Quantification and prediction of enteric methane emissions from Chinese lactating Holstein dairy cows fed diets with different dietary neutral detergent fiber/non-fibrous carbohydrate (NDF/NFC)ratios

DONG Li-feng,JlA Peng,Ll Bin-changWANG BeiYANG Chun-lei,LlU Zhi-haoDlAO Qi-yu

1 Feed Research Institute,Chinese Academy of Agricultural Sciences/Sino-US Joint Lab on Nutrition and Metabolism of Ruminant/CAAS-CIAT Joint Laboratory in Advanced Technologies for Sustainable Agriculture,Beijing 100081,P.R.China

2 Institute of Ruminant Research,College of Pastoral Agriculture Science and Technology,Lanzhou University,Lanzhou 730020,P.R.China

3 College of Biotechnology and Bioengineering,Zhejiang University of Technology,Hangzhou 310014,P.R.China

Abstract Methane (CH4) emissions from ruminant production are a significant source of anthropogenic greenhouse gas production,but few studies have examined the enteric CH4 emissions of lactating dairy cows under different feeding regimes in China. This study aimed to investigate the influence of different dietary neutral detergent fiber/non-fibrous carbohydrate(NDF/NFC) ratios on production performance,nutrient digestibility,and CH4 emissions for Holstein dairy cows at various stages of lactation.It evaluated the performance of CH4 prediction equations developed using local dietary and milk production variables compared to previously published prediction equations developed in other production regimes. For this purpose,36 lactating cows were assigned to one of three treatments with differing dietary NDF/NFC ratios:low (NDF/NFC=1.19),medium (NDF/NFC=1.54),and high (NDF/NFC=1.68). A modified acid-insoluble ash method was used to determine nutrient digestibility,while the sulfur hexafluoride technique was used to measure enteric CH4 emissions.The results showed that the dry matter (DM) intake of cows at the early,middle,and late stages of lactation decreased significantly (P<0.01) from 20.9 to 15.4 kg d–1,15.3 to 11.6 kg d–1,and 16.4 to 15.0 kg d–1,respectively,as dietary NDF/NFC ratios increased. Across all three treatments,DM and gross energy (GE) digestibility values were the highest (P<0.05)for cows at the middle and late lactation stages. Daily CH4 emissions increased linearly (P<0.05),from 325.2 to 391.9 kg d–1,261.0 to 399.8 kg d–1,and 241.8 to 390.6 kg d–1,respectively,as dietary NDF/NFC ratios increased during the early,middle,and late stages of lactation. CH4 emissions expressed per unit of metabolic body weight,DM intake,NDF intake,or fat-corrected milk yield increased with increasing dietary NDF/NFC ratios. In addition,CH4 emissions expressed per unit of GE intake increased significantly (P<0.05),from 4.87 to 8.12%,5.16 to 9.25%,and 5.06 to 8.17%respectively,as dietary NDF/NFC ratios increased during the early,middle,and late lactation stages. The modelling results showed that the equation using DM intake as the single variable yielded a greater R2 than equations using other dietary or milk production variables. When data obtained from each lactation stage were combined,DM intake remained a better predictor of CH4 emissions (R2=0.786,P=0.026) than any other variables tested. Compared to the prediction equations developed herein,previously published equations had a greater root mean square prediction error,reflecting their inability to predict CH4 emissions for Chinese Holstein dairy cows accurately. The quantification of CH4 production by lactating dairy cows under Chinese production systems and the development of associated prediction equations will help establish regional or national CH4 inventories and improve mitigation approaches to dairy production.

Keywords:methane emission,feeding regime,prediction equation,lactating dairy cow

1.lntroduction

In recent decades,the livestock industry has become increasingly important;worldwide,approximately 17% of all kilocalories and 33% of all protein consumed come from livestock products (Steinfeldet al.2006;Lan and Yang 2019). Ruminants,one of the most important sectors of livestock systems,effectively convert human-inedible plant biomass into high-quality human-consumable protein in the form of milk and meat (van Gastelenet al.2019).However,methane (CH4),the final by-product of the ruminal fermentation of dietary structural carbohydrates via methanogenesis,increases the accumulation of atmospheric greenhouse gases,contributing to extreme climate changes (Hristovet al.2013). Globally,ruminants are estimated to produce 5.7 gigatons CO2equivalent per annum,of which 2.5 gigatons are associated with beef production and 2.1 gigatons are associated with milk production (Kamilariset al.2020). Furthermore,due to the continuous increase in dairy cow populations and milk production over the past few decades,it is projected that CH4emissions from Chinese livestock production systems will total 114.4 Mt by 2030 (Höglund-Isaksson 2012).As part of its domestic and international commitment to sustainable development,the Chinese government has announced that carbon emissions will be peaked by 2030,with carbon emissions reduced 60–65% from 2005 levels(Yanget al.2018). Because enteric CH4emissions are a natural energy cost of ruminant production,ranging from 2–12% of the gross energy (GE) intake for cattle (Johnson and Johnson 1995),the quantification and reduction of enteric CH4emissions will substantially improve dairy energy utilization efficiency,alleviate environmental pressure,and fulfil greenhouse gases reduction commitments.

Several recent studies have established enteric CH4prediction equations based on individual emission values to develop national or regional emission inventories and mitigation approaches for the Chinese livestock industry (Elliset al.2007). For example,Donget al.(2019) measured enteric CH4emissions from heifers at various growth stages and developed CH4prediction equations based on nutrient intake parameters. A general methodology aiming to quantify CH4emission inventories in different countries was recommended by the International Panel on Climate Change (IPCC 2006).However,this methodology does not consider variations in geographic conditions,animal physiology stages,and dietary components. For example,Hassanatet al.(2017)found that the average CH4/DM intake was 18.4 g kg–1for lactating dairy cows fed typical Canadian diets (i.e.,with high levels of corn silage). In contrast,Beukeset al.(2010)reported that the CH4/DM intake was 25.1 g kg–1for cows grazing perennial grass/clover pastures in New Zealand.Finally,Yanet al.(2000) measured the CH4emissions of dairy cows fed typical Western European grass silagebased diets,and these data were incorporated into the official GHG modelling system for the UK.

Although several reports on enteric CH4emissions in China are available,little attempt has been made to significantly improve animal genetic merit,feed quality,and management practices in response to these reports. Lageet al.(2021) demonstrated that nutritional composition (e.g.,dietary fiber and starch content) could significantly affect CH4emissions due to variations in digestion and energy utilization efficiency. In addition,as wide discrepancies in dairy cow CH4emission levels have been identified among different production systems,the use of fixed values from the IPCC or studies in other regions may lead to substantial errors in the development of national or regional CH4inventories in China. Based on previous studies,this work hypothesized that dietary neutral detergent fiber/non-fibrous carbohydrate (NDF/NFC)ratios would affect the enteric CH4emissions of Chinese lactating Holstein dairy cows,and that prediction equations developed using local emissions values would be more accurate than those developed using nonlocal emissions values. The objectives of this study therefore were:(1) to examine the enteric CH4emissions of Holstein dairy cows fed typical Chinese diets;(2) to develop prediction equations using data collected with the sulfur hexafluoride (SF6) method;and (3) to compare the newly developed equations to other existing equations developed based on different production systems.

2.Materials and methods

2.1.Animals,experimental design,and diets

This study was conducted in 2018 at the Zhongjiayonghong Dairy Farm (Fangshan District,Beijing,China;latitude 39°39´6´´N,longitude 116°12´21´´E).

A total of 36 lactating Holstein dairy cows were used in this study,with an average days in milk (DIM) of 59±7.2(mean±SD),an average milk yield of (18.7±2.1) kg d–1,and an average body weight (BW) of (587±76.8) kg.Measurements were taken during the early,middle,and late lactation stages. Each measurement period lasted 32 days:18 days of adaptation,ten days of CH4emission measurement,and five days of nutrient digestibility measurement. Before the commencement of each period,animals were divided into three balanced blocks based on initial BW,parity,DIM,and milk yield (n=12),and these blocks were allocated to one of the following three treatments following a randomized complete block design:(1) low dietary NDF/NFC ratio (1.19),(2) medium dietary NDF/NFC ratio (1.54),and (3) high dietary NDF/NFC ratio (1.68). The animals were housed in individual stalls covered with mats and sawdust to measure feed intake and nutrient digestion. Diets were formulated to ensure that nutrient and energy content remained constant throughout the experiment so that the effects of dietary composition on enteric CH4emissions could be investigated. The major forage components were corn silage,oat grass,alfalfa hay,and Chinese wildrye grass,which are common feed items on Chinese dairy farms. The composition and nutrient contents of the feed ingredients are shown in Table 1. TMRs were prepared daily using a feed mixer(Belle Engineering Ltd.,Derbyshire,UK) and distributed to individual cows twice daily:between 0600 and 0800 h,and between 1600 and 1800 h. All cows had free access to water throughout the experiment.

Table 1 Dietary ingredient and chemical composition in the current study (as fed-basis,unless otherwise noted)

2.2.Sampling,measurement,and analysis

Feed intake and nutrient contentCow BW was determined before the commencement of each experimental period. Offered and refused feed was collected and weighted to determine daily feed intake.All samples were dried in a forced-air oven at 65°C for 48 h,ground,and passed through a 1-mm screen in a Cyclotec mill (Tecator 1093,Tecator,Hoganas,Sweden)for chemical analysis. Dry matter and crude protein were determined using AOAC International method 990.03 (2016). Acid detergent fiber (ADF) and NDF were measured as described by Van Soestet al.(1991).Ash content and ether extract (EE) concentration were determined using AOAC International method 942.05(AOAC 2016).

Methane collection and measurementEnteric CH4emissions from each animal were measured using a modified SF6tracer technique,as Moateet al.(2016)described. Measurements were conducted in the early,medium,and late stages of lactation,with each measurement period lasting a maximum of 10 consecutive days. Permeation tubes were filled with 450 mL of ultra-pure SF6gas by immersing them in liquid nitrogen and calibrated in a dry incubator set at 39°C for over 8 wk. The calculated SF6release rates for the early,medium,and late stages of lactation were 3.22–3.75 mg d–1(mean±SD,(3.3±0.15) mg d–1),3.14–3.80 ((3.4±0.21) mg d–1),and 3.19–3.54 ((3.3±0.17) mg d–1),respectively.Stainless canisters (1.85 L) with an initial sampling rate of approximately 0.53 mL min–1were used for continuous sample collection. Background SF6and CH4samples were collected daily by placing six additional canisters in the experimental barn,some mounted on the backs of the animals and some fixed about 2.5 m above ground level.Enteric and background gas samples were measured using a gas chromatograph (GC126,Shanghai Precision Instruments Co.,Ltd.,Shanghai,China) equipped with a flame-ionization detector and an electron-capture detector. The concentrations of the background samples of CH4and SF6were (23±7.5) ppm and (9.7±1.42) ppt,respectively.

Milk production and compositionDuring each CH4measurement period,milk yield was recorded twice daily(at 0600 and 1600 h) for six consecutive days using a DeLaval milk meter (MM25,DeLaval International,Tumba,Sweden). Morning and evening milk samples were combined in equal proportions to determine fat,protein,and lactose contents (Hassanatet al.2017). The fat-corrected milk yield of each cow,standardized to 4.0%fat,was calculated following the recommendations of Feeding Standard for Dairy Cattle (NY/T 34-2004 2004).

Total-tract apparent nutrient digestibilityDuring the final five days of each measurement period,four cows were randomly selected from each treatment for nutrient digestibility measurements using a modified acid-insoluble ash method (Van Keulen and Young 1997). Feces were collected from the rectum of each selected cow at various time points to obtain representative samples (day 1:1000 h and 2200 h;day 2:0200 h and 1400 h;day 3:0500 h and 1700 h;day 4:0800 h and 2000 h;day 5:1100 h and 2300 h). Fecal samples were freeze-dried,ground,and passed through a 2-mm screen. The acid-insoluble ash content in each filtered sample was measured after the addition of hydrochloric acid (Donget al.2015).

2.3.Statistical analyses

The effects of dietary NDF/NFC ratios on the growth performance,nutrient digestion,and enteric CH4emissions of lactating dairy cows were evaluated using two analytical approaches (ANOVAs and linear mixed modelling) in the GenStat statistics system (version 16.2,Payneet al.2013).

ANOVAThe dependent variables,including CH4emissions,DM intake,nutrient digestibility,and milk yield,were measured at each lactation period and analyzed using ANOVAs. Each animal was considered an experimental unit. In these analyses,the treatment groups were fitted as a fixed effects,and the animals in each treatment were fitted as random effects. For evaluating CH4emissions,additional variables were also fitted as covariates,including BW,parity,days in milk,and days of pregnancy. Effects were considered significant atP<0.05,and differences among treatments were identified using post-hoc tests (Fisher’s least significant difference).The effects of different dietary NDF/NFC ratios were evaluated using polynomial linear and quadratic contrasts as follows:

whereYiis the observation made on the ithtreatment,μis the overall mean,Ciis the effect of NDF/NFC,andεiis the residual error.

Model developmentPrediction models for the dependent variable methane energy output (CH4-E)were generated separately using the independent variables (i.e.,feed intake,nutrient content,and milk yield parameters) obtained during each measurement period.Prediction models for CH4-E across the entire lactation period (including the early,middle,and late lactation stages) were also estimated. The significance level was set to P≤0.05 so that the most relevant explanatory variables could be included in the prediction models.The performance of each developed CH4-E model was assessed using the root mean square error of prediction(Moraeset al.2014).

Comparison with extant modelsThree existing models were compared with the model developed herein:the IPCC Tier 2 Model (IPCC 2006),the lactating dairy cow model of Yanet al.(2000),and the animal model for the United States of Niuet al.(2018). Following St-Pierre (2003),the mean square prediction error (MSPE),the square root of MSPE (RMSPE),and the coefficient of determination (R2) were calculated to evaluate the prediction quality of each model (Velarde-Guillénet al.2019). MSPE was calculated as the sum of the squared differences between the observed and predicted CH4emissions,divided by the number of observations:

whereOiis the observed value,Piis the predicted value,andnis the total number of observations. RMSPE was expressed as a proportion of the observed mean,providing a reliable estimate of the overall prediction error.

MSPE was decomposed into three components:the first component was the error due to the overall bias of prediction or the error in central tendency (ECT),representing the deviation of the predicted values concerning the observed values;the second component was the error due to linear bias or regression (ER);and the third component was the error due to random variation or disturbance (ED),representing the variation in observed values unexplained after the mean and the regression biases have been removed. The concordance correlation coefficient (CCC) was calculated to evaluate the precision and accuracy of the predicted versus observed values for each model. The CCC value was the product of two components:the first component was the correlation coefficient (r),which measures precision (i.e.,the deviation of the observations from the best fit line),and the second component was the bias correction factor(Cb),indicating accuracy (i.e.,how far the regression line deviates from the line of unity). Another estimate (μ)measured location relative to scale (i.e.,the difference between the means relative to the square root of the product of the two standard deviations). A negative value ofμindicates over-prediction of observed values by the model,while a positive value indicates under-prediction of observed values by the model (Patra 2014;Patra and Lalhriatpuii 2016).

The IPCC Tier 2 Model incorporates fixed CH4conversion factors (Ym,calculated as CH4-E/GE intake) for different categories of livestock,animal production,and feed utilization efficiencies. TheYmvalue proposed by IPCC (2006) for lactating dairy cows is 6.5% of GE intake.The prediction equation used was as follows:

Yanet al.(2000) summarized energy metabolism and CH4emissions data using calorimetric chambers and developed a range of prediction models for lactating dairy cows. The equation used by Yanet al.(2000) was:

Niuet al.(2018) developed several models to predict enteric CH4emissions from lactating dairy cows using an intercontinental database that included animal performance,dietary nutrient content,and feed intake variables. Several models from Niuet al.(2018) were selected for use in this study as they were developed using an intercontinental database and had relatively small prediction errors and residual variances:

3.Results

3.1.Dietary effects on production performance and nutrient digestibility

During the early,middle,and late stages of lactation,BW did not differ significantly among treatments (P>0.05;Table 2). However,the DM intake of cows in the early,middle,and late lactation stage decreased (P<0.01)linearly,from 20.9 to 15.4 kg d–1,15.3 to 11.6 kg d–1,and 16.4 to 15.0 kg d–1respectively,as dietary NDF/NFC ratios increased. Similarly,cows fed the low NDF/NFC ratio diet had the greatest GE intake across all treatments (P<0.01)(low,368.7 MJ d–1;medium,232.6 MJ d–1;high,271.9 MJ d–1) during the early lactation stage. GE intake was similar (P>0.05) across cows fed medium and high NDF/NFC ratio diets during the middle lactation stage (medium,237.1 MJ d–1;high,204.5 MJ d–1) and the late lactation stage (medium,272.4 MJ d–1;high,267.4 MJ d–1). Milk yield was highest (P<0.05) for cows fed the low NDF/NFC ratio diet irrespective of lactation stage;milk yield was similar (P>0.05) among cows fed medium and high NDF/NFC ratio diets across all lactation stages.

NDF and CP digestibility were significantly higher(P<0.05) for cows fed the low NDF/NFC ratio diet than for medium or high NDF/NFC ratio diet cows;NDF and CP digestibility did not differ (P>0.05) significantly betweenthe medium and high NDF/NFC ratio treatments,irrespective of lactation stage (Table 2). DM and GE digestibility were similar (P>0.05) across cows fed low,medium,and high NDF/NFC ratio diets during the early lactation stage. However,DM and GE digestibility were significantly higher (P<0.05) for dairy cows in the low NDF/NFC ratio treatment group than for the other two treatment groups at the middle and late lactation stages.

Table 2 Growth performance and nutrient digestibility of Chinese lactating Holstein dairy cows fed diets with different dietary neutral detergent fiber/non-fibrous carbohydrate (NDF/NFC) ratios1)

3.2.Dietary effects on enteric CH4 emissions

Overall,daily CH4emissions increased (P<0.05)linearly (from 325.2 to 391.9 kg d–1,261.0 to 399.8 kg d–1,and 241.8 to 390.6 kg d–1) as dietary NDF/NFC ratios increased from 1.19 to 1.68 during the early,middle,and late lactation stages (Table 3). Similarly,CH4emission expressed per kilogram of DM intake was greatest for cows fed the high NDF/NFC diet(P<0.05) irrespective of lactation stage. During the early and late lactation stages,CH4emissions expressed per kilogram of metabolic BW were similar(P>0.05) between cows fed the low and medium NDF/NFC diets. CH4/BW0.75was greater (P<0.01) for cows fed the high NDF/NFC diet than for those fed low and medium NDF/NFC diets during the middle lactation stage. Across all three treatments,methane yield per unit FCM was highest (P<0.05) for cows fed the high NDF/NFC diet during the early and middle lactation stages. CH4/FCM yield was similar between cows fed low and medium NDF/NFC diets,and the CH4/FCM yields were significantly lower (P<0.05) than that of cows fed the high NDF/NFC diet during the late lactation stage.In addition,CH4emission expressed per unit of GE intake (Ym) increased (P<0.05) linearly (from 4.87 to 8.12%,5.16 to 9.25%,and 5.06 to 8.17%) as dietary NDF/NFC ratios increased from 1.19 to 1.68 in the early,middle,and late lactation stages.

3.3.Development and evaluation of CH4 prediction models

A range of linear CH4prediction equations were developed using the dietary and milk yield variables obtained during each lactation stage (Table 3). Overall,the linear relationships between CH4emissions and the production parameters derived from individual experimental periods or across all three periods were significant (P<0.05),withR2andP-values of 0.451–0.786 and 0.014–0.046,respectively (Table 4). Across all experimental periods,there were stronger positive relationships between CH4-E(MJ d–1) and DM intake (kg d–1) (R2values of 0.748,0.782,and 0.761 for the early,middle,and late lactation stages,respectively) than between CH4-E (MJ d–1) and NDF intake (kg d–1),FCM yield (kg d–1),or GE intake (MJ d–1). Across all equations and experimental periods,the weakest correlation was observed between CH4-E (MJ d–1)and FCM milk yield (kg d–1). When data obtained from each lactation stage were combined,CH4emission levels were most strongly correlated with DM intake. However,moderate correlations (R2=0.551) were also observed between CH4-E (MJ d–1) and DM intake (kg d–1),and between CH4-E (MJ d–1) and NDF content (% DM) (eq.(15)).

The values of RMSPE and other evaluation parameters for the equations developed herein and previously published equations are given in Table 5. Of the linear CH4prediction equations developed for cows at the early,middle,and late stages of lactation,the equations based on DM intake were the best predictors of CH4production,as indicated by lower RMSPE (20.4,21.2,and 21.7%for eqs.(1),(5),and (9),respectively),greater precision(CCC=0.86,0.85,and 0.84 for eqs.(1),(5),and (9),respectively),and higher accuracy (Cb=0.92,0.91,and 0.94 for eqs.(1),(5),and (9),respectively). Of the linear CH4predictions developed using the complete lactation dataset,the equation based on DM intake (eq.(13))was the best predictor of CH4production. This equation had the lowest RMSPE (17.7%,with 99.83% error from random sources),greatest precision (CCC=0.88),and highest accuracy (Cb=0.96). The equation based on NDF intake (eq.(14)) was the next best predictor of CH4production,with an RMSPE of 18.2% (89.83% error from random sources),a CCC of 0.85,and a Cbof 0.91. The linear CH4prediction equation based on DM intake and NDF content (eq.(15)) had a relatively high RMSPE(29.6%,with 80.55% error from random sources) and a CCC value of 0.77.

Table 3 Enteric methane emissions of Chinese lactating Holstein dairy cows fed diets with different dietary neutral detergent fiber/non-fibrous carbohydrate (NDF/NFC) ratios1)

Table 4 Prediction equations of enteric methane emissions for Chinese lactating Holstein dairy cows under different lactation stages and the whole lactation stage

Plots of observed versus predicted CH4for eqs.(13),(15),and (17),as well as previously published equations from the IPCC (2006),Yanet al.(2000),and Niuet al.(2018),are presented in Figs.1,2,and 3. Enteric CH4output was overestimated by the linear equations proposed by the IPCC (2006),Yanet al.(2000),and Niuet al.(2018),as indicated by strongly negativeμvalues and high ECT values,which were 21.5–51.5% of the RMSPE (Table 5). However,the linear equation of Niuet al.(2018) based on DM intake and NDF content had a lowμvalue (0.002) and high precision (CCC=0.78) with an RMSPE of 27.5%.

Table 5 Mean square prediction error analysis using developed and extant methane prediction equations

4.Discussion

The hypothesis that the dietary NDF/NFC ratios might affect the enteric CH4emissions of lactating dairy cows was supported by the present study results. These results demonstrated that increasing dietary NDF/NFC ratios decreased feed intake,milk production,and nutrient digestibility across all lactation stages. In addition,the results herein were consistent with the hypothesis that prediction equations developed using local emissions values would have a greater prediction accuracy due to accurate measurements of enteric CH4emissions from individual animals. Therefore,the findings of this study provided an empirical basis for the determination and reduction of CH4emissions from Holstein dairy cows fed diets with a variety of NDF/NFC ratios.

The dietary NDF ratio is considered the best single chemical predictor of ruminant DM intake due to its close relationship to rumen filling.In contrast,dietary NFC content,composed of varying amounts of starch,simple sugars,and soluble fiber,indicates dietary energy density(Allen 2000). Generally,increases in dietary NDF/NFC ratios lower DM intake and production performance.Here,the increase of dietary NDF/NFC ratio from 1.19 to 1.54 significantly decreased DM intake from 20.9 to 15.4 kg d–1,15.3 to 11.6 kg d–1,and 15.4 to 15.1 kg d–1for cows in the early,middle,and late stages of lactation,respectively. Jiaoet al.(2014) found that,compared to cows fed low-NDF diets,cows fed diets with higher NDF/NFC ratios had lower enteric solid passage rates and higher digesta retention times,leading to decreases in DM intake and production performance (Krizsanet al.2010;Obersonet al.2019). In addition,increases in dietary NFC levels improved lactation performance and the efficiency of energy and nitrogen utilization (Aguerreet al.2012).Here,FMC yield increased nearly 3.6 kg d–1in response to higher dietary NFC levels. Consistent with this result,Agleet al.(2010) showed that milk yield increased by approximately 2.8 kg d–1as dietary concentrate content increased from 52 to 72%.

The results of this study indicated that decreases in dietary NDF/NFC ratios increased DM intake and improved nutrient digestibility for cows at any lactation stage,consistent with a previous study using similar NDF/NFC ratios (Donget al.2019). DM digestibility decreased by 3.51,9.45,and 5.03% in the early,middle,and late lactation stages,respectively,as dietary NDF/NFC ratios increased from 1.19 to 1.68. Appuhamyet al.(2016) reported that variations in forage type and diet composition affected nutrient digestibility. In contrast to studies in other regions,which used fresh grass or grass silage,this study used forage common in Chinese dairy production regimes,including corn silage,DDGS,and other local agricultural by-products. These different forage sources might explain the variations in nutrient digestibility among the three treatments across the three lactation periods. Increases in milk and FMC yield are generally considered to result from increased dietary NFC content intake and improved nitrogen digestibility (Lageet al.2020). Indeed,a recent study showed that total tract CP digestibility in beef cattle increased from 53.1 to 58.1% as dietary concentrate levels increased (Trottaet al.2018). However,other researchers have suggested that changes in dietary CP content,not concentrate content,lead to changes in dietary CP digestibility in ruminants (Moateet al.2016). In the present study,CP content was increased to 18.4,17.5,and 16.4% in the low,medium,and high NDF/NFC-ratio diets,respectively,while dietary concentrate content was decreased to 20,17,and 14%,respectively. It was thus difficult to elucidate the direct relationship between CP digestibility and dietary composition. Indeed,cofounding factors may require further consideration in subsequent analyses (Donget al.2019).

Methane production was 241.8–399.8 kg d–1in the present study. These values were consistent with those of cows measured using respiration chambers (67.8–701 g d–1),the GreenFeed System (139.0–728.6 kg d–1),and the SF6technique (109.2–710.7 kg d–1) as part of the Global Network Project (Hristovet al.2018). In previous studies,disparities among CH4emission levels have consistently been attributed to differences in dietary composition,feed quality,and animal performance (Niuet al.2018). Further,daily CH4production increased significantly with increased dietary NDF/NFC ratios in the present study,consistent with previous studies conducted under confinement or grazing conditions (Aguerreet al.2011;Wattiauxet al.2019). However,Aguerreet al.(2011) reported that,as dietary forage-to-concentrate ratios increased from 47:53 to 68:32,daily CH4emissions ranged from 538 to 648 kg d–1,which was much higher than what was found here.Agleet al.(2010) showed that cows fed a 28:72 dietary forage-to-concentrate diet produced less enteric CH4than those fed a 48:52 dietary forage-to-concentrate diet.Similarly,when expressed per kilogram of DM intake or as a proportion of GE intake,cows fed low dietary NDF/NFC ratios produced significantly less CH4than those fed high dietary NDF/NFC ratios in the present study. Hristovet al.(2013) demonstrated that increasing the proportion of concentrate in the diet lowered CH4emissions per unit of feed intake and animal product if production remained the same or was increased. Enteric CH4production is directly associated with the formation of hydrogen.Therefore,diets high in starch lead to increased acetate and hydrogen production,which may enhance CH4production (Moateet al.2016). However,Hristovet al.(2013) suggested that high dietary starch levels might inhibit methanogenesis due to the rapid decrease of pH in the rumen.

It is essential to establish regional or national enteric CH4emission inventories to develop mitigation approaches and implement GHG reduction targets for dairy production systems. Although default values ofYmare provided in the IPCC guidelines for the calculation of emission inventories,the use of these default values may lead to significant errors,asYmcan vary considerably with geographic conditions,cow genetic potentials,dietary characteristics,and production systems (Donget al.2019). Hassanatet al.(2017) showed that theYmof lactating dairy cows fed conventional corn silage in Canada was 5.52%,lower than the defaultYmvalue (6.5%)for adult dairy cows under the IPCC Tier 2 approach. In contrast,van Wyngaardet al.(2018) found that theYmof grazing Jersey cows in South Africa ranged between 7.9 and 9.0%,slightly higher than values reported by recent grazing studies (5.3vs.6.7%;Jiaoet al.2014).Furthermore,a previous study reportedYmvalues as high as 11.4% for cattle fed diets containing tropical grass (Tangjitwattanachaiet al.2015). In this study,Ymvalues varied from 4.87 to 10.25%,attributed to the significant variations in dietary concentrate content and daily DM intake. A recent meta-analysis showed thatYmdecreased substantially when dietary concentrate level increased (Hristovet al.2013). Van Wyngaardet al.(2018) found thatYmdecreased when the concentrate component increased from 0 to 46%. This was consistent with the results of Jiaoet al.(2014),who reported thatYmdecreased when the concentrate component increased from 12 to 46%. Variations inYmmay result from changes in dietary components due to the diet-associated modifications in the enteric fermentation environment and methanogenesis functions,which consequently affect CH4emissions and energy partitioning.

Fig.1 Regression between observed and predicted methane (CH4) emissions derived from the proposed equation in the current study (A) and using the equation of Niu et al. (2018) (B). Dietary dry matter intake was used as a single independent variable in both equations.

Fig.2 Regression between observed and predicted methane (CH4) emissions. A,derived from the proposed equation in the current study. B,using the equation of Intergovernmental Panel on Climate Change (IPCC 2006). C,using the equation of Yan et al. (2000). D,using the equation of Niu et al. (2018). Dietary gross energy intake was used as a single independent variable in the proposed,IPCC (2006),Yan et al. (2000),and Niu et al. (2018) equations.

Empirical models that relate nutrient intake to CH4emissions are becoming a much-needed alternative tool due to their simplicity and low application cost. In the present study,CH4emission data were obtained from Chinese Holstein dairy cows during different lactation periods using the SF6technique,and several key explanatory variables that predicted individual or overall CH4emission levels were identified. Of all the variables identified,DM intake was most strongly correlated with enteric CH4emissions,irrespective of the lactation period. These results were consistent with those of Elliset al.(2007) and Millset al.(2003),who reported that DM intake was the best predictor of daily enteric CH4production (R2values of 0.64 and 0.60,respectively). Similarly,a meta-analysis showed a significant correlation between DM intake and CH4emissions for dairy and beef cattle (Charmleyet al.2016).An examination of more than 40 empirical models suggested that DM intake predicted CH4emissions with satisfactory accuracy (Appuhamyet al.2016). Niuet al.(2018) developed a set of global and regional models for various variables based on data from 5 233 individual dairy cows and reported a robust correlation between CH4emission and DM intake (R2=0.90). Positive correlations between dietary NDF proportion and CH4emissions were also identified (Niuet al.2018),consistent with many previous reports establishing the roles of dietary fiber and carbohydrates in enteric fermentation. However,it is important to note that the coefficients of determination of the prediction models were strongly affected by various factors,including dietary characteristics,the physiological condition of the animal,and management practices (Yanet al.2000). Here,the coefficient of determination for the correlation between DM intake and CH4emission was highest for cows at the middle lactation stage (R2=0.782),whereas prediction models based on data from the early lactation stage did not show a strong correlation between these factors (R2=0.748). Jiaoet al.(2014) reported a strong correlation between CH4-E and GE intake for heifers at 6 and 12 months (R2values of 0.72 and 0.80,respectively),but a weak correlation between CH4-E and GE intake for heifers at 18 and 22 months (0.24 and 0.22,respectively). Moraeset al.(2014) showed that dietary nutrient composition and feeding behavior might affect the emissions of grazing and feedlot animals. Therefore,prediction models developed based on a given production system should be applied to different feeding systems with caution.

Here,significant but moderately weak correlations were also observed between FCM yield and CH4emissions for cows at the early,middle,and late stages of lactation(R2values of 0.518,0.572,and 0.605,respectively). Niuet al.(2018) reported that energy-corrected milk yield and milk component-based models had lower RMSPE values(24.8 and 24.3%,respectively) than a model based on milk yield alone (26.5%). The present study results also suggested that,when DM intake was omitted from the CH4prediction model,ECM could be used instead due to its high correlation with DM intake. However,the use of ECM reduced the predictive ability of the model.Some recent studies have shown that milk fatty acids are generated by enteric fiber fermentation,and that these acids could be used alone or in combination with milk yield to predict enteric CH4emissions (Engelkeet al.2019). In addition,previous studies have suggested that enteric CH4emissions might not follow a linear pattern,and thus nonlinear models might better describe enteric function and fermentation dynamics compared with linear models (Huhtanenet al.2019). Patra and Lalhriatpuii(2016) found that nonlinear models better quantified CH4production across a wide range of production variables;such models might be particularly appropriate when extreme values are obtained during practical applications.

The dataset generated herein evaluated previously published prediction equations for CH4emissions. The previous equations were developed based on the CH4emissions of dairy and beef cattle offered grass silagebased diets in Northern Ireland (Yanet al.2000);on lactating dairy cows from 15 countries in Europe,the USA,Australia,and New Zealand (Niuet al.2018);and on dairy cattle under the IPCC (2006) Tier 2 Model,which is used in most countries. The equations developed herein,which predicted CH4emissions based on DM intake(eq.(13)) and NDF intake (eq.(14)),had lower RMSPE values than any of the previously published equations,indicating superior performance. In addition,compared with the equation developed here,which uses GE intake as a single variable,the equations of the IPCC (2006),Yanet al.(2000),and Niuet al.(2018) overestimated CH4emission levels:(20.42±3.18) MJ d–1(RMSPE=36.4%),(20.15±2.73) MJ d–1(RMSPE=31.1%),and (21.12±1.85)MJ d–1(RMSPE=34.7%),respectively. Consistent with this,Kebreabet al.(2008) reported that the IPCC Tier 2 equation overestimated CH4emissions,with MSPEs of 65,18 and 18%,respectively,for ECT,ER,and ED.Patra and Lalhriatpuii (2016) reported that prediction error decreased considerably when the IPCC Tier 2 (2006)equation was used to predict goat CH4emissions. These discrepancies might be due to differences in animal type,geographical location,and dietary contents. The diets in the present study consisted of medium-to high-quality corn silage and forage from various sources,whereas the diets of cows in the British dairy production system included a high proportion of grass silage. In the present study,the equations based on DM intake and NDF intake estimated annual enteric CH4emissions of 116.5 and 115.6 kg/cow for dairy cows (actual annual CH4emissions were approximately 115.3 kg/cow). Using the equations of Yanet al.(2000) and Niuet al.(2018),annual CH4emission levels were estimated to be 120.9 and 121.5 kg/cow respectively,which were substantially greater than the observed CH4emission levels. Hassanatet al.(2017)reported CH4emissions of around 173.9 kg/cow annually for lactating dairy cows fed Canadian diets,while the IPCC(2006) suggested an annual CH4emissions factor of 68 kg/cow for lactating dairy cows with an average annual milk yield of 1 650 kg/cow in most Asian countries.Therefore,the CH4emission prediction equations developed herein were more accurate than those adopted by the IPCC (2006) using default emission factors and other equations developed in different production regimes. Therefore,the quantification of the CH4emissions of lactating dairy cows under Chinese production systems and the development of prediction equations will be useful for establishing Chinese regional or national CH4inventories and better understanding CH4mitigation approaches for dairy production.

Fig.3 Regression between observed and predicted methane (CH4) emissions derived from the proposed equation in the current study (A) and using the equation of Niu et al. (2018) (B). Dietary dry matter intake and neutral detergent fiber content were used as independent variables in the proposed and Niu et al. (2018) equations.

5.Conclusion

This study examined the effects of variations in NDF/NFC ratios on the production performance,nutrient digestibility,and enteric CH4emissions of lactating dairy cows at different lactation stages. Proposed CH4prediction equations were also compared with previously published prediction models developed based on cows under different production regimes.Consistent with the initial hypothesis,cows fed a diet with a high NDF/NFC ratio had greater DM intake,milk production,and CH4emissions but lower nutrient digestion than cows fed diets with low and medium NDF/NFC ratios. The modelling results presented herein demonstrated that DM intake was the best predictor of CH4emissions (i.e.,the highestR2value)when the complete dataset was used.

Acknowledgements

This study was supported by the Key Program for International S&T Cooperation Projects of China(2016YFE0109000),the National Natural Science Foundation of China (31802085 and 31702133),and the Central Public-interest Scientific Institution Basal Research Fund of Chinese Academy of Agricultural Sciences(Y2021GH18-2).

Declaration of competing interest

The authors declare that they have no conflict of interest.

Ethical approval

The Animal Ethics Committee of the Chinese Academy of Agricultural Sciences (CAAS,Beijing,China) approved all animal care and handling procedures (protocol number 019–2018) prior to the commencement of the experiment.