical fertilizers and pesticides), while emissions from the livestock and poultry industry mainly involve livestock enteric fermentation and manure management (rather than agricultural materials), namely the external consumption of energy. Therefore, if only for the purpose of studying the factors that influence total agricultural carbon emissions, the results would certainly be biased.

The objectives of this study were to: (1) analyze dynamic variation in and development of agricultural carbon emissions, based on measures of emissions from the crop industry and the livestock and poultry industry in Hunan; (2) analyze the relationship between agricultural production and associated carbon emissions in the region; (3) quantitatively analyze the relationship between carbon emissions from those agricultural industries and the factors that influence emissions; (4) analyze the potential and pressure for mitigation of agricultural carbon emissions, thereby providing a theoretical and policy basis for the assessment of such emissions, along with approaches for reduction such as the development of low-carbon agricultural technology.

To achieve these objectives, carbon emissions from the crop industry were assessed independently of those from the livestock and poultry industry. Our approach involved time series analysis, which is a useful tool in environmental and economic research including experimental, modeling, and policy studies (Lee and List 2004; Verbeke and De Clercq 2006). The related potential drivers of emissions were analyzed using principal component analysis (PCA) and multiple regression analysis. PCA is a method of ordination that is often used to extract relevant information from multivariables, complex data sets. It can functionally eliminate multicollinearity among independent variables, generating new variables that can be used as predictors in regression analysis (Rajab et al. 2013). This combination of statistical approaches is therefore used widely to identify the key factors that drive energy, environmental, and economic dynamics (e.g., Hillier et al. 2009; Shu and Nina 2011; Tian et al. 2012a; Braun et al. 2014; Zhang et al. 2014).

An additional step of our approach drew upon decoupling theory, which is used to study the causal relationship between resource consumption and economic development. It is especially useful in energy environment analysis and is thus important in the empirical study of links between economic development and carbon emissions (Guivarch and Mathy 2012; Tian et al. 2012b). Combined, these analyses were used to achieve the overall purpose of this paper, namely to better understand the dynamics of agricultural carbon emissions in Hunan in order to facilitate future efforts to mitigate their contribution to global climate change.

Description of the investigated area

The study identified the key factors that affect agricultural carbon emissions in Hunan, with a view to evaluating the sustainability of agricultural production in the region. Hunan is a province in south-central China, located from 25°–30°N to 108°–114°E. It covers ~211,800 km2, accounting for 2.2 % of China’s land area which places it 10th among the country’s provinces and direct-controlled municipalities (and first, within the central part of the country). Hunan Province has 14 cities, with a resident population of 66.9 million at the end of 2013, ranking seventh among all provinces in China.

Roughly half of the province’s total area consists of mountainous and hilly landscapes. In comparison, its arable area of 3.8 million ha represents 17.9 % of the province’s land and 3.1 % of the national availability of arable soils. The subtropical monsoon climate provides ample heat and concentrated rainfall for crop production, although there is an annual drought during summer. Hunan is an important base of grain production, ranking seventh among the provinces in 2012. It supports the nation’s highest rice acreage and yield, along with a high output of other major agricultural products including cotton, oil, tobacco, and ramie. Hunan is the most important province in terms of pig farming.

Materials and methods

Several steps were involved in the process of estimating carbon emissions from agricultural production in Hunan (Fig. 1). The period of time used in the analysis was 1998–2012.

Fig. 1
figure 1

An overview of the steps in the methodology of this study

Full size image

Methods for calculation of carbon footprints

Agricultural carbon emissions result mainly from the processes of growth and development of cultivated plants, and material and energy recycling in the natural environment. Various studies of agricultural carbon emissions have been conducted based on the framework of Johnson (2007), (IPCC 2006; Schneider et al. 2007; Hillier et al. 2009; Cheng et al. 2011; Min and Hu 2012; Tian et al. 2012a; Tian and Zhang 2013). Based on that body of research, the following four key sources of agricultural carbon emissions are considered in this paper: (1) those caused by the input of agricultural materials (e.g., chemical fertilizers, pesticides, and diesel oil) and other direct and indirect emissions caused by agricultural energy consumption (e.g., direct N2O emissions due to nitrogen fertilization), as shown in Table 1. (2) CH4 and other GHG emissions from rice production; the typical CH4 emissions from rice paddies are shown in Table 2. (3) Carbon emissions from agricultural land use, including loss of the soil organic carbon pool due to soil tillage, disposal of agricultural waste, and CH4 and N2O emissions caused by straw burning (calculated based on Cao et al. 2007). The loss of soil, N2O emissions caused by damage to topsoil, and N2O emissions from soil are shown in Table 3. (4) Carbon emissions caused by feeding animals, especially ruminants. The GHG emission output factors from the livestock and poultry industry are shown in Table 4. In this paper, the definition of agricultural carbon emissions encompasses only the harvesting of crops and livestock/poultry products, excluding any emissions from the processing and transportation of agricultural products.

Table 1 Sources of agricultural carbon emissions from different material inputs, energy consumption, and straw burning
Full size table
Table 2 Methane emissions throughout the growth cycle of rice
Full size table
Table 3 Nitrous oxide emissions from soil for various crops
Full size table
Table 4 Carbon emissions from the rearing of major types of livestock and poultry
Full size table

Agricultural carbon emissions

Using the above limits to the definition of agricultural carbon emissions, the formula for calculating the emissions was established as:

$$e = \sum {\mathop E\nolimits_{i} } = \sum {\mathop T\nolimits_{i} } \cdot \sum {\mathop \delta \nolimits_{i} }$$
(1)

where E represents the total agricultural carbon emissions, Ei represents each source of carbon emissions, Ti is the amount of carbon emissions from carbon source i, and δi is the emission factor for carbon source i. To facilitate the analysis, all values were converted to the units of CO2 equivalents (CE).

The amount of direct N2O emissions resulting from nitrogen fertilization was calculated as (IPCC 2006):

$$CF = \mathop F\nolimits_{N} \times \mathop \delta \nolimits_{N} \times \frac{14}{28} \times 290 \times \frac{12}{44}$$
(2)

where CF represents the direct N2O emissions from nitrogen fertilization (converted to CE), FN represents the amount of nitrogen fertilizer applied, δN is the N2O emission factor (t N2O-Nt−1N fertilizer) caused by nitrogen fertilization, 44/28 represents the molecular weights of N2 compared to N2O, 298 is the net Global Warming Potential (GMP) per century, and 12/44 represents the molecular weights of C compared to CO2 (IPCC 2006).

Annual rearing of livestock

During the measurement of GHG emissions from the livestock and poultry industry, the measure of average annual rearing of livestock must be adjusted based on the slaughter rate (IPCC 2006; Hu and Wang 2010; Min and Hu 2012). Since the slaughter rates for pigs, rabbits, and poultry are all greater than 1, with respective average life cycles of 200, 105, and 55 days, the average annual amount of rearing was adjusted according to the following formula:

$$\mathop N\nolimits_{i} = Days\_alive_{i} \times {{M_{i} } \mathord{\left/ {\vphantom {{M_{i} } {365}}} \right. \kern-0pt} {365}}$$
(3)

where N i is the average annual amount of rearing for the ith kind of livestock, Days_ alive i is the average life cycle of animal type i, and M i is the annual yield of animal type i (i.e., slaughter).

For types of livestock and poultry with a slaughter rate of less than 1, the average amount of annual rearing was adjusted according to the breeding stock at the end of the year:

$$\mathop N\nolimits_{\text{i}} = \left[ {\mathop C\nolimits_{it} + \mathop C\nolimits_{{i\left( {t - 1} \right)}} } \right]/2$$
(4)

where N i is the average annual amount of rearing animal type i; C it and C i (t − 1) are the breeding stock at the end of years t and t − 1 for animal type i.

Carbon intensity and carbon efficiency

In this paper, carbon intensity (CI) refers to the carbon emissions per unit of output, which can be used to estimate the influence per unit of output on carbon emissions and the cost-effectiveness of carbon emissions. It is expressed in the units of t CE−1 USD 10,000 GDP, calculated as:

$$CI = {{CF} \mathord{\left/ {\vphantom {{CF} Y}} \right. \kern-0pt} Y}$$
(5)

where CF is total agricultural carbon emissions (t CE), and Y represents total agricultural output (USD 10,000 GDP).

In contrast, carbon efficiency (CFE, also expressed per USD 10,000 GDP−1 t CE) refers to the ratio of agricultural production (i.e., sum of the crop plus livestock and poultry industries) to agricultural carbon emissions, calculated as:

$$CFE = {Y \mathord{\left/ {\vphantom {Y {CF}}} \right. \kern-0pt} {CF}}.$$
(6)

Decoupling theory

The decoupling elasticity concept (Tapio 2005) was based on the OECD’s carbon decoupling model (OECD 2002). The latter is also known as carbon emissions elasticity, namely the ratio of the magnitude of changes in economic development to the magnitude of changes in CO2 emissions. Compared to OECD decoupling theory, elastic decoupling theory can more accurately reflect the sensitivity of CO2 changes to economic changes. Detailed Tapio elastic decoupling indices are shown in Table 5. In this study, we used the Tapio model for decoupling analysis of Hunan’s total agricultural carbon emissions from the crop industry, the livestock and poultry industry, and their individual economic outputs (USD 10,000 GDP, with 1997 as the base year), separately. The decoupling model was constructed as follows:

Table 5 Classification of Tapio decoupling indices
Full size table
$$t = \frac{{{{\Delta C} \mathord{\left/ {\vphantom {{\Delta C} C}} \right. \kern-0pt} C}}}{{{{\Delta G} \mathord{\left/ {\vphantom {{\Delta G} G}} \right. \kern-0pt} G}}}$$
(7)

where t represents decoupling elasticity, C represents agricultural carbon emissions, and ΔG is industrial GDP.

Statistical analysis

This study involved an analysis of carbon emissions over a 15-year time frame. The problem of multi-collinearity among repeated measurements that exists in time series can be solved by using approaches such as ridge regression and multiple linear regression analysis (MRA) (Mason and Perrault 1991; Grapentine 1997; Zhang et al. 2014). MRA minimizes the observed value and estimated difference with the least squares method, so as to fit linear equations to observed responses (Braun et al. 2014); the recommended correlation coefficient between each independent variable is less than 0.7 (Anderson et al. 2003).

In our approach, PCA was first used to filter all influencing factors that affect agricultural carbon emissions in order to determine the most significant independent variables (i.e., driving factors). We then selected variables from the PCA with high loading as the main components to build the final regression model (Abdul-Wahaba et al. 2005; Al-Alawi et al. 2008) that determined the linear relationship between agricultural carbon emissions and their drivers.

Data sources and processing

Overall, the combination of crop and livestock production is though to represent a balanced approach to economic development in the agriculture sector (China Statistical Yearbook 2013). Relying on agricultural inputs, this sector in Hunan produced a sustained, high investment yield in recent years and showed a huge potential for reduction of emissions. As our aim was to evaluate the sustainability of that approach, in terms of carbon emissions, data sources were selected in order to capture as fully as possible the range of production in this major agricultural province of China.

Data were collected from the following sources: the China Statistical Yearbook, China Rural Statistical Yearbook, China 60 Years of Agricultural Statistics, Hunan Statistical Yearbook, Hunan Rural Statistical Yearbook.

To “process” (i.e., inspect, correct, and standardize) the data, we conducted a check among all the sources to compare the same type of data reported in different sources and made any necessary adjustments during proofreading. For example, to inspect the data on application of nitrogen fertilizer, we first calculated the sum of N applied in various regions of Hunan. Then, that sum of regional data was compared with the value for the whole province, to see whether the numbers made sense. If an error was found during this step of data checking, we adjusted the value in question according to the relative proportion of the overall available data for that variable. Because the values for some agricultural material inputs raw application were lacking for 2003, instead we took the average value of the sum in 2002 and 2004.

All data analyses were performed with SPSS software, version 20.0.

Results and analysis

Total agricultural carbon emissions

The total agricultural carbon emissions in Hunan showed an overall upward trend over time; the value of 7168.89 × 104 t CE in 2012 was 1147.42 × 104 t CE (19.06 %) higher than the total in 1998. The growth rate of carbon emissions from agricultural material inputs was the highest, at 52.17 %, with an increase of 235.93 × 104 t CE (Table 6). The pattern of variation in Hunan’s agricultural carbon emissions had four clear stages: slow growth, rapid growth, rapid decline, and finally a slow increase. Specifically, 1998–2002 was a period of slow growth in emissions; compared to 1998, the total in 2002 increased by only 66.06 × 104 t CE (1.10 %). The growth rate of emissions from enteric fermentation was the highest during that time, at 5.48 %, and that of agricultural material inputs was the second highest at 4.06 %. From 2002 to 2007, carbon emissions demonstrated a rapid growth phase, with an increase of 1084.42 × 104 t CE (17.81 %). Due to national policies, crop and livestock husbandry then expanded rapidly, leading to increased carbon emissions, with CH4 emissions from rice paddies being the largest contributor. As the “three agriculture-related issues” emerged—i.e., agriculture, rural areas, and farmers—farmers were overburdened, and the investment in agricultural materials slowed down, yet the growth rate of emissions from agricultural material inputs was still 13.73 %. From 2007 to 2008, there was a rapid decline in carbon emissions; in 2008 they decreased by 374.16 × 104 t CE (5.22 %), due to a massive snowfall, but the growth rate of those from agricultural material inputs increased by 0.32 %. This shows that carbon emissions from agricultural material inputs are more important in carbon emissions from agricultural production in Hunan. From 2008 to 2012, there was a slow increase in carbon emissions; compared to 2007, because of increased government support for agriculture, production recovered rapidly and developed further, so the amount of agricultural inputs continued to rise. Total agricultural carbon emissions in 2012 increased by 371.10 × 104 t CE (5.46 %) compared to 2008, while the growth rate for emissions from agricultural material inputs increased by 28.17 % (Fig. 2).

Table 6 Variation of agricultural carbon emissions in Hunan from 1998 to 2012 (units = 104 t CE)
Full size table
Fig. 2
figure 2

Carbon emissions from agricultural materials in Hunan during the period 1998–2012

Full size image

Government implementation of a series of energy saving policies and technologies then reduced the intensity of agricultural carbon emissions (CI), year after year (Fig. 3), whereas the efficiency (CFE) gradually increased. However, due to drought conditions in 2012, the development direction of both agricultural CI and CFE was reversed. The agricultural production system in Hunan is sensitive and vulnerable, so it depends heavily on agricultural materials, showing characteristics of “high input, high pollution and high emissions” (i.e., the “three high” characteristics), weak adaptability to climate change, and unstable agricultural carbon emissions. The coupling effect that ties agricultural carbon emissions with economic development and climate change is strong, and creates a negative feedback cycle from years of accumulated climate damage caused by carbon-intensive agriculture. Taking into account population growth and food security strategies, Hunan agricultural carbon emissions will continue to rise. In this study, most of the carbon emission measures were based on domestic experimental data—the calculation boundaries for what was considered as agricultural emissions were determined scientifically, and the calculation method was feasible, so the calculation results can be considered objective and reliable.

Fig. 3
figure 3

Agricultural carbon intensity (CI) and carbon efficiency (CFE) in Hunan from 1998 to 2012

Full size image

Decoupling analysis

The relationship between GHG emissions and the output value of Hunan agricultural production demonstrated mainly weak decoupling, but it was unstable. For example, the snowstorm in 2008 caused a strong decoupling state. The elastic characteristic of total agricultural carbon emissions in 2003 showed expansionary negative decoupling, which was mainly due to the negative decoupling state of crop industry expansion in 2003 along with the expansionary connection of the livestock and poultry industry in 2003. The analysis of decoupling states (Table 7) showed that agricultural emissions grew faster than did agricultural output and production. Production depended on a large number of agricultural material inputs to improve economic efficiency, and this trend eased after 2003. Because of the impact of the “three agriculture-related issues” issues, farmers’ demand for production materials gradually slowed down, which inhibited the rapid growth of agricultural carbon emissions in recent decades. However, it is worth noting that drought in 2012 caused the total agricultural carbon emissions to exhibit expansionary negative decoupling, the elastic characteristics of crop carbon emissions to exhibit expansionary negative decoupling, and livestock and poultry carbon emissions to exhibit expansionary connection—these patterns suggest that Hunan agricultural production has weak adaptability to climate variation. That is, Hunan agriculture is susceptible to suffering from the impact of climate and environmental and economic policy, and other external environmental factors. The decoupling state in the livestock and poultry industry was more unstable than in the crop industry, indicating that carbon emissions from the former grew faster than the output. That in turn induced greater pressure for reduction of carbon emissions produced by the livestock and poultry industry.

Table 7 Decoupling relationship between agricultural carbon emissions and their output values in Hunan from 1998 to 2012
Full size table

Principal component analysis of carbon emissions from crops versus livestock

Previous studies have suggested that the key variables that influence carbon emissions are economic, structural, efficiency, and demographic factors (Bennetzena et al. 2012; Dong et al. 2013; Li et al. 2011). Based on Hunan’s agricultural production process and characteristics, we chose 11 kinds of indices to represent the factors that influence carbon emissions from the crop industry. These included economic factors (per capita annual net income, PI; per capita agricultural GDP, or PAG), structural factors (ratio of crop production GDP to total agricultural GDP, or CRA; ratio of rice production to total crop production, RRP), efficiency factors (crops productivity per capita, PCP; rice productivity per capita, RCP; crop arable productivity, CAP), energy input factors (amount of nitrogen fertilizer applied per acre, NAA; total power of agricultural machinery per capita, PCM), and social development factors (urbanization rate, UR; labor force, LF). For the livestock and poultry industry, we chose nine kinds of indices to represent the factors that influence carbon emissions, including economic factors (PI; PAG; household livestock income, FLI), structural factors (ratio of pork production to livestock production, PRL; ratio of beef production to livestock production, BRL; ratio of animal husbandry GDP to total agricultural GDP, or LRA), efficiency factors (farmland productivity in the livestock industry, LP), and social development factors (UR and LF). Results of the PCA that was used to assess all of these factors are presented in Tables 8, 9, 10, 11, 12.

Table 8 KMO and Bartlett’s Bartle
Full size table
Table 9 Ttotal explained variance of crop industry carbon emissions
Full size table
Table 10 Total explained variance of livestock and poultry industry carbon emissions
Full size table
Table 11 Rotated component matrix of carbon emissions from the crop industry
Full size table
Table 12 Rotated component matrix of carbon emissions from the livestock and poultry industry
Full size table

The KMO values of 0.738 and 0.651 (Table 8) indicate a moderate degree of common variables, and therefore that more variables should be taken into consideration for PCA. Bartlett’s sphericity test was significant, indicating clear structural property and interdependence between the original variables, and hence that factor analysis was suitable. The factors with a characteristic value >1 were extracted, and the explained variables are shown in Table 9. The first two factors had characteristic values of 92.51 and 86.27 % of the total characteristic value, so they were extracted as the main factors. The top two most relevant factors in the crop industry were UR and NAA (Tables 11,12), while in the livestock and poultry industry they were PAG and BRL, all of which were kept as substitution variables.

Regression analysis of carbon emissions from crops and from livestock and poultry

The main factors from the PCA for crops (UR and NAA) and the livestock and poultry industry (PAG and BRL) were employed as independent variables for the multiple regression analysis of the dependent variable (i.e., farming industry carbon emissions and livestock–poultry industry carbon emissions), results of which are shown in Tables 13, 14, 15, 16, 17.

Table 13 Pearson correlation matrix for main factors used in multiple regression
Full size table
Table 14 Statistical performance indicators for different modelsa of crop industry carbon emissions
Full size table
Table 15 Statistical performance indicators for different modelsa of livestock and poultry industry’s carbon emissions
Full size table
Table 16 Coefficients of crop production for carbon emission model 1
Full size table
Table 17 Coefficients of livestock and poultry industry for carbon emission model 1
Full size table

The correlation coefficients between the independent variables were small in the two regression analyses, and there was no multicollinearity of variables (Table 13). There was a high degree of fit in the regression (R 2 = 0. 829, 0.945), and the model residuals did not show auto-correlation (D–W values = 1.523, 1.747), suggesting good performance of the regression model (Tables 14,15).

The regression coefficients of NAA and UR were 5193.164 and 8288.964 (Tables 16, 17, 18, 19), and they were significant (P < 0.05). Thus, for crop industry carbon emissions (C p), the two main factors (UR = x 1, NAA = x 2,) explained nearly 93 % of the information provided by all 11 indices. The regression equation was:

$$\mathop C\nolimits_{p} = 906.823 + 8288.964X_{1} + 5193.164X_{2}$$
(8)
Table 18 Coefficients of crop production for carbon emission model 2
Full size table
Table 19 Coefficients of livestock and poultry industry carbon emission model 2
Full size table

Further, the coefficients of BRL and PAG were 32,934.198 and 0.354, and they were also significant (P < 0.05). Therefore, for livestock and poultry industry carbon emissions (C l), the two main factors (BRL = x 1, PAG = x 2) explained nearly 87 % of the information provided by all nine indices. The regression equation was:

$$\mathop C\nolimits_{1} = 402.834 + 32934.198X_{1} + 0.354X_{2} .$$
(9)

Discussion

Carbon emissions from the crop industry

The Chinese Government has promised to reduce carbon intensity per unit GDP by 40–50 % (versus the 2005 base year) by 2020. In 2005, the CI in Hunan Province was 26.21 t CE/USD 10,000 GDP for agriculture, 32.85 t CE/USD 10,000 GDP for the crop industry, and 19.25 t CE/USD 10,000 GDP for the livestock and poultry industry. However, the average annual decrease was only 3.40, 3.45, and 3.28 % in 2005–2012, respectively. Under the combined effects of both external environmental factors (e.g., natural resource shortages, frequent climate disasters) and socioeconomic factors (e.g., accelerating urbanization, population growth), it will be very difficult to meet the minimum requirements for emission reduction. Thus, China should take measures to promote the development of agriculture that minimize pressures on the environment (Zhang and Cheng 2009).

As urbanization in China has gradually accelerated, negative impacts on the development of agricultural production have become more prominent. Indeed, urbanization is significantly affecting China’s total carbon emissions (Lin and Liu 2010). Specifically, with a hypothetical increase in the rate of urbanization by 1 %, carbon emissions from Hunan’s crop industry will increase by 83 units. While the phenomenon of urbanization is reducing the agricultural labor force, it is also increasing land competition between urban expansion and agricultural production, placing great pressure on the supply of arable land needed to ensure food security. Efforts to ensure agricultural productivity, despite labor shortages and the lack of arable land, will inevitably lead to sustained high levels of agricultural investment in fossil fuels, which has already created a series of climate and other environmental problems. Thus, the protection of prime agricultural land has become a critical measure for the simultaneous development of agricultural production and reduction of agricultural emissions.

Our findings indicated that carbon emissions from the use of nitrogen fertilizers accounted for 51.48 % of the total carbon emissions due to agricultural material inputs—much less than what has been reported for nitrogen fertilizers in the British crop industry (75 %, Hillier et al. 2009), though not much different than measures for nitrogen fertilizers in the total Chinese crop industry (52.5–56.5 %), (Cheng 2011). The discrepancy compared to the UK might be due to the underdeveloped economy of Hunan Province (which is about average when compared to all provinces of China). Carbon-intensive agriculture has become China’s major development model for agricultural production. The carbon intensity of Hunan’s crop industry depends significantly on the use of nitrogen fertilizers (Fig. 4a). In China as a whole, many chemical fertilizers are used in agricultural production though the seasonal utilization rate is low, reportedly 30–35 % for nitrogen, 35–50 % for potash, and 15–20 % for phosphate (Wang et al. 2010). In contrast, the utilization rate in developed countries can reach ~60 % (Janzen et al. 2003). Improving the utilization efficiency of nitrogen fertilizers is thus an important means for developing agriculture in China while reducing carbon emissions from crop production, especially those from nitrogen fertilizers.

Fig. 4
figure 4

a Correlation between total CF of crop production and annual N fertilizer application rate in crop production, b correlation between crop production capacity and carbon intensity of production

Full size image

Overall, our results were similar to other assessments of carbon emissions from the Chinese agriculture sector. A study of the carbon footprint of China’s agricultural production from 2002 to 2011 (Tian et al. 2014), including crops, livestock, and poultry, found that those carbon emissions in 2002, 2005, 2008, and 2011 were 4934.37 wtCE, 5333.18 wtCE, 5329.33 wtCE, and 5622.68 wtCE, respectively, in Hubei. That province is close to Hunan and has similar characteristics of agricultural production. They calculated that agricultural emissions for those years in Hunan were 5944.57 wtCE, 6695.69 wtCE, 6500.81 wtCE, and 6737.26 wtCE. Those estimates were similar to our findings of 6087.53 wt, 6945.97 wt, 6797.79 wtCE, and 6923.49 wtCE. The discrepancy could be due to variation in the method of calculation, the carbon emission coefficient that we used for nitrogen fertilizers, and the fact that our study included emissions from straw burning.

In China, every 1 % increase in the rate of fertilizer efficiency can avoid the use of ~2.5Mt of standard coal (Chen and Zhang 2010). Theoretically, if the amount of nitrogen applied per acre were reduced by 10 %, carbon emissions from crop production in Hunan would be reduced by 519 units. The improvement of nitrogen fertilizer efficiency in crop production is strongly encouraged in the practice of agriculture, via the use of both organic and inorganic fertilizers, controlled release nitrogen fertilizer and imitation release fertilizers, scientific water management, and other measures, thereby reducing the amount of fertilizer needed. In the red soil paddy fields where double cropping of rice occurs, the combined application of organic and inorganic fertilizers can increase production by 30 % and the amount of carbon sinks by 50 %, compared to the application of solely chemical fertilizers (Li et al. 2009).

We found a negative linear correlation between crop productivity and the carbon intensity per yield, suggesting that increasing carbon intensity could not bring about an increase in crop production (Fig. 4b). But the trend for the use of excessive chemical fertilizers and other agricultural materials has been increasing recently in China’s agricultural production (Zhang et al. 2008). Considering the diminishing rates of return from chemical fertilizers along with their side effects on soils, the net amount of fossil energy required will increase in the foreseeable future. This trend adds more pressure for mitigating the environmental impact of crop production practices. In the study of agricultural carbon emissions and the development of policies to reduce them, it is essential to correctly recognize and understand the negative correlation between agricultural production and carbon emissions from agricultural material inputs.

Carbon emissions from the livestock and poultry industry

The emissions of CH4 and N2O from livestock production come mainly from animal gastrointestinal fermentation and manure management; the amount of emissions depends on how livestock are fed, manure handling, and the environment in which they are housed. CH4 emissions from livestock gastrointestinal fermentation account for one third of global agricultural non-CO2 emissions (USEPA 2006). Using the IPCC method to calculate GHG emissions from China’s livestock and poultry industry during 2000–2007, Hu and Wang (2010) found that ruminants caused the most CH4 emissions from gastrointestinal fermentation, while the major sources of CH4 and N2O from livestock feces were pigs and poultry. The 2011 emissions of CH4 and N2O from animal husbandry were 1041.81 and 40.87 × 104 tons, respectively, with gastrointestinal fermentation and manure management contributing most (Chen and Shang 2014). Thus, we used the measures of PRL and BRL to assess livestock and poultry industry GHG emissions. We found that in Hunan Province, carbon emissions from cattle and pig rearing accounted for 39 and 48 % of total carbon emissions from the livestock industry, and that there was a negative correlation between carbon emissions from those industries versus the pork production ratio. Interestingly, the beef ratio was one of the main factors that influenced emissions from the livestock and poultry industry (Fig. 5). The reason may be that CH4 produced by ruminant gastrointestinal fermentation is the main source of GHG in that industry, and the CH4 and N2O emission factor for ruminants is far greater than that of pigs. De Vries and de Boer (2010) calculated the carbon emissions per kg of yield for a variety of livestock and poultry in Europe; for example, they were 15–32 kg CE for beef, 4–11 kg CE for pork, and 4–6 kg CE for chicken. That suggested a huge potential for emission reductions in the rearing of ruminants, similar to our results in this study. Regression analysis showed that if the BRL declined by 1 %, livestock GHG emissions would decline by 329 units.

Fig. 5
figure 5

Correlation between carbon emissions from the livestock and poultry industry and the ratio of pig production to total animal production

Full size image

Per capita GDP is recognized as one of the key factors that influence the growth in total global carbon emissions (Tucker 1995; Taskin and Zaim 2000; Neumayer 2002; RauPach et al. 2007; Wagner 2008). Chen and Shang (2014) confirmed that economic factors are the greatest incentive for livestock industry GHG emissions in China. Here, we showed that if agricultural GDP per capita were to increase by one unit, the livestock and poultry industry’s GHG emissions would increase by 0.354 units. Presumably, economic development factors will remain as the main inducement for increased carbon emissions from livestock and poultry in developing countries, over the long term.

The livestock and poultry industry produces a large amount of carbon emissions, and the demand for the yield and quality of livestock products is steadily rising, so it will be very difficult to reduce GHG emissions from this source, in general. The trend of livestock development lies in large-scale and intensive development, which will bring more favorable conditions for the improvement of livestock quality, diet, disease prevention and control, waste treatment, and other technical aspects.

Thus, in Hunan, across China, and abroad, innovation of crop production technology and improvement of the quality of employees will not only promote unit productivity of the livestock and poultry industry, but will also reduce carbon emissions. Hermansen and Kristensen (2011) argued that, in addition to focusing on livestock manure management to reduce emissions, attention must also be paid to the utilization efficiency of manure as inorganic fertilizers in the crop industry, along with the development of bio-energy applications. In developing countries, this would be an effective measure to reduce emissions from livestock and poultry production, and to strengthen the agriculture industry through research and development of efficient bio-energy sources from livestock manure (e.g., methane).

Conclusion

Our findings show that among carbon emissions from Hunan’s agriculture sector, material inputs were the component with the highest growth rate, at 52.17 %. That source accounted for 50.79 % of total growth in carbon emissions from the province’s crop production during 1998–2012. To a certain extent, emissions from crop production were positively correlated with the utilization rate of nitrogen fertilizer, and crop productivity was linearly and negatively correlated with carbon intensity per yield. That suggested that continually increasing the amount of agricultural material inputs does not lead to a proportional increase of agricultural productivity. We cautioned that agricultural production and the development of policies to reduce agricultural emissions depend on an accurate understanding of the correlation between agricultural production and carbon emissions. Most importantly, they depend on finding the optimal economic combination of agricultural production and reduction in agricultural emissions. There was a weak state of decoupling between Hunan agricultural carbon emissions and their output value, but the decoupling was unstable. Priorities for mitigation should be improving the utilization efficiency of nitrogen fertilizers, protecting farmland to reduce emissions, and developing the crop industry in Hunan. Optimizing the production structure and enhancing low-carbon production methods and the scale of operation in order to improve productivity in the livestock and poultry industry will be key tasks for achieving emissions reduction and the sound development of Hunan’s livestock industry.