Figure 2 shows the values relating to the environmental footprint per kg FPCM of the farms in the three scenarios previously described. Considering the GWP without allocation and in the BASE scenario, the emission of CO₂eq per kg of FPCM was 1.43 ± 0.106 kg, while the emission of CO₂eq per m² of UAL was 1.64 ± 0.130 kg (data not shown). These values are very similar to those reported by Thomassen et al. [50] who, considering 10 Dutch farms that had an average consistency of 81 lactating cows with a production of about 8 t per animal, obtained an average value of 1.40 kg of CO₂eq/kg FPCM. These emissions values were recently confirmed by the study of Lovarelli et al. [18], where the average value of 1.46 kg CO₂eq/kg FPCM was calculated for farms whose milk was used for the production of Parmigiano Reggiano cheese. Interestingly, the farms analyzed in Lovarelli et al. [18] had a milk yield per cow of 25.0 kg FPCM/d and a feed efficiency of 1.15 kg FPCM/kg DMI, similar to that of the farms considered in the present study. In the study of Guerci et al. [51], five Danish, two German and five Italian farms, with very variable conditions in terms of herd size, from 36 to 350 cows bred, in terms of production potential, from 6.2 to 10.9 t of milk per animal, and in terms of agricultural area, from 21 to 225 ha, were considered. In this study, the emissions in terms of CO₂eq per kg of energy-adjusted milk (ECM) ranged widely from 0.55 to 1.91. However, if only the Italian farms were taken into account, the emissions ranged between 1.11 and 1.91 kg CO₂eq/kg FPCM, with the highest value being observed in the farm and with the lowest production level per cow and the lowest feed efficiency. Salvador et al. [22], Guerci et al. [52] and Kiefer et al. [21] showed that the highest emission levels happened in more extensive production systems; conversely, this result was not confirmed by Chobtang et al. [53] or by Morais et al. [54]. However, usually, the greater the productive level of the farm, or its productive efficiency, the lower the environmental footprint per kg of milk [55]. Moreover, Price and Bell [56] observed that the improvement of the genetic selection practices of dairy cows in Australia in the last 10 years led to a reduction in CO₂eq emissions of about 1%, and Wattiaux et al. [57] explained that the factor most associated with farms’ emissions was the management practices. It can be argued that the emissions values obtained in the present study are in line with those reported in most studies considering farms with similar production efficiency, but irrespective of the breed reared. Different considerations could be drawn if the m² of UAL was considered as the functional unit instead of milk. In fact, in comparison to the extensive mountain dairy farms considered by Salvador et al. [22], the present study showed a lower environmental footprint per kg of FPCM—1.43 vs. 1.46 kg CO₂eq—but higher per m² of UAL—1.64 vs. 0.69 kg CO₂eq. Bava et al. [58] explained that the environmental footprint per ha can be positively related with the intensification of the productive process. The carbon footprint per kg of FPCM was 1.12 ± 0.085 and 1.23 ± 0.096 kg CO₂eq when using a physical and economic allocation, respectively (Figure 2). These values were lower by 22% and 14% than that obtained without allocation. Kiefer et al. [21], adopting the physical allocation procedure, showed a reduction in the emissions allocated to milk by 19% in farms that mainly reared Holstein animals and 24% in farms that mainly reared animals of the dual-purpose Simmental strain. These data show how the method of physical allocation, which also considers the co-products to milk production, is particularly relevant in farms rearing dual-purpose breeds. The average on-farm contribution to GWP was 68.7% (data not shown) while the contributions of different emissions sources are shown in Table 2. Enteric emissions and manure storage together represented, on average, 56% of total GHG emissions. From this point of view, the most important gas is CH₄; therefore, maximizing the synthesis efficiency of ruminal microbial fermentations can be considered a possible strategy to reduce the emission of gases that contribute to GWP. The other main contribution to GWP is represented by off-farm emissions related to the production of feed purchased by farms, constituting 20% of the total emissions.

As reported in Figure 2, considering AC without allocation and in the BASE scenario, the emission of SO₂eq/kg FPCM was 18.99 ± 1.408 g. This value falls within the range of emissions calculated by Guerci et al. [51]—7.4–25.6 g SO₂eq/kg ECM—and is very similar to the value obtained by Cederberg and Mattsson [59]—18.0 g SO₂eq/kg ECM—in intensive dairy farms. The acidification footprint per m² of UAL was 21.69 ± 1.764 g SO₂eq (data not shown). However, even in this case, if we compare the results of the present study with those obtained by Salvador et al. [22], who considered dairy cows of dual-purpose breeds but in more extensive husbandry systems, the AC values are lower if referring to the kg of FPCM—19.0 vs. 27.2 g SO₂eq—but higher, if referring to the m² of SAU—21.7 vs. 12.4 g SO₂eq. The acidification footprint per kg of FPCM was 14.84 ± 1.091 and 16.22 ± 1.246 g SO₂eq using physical and economic allocation, respectively (Figure 2). These values were lower by 22% and 15% than that obtained without allocation. On average, 90% of the emissions related to acidification potential derive from on-farm activities (data not shown). It is clear that, for this category of emissions, the farm’s own activities have a crucial role. As reported in Table 2, 75% of the emissions are derived from the management of manure, which, together with crop production, represent about 90% of the total emissions. This percentage falls within the range found by Guerci et al. [51]—83–96%. These high values can be explained when considering that the main thing responsible for acidification in dairy livestock systems is the volatilization of ammonia that occurs during the storage, treatment and distribution of manure. In fact, according to the study by Cederberg and Mattsson [59], almost 90% of the acidification potential would be due to ammonia losses.

Considering EUP without allocation and in the BASE scenario, the emission of PO₄³⁻eq/kg FPCM was 6.36 ± 0.486 g (Figure 2). Guerci [51] reported EUP values between 4.6 and 11.1 g PO₄³⁻eq/kg FPCM. Bava et al. [58], considering dairy farms in the Po Valley with an average area of 41 ha and an average consistency of 90 cows, and Guerci et al. [60] obtained values higher than those of the present study—7.3 and 9.0 g PO₄³⁻eq/kg FPCM, respectively. The eutrophication footprint per m² of UAL was 7.10 ± 0.394 g PO₄³⁻eq (data not shown). Guerci et al. [60] calculated that the incidence of on-farm emissions compared to the total was 74–78%, a value only slightly lower than the 82% obtained by the present study. It is interesting to note that, in the farms considered, the EUP is mainly due to manure management (45%) and to crop production (37%) (Table 2), and the gas mainly responsible is NH₃, which develops during the manure management and nitrogen losses on land, in particular as N leaching. The incidence of emissions linked to the production of the feed purchased is 13%, which also can be considered relevant. Therefore, the improvement of the efficiency of the use of concentrates in feeding and the equilibrate use of organic and mineral fertilizers in soils can be considered valid strategies to reduce the environmental footprint, in terms of EUP, of the farms.

As previously reported, the environmental footprint of the farms was mainly due to CH₄ and NH₃ emissions and N leaching in the soil. These emissions were mainly assessed with a Tier 2 method, which leads to some uncertainty in the results, as it only includes country-specific emission factors. Indeed, to improve the reliability of the results, a Tier 3 method should be used; however, in the case of the present study, it was not possible because of the absence of emission factors disaggregated at a sub-national level. The development of appropriate field and laboratory methods could contribute to overcoming this problem, e.g., the use of rumen continuous fermenters [61] could allow one to obtain a reliable assessment of the rumen methane emissions of each single farm.

Figure 3a shows the PCA of farm characteristics and GWP expressed as CO₂eq/kg FPCM calculated with the no allocation method. The first and the second principal component explained 28.1% and 26.6% of the total variance, respectively. All the impact categories considered—GWP, AC and EUP—were characterized by positive loadings on the first component; conversely, on the same component, milk production per cow and production intensity were characterized by negative loadings. This means that, as previously discussed, the environmental emissions of the farms could be reduced by increasing milk production per cow and production intensity. Interestingly, the emissions, especially in terms of GWP and AC, were closely associated with the culling rate, highlighting the importance of improving the efficiency, not only in milk production performances, but also in the reproductive performances of the animals, for reducing the environmental footprint of the farms. The results change using UAL instead of FPCM as the functional unit (Figure 3b), as also explained by Bava et al. [58]. Indeed, in this case, GWP and AC were positively associated with production intensity. Moreover, the first principal component, which explained 34.3% of the total variance, highlighted a negative relationship between the impact categories GWP and AC and the feed self-sufficiency of the farms, while the second principal component, which explained 23.4% of the total variance, highlighted a negative relationship between EUP and feed self-sufficiency and a highly positive relationship between EUP and the percentage of the concentrates in the diets of animals.