Variability in greenhouse gas emission intensity of semi-intensive suckler cow beef production systems

Emission intensities from beef production vary both among production systems (countries) and farms within a country depending upon use of natural resources and management practices. A whole-farm model developed for Norwegian suckler cow herds, HolosNorBeef, was used to estimate GHG emissions from 27 commercial beef farms in Norway with Angus, Hereford, and Charolais cattle. HolosNorBeef considers direct emissions of methane (CH4), nitrous oxide (N2O) and carbon dioxide (CO2) from on-farm livestock production and indirect N2O and CO2 emissions associated with inputs used on the farm. The corresponding soil carbon (C) emissions are estimated using the Introductory Carbon Balance Model (ICBM). The farms were distributed across Norway with varying climate and natural resource bases. The estimated emission intensities ranged from 22.5 to 45.2 kg CO2 equivalents (eq) (kg carcass). Enteric CH4 was the largest source, accounting for 44% of the total GHG emissions on average, dependent on dry matter intake (DMI). Soil C was the largest source of variation between individual farms and accounted for 6% of the emissions on average. Variation in GHG intensity among farms was reduced and farms within region East, Mid and North re-ranked in terms of emission intensities when soil C was excluded. Ignoring soil C, estimated emission intensities ranged from 21.5 to 34.1 kg CO2 eq (kg carcass)−1. High C loss from farms with high initial soil organic carbon (SOC) content warrants further examination of the C balance of permanent grasslands as a potential mitigation option for beef production systems.


Introduction
Globally, the agricultural sector accounts for 10-12% of greenhouse gas (GHG) emissions (Tubiello et al., 2014) with livestock production contributing a significant portion. It is estimated that food production will need to increase by 50% compared with 2012 levels to feed the global population in 2050 (FAO, 2017). As a consequence, beef consumption is expected to increase in both developed and developing countries (OECD/FAO, 2018) and, thus greenhouse gas (GHG) emissions from beef production are also likely to increase.
Beef products have been shown to have a relatively high GHG emission per kg food (Mogensen et al., 2012). However, there is substantial variation in emission intensities among countries (Gerber et al., 2013), and among farms within a country (Bonesmo et al., 2013). This variation in GHG intensity is partly due to methodological differences among studies, but fundamental differences in natural resource availability and farm management practices also contribute significantly (Alemu et al., 2017a;White et al., 2010). Exploring differences between farm systems in GHG intensity may help identify beef production systems and practices that are more efficient, which could lead to the development of mitigation options at farm level. Hristov et al., (2013) reviewed different management practices such as diet formulation, feed supplements, manure management, improved reproductive performance, and enhanced animal productivity to reduce GHG emissions from ruminant production and showed potential long term mitigating effects.
Globally, approximately 44% of livestock GHG emissions are in the form of CH 4 (Gerber et al., 2013). In Norway, enteric CH 4 accounts for 44-48% of total farm emissions from beef cattle production systems (Samsonstuen et al., 2019). The diet influences CH 4 emissions through the digestibility and fibre content of the feed. A high proportion of fiber in the diet yields a higher acetic:propionic acid ratio in rumen fluid, which leads to higher CH 4 emissions (Sveinbjörnsson, 2006). Enteric CH 4 emissions can be lowered through improved feed quality, use of inhibitors and by breeding animals for lower emissions (Difford et al., 2018). Legesse et al. (2011) investigated the effect of management strategies for summer and winter feeding and found a 3 to 5% difference in CH 4 emissions across production systems. Concentrate-based beef production systems show lower GHG intensity compared with roughage based systems (de Vries et al., 2015). However, to ensure future food supply, grasslands less suitable for crop production might be preferred over highly productive cropland for production of feed for beef cattle. Beef production in Norway relies on use of pasture and forages because the total land in Norway is 90% "outfields" (i.e. rough grazing in forest, mountain and coast areas), with half the outfield area suitable as pastures or for forage production (Rekdal et al., 2014). According to Norwegian laws and regulations, all cattle must be kept on pasture for at least 8 weeks during the summer (Landbruks-og Matdepartementet., 2004). Grasslands have a large potential of storing C in plant biomass and soil organic matter through C sequestration (Wang et al., 2014). Grazing management influences the GHG emission intensity from beef production through diet quality (McCaughey et al., 2010), animal performance (Thornton and Herrero, 2010), nitrogen (N) fertilizer use (Merino et al., 2011), and soil C change (Alemu et al., 2017b). The effect of grazing management and stocking rate on C balance have been investigated by a number of studies (Reeder and Schuman, 2002;Soussana et al., 2007;Wang et al., 2014). Reeder and Schuman (2002) found significantly greater soil C content with light to moderate stocking rates compared with no grazing due to a more diverse plant community with fibrous rooting systems. Soussana et al. (2007) reported that managed grasslands in Europe are likely to act like atmospheric C sinks. However, when the study included C exports through grazing and harvesting and related emissions of CH 4 and N 2 O, total GHG emissions from grazed European grasslands were not significantly different from zero. Alemu et al. (2017b) concluded that a whole-farm approach is important to evaluate the impacts of changes in farm management aimed at decreasing the environmental impact of beef production systems. Yet, soil C is not included in most whole-farm GHG studies . Samsonstuen et al. (2019) developed a whole farm model, Ho-losNorBeef, adapted to Norwegian conditions and estimated GHG emission intensities for average Norwegian beef cattle farms in two distinct geographical locations (low altitude flatlands suitable for grain production and high altitude mountains not suitable for grain production). The emission intensities in flatlands and mountains were 29.5 and 32.0 kg CO 2 eq kg −1 carcass for British breeds, and 27.5 and 29.6 CO 2 eq kg −1 for Continental breeds, respectively. However, the use of average farm scenarios did not account for variation in production systems, differences in resource base, breed differences, management practices, selection strategies, feed composition and feed quality that typically prevail among farms.
Thus, the aim of this study was to use the HolosNorBeef model to evaluate commercial herds of Aberdeen Angus, Hereford, and Charolais cattle in geographically different regions of Norway with different management practices, resources, and quality of feed available to establish the variability in emission intensities and corresponding soil carbon (C) balance from suckler cow beef production under Norwegian conditions.

Materials and methods
This analysis was based on a study of suckler cow efficiency and genotype × environment interactions. The project (Optibeef -Increased meat production from beef cattle herds) gathered comprehensive information from 2010 to 2014 on farm structure, herd management, animal production and economics for suckler cow herds with the breeds Aberdeen Angus (AA), Hereford (H) and Charolais (CH). To be included in the study the farms had to record a minimum of 60% of weaning weights (WW) and have a minimum of 10 purebred cows per herd. The requirements were met by 188 herds, and 27 farms (nine of each of the three breeds) were finally selected based on variety in geographical locations. The farms provided sufficient information to quantify wholefarm GHG emissions. Through market regulation and subsidies, farmers are encouraged to buy concentrates and sell grains produced on farm, rather than using it as feed in livestock production (LMD, 2018). Hence, other production enterprises on the farms not related to the cow-calf operation, such as production of natural resources, use of farm inputs (i.e. area, fertilizer, and pesticides) for grain production, ley area for horses, and finishing of calves not born on the farm, was excluded from the analysis.
The farms were distributed across Norway from Rogaland in the South to Troms in the North within climatic zones varying from 3 (good) to 8 (harsh) on the scale developed by the Norwegian Meterological Insitute and Det norske hageselskap 2006. The farms had a wide range of farm characteristics such as herd size, management practices, resource base and areas available for forage production. Thus, the farms were considered representatives of the broad spectrum of suckler cow farms in Norway.

Farm characteristics
The input data were farm specific production data, farm operational data and soil and weather data for the specific locations. The farm specific animal production data from the period 2010-2014 were obtained from the Norwegian Beef Cattle Recording System (Animalia, 2017; Table 1). Calving typically occurred in the period January-July, with an average calving date April 1 st . However, three farms had a small proportion of the cows (0.18-0.41) calving during the autumn, with an average calving date October 1 st .
The feeding of each group of cattle throughout the year including type and proportion of concentrates, forage type and quality and time spent on pasture, were available through interviews with the respective farmers. The nutritive values of all forages, concentrates, and pastures (Table 2) were estimated using laboratory analysis information for the specific municipalities (Eurofins, Moss, Norway), information from the two largest feed manufacturers in Norway (Felleskjøpet SA, Oslo Norway; Norgesfor AS, Oslo Norway) and from the chemical composition of forage, grains and pasture (NMBU, Norwegian Food Safety Authority 2008).
The manure was assumed to be deposited on pasture during the grazing period and during housing the manure handling system was deep bedding, solid storage or a combination set according to the management practices on the specific farm. All manure collected through the housing period was used for fertilizing ley areas. The areas (ha) and yields (kg ha −1 ) of forage and use of fertilizers (kg N ha −1 ; Table 3), were obtained through interviews with the farmers and the farm accounts. However, two farms had no grass silage production on the farm and buy grass silage from farms within the same area. Thus, the forage yield of the individual farms was assessed as the calculated forage requirement plus an additional 10% (DM basis) to account for losses due to ensilaging (DOW, 2012). The areas required for forage production on these specific farms were estimated based on yield statistics for the specific area (Statistics Norway, 2017) and the use of fertilizers was based on the Norwegian recommendations for N application levels for forage production (NIBIO, 2016).
The use of energy, fuel, and pesticides was calculated based on information from the respective farm accounts (Table 3). For each of the individual farms a cultivation factor (r w × r T ) was calculated based on annual mean indices of soil temperature (r T ) and soil moisture (r w ) according to Skjelvåg et al. (2012; Table 4). The cultivation factor was used together with initial soil C content in the Introductory Carbon Balance Model (ICBM; Andrén et al., 2004) to account for external effects such as soil moisture and temperature, and variation in resource base. Water filled pore space (WFPS) and soil temperature at 30 cm depth (ts30) for each individual farm were used for estimation of N 2 O emissions. WFPS to saturation was calculated according to Skjelvåg et al. (2012) using detailed soiltype recordings available through NIBIO, whereas ts30 was calculated based on air temperature according to Kätterer and Andrén (2009). Due to expansion of the herd and/or sales of breeding stock, the herd size was not stable in most of the farms. Thus, carcass production assuming a constant herd size was calculated based on the corresponding replacement rate, farm specific slaughter weights, and dressing percentages from culled cows, surplus heifers and finishing bulls. Bulls not born on the farm were excluded as they were purchased and sold for breeding purposes, and did not contribute to carcass output.

The HolosNorBeef model
The GHG emissions were estimated using HolosNorBeef developed by Samsonstuen et al. (2019). HolosNorBeef is an empirical model based on the HolosNor model (Bonesmo et al., 2013), BEEFGEM , HOLOS (Little et al., 2008), and the Tier 2 methodology of the Intergovernmental Panel on Climate Change (IPCC, 2006) modified for suckler beef production systems under Norwegian conditions. The model estimates the GHG emissions on an annual time step for the land use and management changes and on a monthly time step for animal production, accounting for differences in diet, housing, and climate. HolosNorBeef estimates the whole-farm GHG emissions by considering direct emissions of methane (CH 4 ) from enteric fermentation and manure, nitrous oxide (N 2 O) and carbon dioxide (CO 2 ) from on-farm livestock production including soil carbon (C) changes, and indirect N 2 O and CO 2 emissions associated with runoff, nitrate leaching, ammonia volatilization and from inputs used on the farm (Figure 1; adopted by Samsonstuen et al., 2019). All emissions are expressed as CO 2 eq to account for the global warming potential (GWP) of the respective gases for a time horizon of 100 years: CH 4 (kg) × 25 + N 2 O × 298 + CO 2 (kg) (IPCC, 2007). Emission intensities from suckler cow beef production are related to the on farm beef production and expressed as kg CO 2 eq (kg beef carcass) −1 .
2.2.1.1. Methane emissions. Enteric CH 4 emissions are estimated for each age and sex class of cattle using an IPCC (2006) Tier 2 approach. Estimation of gross energy (GE) intake is based on energy requirements for maintenance, growth, pregnancy, and lactation according to Refsgaard Andersen (1990). The DM intake (DMI; Table 5) depends on both the energy requirements of the animal and the animals' intake capacity. The intake capacity is dependent on the fill value of the    (2014) forage, as well as the substitution rate of the concentrates (Refsgaard Andersen, 1990). The GE intake to meet the energy requirements was estimated from the energy density of the diet (18.45 MJ kg −1 DMI; IPCC, 2006; Table 6). Enteric CH 4 was estimated from monthly GE intake using a diet specific CH 4 conversion factor for each cattle group (Ym = 0.065; IPCC, 2006; Table 6). The Ym factor is adjusted for the digestibility of the diet ( × DE 0.1058 0.0006 ) as suggested by Beauchemin et al. (2010 ; Table 6).
Manure CH 4 emissions are estimated from the organic matter (volatile solid; VS) content of the manure. The VS production is calculated according to IPCC (2006), taking the GE content and digestibility of the diet into account. The VS are multiplied by a maximum CH 4 producing capacity of the manure (B o =0.18 m 3 CH 4 kg −1 ), a CH 4 conversion factor (MCF=0.01, 0.02, 0.17 kg CH 4 VS −1 for manure on pasture, solid storage manure and deep-bedding, respectively) and a conversion factor from volume to mass (0.67 kg m −3 ; IPCC, 2006; Table 6).

Nitrous oxide emissions.
Direct manure N 2 O emissions are calculated based on the N content of manure and an emission factor for the manure handling system (0.01, 0.02, 0.05 kg N 2 O-N (kg N) −1 for deep-bedding, pasture manure, and solid storage, respectively; IPCC, 2006; Table 6). The N content of the manure is estimated according to IPCC (2006), based on the DMI, crude protein (CP; CP = 6.25 × N) content of the diet and N retention by the animals (Table 6).
Direct soil N 2 O emissions are estimated by multiplying the total N inputs with an emission factor of 0.01 kg N 2 O-N kg −1 N according to IPCC (2006). The total N inputs include above-and below ground crop residue N, using crop yields of Janzen et al. (2003), and mineralized N in addition to application of N fertilizer and manure. The derived C:N ratio of organic soil matter (0.1; Little et al., 2008) is used to calculate mineralization of N inputs ( Table 6). The effect of location and seasonal variation was taken into account by including four seasons based on the local weather conditions and growing season; spring (April-May), summer (June-August), autumn (September-November) and winter  Table 6).
Indirect N 2 O emissions from soil are estimated from the assumed losses of N from manure, crop residues, and fertilizer according to IPCC (2006). The emissions from run-off, leaching and volatilization are estimated based on the fraction of the loss for the manure handling system adjusted using emission factors (0.0075 and 0.01 kg N 2 O-N kg −1 ) for leaching and volatilized ammonia-N, respectively (IPCC, 2006; Table 6). The emissions were based on the assumed fraction of N lost adjusted for emission factors for leaching (0.0, 0.0, 0.3, 0.3 kg N (kg N) −1 for deep bedding, solid storage, pasture manure and soil N inputs including land applied manure, grass residue, synthetic N fertilizer and mineralized N, respectively; IPCC, 2006; Table 6). Emissions from volatilization were adjusted for the emission factors for volatilized ammonia-N (0.1, 0.2, 0.3, 0.45 kg N (kg N) −1 for soil N inputs, pasture manure, deep bedding, and solid storage, respectively; IPCC, 2006; Table 6).   Andrén et al. (2004), which estimates the change in soil C from total C inputs (i) from grass residues and manure. The fraction of the young (Y) C pool entering the old (O) C pool is estimated based on a humification coefficient of grass residue (h= 0.13; Kätterer et al., 2008; Table 6) and a humification coefficient of cattle manure (h= 0.31; Kätterer et al., 2008; Table 6). The degradation of the pools is determined by the respective decomposition rates (k y = 0.8 year −1 and k o =0.007; Andrén et al., 2004; Table 6). The change in Y and O soil C stocks is estimated based on the humification rates and decomposition rates together with the relative effect of soil moisture and temperature r w × r T to account for regional differences due to soil type and climate. The yearly fluxes of Y and O soil C are given by the differential equations of Andrén and Kätterer (1997): hk rY k rO 1 2 2.2.1.4. Carbon dioxide emissions. Direct CO 2 emissions are estimated from on-farm use of diesel fuel using an emission factor (2.7 kg CO 2 eq L −1 ; The Norwegian Environment Agency, 2017; Table 6). Off-farm emissions from production and manufacturing of farm inputs are estimated using emission factors for Norway or Northern-Europe; pesticides, 0.069 kg CO 2 eq (MJ pesticide energy) −1 (Audsley et al., 2014); electricity, 0.11 kg CO 2 eq (kWh) −1 (Berglund et al., 2009); diesel fuel, 0.3 kg CO 2 eq (L) −1 (Öko-Instititut, 2010); silage additives, 0.72 kg CO 2 eq (kg CH 2 O 2 ) −1 (Flysjö et al., 2008); and N-based synthetic fertilizer, 4 kg CO 2 eq (kg N) −1 (DNV, 2010; Table 6).  Emissions related to the use of concentrates are estimated according to Bonesmo et al. (2013). The concentrates are assumed to be supplied by barley and oats grown in Norway (0.62 kg CO 2 eq kg DM −1 ; Bonesmo et al., 2012; Table 6) and soybean meal imported from South America (0.93 kg CO 2 eq kg DM −1 ; Dalgaard et al., 2008; Table 6). Emissions from on-farm production of field crops are not included in the total farm emissions as they are sold and not used as feed by the beef enterprise.

Sensitivity analysis and comparisons
A sensitivity analysis was performed to investigate the evaluate possible errors in the estimated soil C balance. The sensitivity of the yearly effect of temperature and soil moisture (r W × r T ) and initial soil organic carbon (SOC) was estimated by changing the factors 1% and recalculating the emission intensities.
Breeds and regions were compared through mean comparison of the estimated emission intensities (CO 2 eq (kg beef carcass) −1 ) using the PROC GLM procedure of SAS® software, V9.4 (SAS Institute Inc., Cary,

Results
The total farm GHG emission intensities showed no significant difference across breeds (Table 7). However, N 2 O emissions from manure (P≤0.01) and emissions related to off-farm production of barley (P≤0.05) and soya (P≤0.01) differed across breeds. Angus showed most variation in total emission intensities. This variation decreased when soil C balance was ignored.
The farms showed wide variation in emission intensity (including soil C) with a mean estimate of 29.2 CO 2 eq (kg carcass) −1 (median= 29.5, range 22.5 to 45.2; Table 7). Enteric CH 4 contributed most to the total GHG emissions, accounting for 44% of the total emissions. N 2 O from soil and manure was the second largest source, accounting for 13% and 11%, respectively. Soil C balance accounted for 6% of the total emissions and had the largest variation across farms, ranging from -2.7 to 14.1 CO 2 eq (kg carcass) −1 depending on location. On-farm emissions from burning of fossil fuels accounted for 9% and the indirect CO 2 emissions from manufacturing of farm inputs (i.e. N-fertilizers, fuels, electricity, pesticides) accounted for 8%.
Regions East and Mid had lowest mean emission intensities, whereas Southwest and North had greatest mean emission intensities (Table 8). Soil C differed across regions (P≤0.05) and was the largest source of variation, on average accounting for 0.1 to 1.4 CO 2 eq (kg carcass) −1 of the total emissions in East and Mid, and 3.4 to 6.2 CO 2 eq (kg carcass) −1 of the total emissions in Southwest and North. North had greater emissions from indirect and direct energy. By excluding the soil C balance, the variation between individual farms decreased and the emission intensity across all farms had a mean estimate of 27.5 CO 2 eq (kg carcass) −1 (median= 26.9, range 21.5 to 34.1). Excluding soil C led to re-ranking of individual farms in terms of GHG emission intensity ( Table 9).
The comparison of the least square mean (LSM) differences of emission intensities showed that the differences in manure N 2 O emissions were significant both across breeds and regions (P≤0.01). Soil C differed across regions and direct energy differed across breeds (P≤0.05 and P≤0.05 respectively), while the difference between breeds and locations for other sources of emissions was not significant (Table 10).
Estimated GHG were moderately sensitive to changes in initial SOC and the yearly effect of soil temperature and soil moisture (r W × r T ). The sensitivity elasticity had a linear response ranging from 0.14 to 0.23 CO 2 eq (kg carcass) −1 across region, caused by 1% change in initial SOC (Table 11). Changing the r W × r T 1%, caused a 0.12-0.19 CO 2 eq (kg carcass) −1 across regions (Table 11).

Animal production
Our study investigated the GHG emissions from commercial Norwegian farms from different geographical regions, compared with simulated farms used in other studies (e.g. Mogensen et al., 2015;White et al., 2010) with different management practices, cattle breeds, and natural resources. The farms investigated were distributed across Table 7 Mean, minimum (Min), maximum (Max) and standard deviation (SD) estimates for greenhouse gas emission intensity (kg CO 2 eq kg −1 carcass) (n=9 for each breed).  the country and had a wide range of farm characteristics, representing the broad spectrum of suckler cow farms in Norway. Carcass weights used for estimating emission intensities from herds of Angus, Hereford, and Charolais were similar to carcass weights from intensive and extensive beef breed farming systems in Sweden and Denmark (Mogensen et al., 2015).

Greenhouse gas emissions
Under the current conditions for beef production in Norway, HolosNorBeef estimated mean emission intensities, including soil C, of 29.2 CO 2 eq (kg carcass) −1 (median= 29.4, range 22.5 to 45.2) for 27 herds of Angus, Hereford, and Charolais. This range of emission intensities is similar to reports for other Nordic countries; Denmark 23.1 to 29.7 CO 2 eq (kg carcass) −1 and Sweden 25.4 CO 2 eq (kg carcass) −1 (Mogensen et al., 2015). Emissions related to off-farm production of soya differed in terms of emission intensities across breeds. Observed feed intake and use of concentrates showed variation both across breeds and between farms within breed as a consequence of diet composition and feed requirements. In general, farms with lower quality forage fed a larger proportion concentrates to the replacement heifers. Bulls were on average fed 33% concentrates and were usually fed good quality silage. However, as increased production follows increased feed intake, the observed variability did not cause differences in total emission intensities across breeds.

Methane emissions
Enteric CH 4 contributed most to the total GHG emissions, accounting for 44% of the total emissions on average. HolosNorBeef estimated enteric CH 4 emissions based on the GE intake while adjusting the Ym for the digestibility of the diet (i.e. DE%). Hence, as shown by Samsonstuen et al. (2019), variation in Ym would cause a linear change in emission intensities. At equal GE intake, increased DE% would result in a linear decrease in Ym and a corresponding decrease in enteric CH 4 emissions. Within breed, Angus showed the largest variation in both % DE, DMI and enteric CH 4 emissions. Enteric CH 4 emissions are mainly related to variation in DMI (Herd et al., 2014) and feed quality (Ominski et al., 2011), with improved quality associated with lower emissions as the proportion of easily digested organic matter in the feed increases (Wims et al., 2010). Diets with more starch and less fiber produce less CH 4 per kg DM (Haque, 2018). In Sweden and Denmark, enteric CH 4 was reported as the largest source of emissions, accounting for 45.1-50.4% of total GHG emissions (Mogensen et al., 2015), depending on feeding intensity. In the present study, the DMI varied between and within farms dependent on the production and diet composition as the location of the farm dictated the available feed resources and use of pastures. Diet composition and forage quality changed throughout the year due to differences in animal requirements (e.g. for maintenance, growth, pregnancy, lactation) and availability of feed resources (e.g. pasture, silage, concentrates). For suckler cows, the variation in DMI within breed is mainly due to forage quality and use of Table 9 Ranking of farms with Aberdeen Angus (AA), Hereford (H) and Charolais (CH) in different regions in terms of GHG emission intensities including and excluding soil C balance. East (n=16) Southwest ( H1  AA3  H17  H17  CH19  AA22  CH23  AA25  CH2  H11  H18  H18  AA20  CH21  H24  H26  AA3  H1  CH21  AA20  AA25  CH23  AA4  A10  AA22  CH19  H26  H24  CH5  CH2  AA27  AA27  H6  H6  AA7  AA4  CH8  CH5  CH9  CH8  AA10  CH14  H11  AA7  H12  CH9  A13  AA13  CH14  H12  H15 H15 CH16 CH16 n = number of farms in each region.

Table 10
Least square means (LSM) of greenhouse gas (GHG) emission intensities and proportion of total emissions (in parenthesis) from average herds of Aberdeen Angus (AA), Hereford (H), and Charolais (CH) in four regions of Norway (kg CO 2 eq kg−1carcass). concentrates, as the digestibility of the forage and proportion concentrates influences the forage intake capacity. Use of pasture also influenced the DMI as the cows were assumed to have a higher DMI from cultivated pastures than outfield pastures due to the availability of the feed. Feed requirements varied both between breeds and within breeds due to differences in weights at different ages. The variation in DMI from birth to slaughter is influenced by slaughter age and slaughter weight as it influences the feed required for growth. The DMI of heifers from birth to calving is influenced by the diet composition and requirements for growth. Surplus heifers were fed the same diet until they reached slaughter weight. Manure CH 4 emissions varied from 2-8% of total emissions depending upon diet composition, housing conditions, and manure storage. HolosNorBeef calculated the manure CH 4 emissions on a monthly basis for each cattle class and determined the organic matter (i.e. VS) content of manure based on GE intake and the digestibility (i.e. DE%) of the diet. The DE% were variable, ranging from 59 to 71% among the farms leading to a large variation in manure CH 4 emissions between farms. This is similar to the range in DE% (49 to 81%) reported by Hanigan et al. (2013). Diet composition and DMI influence manure CH 4 emissions as increased organic matter (i.e., VS) content of manure increases the emissions from degradation (Monteny et al., 2001). Farms with low quality forage (e.g. straw or low quality silage) had lower manure CH 4 emissions as both the digestibility of the diet and the VS content of manure decreases. Crude protein (CP) and fiber content of the diet is significantly related to VS (Appuhamy et al., 2017), and Amon et al. (2007) showed that increased lignin and cellulose content in the manure reduces the CH 4 emissions as the digestibility decrease. However, manure management influence the manure CH 4 emissions as the CH 4 conversion is greater in deep bedding, compared with solid storage, due to anaerobic conditions. Thus, the greater CH 4 manure emissions were for farms using deep bedding during the housing period.

Regional variation
Soil C (discussed in section 4.2.3) differed across regions. By excluding the soil C balance, the variation between regions and individual farms decreased and the emission intensity across all farms had a mean estimate of 27.5 CO 2 eq (kg carcass) −1 (median= 26.9, range 21.5 to 34.1). East and Mid had lowest mean emission intensities, whereas Southwest and North had greatest mean emission intensities. Direct comparisons across and within regions are challenging as not all breeds were represented in all regions. Unequal distribution of breeds might cause confounding of breed within region. As all breeds were represented in both East and North and two breeds were represented in the region Mid, the confounding of region with breed was of greatest concern in the region Southwest, which was confirmed by region*breed LSM solutions for all regions except Southwest. Two farms and only one breed within this region Southwest might suggest that this region should have been omitted and the two farms included in a "South Region". However, an increase in geographical area could have concealed differences across regions caused by differences in the resource base, such as initial SOC.
The use of input factors is to a large extent influenced by the resource base, as the use of e.g. pesticides, fertilizer, and diesel fuel is related to the areas available for forage production, as pastures, and the distance from the field to the farm. Nevertheless, the least square mean comparison reveals a significant difference in emission intensities between breeds from direct energy, suggesting that there are differences in use of diesel fuel between breeds, within regions. In general, the Southwest and North have smaller areas available, with a greater distance between farm and field and greater variation in climatic conditions. A large proportion of the farms were located in the East, which also had most variation within region. Differences in feed requirements between breeds increases the difference between individual farms within the region. The resource base in the East facilitates both good quality silage and the use of straw as forage due to grain production in the region, resulting in a great variety in diet composition and corresponding emissions between farms.

Soil C balance
The GHG contribution from soil C balance accounted for 6% of the total emission intensities on average and had the largest variation across farms, ranging from -2.73 to 14.11 CO 2 eq (kg carcass) −1 depending on location. HolosNorBeef estimated the C balance between the soil and atmosphere using the two-compartment ICBM model (Andrén et al., 2004). The GHG contribution of soil C balance was influenced by the level of the initial SOC content, temperature and moisture in addition to forage production, application of manure, and N fertilizer. Inputs into ICBM are used to adapt the model to the local management and weather conditions (Bolinder et al., 2011). This model was previously calibrated to Norwegian conditions and used to estimate soil C change in the 100 th year with continuous grass and arable cropping (Bonesmo et al., 2013;Skjelvåg et al., 2012). Skjelvåg et al. (2012) investigated the farm specific natural resource base in six municipalities in different parts of Norway and found a wide range in initial SOC content in top soil varying from 56.1 to 116.8 Mg ha −1 . The 30 Norwegian dairy farms investigated by Bonesmo et al. (2013) had an average initial SOC of 71.3 Mg ha −1 , ranging from 40.3 to 99.5 Mg ha −1 . In comparison, the current study had an average initial SOC of 75.7 Mg ha −1 , ranging from 44.8 to 168.4 Mg ha −1 .
On average, the C balance accounted for 0.1 to 1.4 CO 2 eq (kg carcass) −1 of the total emissions in East and Mid, whereas in Southwest and North the average C balance accounted for 3.4 to 6.2 CO 2 eq (kg carcass) −1 of the total emission. The resource base of the regions varies, whereas the East and Mid are regions with a climate suitable for grain production. The regions Southwest and North are less suitable for grain production, and the arable lands have been used for forage production or as pastures for decades, resulting in high initial SOC. The initial C in topsoil is crucial for estimating C balance as a high initial SOC content will lead to a decrease, and a low initial SOC will lead to an increase (Andrén et al., 2015). Hence, the estimated C loss from farms in Southwest and North is a result of high initial SOC. As the soil C content is difficult to measure, Andrén et al. (2015) suggested to modify the initial SOC if the changes between samplings are unrealistic. However, in the present study there is only a single estimate of the SOC content and modifying the initial SOC is not possible.
The ICBM model has been further developed into a multi-

Table 11
Sensitivity elasticities for the effect of 1% change in soil C change external factor (rw × rT) and initial soil organic carbon (SOC) on the greenhouse gas (GHG) emission intensities CO 2 eq (kg carcass) −1 .  (Kätterer and Andrén, 2001). ICBM/3 divides the Y SOC pool into above ground residues, below ground residues and addition of manure and other organic matter. Multi-compartment models have pool-specific decomposition rates and humification factors, making the model more dynamic and adapted to various management practices. Future soil C balance estimations could possibly be improved by incorporating the newest version of ICBM/3 to HolosNorBeef, or by calibrating the existing ICBM model with multiple soil samples from areas with large initial SOC. However, the complexity of multi-compartment models (e.g. ICBM/3) increases the amount and detail level of the required input and decreases the transparency of the model. Such detailed input data for use in the multi-compartment model are not available at this point. According to Bolinder et al. (2011), single-and two-compartment models such as the ICBM model may replace more complex models in whole farm modelling and life cycle assessment (LCA) approaches as they are simple, transparent and can be programmed in a spreadsheet format. Kröbel et al., (2016) investigated the inclusion of both the twocompartment ICBM model and the multi-compartment Century model in the Canadian Holos model. The study indicated that the ICBM model allowed a more dynamic output of management and climate, increasing the flexibility and allowing more farm specific estimation compared with the more complex Century model (Kröbel et al., 2016). Hence, the two-compartment ICBM model may be sufficient for whole farm modelling of GHG emissions as it reflects the dynamics of the SOC stocks while taking the influence of crop yield, management, soil moisture and temperature into account. Sensitivity elasticities showed an average change in emission intensities of 0.10 to 0.23 (SOC) and 0.12 to 0.19 CO 2 eq (kg carcass) −1 (r W × r T ) across regions. However, there were no significant different response in sensitivity elasticities between regions, implying that the estimated difference in soil C balance occurs due to more than just variation in the initial SOC and r W × r T .
Grazing influences plant production (Lee et al., 2010), plant diversity (Limb et al., 2018) and adds organic matter through manure (Baron et al., 2007). The influence of grazing management on C sequestration has been investigated in various studies (Pelletier et al., 2010;Reeder and Schuman, 2002;Soussana et al., 2007Soussana et al., , 2010Wang et al., 2015). The influence of grazing is complex, as the soil C dynamics are influenced by the animal, climate, soil, plant, management and their interactions (Bolinder et al., 2011;Schuman et al., 2002). HolosNorBeef does not include the effect of grazing management on C balance as the ICBM model does not account for the effect of grazing or stocking rate. Norwegian land contains approximately 60,000 (arable) to 100,000 kg C ha −1 (pastures; NIBIO, 2019) and the potential for mitigation by sequestering C in outfield pastures under Norwegian conditions has not been scientifically documented. Applying Norwegian conditions to US studies, the estimated potential for C sequestration is 1000 to 6000 kg CO 2 ha −1 year −1 (NIBIO, 2019). When considering pasture management strategies, the corresponding ecosystem services directly or indirectly influenced by pasture management should be taken into account.

Conclusions
A whole-farm approach that included changes in soil C estimated GHG emission intensities of 22.5 to 45.2 CO 2 eq (kg carcass) −1 from representative suckler cow beef farms in Norway with Angus, Hereford, and Charolais cattle. The variation in DMI and diet composition between farms influenced both enteric and manure CH 4 emissions, and contributed to variation in emission intensities between individual farms. Including soil C balance in the emission intensity of beef production increased variability in GHG emissions among individual farms and caused a significant difference in estimated GHG intensities between regions. In addition to level of forage production, application of manure, and N fertilizer, the soil C balance was influenced by the level of the initial SOC content, temperature and moisture. Arable lands used for forage production or as pastures for decades result in high initial SOC in soils of some farms, which warrants further examination and additional measurement as the ICBM model is sensitive to high initial SOC and does not account for the effect of grazing or stocking rate.

Author Statement
This manuscript has not been published and is not under consideration for publication elsewhere. All authors have approved the manuscript and agree with its submission to Livestock Science. We have no conflict of interest to disclose.

Declaration of Competing Interest
The authors have no conflict of interest to disclose.