Exploring the impact of replacement rates on carbon footprint in south-east Australian dairy herds

Case study farms

Five case study farms from northern Victoria/southern New South Wales, and three study farms from Gippsland region were selected for this study (Fig. 1). The farms’ demographic information can be found in the Table 1. The farms included in this study were selected through industry contacts (Murray Dairy and Gipps Dairy). The selection process involved purposive sampling to ensure representation of diverse feeding systems, specifically System 2 and System 4 (Wales and Kolver 2017), as well as all three calving systems (seasonal, split and year-round calving), farmers cooperativeness and quality of the records. In addition to the variation in feeding and calving systems, farms were selected to ensure differences in other key factors. These factors included the extent of supplementary feeding, level of milk production, reliance on irrigation water for pasture and crop production, and use of nitrogen fertiliser on pastures.

Fig. 1.

Locations of the case study farms of northern Victoria/southern New South Wales (Farms A–E) and Gippsland (Farms F–H).

Feeding systems

The case-study farms were categorised by feeding system according to the Australian five category classification scheme described by Wales and Kolver (2017). In all the farming systems studied, pasture, silage, and hay were all fed. Concentrate feeds were used in all farms except for Farm H. Concentrate feeds (used in the case-study farms) included grains such as barley and wheat, as well as canola meal. Across the farms, both annual and perennial ryegrass were commonly used, often in combination with other species such as white clover (Trifolium repens), sub-clover (Trifolium subterraneum), wheat (Triticum aestivum L.), oats (Avena sativa), and fescue (Festuca arundinacea). However, Farm F exclusively utilised perennial ryegrass in its grazing system. Moreover, the specific types of silage and hay used varied among the farms, depending on the pasture species, their availability and cost. A comprehensive overview of the different silage and hay types used by each farm in the study has been provided in Table 2. Concentrate feed was sourced from local suppliers. Hay and silage production varied across the farms. Some farms relied on homegrown hay/silage, whereas others used a combination of homegrown and locally purchased hay/silage. However, details on farm gate production matrices for specific homegrown feed were not included in this study because of the variability in scale and magnitude of the homegrown feed.

Data collection

Farm data were collected from July 2020 to June 2021 via email and phone calls. No sensitive data were collected, and no farmer interviews were conducted, so human ethics approval was not required for this study. Herd-level average data were collected for the study year 2020/2021. Farm records for the financial year 2020/2021 were also utilised as a source of information. In cases where certain data, such as feed-quality data, were missing, estimates from literature were used to supplement the information. Additionally, communication with the farmers was maintained through telephone and email exchanges to clarify any uncertainties, and to gather additional details. Milking herd composition by cow age, numbers of cattle (heifers, milking cows and bulls), typical cow liveweight (kg), typical liveweight gain of replacement heifers (kg/head.month), typical age at first breeding and calving and herd average milk production (L/day) by season were collected. Except for Farm D, all farms used a weighing platform in the milking area to weigh the animals. For Farm D, bodyweight adjustments were made on the basis of literature and farmers’ assumptions. Details of feed types, allocation per feed type (Mg/month), DM percentage (DM %), metabolisable energy (ME) content (MJ/kg DM), crude protein percentage (CP %), and the quantity of purchased feed. Furthermore, the data collection process extended to include calving system, area of farmland used for pasture or crops, energy, fuel, fertiliser application, and chemical usage. A comprehensive summary of the different fertilisers used and grain/hay purchased by each farm in the study has been provided in Table 3. Replacement rate was calculated for each herd as number of cows calving for the first time in the study year, divided by total number of calvings in that year.

Modelling of steady-state base farm

For each farm, a steady-state herd structure has been modelled, allowing for a systematic assessment of changes in system parameters at the farm level. This modelling approach enabled the representation of different feeding systems across various regions within the chain. The calving data from the herds were integrated into the model, with an average cow being modelled on the basis of the farm average data. The number of milking herd replacements was calculated as the required number of animals to maintain the steady-state herd size (i.e. a constant number of heifers and cows calving each year). It was assumed that replacement animals give birth first at the age of 2 years and recalve every 15 months until dying or being culled. Male calves were assumed to be immediately removed from the farm after birth. This modelling framework provided a structured and quantitative approach to evaluate the impact of various factors on the steady-state herd, considering calving system and the management of replacement animals within the dairy system.

Modelling effects of decreases in replacement rate

‘Baseline’ GHG emissions were calculated separately for each farm by using the farm actual data, including their replacement rate. To analyse the impact of reducing the replacement rate on GHG emissions, emissions were also modelled for replacement rates decreased by each of 5%, 10% and 15% (all percentage points), while keeping herd size (number of calvings annually) at the herd’s actual baseline value. For example, for Farm A where replacement rate was 25% in the study year, a decrease of 15% would result in a replacement rate of 10%. The 5%, 10% and 15% reductions in replacement rates were not applied to all farms, as some farms (Farms D, F and H) already had lower replacement rates. Within each farm, it was assumed that average daily milk production per cow for first-calving and second-calving cows were 20% and 10% respectively, lower than the herd’s average daily milk production per cow in each season (Dairy Australia InCalf 2017). Relative milk production of mature cows (i.e. third-calving and greater-parity cows) at baseline was calculated for each farm. This was adjusted upwards to ensure that the herd’s average milk production matched the base farm level. The adjustment was made using an iterative method, considering different ratio values until the calculated herd average closely aligned with the specified level for the farm.

These relative milk production values were assumed to not change with decreases in replacement rate. Thus, with decreases in replacement rate, herd average daily milk production was assumed to increase because the necessary additional feed was available or supplied to support this increased herd milk production. Dairy GHG Accounting Framework (D-GAF, ver. 14.5; Ekonomou and Eckard 2022) tool calculated the additional ME required to support this increased herd milk production (details in GHG estimation section).

For each of 5%, 10% and 15% reduction in replacement rate, all input parameter values were kept the same as the farm baseline values, except for numbers of replacement heifers and milk production (resulting from changes in the herd’s replacement rate and consequent distribution of calvings by cow parity), and herd ME requirements. The further assumption was that the number of heifer calves reared does not decrease when the replacement rate is reduced. Instead, more heifers are sold during their first winter, as well as during the following summer, autumn, and winter. These calculated values were used in the overall calculation of farm GHG emissions by using the built-in equation provided by the D-GAF tool. Effects of changes in replacement rate on herd genetics were not modelled in this study.

GHG estimation

These calculated values were used in the overall calculation of farm GHG emissions by using the built-in equation provided by D-GAF (ver. 14.5; Ekonomou and Eckard 2022) tool. Once their baseline data were refined by farm record, literature and cross-check, GHG emissions from each farm were estimated using the D-GAF, a modelling tool specifically designed for this purpose. D-GAF utilises various data inputs, including the number of animals, liveweight of animals, dietary composition (CP and DM digestibility), area of farmland used for pasture and crops, and use of fertilisers, chemicals, energy, and fuel. This tool follows the approaches outlined by the Australian government to estimate emissions from Scope 1 (on-farm CH₄, N₂O and CO₂ from diesel consumption and animals), Scope 2 (electricity usage), and Scope 3 (purchased farm inputs such as feed, fertilisers, herbicides/pesticides, lime, and fuel) emissions from the farms. D-GAF calculates the net farm emission (NFE; expressed in Mg of carbon dioxide equivalents, Mg CO₂e) as a sum of the emissions of those three scopes. Also, D-GAF tool considers co-product (Gollnow et al. 2014) allocation to account for additional products generated by dairy farms, such as surplus calves, culled dairy cows, and other cattle removed from the farm. The distribution of carbon footprint allocation among these co-products was based on the physiological feed requirement of the animal to produce milk and meat (Gollnow et al. 2014). In this study, an average allocation percentage of 85% was utilised (i.e. 85% for dairy and 15% for the additional products generated from dairy farm). The calculation for NFE with co-product allocation involved considering 85% of the actual NFE. Carbon sequestration in soils and trees was not considered in this study. Emission intensity (EI) was calculated by dividing NFE by the amount of product, measured as two separate indicators, (1) kilograms of carbon dioxide equivalent per kilogram of milk (kg CO₂e/kg of milk) and (2) kilograms of carbon dioxide equivalent per kilogram of milk solids (kg CO₂e/kg of MS).

GHG emissions of case-study farms

GHG emissions for the eight farms are presented in Table 4. The average EI per kilogram of MS for the eight study farms was 15.8 ± 3.5 kg CO₂e/kg MS. With coproduct allocation, the average EI was calculated to be 13.4 ± 3.0 kg CO₂e/kg MS (Table 4). Among the study farms, the greatest EI per kilogram of MS was found in Farm G (24.33 kg CO₂e/kg MS), whereas the least was observed in Farm D (13.79 kg CO₂e/kg MS). The EI/kg milk ranged between 0.98 and 1.72 kg CO₂e/kg milk. With co-product allocation, the EI/kg milk ranged between 0.83 and 1.47 kg CO₂ e/kg milk. The greatest EI per kilogram of milk was observed in Farm G, whereas the least was found in Farm D. Additionally, per cow emissions were greatest in Farm A and least in Farm H. As a result of variations in farming system sizes, the net annual GHG emissions per farm exhibited large variability across the study farms. Specifically, the NFE ranged from 1244 to 13,219 Mg CO₂e per farm (Table 4).

Within the on-farm emissions, enteric fermentation was found to be the largest contributor, accounting for approximately 55–70% of the total emissions in the case study farms. Pre-farm Scope 3 emissions, which include emissions from purchased farm inputs such as feed, electricity, fertilisers, herbicides/pesticides, lime, and fuel, were the second-largest source of emissions. This was particularly evident in the northern Victorian farms that practiced hybrid and System 2 feeding, as they relied on high amount of purchased feed, resulting in increased emissions in this category.

Replacement rate, GHG emissions and MS production

Reducing the replacement rate resulted in an average cow undergoing more lactations and dry-off periods than in the base farm situation. Consequently, this not only increased the number of surplus female calves but also affected the herd age structure (and parity) of the cow herd. This change led to an increase in average milk production per cow per day. The reduction of the replacement rate led to linear increase of milk production and greater NFE for some farms but lower EI per kilogram of MS (Figs 2 and 3). Farm G exhibited the greatest reductions in EI under each replacement rate scenarios, whereas Farm D demonstrated the least reduction in EI for the 5%, 10%, and 15% replacement rates respectively (Fig. 3). In contrast, NFE showed a small increase with the changes in the replacement rate in Farms B, C, E, F and G compared with the base farm scenario. A slight decrease in NFE was observed in Farms A, D and H with the reduction of the replacement rate (Fig. 4).

Fig. 2.

Milk solids production (kg/cow.year) of case study farms in northern Victoria/southern New South Wales (Farms A–E) and Gippsland (Farms F–H) regions, with a reduction in replacement rate (RR) compared with the base farm. Farms D, F and H had lower replacement rates, and thus not all percentage reductions in replacement rate were tested for these three farms.

Fig. 3.

Emission intensity (EI) of case study farms in northern Victoria/southern New South Wales (Farms A–E) and Gippsland (Farms F–H) regions, with a reduction in replacement rate (RR) compared with the base farm, (a) excluding and (b) including the co-product allocation to the red meat industry. Farms D, F and H practised lower replacement rate, and thus not all percentage reductions in replacement rate have been tested for these three farms.

Fig. 4.

Net farm emission (NFM) of case study farms in northern Victoria/southern New South Wales (Farms A–E) and Gippsland (Farms F–H) regions, with a reduction in replacement rate (RR) compared with the base farm, (a) excluding and (b) including the co-product allocation to the red meat industry. Farms D, F and H practised lower replacement rate, and thus not all percentage reductions in replacement rate have been tested for these three farms.

GHG emissions of case study farms

In this study, the GHG EI was found between 15.8 ± 3.5 kg CO₂e/kg MS, which aligns with industry reports of 13.5 kg CO₂e/kg MS (Dairy Australia 2016). A recent study by Lahart et al. (2021) in Ireland, which examined different feeding systems ranging from high pasture allowance to high concentrate feeding, reported an EI of 11.4–13.6 kg CO₂e/kg MS.

To compare the findings of this study with other national and international studies, we calculated the EI per kilogram of milk. The EI per kilogram of milk ranged between 0.87 and 1.72 kg CO₂e/kg milk. One of the limitations of this study was that milk quality data were not available for all the farms, so total milk production was used instead of fat- and protein-corrected milk in reporting EI. International studies reported emission intensities by using the unit kg of CO₂ equivalent/kg of fat protein-corrected milk as follows: 0.75–0.81 (New Zealand; Ledgard et al. 2020), 0.44–1.73 (Canada; Jayasundara et al. 2019), 1.53 (Germany; Kiefer et al. 2015) and 0.98–1.67 (Ireland; O’Brien et al. 2016). In a study comparing 60 dairy farms in Tasmania, Australia, Christie et al. (2011) reported an EI of milk production ranging from 0.83 to 1.39 kg CO₂e/kg milk. These comparisons affirm that the findings of this study are consistent with previous national and international studies. Those international studies (Kiefer et al. 2015; O’Brien et al. 2016; Jayasundara et al. 2019; Ledgard et al. 2020) allocated between 85% and 90% of total farm GHG emissions to milk production, with the remaining portion being allocated to meat production from cull cows, surplus heifers, and bull calves. This allocation is similar to the present study. Conversely, the study by Christie et al. (2011) allocated all farm GHG emissions to the primary product, milk. Therefore, it is important to note these differences in the allocation of farm GHG emissions to milk and meat when comparing results across studies.

The estimated per cow GHG emissions ranged between 5.87 and 10.30 Mg CO₂e, which aligns with a national study reporting per cow emissions of 6.9 ± 1.96 Mg CO₂e (Christie et al. 2012). Production traits were positively related to per cow GHG emissions, owing to the positive relationships of milk production and liveweight with feed intake, which is the main driver of GHG emissions (Charmley et al. 2016). The greatest and second-greatest per cow emissions in Farms A and B respectively, might be related to hybrid feeding, meaning the animals were fed a relatively high level of grain compared with other feeding systems. The increased grain feeding in these farms corresponded to an increase in per cow milk production (28 and 26 L/day in Farms A and B respectively), resulting in increased DMI. Because the algorithms and equations used in the D-GAF tool are adjusted for greater DMI on the basis of milk production, increased milk production and DMI directly contribute to greater enteric CH₄ production and per cow GHG emissions, as observed in Farms A and B. The increased milk output in the hybrid feeding system might not be sufficient to dilute the increased emissions associated with the grain feeding. Refining pasture-based management systems to increase grass utilisation and reduce concentrate supplementation can reduce both GHG emissions and GHG EI (O’Brien et al. 2015; Lahart et al. 2021).

Additionally, the D-GAF tool considers the liveweight and liveweight gain when calculating DMI (Ekonomou and Eckard 2022). As a result, animals with greater liveweight and weight gain tend to have high DMI as maintenance requirements, contributing to greater CH₄ emission. The average liveweight of milking cows varied significantly across the study farms, with weights ranging from 550 to 625 kg in northern Victoria and from 450 to 550 kg in Gippsland. The lower liveweight of milking cows observed in some farms may contribute to lower per cow emissions. In considering industry targets for emissions reduction, it is essential to acknowledge potential implications on herd management and animal selection. Targets based solely on EI may inadvertently promote the breeding of larger cows, which could lead to increased total emissions. Conversely, targets based on absolute emissions may encourage the adoption of smaller, more efficient breed type reminiscent of those commonly found in New Zealand.

Net farm emission across the farms in northern Victoria and Gippsland varied considerably. According to industry report (Dairy Australia 2016), average farm emissions for a pasture-based, 400–500-cow dairy farm are ~2500 Mg CO₂e per annum, which was consistent with the observations in the pasture-based Gippsland region farms included in our study. Despite Farm G NFE being average for a pasture-based farm, the EI was found to be the greatest. One possible explanation for this discrepancy could be the low milk production of 10 L/cow.day, which might be a result of practicing once-a-day milking. Among the case-study farms, Farm D had the greatest NFE owing to its larger herd size of 1625 milking cows and the use of purchased grain during pasture deficit months. However, the per cow emissions and EI per kilogram of MS were low on that farm. It is important to note that when comparing the results, it is crucial to consider the analytical methods, system boundaries, and metrics used to measure emissions profile. Additionally, the results of this study are based on data from the 2020/2021 period and may not represent a long-term average.

Replacement rate, GHG emissions, and milk production

In this study, the EI showed a downward trend with the reduction in replacement rate. The increased milk production resulting from a change in the parity of the milking herd improved the EI through a dilution effect. However, the reduction in replacement rate resulted in greater NFE in Farms B, C, E, F and G. This can be attributed to the increased DMI by cows for greater liveweight and milk production than for heifers. The discrepancies in NFE results across the case study farms may be due to the balance between the reduced DMI from reduced replacement heifers and the average increase in milk production for different parity groups.

Among the three scopes accountable for NFE, only Scope 1 (on-farm emissions), particularly enteric fermentation, changed with different proportions of replacement rate across the farms. Although only the milkers produce milk, replacement heifers (at different scales) are part of the dairy farm and contribute to Scope 1 emissions. Additionally, the case-study farms varied across their input. High-input systems with greater cow numbers and larger animals tend to have greater NFE. Conversely, low-input systems have lower NFE but, owing to reduced yields (both in meat and milk), factors such as animal maintenance per megagram of DM consumed often result in greater EI.

It is worth noting that the percentage reduction in GHG emissions in this study was minimal. Weiske et al. (2006) noted a reduction of 13% when reducing replacement rates from 40% to 30% in modelled dairy systems. Beukes et al. (2010) highlighted that reductions of 27–34% in GHG emissions can be achieved by combining strategies such as reducing replacement rates, improving cow genetics, and enhancing pasture management. To minimise GHG emissions, increasing the farm productivity while maintaining constant livestock DMI remains crucial. It is possible that the current replacement rates in the case study farms were not high enough to observe a significant reduction in GHG emissions. Beukes et al. (2010) reported that potential reductions in GHG emissions of 27–34% were found by combining strategies such as reducing animal replacement rates, improving cow genetics, and enhancing pasture management. Increasing the farm production while keeping the livestock DMI constant is crucial to minimise GHG emissions.

The current replacement rate of the case study farms ranged from 17% to 33%. Most farms raised heifers and used them as replacement stock for breeding, ranging from 76% in Farm F to 90% in Farm E, except for Farm C. Farm C had a greater replacement rate of 33% owing to a high incidence of twin calves and surplus heifers available for sale. Only 52% of the heifers reared on this farm were used as replacement stock, which may contribute to a greater emission profile for that farm. Farm B, which practiced year-round calving, had the greatest replacement rate. This farm is also looking to expand its size and production, which could potentially contribute to greater emissions. A previous study on modelled farms (Zehetmeier et al. 2012) reported that GHG emissions during the rearing phase of replacement heifers accounted for up to 20% of total GHG emissions from the modelled dairy farms. Therefore, the replacement rate plays a significant role in the total GHG emissions of dairy systems. However, these emissions can be offset by lower GHG emissions associated with rearing fewer replacements (Browne et al. 2014).

One drawback of reducing the replacement rate is its potential to decelerate the pace of genetic gain within a herd by selectively culling poorly performing cows (Browne et al. 2014). For farms with a greater replacement rate, further research could focus on reducing the replacement rate by improving overall herd efficiencies and increasing the length of productive life. This can be achieved through strategies such as enhancing body condition scores, improving pregnancy rates through better calving and mating management, and implementing effective reproductive practices. These approaches can contribute to optimising the herd structure and reducing the need for a high number of replacements, ultimately leading to more sustainable and efficient dairy systems.

There are few limitations to consider in this study. First, in some of the case study farms, data regarding DM digestibility, liveweight, and liveweight gain were not available. In such cases, feed-quality data were sourced from local government websites, and farmers’ assumptions and literature were used to estimate the average of the animals’ liveweights. As a result, the data input used in this study to calculate GHG emissions may not provide an assessment of the actual on-farm GHG emissions, but rather an estimation. The accuracy of data collection could be improved by obtaining precise farm-based records. Second, estimating carbon sequestration from trees and soil was beyond the scope of this project, and including these factors could provide a more accurate estimation of GHG emissions. In addition, there may have been some effects of COVID-19 on overall farm management, but these effects were not considered in this study.

Further research should consider extending the analysis to include full or partial life-cycle assessments, particularly from cradle to farm-gate. This comprehensive approach provides a holistic view of the GHG profile of a farming system, enabling the implementation of targeted mitigation strategies to address emissions effectively. Additionally, investigating the quantification of GHG emissions per unit of product, considering the productive life of cows and associated profitability would be valuable. This research could focus on reducing the proportion of cows that leave the herd before completing their first lactation, because it can have a significant impact on both GHG emissions and profitability.