Milk sustainability in specialized farms as affected by farm size and culling rates: A new perspective for allocation

Highlights

• •We compared two farms, a small and a large one, in the Italian po valley.
• •Kilograms of fat protein corrected milk and hectare of occupied area were the functional units.
• •The different functional units led to different results.
• •This study proposes allocation as a method to investigate herd management.

Abstract

Several studies investigated the sustainability of dairy cattle systems. Global warming potential (GWP) is a “touchstone impact category” of LCA, whereas fossil depletion (FD) remains a scarcely studied environmental indicator. This study aims to help fill the gap of knowledge on FD in dairy cattle systems. We compared two highly specialized dairy cattle farms equipped with precision technologies: a large (L) and a small (S) farm. The discriminants were the number of lactating cows and the agricultural area, with L having approximately twice the heads and three times the area of S.

In LCA we used the main product (FPCM, i.e., 1 kg of milk normalized for fat and protein), as the functional unit (FU). A second FU was used, i.e., hectare of occupied area. By changing the FU, the study showed different results, because S performed better when FPCM was used while L performed better when the agricultural area was used.

Allocation of culled cows affected the environmental impacts of both farms improving their sustainability. A high culling rate provides information on herd management, and it can result in lost income. We proposed the allocation as helpful to quantify the culled cows, a potentially undesirable product, showing a significant difference between farms, i.e., 30 and 38 % of culling rate in the L S, respectively. In S, this rate led to a higher income percentage provided by culled cows (6.3 vs. 4.2 % of L). Allocation decreased the equivalents of some LCA environmental indicators, showing an oversized replacement of cows.

Graphical abstract

Glossary. FPCM=fat protein corrected milk; ha=hectare; GWP=global warming potential; NREU=non-renewable energy use; FD=fossil depletion; ALO=agricultural land occupation

1. Introduction

The world has achieved great success in increasing agricultural production and food safety during the past half-century.

Livestock systems play an important role since 14.5 % of anthropogenic greenhouse gases (GHGs) are due to zootechnical activities (Gerber et al., 2011). Domestic ruminants contribute to global warming potential (GWP) mainly with methane emissions (CH₄) followed by nitrous oxide (N₂O), and the sources are enteric fermentations (CH₄) and manure management (CH₄, N₂O). In particular, European dairy systems have undergone massive intensification to increase milk production and feed efficiency, in terms of nutrient ratio between output/input (Berton et al., 2020). Gerber et al. (2011) and Lorenz et al. (2019) showed that as milk production increases, production efficiency improves (measured as feed conversion ratio, for example). Moreover, these authors stated that this efficiency improves the sustainability per kg of milk, i.e., the functional unit (FU) of many studies.

This scenario led the LCA studies focused on the environmental impact, and GWP is the most investigated impact category (Baldini et al., 2017). The same review (Baldini et al., 2017) emphasised the interest in the wide cluster of land use indicators. Although in dairy cattle systems almost 80 % of the contributors to GWP are due to emissions and feeding sources (Pirlo & Lolli, 2019), services, machinery, transport, energy sources and energy emissions also make a significant contribution. Zervas and Tsiplakou (2012) suggested that in a livestock system, CO₂ emissions from fossil energy use come from activities related to the arable land (e.g., ploughing, sowing, harvesting), transport by farm-owned trucks, manufacture and supply of farm machinery, fuel for heat production (including grain drying), electricity consumption and synthetic fertilizers production. Other LCA categories obviously have a close relationship with the use of energy sources, i.e., NREU and FD.

Non-renewable energy use (NREU) and fossil depletion (FD) are closely related environmental indicators. Baldini et al. (2017) stated that, in dairy cattle, the NREU was assessed in 17 studies, a large share, comparable with global warming potential, but few studies are focused on FD.

Among the examples in LCA studies, Romano et al. (2021b) compared three dairy cattle farms and investigated FD, and recently Bragaglio et al. (2023) studied dairy cattle and considered also FD and NREU.

Land occupation is a critical environmental indicator in the Po Valley framework. Land occupation is a helpful sustainability indicator of the main product (the milk, in this case) because both farms are located in a geographical scenario in which the land to be used for agricultural purposes competes with other human activities (civil dwellings, industries, travel routes, logistics spaces).

Given the growing interest in renewable energy sources and precision agriculture (PA), this manuscript focuses on two precision agriculture-based farms, a small one (S) and a large one (L), characterized as follows: both are specialized dairy farms located in Po Valley (Northern Italy) and they rear Holstein-Friesian lactating cows in a free stall barn with cubicles in the resting area (no access to pasture). The use of precision agriculture devices involves the presence of high automation tools, i.e., automatic milking systems (AMS) and automatic feeding systems (AFS). In particular, the latter manage all the livestock categories (heifers, dry cows and lactating cows). This is remarkable because, on most farms with AFS, only lactating cows are automatically fed. This study is a comparison of the sustainability of the two farms (L vs. S) that are very similar in terms of technology, management and forage self-sufficiency, but differ in terms of herd size (250 vs. 120 lactating cows) and crop area (200 vs.70 hectares).

The first criterion of representativeness of the two farms is automation, particularly AFS are widespread in the lowland areas (Po Valley) of the Veneto and Lombardia regions. Veneto, where the L farm is located, is one of the four Italian regions with the highest number of farms with >500 heads (ismeamercati.it/lattiero-caseari, 2024; lattiero-caseari, 2024), whereas in Lombardia (S) 56 % of the farms have a herd size between 100 and 499 heads (ismeamercati.it/lattiero-caseari, 2024; Tangorra et al., 2022).

According to our census, 12 and 26 cattle farms are managed with AFS in lowland areas of Veneto and Lombardia, respectively. AFS distribution decreases in the other two dairy cattle regions of the Po Valley, Emilia Romagna and Piemonte, where we found 8 and 3 AFS, respectively (Romano et al., 2023).

Having assessed the significant representativeness of the two farms, we verified the willingness to provide sound data for inventory analysis.

The study also aimed to assess the share of by-products, which are exclusively the culled cows. In fact, all the forages (silages and hays) are re-used on the farm, entirely for animal feeding. Moreover, male calves spend a very short time in the sheds before being sold due to the high degree of specialization on both farms.

This research evaluates the impact of by-products with the allocation criterion, as usual in the LCA. Allocation is widely used to subtract pollutants from the main product, distributing the environmental burdens also across the by-products. The present study, in addition to the sustainability comparison of highly specialized farms, introduces a novel item: estimating the share of the undesirable product (i.e., culled cows), which is the unique by-product, per the allocation criterion. This model is based on the fact that, in dairy systems, a high percentage of culled cows will result in reduced income. To the best of our knowledge, only two studies used a similar approach: on the one hand, Romano et al. (2021b) investigated dairy cattle and found significant differences between farms in the share of culled cows; on the other hand, Bragaglio et al. (2022) performed a LCA in buffaloes using the allocation as a marker of undesirable products, but the study did not find statistically significant results.

This study combines the evaluation of the environmental sustainability of two highly specialized farms with high automation levels with the estimation of the economic impact of their culling rate. To this aim, we identified the farm size as a cut-off useful to verify the environmental impact, and the culling rate as a marker to identify the undesirable product, because the farms show the same management and a similar forage self-sufficiency.

2. Methods

2.1. Life cycle assessment and goal definition

This study was carried out following ISO standards (ISO 14040, 2006; ISO 14044, 2006). According to these standards, a life cycle assessment (LCA) study investigates, considering a cradle-to-gate approach, four environmental indicators: global warming potential (GWP), non-renewable energy use (NREU), fossil depletion (FD) and agricultural land occupation (ALO). These indicators were estimated using the software SimaPro 8.03 (Pré Consultants, Amersfoort, The Netherlands).

2.2. Functional unit

In these highly specialised milk-producing farms animals are reared under intensive conditions. Two functional units (FUs) were chosen to assess the farms’ sustainability: (i) 1 kg of fat and protein-corrected milk (FPCM) and (ii) 1 hectare of total (on- and off-farm) occupied area. For the first FU the milk has been corrected using the formula applied to adjust the raw milk nutrient contents from the specific farm to a quantity of milk with standardized quality (4.0 % of fat and 3.3 % of protein), also used by Baldini et al. (2018). This approach has been previously used by other authors under similar environmental conditions (Baldini et al., 2018 and Bragaglio et al., 2023). The second FU has been chosen in agreement with Pirlo and Lolli (2019), which also estimated milk sustainability in a comparable geographic scenario. This FU includes both on-farm and off-farm land areas required to provide all inputs from the technosphere. The term technosphere is defined as the summed material output of the contemporary human enterprise, i.e., according to ISO 14,044, the elementary flows or intermediate flows (ISO 14044, 2006). As was the case in a recent study (Biagetti et al., 2023), not all impact categories refer to the second FU. Given the small number of comparison items, the present study, only GWP and NREU refer to 1 hectare of occupied area.

2.3. System boundaries definition

The cradle-to-gate analysis indicates the physical and chronological boundaries in which the study was performed. According to a widespread practice, the system's chronological boundary is a calendar year (365 days). Within the physical boundary, FPCM is the main product available at the farm gate. Given the highly specialized “dairy profile” of both farms, as suggested by Bragaglio et al. (2023), the main by-product is the meat, provided by culled cows only, male calves were not considered in this study. Both farms use almost exclusively sexed semen, minimising the birth of male calves. In addition, these animals are sold as soon as possible.

We considered all the operations pertaining to animal care, i.e., rearing, housing, weaning, feeding and milking. Feedstuffs and concentrates for the diet of lactating cows are purchased in both farms, and both showed full self-sufficiency for forages. Concentrates used in the L farm are richer in extra-oceanic (mainly soybean) sources than those of S. According to the literature drugs, detergents and disinfectants have been excluded (Chirone et al., 2022). One of the studies (Chirone et al., 2022) included detergents, given their impact on water scarcity and human health, for example). The analysis also involved the burdens directly attributable to farming procedures, such as energy requirements (fuel and electricity for all the operations) and those from outside the upstream flows, for example, energy sources and emissions related to the production of chemical fertilizers, seeds, and pesticides. This scheme, shown in Fig. 1 (separate file, Fig._1), clarifies the framework. The blue dotted line defines the system boundaries (from cradle to farm gate), with the main product (1 kg of FPCM) and the by-product (the culled cows). The light pink area shows (i) on-farm self-produced feed inputs, and (ii) the output, split in output to the technosphere (milk and meat), and output to the environment (the pollutants). The light green area shows off-farm activities (i.e., data about feed, agrochemicals, energy, and transportation).

Fig. 1.

System boundaries, representing on-farm self-produced feed inputs (pink), off-farm activities (green), outputs to the environment (i.e., pollutants), and outputs to the technosphere, (i.e., the main product -milk- and by-products -the culled cows-), at the farm gate.

2.4. Allocation

ISO 14,044 standards (ISO 14044, 2006) provide standards for partitioning the input and output flows. According to these standards, dairy systems cannot be split, because milk and meat cannot be produced separately. Ardente and Cellura (2012) supported this idea, suggesting economic allocation as the most adequate for dairy systems. The same approach has been suggested also for fishery (Ziegler et al., 2003 and Kok et al., 2020).

Economic allocation is an agreed-upon and widespread method for dairy studies. Applications can be found in Pirlo et al. (Pirlo et al., 2014a, 2014b)(who carried out two LCA studies on buffalo milk), Pirlo and Lolli (2019) (who compared conventional and organic dairy cattle farms in Italy), and also in a Danish study (Mogensen et al., 2022).

The equation below, previously used by other authors (Ardente & Cellura, 2012; Kok et al., 2020), leads to the economic allocation partitioning in the present study:

$$P_i=\frac{\left(n_i*x_i\right)}{\sum \left(n_{ij}*x_{ij}\right)}$$ (1)

(Eq.1), where:

P = partitioning factor;

n = amount of the product (kg);

x = price of the product (€);

i = main product;

j = by-product.

In this study economic allocation was used to estimate the sustainability of the FU, attributing some of the burdens to the by-product. The same mathematical criterion was used for the share of culled cows. In fact, it is known that a high culling rate indicates (i) cows with poor longevity (low number of lactations), and (ii) the realistic possibility of reduced income (Fetrow et al., 2006; De Vries, 2017; Rilanto et al., 2020). The culling rate is the component that mostly affects this scenario (Cockram, 2021; De Vries, 2020; Hadley et al., 2006).

Some studies recently proposed the use of the term “reference allocation” rather than “no allocation” when the main product is loaded with all pollutants of the system (Bragaglio et al., 2023, 2024), and we followed this indication. The same approach was used by Lauri et al. (2020) in a LCA study on timber.

With respect to the choice of economic allocation, the main options for allocating the environmental load of a process are 1) the mass of products and by-products obtained, or 2) their economic value; the choice discretion is linked to the reasons why the LCA study is being carried out. In the present study, we are evaluating and comparing the environmental profile of two livestock farms for their technical, and consequently economic, dimensions.

Allocation carried out according to the mass of products and by-products of the farms has certainly the advantage of providing stability over time to the results obtained because, unless there are strong innovations in production technologies, the relationships between products and by-products remain very similar over time; on the contrary, the allocation in terms of economic value is based on prices and therefore highly variable over time given the high volatility of prices, especially for primary products (Ardente & Cellura, 2012).

However, for some production processes the mass of by-products is equal or even greater than that of the main product, thus generating inequality in the attribution of environmental loads: greater "environmental responsibility" for by-products compared to the main product (which is the main study object and cause of the production of the by-products) (Kneese, 1970). In the economic allocation instead, the economic value determines the distribution of environmental loads. Consequently, the market (i.e. the price) attributes to the main product an objectively higher value than to all by-products. This is the reason that led us to use the economic criterion in allocation rather than others (Guinée et al., 2004).

2.5. Life cycle inventory

Primary data was collected from January to December 2023 in both the large (L) and small (S) farm through interviews with farmers and data collection in the field. S farm was already investigated in a previous study (Bragaglio et al., 2023) and data have been updated. Slightly different results were obtained, in particular by changing the buildings and the use of agrochemicals. Both farms use sexed semen, mainly for heifers, moreover, according to Bragaglio et al. (Bragaglio et al., 2023) the system boundaries of both farms exclude male calves (see Table A1), which are sold as soon as possible (as soon as the navel is healed, after the 15th day of life). The only other study (Romano et al., 2021b) proposing allocation as a marker of culled cows (Bos taurus) also excluded male calves.

The inventory involved the sheds (Shed {GLO}, market for, Alloc, Def, S) for all livestock categories (young animals, dry cows, and lactating cows). Similarly to other studies (Bragaglio et al., 2023, 2024; Romano et al., 2021b), we assumed for these buildings a productive life of 50 years, while keeping the default options for the construction (Ecoinvent database). Data collection for other concrete buildings (bunker for silages, pits, shelter for tractors and implements or trucks and trailers) was not carried out. These facilities have been excluded from the inventory. As shown in Fig. 1 (separate file, Fig._1), the inventory follows the distinction between on-farm and off-farm resources, summarized in Table 1, Table 2 respectively.

Table 1.

On-farm resources, main product, by-products and allocation framework. Description of arable area (AA), forage availability, herd composition, sheds area and number of automated milking systems (AMS). Sizes refer to sheds housing all livestock categories except the younger calves, which are kept in cages. Female calves = up to 6 months; Heifers = 6–24 months. Floor type is shown for each livestock category (Heifers -H-, Dry Cows -DC- and Lactating Cows -LC-). *Triticum aestivum L.; **Triticale (x Triticosecale Wittmack) and Triticum aestivum L.; ***, Lolium perenne, Medicago sativa and Triticum aestivum L.; ^a(CLAL.it, 2024a); ^bCLAL.it, 2024b; ^cCamera di Commercio di Cremona, 2024.

	L	S
Arable area (AA), Ha
Total	200	70
Irrigated	150	40
Not irrigated	50	30
Forages, tons year−1
Corn silage	1650	1500
Wheat silage*	–	450
Sorghum silage	–	600
Alfalfa silage	230	–
Mixed silage**	1180	–
Alfalfa hay	490	195
Italian ryegrass hay	–	205
Meadow hay***	335	–
Herd composition, heads
Total, Livestock Units (LUs)	710	287
Female calves	170	30
Heifers	240	100
Dry cows	50	35
Lactating cows	250	120
Bulls	–	2
Sheds, m2	7800	2720
Solid concrete floor	H, DC, LC	H, DC
Perforated concrete floor	–	LC
Automatic Milking System, n°	4	2
Milk yield year−1, tons	2920	1300
Income from milk, € ton−1	538a	522b
Total income from milk, €	1570,960	678,600
Live weight of culled cow, tons head−1	0.65
Income from culled cows, € ton−1	1180c
Culled cows year−1, heads	90	60
Live weight of culled cows year−1, tons	58.5	39
Total income from culled cows, €	69,030	46,020
Income from milk %	95.8	93.7
Income from culled cows %	4.2	6.3

Table 2.

Off-farm resources. External inputs are divided into (i) simple feeds, (ii) concentrate meal, for each livestock category, (iii) energy, (iv) herbicides, (v) synthetic fertilizers, and (vi) straw for litter. Lactating cows are fed with “a” and “b” concentrates. The chemical composition of concentrates provided to heifers and cows, specified for L and S farms, is shown in Table S.1 (please see the Supplementary materials).

	L		S
Feedstuffs	km travel	tons year−1	km travel	tons year−1
Wheat straw	40	90	450	150
Corn (flower)	100	530	100	160
Corn (flaked seeds)	100	180	–	–
Milk replacer powder	1000	3	1000	0.5
Soybean meal	200+10,000	370	–	–
Concentrates per each animal category	tons year−1		tons year−1
Calves	43.4		7.8
Heifers	95.7		90
Dry cows	36.5		30
Lactating cows a	185		160
Lactating cows b	290		170
Energy
Diesel, tons year−1	114		60
Electricity, MWh year−1	418		120
Synthetic fertilizers, tons year−1
Urea (N 46 %)	70		32
Ammonium nitrate (N 27 %)	25		16
P2O5 (P 31 %)	30		2
Straw for litter, tons year−1	370		150

Table 1 shows the arable area (AA), the quantity of forages produced in both farms (distinguishing between silage and hay), and the shed overview. In agreement with other studies (Romano et al., 2021b; Bragaglio et al., 2023), forages produced on farm are assumed to be transported for 1 km by tractor (Transport, tractor and trailer, agricultural {GLO}, market for, Alloc, Def, S). The tractor, available in Ecoinvent database, is computed without changing its default options, whereas forages characteristics have been modified from those suggested by the database (Agri-footprint). With respect to manure management, in L all floors are solid, thus manure is removed with V-blade scrapers (Buck et al., 2013) and tractor operations. In S, only lactating cows are on a perforated concrete floor with manure collection underneath, while manure from heifers and dry cows is removed with a tractor.

Cows of L and S farms are milked with 4 and 2 Automatic Milking System (AMS) respectively.

Emissions are considered as on-farm, however, given the peculiarity of the topic, a more extensive description will be provided in Section 2.6.

Although both farms show a high level of self-sufficiency, some inputs were purchased. Neither L farm nor S farm has energy production from renewable sources, therefore, besides diesel, also electricity originates off-farm. Both farms purchase all the straw, used both as diet ingredient and litter.

The off-farm inputs, summarized in Table 2, are divided into materials and energies. Another significant division splits simple feeds (e.g., straw) from concentrates (sold as a mixture after industrial processes, i.e., milling, crushing, heating). Similarly to on-farm flows, the means of transport have been considered also for external inputs. From the Ecoinvent database, transoceanic cargo was used for the shipping of soybean and palm oil (Transport, freight, sea, transoceanic ship {GLO}, market for, Alloc, Def, S) and trucks were used for road travel (Transport, freight lorry >32 t, EURO 5, RER); (Transport freight lorry 3.5–7.5 t, EURO 5, RER). The travel distances and input quantities are shown in Table 2.

2.6. Emissions

The LCA environmental indicators lead the choice of guidelines and equations helpful to assess the emissions. IPCC guidelines (IPCC, 2019b) are commonly used in livestock systems, and suggest several assessing approaches, depending on the desired accuracy level, i.e., Tier 1, Tier 2, or Tier 3. According to Bava et al. (2014) a special attention is paid to the greenhouse gases (GHGs, i.e., CO₂, CH₄ and N₂O) that are produced on-farm. The main source of on-farm GHGs are animals however, crops, synthetic fertilizers, and energy also contribute to the GWP impact category. In agreement with other studies (Bragaglio et al., 2018, 2023) and IPCC guidelines (IPCC, 2019b), carbon dioxide emissions from livestock are negligible. Emissions from the soil carbon stock variation were not taken into account.

The owners and hired workers of L and S live, as usually happens, close to the farm. Thus, carbon dioxide emissions provided by their travel have been excluded. On farm L the higher demand for hired labour is compensated by more family members than S.

Manure, dung, livestock excreta and synthetic fertilizers affect NH₃ and NOx emissions. Acidification and eutrophication are not evaluated in this study, so we considered only greenhouse gases (methane and nitrous oxide) provided by manure and synthetic fertilizers .

The evaluation of NREU, FD and ALO, requires (i) an accurate inventory data collection, (ii) a careful choice and entering of the materials available in the software database, (iii) correct data processing.

Table 3 provides a synthesis of emissions sources, equations, emission factors and references. The equations used to assess enteric emissions and methane manure emissions are explained in detail, in Appendix A.

Table 3.

N₂O emissions from manure management. Equations, emission factors and references. Nex=nitrogen excreted; AWMS= fraction of manure from the livestock category, handled with the specific manure management system. The values used for dairy cattle in Western Europe are 43 % for slurry/liquid storage (in L and S farms) and 29 % for solid manure (L farm only). EFa,b=IPCC 2019a, Tab.10.21; EFc=IPCC 2019b, Tab. 11.1; EFd,e=IPCC 2019b, Tab. 11.3; Frac_Gas(T,S)=0.3 volatilising default value for storage origin (liquid with natural crust and solid), Tab. 10.22, IPCC 2019a; Frac_Gas(T,S)=0.21 volatilising default value at field, Tab. 11.3, IPCC 2019b; Frac_LeachMS(T,S)=0.24 leaching default value at field, Tab. 11.3. 44/28=conversion of N₂O—N emissions to N₂O emissions.

	L	S
Housing,N2Okghead−1year−1
Calves	0.40 (Baldini et al., 2018; Bragaglio et al., 2023)
Heifers	1.05 (Baldini et al., 2018; Bragaglio et al., 2023)
Dry cows	1.30 (Baldini et al., 2018; Bragaglio et al., 2023)
Lactating cows	- (Baldini et al., 2018; Bragaglio et al., 2023)
Direct	Nex * AWMS * EF * 44/28
Storage	AWMS=43 %, 29 %; EFa,b = 0.005, 0.01 (IPCC, 2019b)	AWMS=43 %; EFa=0.005 (IPCC, 2019b)
Field	AWMS=43 %, 29 % (IPCC, 2019b); EFc=0.004 (IPCC, 2019a)	AWMS=43 % (IPCC, 2019b); EFc=0.004 (IPCC, 2019a)
Indirect, volatilization	Nex * AWMS * FracGas(T,S) *EF* 44/28
Storage	AWMS=43 %, 29 %; EFa,b = 0.005, 0.01 (IPCC, 2019b)	AWMS=43 %; EFa=0.005 (IPCC, 2019b)
	FracGas(T,S)=0.3 (liquid with natural crust / solid) Tab 10.22 (IPCC, 2019b)
Field	AWMS=43 %, 29 % (IPCC, 2019b); EFd=0.01 (IPCC, 2019a)	AWMS=43 % (IPCC, 2019b); EFd=0.01 (IPCC, 2019a)
	FracGas(T,S)=0.21 Tab 11.3 (IPCC, 2019a)
Indirect, leaching	Nex * AWMS * FracLeachMS(T,S) *EF* 44/28
Field	AWMS=43 %, 29 % (IPCC, 2019b)	AWMS=43 % (IPCC, 2019b)
	FracLeachMS(T,S)=0.24, EFe=0.011 (IPCC, 2019a)

2.6.1. Emissions from manure management, synthetic fertilizers, soils and energy sources

Nitrogen balance (N) data are required to estimate nitrous oxide (N₂O) emissions. According to Romano et al. (2021b) and Bragaglio et al. (2023) we used nitrogen balance data previously calculated by Nennich et al. (Nennich et al., 2005), i.e., 0.491, 0.228, 0.117 and 0.063 N kg/day for lactating cows, dry cows, heifers and calves respectively. Afterwards, following IPCC guidelines (Tier 2), we considered housing and direct and indirect N₂O emissions, calculating direct and indirect emissions during storage and in the field. Table 3 reports the equivalents, origin, formulas, emission factors and references for N₂O emissions. In the S farm manure is managed as liquid (i.e., slurry), while in the L farm both liquid and solid manure are handled. Nitrous oxide emissions from housing were taken into account as reported in the literature (Baldini et al., 2018; Bragaglio et al., 2023), whereas emissions due to manure management have been calculated according to IPCC guidelines (IPCC, 2019a, 2019b).

Chirone et al. (2022) performed a LCA of dairy buffalo farms and found a low contribute to GWP from synthetic fertilisers (2.4 %). In this study, the synthetic fertilisers applied are urea (46 % N) and ammonium nitrate (27 % N). Agri-footprint, Ecoinvent 3 and USLCI databases suggest using these agrochemicals as inputs from the technosphere, including (i) burdens from industrial processes, (ii) emissions at field. (IPCC, 2019a). Default values that affect livestock emissions, classified as (i) direct, (ii) indirect volatilization, (iii) indirect leaching, are described by the following emission factors: (i) 0.005 and 0.01 for liquid and solid manure storage, 0.004 for field emissions; (ii) a specific Frac_GAS default value (0.3), in addition to 0.005 and 0.01 for liquid and solid manure storage, moreover a specific “on field” equation matches Frac_GAS default value (0.21) with 0.01 emission factor; (iii) field spreading only concerns leaching (Frac_LEACH 0.24 default value) combined with 0.011 emission factor. Emission factors for Frac_GASF are 0.15 and 0.05 for urea and ammonium nitrate respectively. Frac_LEACH emission factor for N addition sensu lato has a default value of 0.24 (IPCC, 2019a).

Diesel fuel density was set at the standard value of 850 kg/m³ and, based on the findings from Romano et al. (Romano et al., 2021b), a 3.13 eq. emission factor was used to estimate CO₂ release from the combustion of 1 kg of diesel, together with the emission factor of 0.47 eq. for 1 kWh of electricity (energetic national mix).

2.7. Impact assessment

This study followed the EPD 1.04 and the ReCiPe Midpoint (H) methods. We used the first one to estimate Global Warming Potential (GWP in the 100 years perspective, kg CO₂) and Non-Renewable Energy Use (NREU, MJ), while the second method was chosen for Fossil Depletion (FD, g oil) and Agricultural Land Occupation (ALO, m² year⁻¹) environmental indicators . Agri-footprint, Ecoinvent 3 and USLCI databases, provided by software SimaPro 8.03 (SimaPro, 2020), ensured several materials, were considered with or without changing their default options.

2.8. Statistical analysis

Using fat protein corrected milk (FPCM) and hectare as functional units, the values resulting from the LCA process included four (GWP, NREU, FD, ALO) and two (GWP and NREU) impact categories respectively. Values were processed by LSD (Least Significance Difference) statistical test (Fisher, F., 1937) using in R (R Core Team, 2023), considering a t-critical value from the t-distribution of α =0.05. The small sample size was the reason why the LSD test was used, as done also by Bragaglio et al. (2024).

The FactoMineR package (Lê et al., 2008) of R software was used to perform a principal component analysis (PCA). The PCA, where the first dimension explains 53.6 % of the total variance and the second dimension explains 21.3 %, helped showing the “efficiency level” of L and S farms, identifying three clusters, as follows:

• (i)farm area and crop, named ha_crop, with 7 variables, i.e., the ratios between: irrigated area and total arable area; dry matter of corn silage and total arable area; dry matter of corn silage and irrigated area; dry matter of non-corn silages and total arable area; dry matter of non-corn silages and irrigated area; dry matter of hay and total arable area; dry matter of hay and irrigated area.
• (ii)cattle and milk, named cattle_fpcm, with 7 variables, i.e., the ratios between: total amount of FPCM and total arable area; herd (all livestock categories) and total arable area; lactating cows and total arable area; dry cows and total arable area; heifers and total arable area; female calves and total arable area; culled cows and total arable area.
• (iii)energy sources and milk/cattle, named energy, with 6 variables, i.e., MWh year⁻¹ and FPCM; MWh year⁻¹ and herd (all livestock categories); MWh year⁻¹ and lactating cows; diesel tons year⁻¹ and FPCM; diesel tons year⁻¹ and herd (all livestock categories); diesel tons year⁻¹ and lactating cows.

3. Results and discussion

In literature, reference allocation is largely identified as “no allocation”, and all the burdens are attributed to the milk, which is the main product. Although the comparison is only speculative, in absence of statistical validation, the following considerations are helpful to describe the role of the different contributors to the four impact categories in L and S. This study considers four LCA environmental indicators, the framework in which the data were collected (Po Valley) and the current sensitivity to some issues guided their choice. We performed a LCA evaluating (i) the global warming potential (GWP) quantified as CO₂ kg and tons equivalents, with reference to FPCM and hectare respectively; (ii) the non-renewable energy use (NREU) quantified as MJ and GJ with reference to FPCM and hectare respectively; (iii) the fossil depletion (FD) and (iv) the agricultural land occupation (ALO) per FPCM, identified by g oil and m² year respectively. For each indicator, values were assessed with reference allocation and after economic allocation. A summary of the results is shown in Table 4, which will be described and discussed in detail in the next sections.

Table 4.

Life cycle assessment analysis results for cow milk production, with reference and economic allocation. Different superscripts on the same line indicate a significant difference (P < 0.05). Abbreviations: L = Large farm; S = Small farm; GWP = Global Warming Potential (kg and tons CO₂ eq.); NREU = Non-Renewable Energy Use (MJ and GJ eq.); ALO = Agricultural Land Occupation (m² year eq.); FD = Fossil Depletion (g oil eq.).

	Reference allocation		Economic allocation
	L	S	L	S
Impact category per FPCM
GWP (kg CO2)	1.41a	1.37a	1.35a	1.28b
NREU (MJ)	7.05a	6.42b	6.75b	6.02c
FD (g oil)	162a	156a	155a	146b
ALO (m2 year)	2.05b	2.23a	1.96b	2.09b
Impact category per ha of total occupied area
GWP (tons CO2)	6.8b	8.5a	6.5b	8.0a
NREU (GJ)	36.1b	37.6a	34.6b	35.2b

Automatization level and farm size farms led this study. The impact categories most affected by automation systems (NREU and FD) showed better sustainability in S. Although AMS is usually considered a more viable technology for large farms, actually the main target group of AMS is medium-sized family farms with 60 to 180 cows (Bernhardt et al., 2019), and this may explain a better compatibility of AMS with the S farm also in terms of sustainability. Interestingly, also as concerns AFS, the highest density of AFS-equipped dairy farms in Italy is in the mountain province of Alto Adige/Südtirol (Romano et al., 2023) and has a smaller average size compared to the Po Valley (20 lactating cows per farm) (Rete di Informazione Contabile Agricola, RICA).

This considerable spread of AFS on small farms could explain the better energy efficiency (NREU and FD) of S farm.

According to Lovarelli et al. (Lovarelli et al., 2020) future studies should be done to evaluate the role of precision technologies in sustainability, also in relation to farm size.

3.1. LCA categories with reference allocation, fat protein corrected milk as functional unit

3.1.1. Global warming potential

Values of kg CO₂ equivalents are similar in the two farms. In addition, they fall in the range provided by literature (Baldini et al., 2017). As indicated in Fig. 2, the farms show an analogous profile for synthetic fertilizers, nitrous oxide emissions, and agricultural areas, sheds, and energy sources (AASES). The share of AASES is comparable to the one indicated by Berton et al. (2020), in the sum of “materials and herd handling” (almost 20 %). AASES share is also similar to the carbon dioxide weight (25.4 %) reported by Bava et al. (2014).

Fig. 2.

Percentage of contributors to Global Warming Potential (GWP, indicated as CO2 equivalents) of Fat Protein Corrected Milk (FPCM), with reference allocation, in L and S farms.

The marginal contribution of synthetic fertilizers is confirmed by Guerci et al. (2013).

As concerns the GWP profile of the farms, the contribution of nitrous oxide emissions, mainly due to manure management are under 10 %. The percentage is comparable with that reported by Baldini et al. (2018) and De Boer et al. (2011), however, these authors state that these emissions show high variability.

Methane emissions show differences between the two farms. According to Pirlo and Lolli (2019) the better production efficiency (32 and 29 kg milk yield cow⁻¹ day⁻¹, in L and S respectively) could reduce greenhouse gas emissions, in addition to the different management of manure management from lactating cows.

The CO₂ equivalents from feed describe a contrary trend, with L having higher values than S. This is explained by an effect of the proportions between the other items, mainly affected by the share of methane emissions. Baldini et al. (2018) compared three farms obtaining a contribute of feed to GWP ranging from 27 to 60 %. These farms, similarly to ours, have almost reached self-sufficiency in terms of forages, whereas concentrates are purchased.

3.1.2. Non-renewable energy use and fossil depletion

The main contributors to NREU and FD are synthetic fertilizers (SF), feed (F), agricultural areas, sheds and energy sources (AASES), and show very similar percentages. Their proportions are due to the absence of renewable energy sources in both farms. Ecoinvent, USLCI and Agrifootprint databases provide materials that have been used without changing the default options (for example Electricity, medium voltage {IT}, market for, Alloc Def. S; Transport, tractor and trailer, agricultural {GLO}, market for, Alloc Def. S; Transport, freight, sea, transoceanic ship {GLO}, market for, Alloc Def. S) and materials adapted to the specific scenario. For the latter, some characteristics have been changed, such as hectare yields of winter cereals.

The small differences in the percentages are mainly because the two indicators refer to two different methods.

As noted above, about 90 % of the values attributable to FD and NREU, concern F and AASES. Salou et al. (2017) found a comparable percentage for the sum of similar categories, named “Feed, Energy and Buildings”.

Battini et al. (2016) compared four dairy farms in the Po Valley and calculated the energy (MJ) used for feeding. Although the data concerned all the animals, the results of two of the four farms are comparable with those of the L farm of the present study (27 % and 73 % of MJ for on-farm and off-farm feed respectively), and the values of the other two farms are similar to the S farm (40–45 % and 60–55 % of MJ for on-farm and off-farm feed respectively).

The present study focuses on the feeding of two livestock categories: lactating cows and calves. The first category was included because it is directly involved in milk production, and the second one because of the higher proportion of female calves in the L than in the S farm.

Lactating cows category. In the L farm (Fig. S1, Supplementary materials), requirements of non-renewable energy are poorly covered by on-farm feed (about 20 %). According to this matter, Guerci et al. (2013) attributed 16 % of energy requirements to on-farm crop production. Berton et al. (2020) estimated the energy demand of mountainous dairy farms, finding a mean value below 30 % for on-farm sources and highlighting the role of purchased concentrates.

Fig. S2 (Supplementary materials) shows MJ equivalents for feeding split across ingredients. Bragaglio et al. (2023) in a previous study carried out at the S farm calculated for “Corn silage, at farm” about 0.35 MJ. In the present study S shows about 60 % of the MJ for lactating cows deriving from corn silage. Non-renewable energy required by on-farm forages shows a larger share (44 %) in S than L farm. Although the percentage is included in the wide range reported in the literature (Basset-Mens et al., 2009; Battini et al., 2016), the value is probably due to the low efficiency (diesel fuel per dry matter) of wheat and sorghum silages (absent in the diet for lactating cows of the L farm), in comparison with corn silage (Bacenetti et al., 2015).

When assembling the materials processed with the software, the energy share of hay and other silages (wheat and sorghum), is represented by crude oil and diesel, as shown by Bragaglio et al. (2023).

Table S.2 and Fig. S3 (Supplementary materials) show an inventory of non-renewable energy sources, explaining the different profiles of contributors in L and S farms.

Female calves category. The farms show a different quantity of these animals, more than double in proportion in L compared to S. In the young animals’ diets, concentrates (such as milk replacers or weaning meal) are of particular interest in the assessment of feeding sustainability. In the L farm, a larger percentage of calves and a careful formulation of their diet results in a different share of NREU and FD. The burdens of forages are very low and negligible in both farms, whereas the L farm showed higher MJ equivalents than S. This is because (i) there are almost six times as many animals in L than in S, (ii) concentrates of the L farm are richer in extra-oceanic (mainly soybean) sources than those of S. About this second point, according to Raucci et al. (2015), oil, energy derived from it and transport, are considerable inputs during soy cropping. In addition, transoceanic travel is an important contributor to the NREU category in terms of fossil fuel use (Pérez-Neira et al., 2020). Table S.3 and Fig. S4 (Supplementary materials) show an inventory of non-renewable energy sources, explaining the different profiles of contributors in L and S farms.

3.1.3. Agricultural land occupation

The values of ALO for the L and S farms are similar. Land use environmental indicators are frequently investigated, and a review (Baldini et al., 2017) provides a comparable range (0.8–2.8 m² per year) for raw milk. Low equivalents refer to “high-intensity level” farms(Bava et al., 2014). Romano et al. (2021b) found larger values (3.9 m² year per FPCM) in farms without corn silage, with few dairy cows/ha. Pasture-based systems and organic farms often show lower sustainability in terms of land use, as reported by Romano et al. (2021b) and Penati et al. (2016).

Battini et al. (2016) compared the FPCM sustainability of four dairy farms in the Po Valley, and found that those farms that purchased 100 % of the feed (similarly to those of the present study) were not the ones where the percentage contribution of off-farm feed was the highest. A third farm, with a more pronounced self-sufficiency of concentrates (31.8 %) showed the highest value of land occupation, due to off-farm feed (concentrates and purchased forages).

Several authors, performing LCA studies (Salou et al., 2017; Bragaglio et al., 2018) found a strategic role of feed (both on-farm and off-farm) in land use. Similarly, Bragaglio et al. (2023), attributed to feed the largest contribution to land occupation. Fig. 3 (separate file, Figure_3) shows (i) the role of feed on ALO; (ii) the high share of concentrates in calf feeding (L farm), in terms of the land occupied to cultivate transoceanic soybean; (iii) a more precise heifer feeding in L farm compared to S.

Fig. 3.

Contributors to ALO (Agricultural land Occupation). Feed (off-farm and on-farm) contributions are shown by livestock category. Agricultural area, shed and energy also shown.

3.2. LCA categories with reference allocation, hectare of total occupied area as functional unit

Baldini et al. (2017) described a large use of milk (sensu lato), as the functional unit. The “land” as the functional unit was used in fewer studies, defined as the global hectare, representing “a standardized hectare with world average productivity” (Biagetti et al., 2023) and taken into account as the hectares (ha) of the total occupied area, i.e., on-farm plus off-farm land (Salou et al., 2017).

3.2.1. Global warming potential

In this study the L and S farm showed 6.8 and 8.5 tons of carbon dioxide equivalents respectively. Ross et al. (2017), performing a LCA on bovine milk, adopted different functional units, including Land_farm (ha) and Land_total (ha). They compared four farming systems and found a range of carbon dioxide equivalents similar to our results. Salou et al. (2017) carried out a LCA comparing 7 clusters of French lowland dairy farms and found a larger variability in GWP (4.4–8.8 tons CO₂ eq.). According to Pirlo and Lolli (2019), GWP increased together with intensification of milk production. Salou et al. (2017) demonstrated how many variables determine the intensification of production (cattle breed, energy sources, synthetic fertilisers, purchased feeds and crop yield per hectare). In our study, L shows a higher milk yield than S, and this is a peculiarity of intensification. However, carbon dioxide equivalents are lower in L. Interestingly, other low-intensification traits characterize this farm, for example stocking and diesel density have lower values compared to S.

3.2.2. Non-Renewable energy use

As for the GWP, the NREU values show a similar trend, although the differences are less accentuated.

Bava et al. (2014) classified dairy farms based on their intensity level (low, medium, high) and found that high-intensity farms had larger energy use per hectare of land compared to low and medium-intensity farms. Similarly, Salou et al. (2017) compared seven dairy farms finding the highest energy demand in the corn-based ones.

Despite the higher milk yield, the L farm is “less intensive” for other traits, thus explaining the more sustainable results for energy use in the reference allocation framework.

3.3. LCA categories with reference and economic allocation, fat protein corrected milk and hectare of total occupied area as functional unit

Allocation improves the sustainability of the main product (Table 4). However, in this study, only culled cows are allocated, and this criterion could be used as a marker. Romano et al. (2021b) compared three dairy farms attributing to the allocation the same role of marker. Culled cows can also be an undesirable product, and a high culling rate can be a sign of inappropriate management and lost income. Bragaglio et al. (2022) performed a LCA investigating (i) the role of corn silage in the sustainability of buffalo milk, and (ii) proposing the allocation criterion of involuntary culling rate. Corn silage-based diets improved acidification and eutrophication potential whereas economic allocation did not affect any impact category.

In the present study, the S farm showed a larger culling rate than L (38 % vs. 30 %). Moreover, we computed the same price of culled cows in both farms according to the Chamber of Commerce of the city of Cremona. On the other hand, the price of milk was computed by referring to the different markets (Veneto region for L and Lombardia for S, see Table 1).

3.3.1. Global warming potential

FPCM as DFU: There are no differences between GWP values, except for CO₂ eq. being the lowest in S after culled cows allocation (P < 0.05).

Romano et al. (2021b) also found that in the farm with the lowest culling rate, after economic allocation, fewer impact categories changed (i.e., GWP and FD) compared to the other farms. (Baldini et al., 2017) in a review about dairy cattle found more studies based on the functional unit (124) than the allocation method (75). In addition, these authors stated that the economic criterion was the most largely used.

Hectare of total occupied area as DFU; The allocation did not improve farm sustainability, however, LSD statistical test highlighted differences (P < 0.05) between L and S, showing improved sustainability of L compared to S . Salou et al. (2017) also used FPCM and hectare as functional units. The authors allocated by-products (calves and culled cows), but they did not compare LCA environmental indicators before and after economic allocation. Notwithstanding this, the study investigated milk yield per area. Interestingly, GWP showed a significant and positive correlation (r) with milk yield per hectare. In addition, CO₂ eq. was correlated with calves and culled cows.

3.3.2. Non-renewable energy use

FPCM as DFU: L after reference allocation showed the highest NREU value while S after economic allocation showed the lowest one. Intermediate values, without statistically significant difference between each other, were observed in S farm after reference allocation and L farm after economic allocation. These values are lower (P < 0.05) than those found in the L farm with reference allocation, but higher (P < 0.05) than those of the S farm after economic allocation. This is the category most affected by allocation when FPCM is used as DFU.

The L farm has lower forage efficiency (yield/ha) than S, and in similar conditions (organic vs. corn-based farm), Salou et al. (2017) highlighted the larger share of energy use in the organic farm (50 vs. 30 %). In addition, Bava et al. (2014) and Nguyen et al. (2012) found that increased feed efficiency on farm decreased energy requirements.

Thus, according to allocation logic, the ratio between prices and amounts of FPCM and by-products decreased the MJ eq. in the S farm. Milk, which is the main product, has higher yield and better price in L compared to S (32 vs. 29 kg cow⁻¹ year⁻¹, and 538 vs. 522 € ton⁻¹). Moreover, despite culled cows of L and S having the same economic value, in the S farm the culling rate is higher (38 vs. 30 %) leading to a higher income from the sale of culled animals.

Hectare of occupied area as DFU: This DFU changed energy values less than FPCM. In reference allocation, the equivalents of the S farm are higher (P < 0.05) than all other values. As mentioned for reference allocation, the high intensity of the S farm in comparison to L, resulted in a higher value of MJ eq. after economic allocation. Similarly, Bava et al. (2014) described the “high intensity farms” as less sustainable based on their environmental profile per hectare. Contrary to what happened for GWP, Salou et al. (Salou et al., 2017) found no significant correlation between NREU and by-products (culled cows and calves).

3.3.3. Fossil depletion (FPCM as DFU only)

This environmental indicator shows the same profile of GWP when FPCM is used as DFU. FD remains a poorly studied descriptor, and a dairy cattle review (Baldini et al., 2017) reported no studies discussing it. Bragaglio et al. (2023) compared two farms and found that, in the precision agriculture-based farm, oil eq. were significantly decreased by allocation. In the conventional farm, the equivalents were not changed (P > 0.05) by allocation.

Romano et al. (2021b) investigated several LCA indicators in a dairy Holstein-Friesian farm. They found that economic allocation affected GWP and FD.

A study by Ghinea and Leahu (2023) performed a LCA of sheep cheese. They found that the characterization of FD is mainly represented by diesel (95 %). The main contributor to GWP is “cheese production” (82 %), followed by diesel and electricity. According to the authors, the main contributors to “cheese production” are energy sources, thus in GWP and FD the energy contribution is close to 100 %. Höök and Tang (2013) described a CO₂ emission trend from fossil sources (gas, oil and coal) from 1971 to 2009, highlighting that as fossil-based energy sources grow, so do carbon dioxide emissions. This proportionality would explain the same response of GWP and FD to allocation.

3.3.4. Agricultural land occupation (FPCM as DFU only)

The S farm with reference allocation shows the highest ALO value (P < 0.05). There are no differences between ALO eq. before and after economic allocation in L (P > 0.05), and these values do not differ from S farmland occupation with economic allocation. Romano et al. (2021b), compared two conventional and one organic, pasture-based farm. Feed production as contributor to land occupation was higher in the conventional than in the organic farm. According their findings, on-farm and purchased feed for lactating cows and heifers account for 50–66 % of ALO. On the other hand, the extent of grazing land (about half of the agricultural area) reduced the contribution of feeding to ALO in the organic farm. In the conventional farms, allocation (economic and mass modes) did not affect land occupation, whereas in the organic farm mass allocation affected this category. This is explained by the high mass of culled cows, due to the size of Simmenthal cattle.

3.4. The role of farm size and new perspectives for allocation

As suggested by Ren et al. (2019), farm size has an effect on sustainability, as partially observed also in the present study. Using the hectare as DFU, the L farm showed a better profile, mainly in terms of GWP. The S farm, on the other hand, showed lower values when using milk as DFU. The different degree of intensification of the two farms can explain this result.

As previously mentioned, PCA was a helpful tool for estimating the profile and intensification level of the two farms. In Fig. 4 (separate file, Fig. 4) the environmental indicators (divided in three clusters, i.e., (i) ha_crop; (ii) cattle_milk; (iii) energy), showed some interesting differences between L and S.

Fig. 4.

Principal component analysis of farm efficiency. Farm efficiency is described with three clusters: (i) ha_crop; (ii) cattle_milk; (iii) energy. Each cluster displays some environmental indicators descriptors, each expressing a ratio between two variables, as follows:

(i) ha_crop_1: irrigated area and total arable area, ha_crop_2: dry matter of corn silage and total arable area, ha_crop_3: dry matter of corn silage and irrigated area, ha_crop_4: dry matter of non-corn silages and total arable area, ha_crop_5: dry matter of non-corn silages and irrigated area, ha_crop_6: dry matter of hay and total arable area, ha_crop_7: dry matter of hay and irrigated area.

(ii) cattle_milk_1: total amount of FPCM and total arable area, cattle_milk_2: herd (all livestock categories) and total arable area, cattle_milk_3: lactating cows and total arable area, cattle_milk_4: dry cows and total arable area; cattle_milk_5: heifers and total arable area, cattle_milk_6: female calves and total arable area, cattle_milk_7: culled cows and total arable area.

(iii) energy_1: MWh year⁻¹ and FPCM, energy_2: MWh year⁻¹ and herd (all livestock categories), energy_3: MWh year⁻¹ and lactating cows, energy_4: diesel tons year⁻¹ and FPCM, energy_5: diesel tons year⁻¹ and herd (all livestock categories), energy_6: diesel tons year⁻¹ and lactating cows.

Indicators described in the clusters are ratios specified as follows:

• (i)ha_crop_1: irrigated area and total arable area; ha_crop_2: dry matter of corn silage and total arable area; ha_crop_3: dry matter of corn silage and irrigated area; ha_crop_4: dry matter of non-corn silages and total arable area; ha_crop_5: dry matter of non-corn silages and irrigated area; ha_crop_6: dry matter of hay and total arable area; ha_crop_7: dry matter of hay and irrigated area.
• (ii)cattle_milk_1: total amount of FPCM and total arable area; cattle_milk_2: herd (all livestock categories) and total arable area; cattle_milk_3: lactating cows and total arable area; cattle_milk_4: dry cows and total arable area; cattle_milk_5: heifers and total arable area; cattle_milk_6: female calves and total arable area; cattle_milk_7: culled cows and total arable area.
• (iii)energy_1: MWh year⁻¹ and FPCM; energy_2: MWh year⁻¹ and herd (all livestock categories); energy_3: MWh year⁻¹ and lactating cows; energy_4: diesel tons year⁻¹ and FPCM; energy_5: diesel tons year⁻¹ and herd (all livestock categories); energy_6: diesel tons year⁻¹ and lactating cows.

Where: FPCM=fat protein corrected milk; MWh= mega-watts per hour; ha=hectare

In particular, the larger irrigated area of L is highlighted by ha_crop_1, in fact, S has an arable area almost equally divided between irrigated and non-irrigated. Cattle_milk_7 shows the higher culling rate of S, a notable theme for allocation.

Two significant critical points of L are shown by ha_crop_6 and cattle_milk_6. The inadequate forage efficiency is mainly observable in the hay yield per hectare, displayed by ha_crop_6. The large number of calves is indicated by cattle_milk_6. This is another example of low efficiency, despite the large share of this livestock category being intrinsically related to the ongoing growing size of the farm.

As expected, and according to literature, economic allocation improved the sustainability of farms, subtracting burdens from milk and parting them to the culled cows, the unique allocated by-product.

Interestingly, the culling rate provides indirect information on herd management, as shown by Romano et al. (2021b) in the only study carried out in dairy cattle reporting significant results with this respect. Their study quantified the culling rate using allocation. Bragaglio et al. (2022) in a similar study on buffalo farms, did not find statistical significance, probably due to the high economic value of buffalo milk and the low culling rate of Bubalus bubalis cows.

Precision agriculture techniques still have a circumscribed diffusion. Both farms involved in the study have a high level of automation, while the difference in size (large vs. small) is mainly related to their geographic location. The L farm is placed in the Veneto region. Here, husbandry is not affected by intense competition from other human activities, such as other farms, industry, urban centres, communication routes and infrastructures. The S farm is located in Lombardia, where the areas for agricultural activities are less available, given the high land value. In addition, Lombardia produces almost one-third-of the Italian milk, and is the second most densely populated region in Italy.

Results achieved with milk and hectare as DFU showed a lower intensity profile in the L farm and a higher production intensity in S. Although the L farm has a higher milk yield, other indicators (e.g., the number of calves) describe this farm as less performing. A large number of calves testifies that the farm does not yet have the desired profile and is increasing its size, whereas the S follows a more stable management.

Bava et al. (Bava et al., 2014) indicate that increased amounts of the main product (raw or normalized milk), improve several impact categories, however this study shows a different trend. Despite a lower milk yield per cow, the better efficiency of the S farms improves its sustainability.

According to the literature, the trend changes when the hectare is chosen as functional unit. In particular, a large farm size usually mitigates the environmental impacts of pasture-based and organic systems, often characterized by low milk yield.

4. Limitations of the study

The aim of this study was to make a first step towards the sustainability assessment of two farms having different sizes but the same highly technological profile. While the number of farms assesses is limited, their representativeness can be considered satisfactory due to the soundness of primary data collection, and conclusions from the present study are potentially translatable or applicable, with due adaptations, to larger scenarios.

The incomplete comparability between the large and the small farm was mainly arose from the different management. In the small farm, although automatic milking and feeding systems were present, the management does not fit well with precision farming practices and resembles instead that observed on more traditional farms. For example, the ensured income from the sale of cows satisfies the farmer, who does not appear to be worried by the high culling rate.

Lastly, the LCA method could involve socio-cultural themes (i.e., externalities), which are also helpful in assessing whether age and education can affect management and farm sustainability. The economic allocation could involve (i) subsidies received for the adoption of precision technologies, and (ii) ecosystem services (landscape conservation, wildlife protection).

However, this fell outside the aim of the present study, also because the assessment of socio-cultural externalities would require the involvement of a large number of farms. Given the adoption of automatic feeding systems and the complexities in data collection for this kind of LCA studies, future research in this field may be supported by machine learning techniques, helpful to obtain satisfactory data.

5. Conclusions

Although L and S have different characteristics, they are representative of the Po Valley farms. Primary data of the reference year (helpful to inventory analysis) may also be recordable at later times. The differences between the two farms are partly explained by the location in two different areas of the plain. Moreover, L is a “young” farm.

In addition to the farm size, efficiency and self-sufficiency, as items involved in dairy systems sustainability, a critical concern of this study is the culling rate. The replacement percentage in the L farm (30 %) is comparable with those reported by other authors for the Holstein-Friesian breed, in corn and silage-based farms of Northern Italy. The culling rate in S is quite high (38 %), suggesting a low number of lactations per cow (2–3 on average). Involuntary culling before the third lactation (usually due to mastitis, infertility, lameness) is the main cause of high culling rates and often determines lost incomes.

The high culling rate of the S farm is partially explained by involuntary replacement, and the presence of Holstein-Friesian bulls, in a highly specialized dairy farm, suggests hypofertility. Moreover, the perforated floor in farm S, probably less comfortable for lactating cows than the solid one in L, possibly resulting in increased lameness.

Another reason for the high replacement rate is the price at which culled cows are sold. Although it is the same on both farms, the owner of the S farm believed the price was satisfactory. Many factors, including genetic, affect fertility, and the livestock market cannot be controlled by the farmer. The longevity of cows could be extended by improving in the comfort of the barn, therefore decreasing the culling rate.

The L farm has an acceptable culling rate, but its environmental impacts can be improved. It would be helpful to re-evaluate its sustainability in the near future, once the increase in heard size will be achieved.

Although reasonable, the milk yield per cow in L can be increased, for example with genetic improvement. The critical aspect of L, however, is its feed efficiency. The arable area is large and the forage yield per hectare is certainly increasable. In addition, there are also margins to produce grains or flour to be used on farm, instead of purchasing barley and corn as ingredients of commercial concentrates. Corn is also purchased as a feedstuff (flour and flaked seeds). This self-production would improve the sustainability of L, since feedstuffs and concentrates, in addition to requiring industrial treatments, are also transported.

Funding

This research has been supported by the Italian Ministry of Agriculture, Food Sovereignty and Forests (MASAF) as part of the “AGRIDIGIT” project, “AGROFILIERE” sub-project (Decree n. 36503 of 20/12/2018).

Ethical statement

The animals involved in the study were only observed. These data refer only to production and performances, therefore no ethical authorization for animal use was needed.

CRediT authorship contribution statement

A. Bragaglio: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. E. Nannoni: Writing – review & editing, Writing – original draft, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. E. Romano: Supervision, Software, Resources, Methodology, Investigation, Formal analysis, Data curation. A. Lazzari: Supervision, Software, Resources. R. Roma: Visualization, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. C. Bisaglia: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources, Project administration, Funding acquisition. M. Cutini: Resources, Project administration, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A

Enteric emissions

As indicated by different sources (Bragaglio et al., 2023), the feeding regimen affects methane enteric emissions in ruminants. In the highly specialised dairy cattle systems, specific diets are administered to each livestock category: lactating cows, dry cows and heifers. Calves are fed with milk replacer powder, weaning meal, and growing amounts of forages. According to Baldini et al. (2018) a default value of 23 kg head⁻¹ year⁻¹ was attributed to methane from calves (females only, 0–6 months of age).

Methane emissions of cows and heifers have been calculated considering gross energy (GE) intake per head, as affected by ingredient percentage, following IPCC 10.21 equation (IPCC, 2019b). The equation below (Tier 2 method) allows to estimate methane enteric emissions:

$$EF=\left[GE*\left(\frac{Y_m}{100}\right)*\frac{365}{55.65}\right]$$ (A1)

(Eq. (A1)), where:

EF = emission factor (kg CH₄ head⁻¹ year⁻¹);

GE = gross energy intake (MJ head⁻¹ year⁻¹) given by different percentages of ingredients, expressed as dry matter. INRAE (INRAE, 2024) provides GE data for different diet ingredients;

Ym = methane conversion factor (percentage of GE in feed converted to methane). Ym in dairy cows is assumed to be 6.5 ± 1.0 %. According to Bragaglio et al. (2023), the high dairy performances suggest a Ym value of 5.5;

55.65 = energy content of methane (MJ kg CH₄⁻¹).

Methane emissions from manure management

L and S farms show different manure management. In both farms, a cubicle resting area hosts lactating cows In L cubicles are layered with straw, whereas in S a sand layer is provided. Dry cows of L farm are housed like the lactating ones, and the resting area of heifers is straw-bedded. Heifers and dry cows of S farm are also kept in a straw-based resting area.

Manure handling is affected by management practices, and the larger use of straw leads L farm to handle solid manure and liquid slurry separately, while in the S farm manure is managed as liquid slurry only. Methane (CH₄) and nitrous oxide (N₂O) are the main GHGs emissions from manure.

According to other studies (Romano et al., 2021a, 2021b), we followed the IPCC guidelines (Tier 2) to assess the CH₄ emissions. Methane emissions could be divided into barn and housing emissions. Baldini et al. (2018) measured lactating cows' barn emissions of three farms in Po Valley and found very low values (1.11–5.84 CH₄ kg head⁻¹ year⁻¹). On the basis of these results, as done by Bava et al. (2014), and by Bragaglio et al. (2023), this study considered storage emissions only.

As indicated by IPCC, volatile solids (VS) are estimated using the following formula:

$$VS=\left[GE*\left(1-\frac{DE\%}{100}\right)+\left(UE*GE\right)\right]*\left(\frac{1-ASH}{18.45}\right)$$ (A2)

(Eq. (A2)), where:

VS = volatile solid excretion per day on a dry-organic matter basis, kg VS day⁻¹;

GE = gross energy intake, MJ day⁻¹;

DE% = digestibility of the feed (expressed as a percentage). The different amount of feed was also considered for the assessment of DE. INRAE (2024) data were used for digestibility;

(UE*GE) = urinary energy expressed as a fraction of GE. Typically, 0.04 GE can be considered as the urinary energy excretion by most ruminants. This value was used in the current study;

ASH = the ash content of manure, expressed as a fraction of the dry matter feed intake, suggested as 0.08 for cattle by IPCC (IPCC, 2019b) and obtained also in other studies (Romano et al., 2021b).

18.45 = conversion factor for dietary GE per kg of dry matter (MJ kg⁻¹).

Afterwards, VS values are used to calculate CH₄ manure emissions, as follows:

$$EF=\left(V{S}_T*365\right)*\left[B_{0T}*0.67*\sum _{S,k}\frac{MC{F}_{S,k}}{100}*AWM{S}_{T,S,k}\right]$$ (A3)

(Eq. (A3)), where:

EF = emission factor of the livestock category T (kg CH₄ head⁻¹ year⁻¹);

VS_T = daily volatile solid excreted by the livestock category T, kg dry matter animal⁻¹ day⁻¹. VS was calculated according to Eq. 3 (IPCC, 2019b);

365 = basis for calculating annual VS production, days year⁻¹;

B_0T = maximum methane-producing capacity from manure produced by the livestock category T, m³ CH₄ kg⁻¹ of VS excreted, the estimated value is 0.24 in Western Europe;

0.67 = conversion factor of m³ CH₄ to kilograms CH₄;

MCF_S,k = methane conversion factors for the manure management system S in the climate region k, percentage. 39 % is the mean value between moist and dry warm temperate climate zones and for 180 days of storage of liquid slurry (in L and S farm). 4 % is the default value in temperate climate zones for 90 days of storage of solid manure (L farm only);

AWMS _T,S,k = fraction of manure from the livestock category T handled with the manure management system S in climate region k. The value used is 43 % for slurry/liquid storage for dairy cattle in Western Europe (in L and S farms) and 29 % for solid manure storage for dairy cattle in Western Europe (L farm only).

Ratio between incomes

This table highlights the negligible economic value of male calves, even though their overestimation in terms of heads and income.

Table A1.

L and S are the Large and the Small farm, respectively. We hypothesised a unit price of 100 € per male calf. Moreover, we hypothesised 85 and 10 male calves born and sold, for L and S, respectively.

	L	S
Total income from milk, €	1570,960	678,600
Total income from culled cows, €	69,030	46,020
Total income from male calves, €	8500	1500
Total farm income, €	1,648,490	726,120
% milk income	95.3	93.5
% culled cows income	4.2	6.3
% male calves income	0.5	0.2