Prediction of methane emission from sheep based on data measured in vivo from open-circuit respiratory studies

Objective The current study analysed the relationships between methane (CH4) output from animal and dietary factors. Methods The dataset was obtained from 159 Dorper×thin-tailed Han lambs from our seven studies, and CH4 production and energy metabolism data were measured in vivo by an open-circuit respiratory method. All lambs were confined indoors and fed pelleted diet during the whole experimental period in all studies. Data from two-thirds of lambs were used to develop linear and multiple regressions to describe the relationship between CH4 emission and dietary variables, and data from the remaining one third of lambs were used to validate the established models. Results CH4 emission (g/d) was positively related to dry matter intake (DMI) and gross energy intake (GEI) (p<0.001). CH4 energy/GEI was negatively related to metabolizable energy/gross energy and metabolizable energy/digestible energy (p<0.001). Using DMI to predict CH4 emission (g/d) resulted in a coefficient of determination (R2) of 0.80. Using GEI, digestible energy intake, and metabolizable energy intake predict CH4 energy/GEI resulted in a R2 of 0.92. Conclusion the prediction equations established in the current study are useful to develop appropriate feeding and management strategies to mitigate CH4 emissions from sheep.


INTRODUCTION
There are increasing concerns about the impact of livestock production on the environment. Methane (CH 4 ), a greenhouse gas that remains in the atmosphere for approximately 9 to 15 years, is over 25 times more effective in trapping heat in the atmosphere than carbon dioxide (CO 2 ) [1]. Livestock farming is a major contributor to atmospheric CH 4 accumulation. In ruminants, approximately 95.5% of CH 4 generation is produced by fermentation of feed in the rumen [2], which causes a loss of 2.3% to 10.8% of feed energy depending on the diet and animal [3]. Therefore, national inventories of greenhouse gas emissions are essential for the quantification of these emissions from individual countries and the elaboration of country level mitigation strategies [4]. However, due to the complexity in determining CH 4 production in vivo, prediction equations are essential to accurately estimate CH 4 emission from ruminants, which is necessary to provide useful strategies for the feeding and management of animals. Although the prediction equation of CH 4 emission has been established for sheep by different researchers [5][6][7][8], it should be noted diet, animal breed, and management system could all contribute to errors when developing national CH 4 emission inventories [9].
The sheep production in China differs from that in Australia and in Europe, which are almost exclusively dependent upon pasture [10,11]. China is featured by vast population and scarce land per capita, as well as limited forests and pastureland [12]. Since the beginning of the 21st century, the Chinese Government has further implemented polices including "Control grazing for grassland recovery" to conserve grasslands, mitigate degradation, and promote economic development in pastoral regions [13]. On the other hand, China has long history of sheep domestication and currently there are 162.06 million sheep in China, accounting for about 14% of total sheep population in the whole world. Consequently, modern sheep production system in China features limited or even no grazing and the sheep are mainly fed on crop residues. Therefore, it is uncertain if those prediction model based on CH 4 production from grazing sheep can also accurately predict CH 4 production from sheep under current feeding conditions in China, which contribute substantially to the world greenhouse emissions due to the large sheep population.
The indigenous breeds such as Hu [14] and thin-tailed Han (small tail Han) [15] sheep are famous for high prolificacy and non-seasonal ovulatory activity. With the introduction of Dorper sheep from Australia in 2001 [16], Dorper sheep×thintailed Han crossbred sheep has become a predominant breed for meat-producing in China. In recent years, the nutrient requirements of Dorper×thin-tailed Han crossbred sheep in terms of energy [17][18][19][20], protein [21][22][23], and minerals [24] have been extensively studied and reported. Based on those conditions, dataset used in the current study to provide basic CH 4 emission parameters were obtained from the same breed (Dorper×thin-tailed Han), feeding conditions (confined indoors), and feeding regime (pelleted diet), which largely mirrors the current characteristics of sheep production system in China. Therefore, in current study, practical equations were established using dataset from our seven previous studies to predict CH 4 production from Dorper×thin-tailed Han crossbred lambs. Our objective was to provide basic information for the establishment of robust national CH 4 inventories and practical mitigation strategies to reduce the environmental impact of sheep production systems.

Animals and diets
The dataset used in the present study was obtained from 159 lambs in seven energy metabolism studies undertaken as part of the National Technology Program for the Meat Sheep Industry of China from 2010 to 2015 [17][18][19][20][25][26][27]. The animals were offered pelleted diet in all studies with concentrate:forage ratio ranging from 12:88 to 92:8. The concentrate included corn, soybean, barley, oat, wheat, sorghum, soybean meal, rapeseed meal, cottonseed meal, peanut meal, and dry distillers grains with solubles, and the roughage included Chinese wild rye hay (Leymus chinensis) and corn stalk. The data, composed of means, standard deviations and ranges for animal and dietary variables, are presented in Table 1. In each experiment the animals were offered the experimental diets for 4 weeks in group-housed pens before conducting the digestion and respirometry trial to measure energy metabolism. The digestion and respirometry trial comprised a 10-d collection period after a 10-d adaptation period. During the 10-d collection period, feed offered, orts, and faeces were weighed and sampled (10% of total weight) daily. Urine was acidified with 100 mL of 1.8 M H 2 SO 4 daily and measured for volume, and 1% was sampled daily. As outlined in Deng et al [17], methane production was measured using an open-circuit respirometry system (Sable Systems International, Las Vegas, NV, USA) integrated with 3 metabolism cages each equipped with a polycarbonate head box. On d 0, 2, 4, 6, and 8 of the 10-d collection period, each group of lambs was moved into the metabolism cages for methane assessment. After a 24-h adaptation period, individual methane production was measured over a 24-h period. Methane concentration as well as temperature, humidity, dew point and air flow rate were recorded and processed using the Sable Systems software to calculate individual sheep methane production.

Statistical analysis
Prediction equations for methane emission were developed using dry matter intake (DMI), neutral detergent fibre intake (NDFI), gross energy intake (GEI), digestible energy intake (DEI), metabolizable energy intake (MEI), dietary metabolisable energy/digestible energy (ME/DE), DE/gross energy (GE) or ME/GE as predictors in multiple regressions. A stepwise multiple regression technique was used to develop multiple prediction equations, and the technique automatically selects the best and significant predictors to fit the prediction equations. Experimental effects on these relationships were removed by the following model: where, a i represents the effect of experiment i for i = 1 to 5,

Correlation between methane emission and feed intake as well as energy digestibility and metabolisability
The relationships between CH 4 and feed intake as well as energy digestibility and metabolisability are shown in Table 2. Total CH 4 output (L/d) was linearly correlated with feed intake (DMI and NDFI) and energy intake (GEI, DEI, and MEI) (p<0.001). CH 4 emission expressed as L/kg DMI was linearly correlated with DMI, GEI, DEI (p<0.05), and NDFI (p<0.01).

Prediction equations for methane emission and validation of the equations
Using two-thirds of the data, linear regression was established between total CH 4 emission (L/d) and DMI (g/d) (R 2 = 0.80) and NDFI (g/d) (R 2 = 0.76), respectively ( Table 3). As the variation in CH 4 production was best predicted by these two parameters, multiple linear prediction equations were developed using DMI and NDFI (R 2 = 0.85). Linear regression was established between total CH 4 energy (MJ/d) and GEI (MJ/d) (R 2 = 0.80). Multiple regression was established between total CH 4 energy (MJ/d) and combination of GEI, DEI, and MEI (MJ/d) and the R 2 of the regression was 0.92. Further validation of those regression models was conducted using the remaining one third of the data ( Table 3). The results showed

Validation of previously published prediction equations for sheep emissions
The present study used datasets from seven studies (n = 159) to validate previously published prediction equations for CH 4 emission from sheep ( Table 4). The CH 4 emission (g/d) was under-predicted by Zhao et al [8] but over-predicted by Bell et al [7]. The R 2 for the relationship between predicted and actual CH 4 emission (g/d) was close to 0.70. The CH 4 -E was over-predicted using either DMI and GEI by Patra et al [6], or GEI, DEI, or MEI by Zhao et al [8]. The R 2 for the relationship between predicted and actual CH 4 -E was greatest in Patra et al [6] using DMI (R 2 = 0.70) or GEI (R 2 = 0.71) and in Zhao et al [8] using GEI (R 2 = 0.71), while the lowest R 2 was observed using MEI as the prediction factor (R 2 = 0.44). A lower CH 4 /DMI was obtained from the predicted value of Zhao et al [8] and our results (20.8 vs 27.2 g/kg). However, the CH 4 energy/GEI predicted by Zhao et al [8] was only 71% (5.95/ 8.37×100) of that measured in the current study. The R 2 in the relationship between predicted and actual CH 4 /DMI and CH 4 -E/GEI was the 0.62 and 0.59, respectively.

DISCUSSION
In the current study, average CH 4 emission was 39.9 L/d or 28 [29]. The average CH 4 scaled to DMI was 37.6 L/kg or 27.2 g/kg in the current study, which was considerably greater than that (16.5 to 21.1 g/kg) reported for sheep fed perennial ryegrass [8,30]. Furthermore, the lower limit of CH 4 emission scaled to DMI (18.8 L/kg or 13.6 g/kg) in the current study was close to that of Welsh Mountain sheep fed on permanent pasture (14.4 g/kg) or Molinia sheep (14.1 g/kg) [31]. In the current study, pelleted diets were used in all experiments, which theoretically can be more rapidly digested and thus promoting feed intake [32]. A study suggested that pelleting could increase DMI by 45% in sheep, especially for young animals [33], compared with grass. Similarly, we also observed higher DMI (1.04 kg/d) compared with others [8,30,31], which could be attributed to the pelleted diet used in our series of studies. Although it was reported that increasing feed intake can reduce CH 4 production per unit of feed intake [8], the substantial higher CH 4 emission (28.9 g/d) compared with others [8,30,31] could be responsible for the higher CH 4 emission scaled to DMI in the current study. Pinares-Patiño et al [34] reported a lower CH 4 emission (22.0 g/kg DMI) from ewes also fed pelleted diet   [35]. It was unexpected that CH 4 emission measured using SF 6 ranged from 26.7 to 27.9 g/kg DMI for grazing sheep reported by Savian et al [29], which was almost identical to our result. This might be due to the high NDF content in Italian ryegrass (from 586 to 606 kg/kg DM) used in their study. Despite of the dietary factors mentioned above, animal factors (breed, sex, and growth stage) as well as measurement technique can also have influence on CH 4 emission and therefore should be taken into consideration in the development of mitigation strategies. CH 4 emission as a proportion of energy losses accounts for 3.79% to 12.0% of GEI in the current study, which was comparable to the range reported in cattle (2% to 15%) [36]. The average ratio of CH 4 to total GEI in this study (8.4%) was higher than the average value reported for grazing sheep (6.2%) [8,29,30]. The lower energy utilization efficiency could be again explained by the high passage rate and low nutrient digestibility of sheep fed pelleted diet in the current study. For example, dry matter digestibility (DMD, 61.6%) and organic matter digestibility (OMD, 61.8%) observed in the current study were significantly lower compared with DMD reported by Moorby et al [31] (72.2%) and Zhao et al [8] (73%), and OMD reported by Fraser et al [30] (66.2%). Intergovernmental Panel on Climate Change Tier 2 methodology [37] currently uses GEI along with a standard CH4 conversion factor (CH 4 energy/GE = 6.5%) to calculate CH 4 emissions, thus, probably underestimating CH 4 emission from sheep under the experimental conditions in the current study.
In the current study body weight (BW) was not significantly correlated with CH 4 emission from sheep. Similarly, it is reported that BW alone is a poor variable for predicting CH 4 emission in grazing beef cattle (R 2 = 0.27) [9] and sheep (R 2 = 0.25) [6], and it was found that metabolic BW was marginally correlated with CH 4 energy (R 2 = 0.49) in goats [38], indicating that the accuracy of using BW to predict CH 4 emission might be affected by feeding conditions.
Feed intake is often used to predict CH 4 production in inventory models. In the present study, DMI is the main determinant of total CH 4 emission (R 2 = 0.80), a result similar to that obtained by Patra et al [6] in sheep (R 2 = 0.83). It is well documented that CH 4 emission (L/d) from enteric fermentation in sheep is closely related to total feed intake [8,39]. A strong relationship between DMI and CH 4 emission was also reported in beef and dairy cattle (R 2 = 0.68) [40]. A quadratic relationship between CH 4 energy and DMI in dairy cows was also observed [41]. However, a study suggested that the prediction equations based on DMI as primary pre-dictors of CH 4 output resulted in a relatively weak R 2 (0.44) in beef cattle [40]. This might suggest that the inclusion of other variables, such as BW and dietary nutrient concentrations, may be important to improve the predictive accuracy of regression models. Nevertheless, Ellis et al [40] reported that NDFI (kg/d) was the best predictor of CH 4 production (R 2 = 0.66) in beef cattle, and further combination of DMI and NDFI could also robustly predict CH 4 emission from cattle (R 2 = 0.67), which was in accordance with the regression models established in our study. The NDF fraction contains cell-wall fractions such as cellulose, hemicellulose, and lignin [42]. The positive relationship between NDFI (kg/d) and CH 4 production in the current study along with the study by Ellis et al [40] might be explained by the dietary NDF concentration, which could improve ruminal fermentation and lead to preferable high acetate:propionate ratio that facilitates CH 4 production [43], making it an easily measured predictor of CH 4 production within a regression model.
Energy intake (GEI alone or GEI, DEI and MEI) is also effective prediction factors of CH 4 emission in the current study, which are in accordance with those observed in cattle [9,40,44]. In agreement with Molano and Clark [45], the quantity of CH 4 emission, per unit of DMI or GE losses as CH 4 was not affected by the level of DMI. In the current study, there was a negative relationship between CH 4 /GEI and dietary ME concentrations or ME/DE, which is similar to that reported in beef [9] and dairy cattle [46], indicating that an improved feed utilisation efficiency could reduce CH 4 emissions. On the other hand, we observed a positive correlation between DMI and CH 4 /GEI, which is inconsistent with previous result in dairy cow [46] and sheep [8]. Indeed, an increase in feeding level (DMI) increases the outflow rate of digesta and thus reduces ruminal nutrient digestion, leading to decrease in CH 4 [46]. However, in the current study, sheep with higher DMI also consumed relatively more concentrate than those with lower DMI. Previous study suggested that ruminal nutrient digestion increased with increasing concentrate intake [47], which in turn result in the increase in CH 4 output. Therefore, the positive correlation DMI and CH 4 / GEI observed in the current study can be expected.
Due to the scarcity of relevant studies for sheep, predicted CH 4 emission parameters using equations from 3 published papers were compared with the actual CH 4 production in the current study. Both Bell et al [7] and Zhao et al [8] developed prediction models for enteric CH 4 emissions using sheep in UK, where the sheep production is featured by long grazing seasons [48]. Therefore, the use of those equations in confined-feeding animals must be with caution. Patra et al [6] established prediction model for CH 4 emission based on the results of more than 1,500 sheep. Although the equations in their study might be more inclusive, it should be noted that the predicting equations established in the current study were more specific in the method (respiratory chamber) used and feeding conditions (confined and fed pelleted diet), which could be more accurate to calculate the CH 4 inventory under similar conditions.

CONCLUSION
In the present study, a range of prediction equations for methane production from sheep was based on in vivo data from open-circuit respiratory studies. Strong relationships were found between methane production and animal or dietary factors including DMI, NDFI, and GEI. These equations are useful to develop appropriate feeding and management strategies for mitigating methane emission from sheep under current feeding system in China.

CONFLICT OF INTEREST
We certify that there is no conflict of interest with any financial organization regarding the material discussed in the manuscript.