Introduction

Globally, beef and milk production from cattle contributes 41% and 29% of greenhouse gas (GHG) emissions from the livestock sector, respectively (Gerber et al. 2013). In Latin America, the main sources of GHG emissions are from land use, land use change, and forestry (35%) and agriculture (23%) (Wellenstein and Hickey 2021). In this context, one sustainable production alternative is the implementation and expansion of silvopastoral systems (SPS). GHG emissions can be reduced in SPS versus conventional grazing systems; hence, lower environmental burdens per product are generated (Rivera et al. 2016; Murgueitio et al. 2012).

Silvopastoral systems contribute to reduce deforestation, furnish a diversified source of income to farmers, provide ecosystem services (water, carbon sequestration, nutrient recycling, biodiversity), increase welfare and animal production, as well as quality of pastures, and contribute to mitigate GHG emissions (Alonso 2011; Fernández 2008; Alegre et al. 2012; Fluker et al. 2016; Montagnini et al. 2013; Pérez et al. 2005; Pezo et al. 2019; Pizarro et al. 2019).

Livestock in Peru is characterized by small-scale productions (< 10 head of cattle), which represent 85.9% of the total nationally. Additionally, 39.4% of the national milk production comes from these cattle herds (MINAGRI 2017). In the same way, the Peruvian Amazon is characterized by dairy cattle productions under SPS as “living fences” and “scattered trees in pastures” (Pizarro et al. 2019). However, the environmental impact from this type of system is unknown.

Carbon footprint (CF) is an environmental impact indicator, which estimates direct and indirect GHG emissions generated and emitted into the atmosphere during the life cycle of a product along the production chain (Vistoso et al. 2015). It is normally expressed in kg of CO2 equivalent (CO2e) per kg of product (IPCC 2019a). Some methodologies used for its evaluation are as follows: ISO 14040, 14044, 14067, PAS 2050, GHG Protocol, IDF Common Carbon Footprint methodology (IDF 2015). It is crucial to know the CF value for the implementation of mitigation strategies (Mainar 2019).

Therefore, the objective of this study was to estimate the CF of milk production (kg CO2e/kg of FPCM) on silvopastoral dairy systems in the Peruvian Amazon.

Materials and methods

Study location

The study was carried out on eight farms in Juan Guerra, San Martín, Peru. Juan Guerra is located at 230 m above sea level, latitude 6° 35′ West, longitude 76° 19′ South. San Martin region is characterized by farms dedicated to milk production (32,697 t/year) and beef (5443 t/year) (MIDAGRI 2020). The Juan Guerra district has a semi-dry and warm climate, a tropical dry forest ecosystem, an average temperature of 26.2 °C, and an annual rainfall of 1213 mm (IDERSAM 2016).

Carbon footprint analysis

The methodology used was described by the methodological guide of the International Dairy Federation (IDF) 479/2015, which uses the methodological structure of ISO 14040, 14044, 14067 standards, PAS 2050, and GHG Protocol (IDF 2015). This study used the life cycle analysis (LCA) methodology from “cradle to farm gate” to estimate and describe the carbon footprint of milk production for each of the 8 farms evaluated. The LCA steps followed for its calculation are detailed below.

Goal and scope

System limits

Figure 1 shows the incomes and outcomes from the system. GHG emissions such as methane, nitrous oxide, and carbon dioxide produced from “cradle to farm gate” were estimated.

Fig. 1
figure 1

Limits of milk production system for each dairy farm in a “cradle to farm gate” approach in the San Martín region

Full size image

Functional unit

A total of 1 kg FPCM and 1 kg of beef were used as functional units. For these estimates, equations provided by the IDF (2015) were used.

$$\mathrm{FPCM }\left(\mathrm{kg}/\mathrm{year}\right)=\mathrm{production }\left(\mathrm{kg}/\mathrm{year}\right)\times \left[0.1226\times \mathrm{milk\:fat\:\%}+0.0776\times \mathrm{milk\:protein\:\%}+0.2534\right]$$

Allocation

The economic allocation was carried out following the equation given by Thoma et al. (2013), considering 4% fat and 3.5% protein. The mass allocation was based on FPCM (kg) and beef (kg) from surplus animals produced per year.

Data collection and feed samples

The data collected came from surveys conducted on eight farms located in the Juan Guerra district. The province and district were selected following the Qualitative Factors Assessment Methodology. It was a non-probabilistic selection. Also, the principal criteria to include farms was the farmer’s availability to work with the researchers, with at least 10 to 20% trees in the grazing area, milk production as the leading activity (60% milking cows), similar edaphic and environmental characteristics, and accessibility to make easer the sampling.

The survey consisted of questions at the farm level: landowner name, farm name, area (for crops, pastures (native and/or cultivated), forage bank, primary forest, secondary forest, and infrastructure), type of silvopastoral arrangements, and predominant tree species. At the livestock level: herd composition (number of animals per category), predominant genotype (%), milk production (kg/cow/day), feed inputs, fertilizers, machinery, fuel, energy, manure management, and hours of grazing.

Subsequently, pasture and feed samples provided to animals were taken from each farm for each period of the year (rainy and dry season). The grass samples consisted mainly of Brachiaria brizantha, using the “hand sampling” methodology (Flores 1983; Austin et al. 1983).

Greenhouse gas estimation

The CF estimation was performed using the refined equations given by the Intergovernmental Panel on Climate Change (IPCC 2019b, c). The global warming potentials (GWP) used were the following: 28 for methane, 265 for nitrous oxide, and 1 for carbon dioxide for a time horizon of 100 years (IPCC 2014). Microsoft Excel 2019 was used for the calculations. Table 1 details emission sources and equations used for their estimation.

Table 1 Emission sources and equations used to calculate carbon footprint from each farm evaluated in the San Martín region
Full size table

Data analysis

This study used a descriptive statistical analysis such as medium and standard error.

Results

Characterization of the farms

The main characteristics of the eight farms (F1–F8) are in Table 2. The average area was 51.5 ± 26 ha, and 30.1 ha of these was used for grazing. The average tree cover was 22.5%. The total size of dairy cows was 32 ± 6.3, of which 19 ± 6.2 were in the lactating stage. An average number of replacements and remnant animals (calves and bulls) were 30 and 13, respectively. Each farm owned one male breeder since natural breeding predominates (Table 2).

Table 2 Overall characteristics from each farm evaluated in the San Martín region
Full size table

The average milk production was 9.7 ± 0.82 L of milk/cow/day (Table 2). Gyr x Holstein crosses were the predominant breed on farms evaluated. In regard to the grazing area, the main cultivated pasture was Brachiaria brizantha and living fences with Guazuma ulmifolia Lam leading was the main silvopastoral system arrangement. In addition, external inputs such as feed additives or cereals were used in negligible quantity. Finally, it should be highlighted that none of the farms used synthetic fertilizers into the pasture.

Greenhouse gas emissions

The major source of emissions was enteric CH4 (80%), followed by N2O from manure handling (10%). CH4 emissions from milking cows were 312 ± 32 g CH4/cow/day. Meanwhile, CO2 emissions from fuel only represented 2% of total emissions. Only farms 2, 4, and 5 showed CO2 emissions from the use of electrical energy because the remaining used solar panels (Table 3).

Table 3 Emission sources in kg of CO2e for each farm evaluated in the San Martín region
Full size table

Greenhouse gas allocation

When a mass assignment was used, the average GHG emissions were 2.26 kg CO2e/kg FPCM in a range of 1.76 to 3.09 kg CO2e/kg FPCM (Table 3).

Regarding economic allocation, the allocation factor (AF) for milk was higher than the AF for meat (0.75 vs 0.25). The results showed an average value of 1.68 ± 0.41 kg CO2e kg/FPCM; enteric fermentation was the most important source with 1.34 ± 0.34 kg CO2e/kg FPCM, followed by manure management (0.19 ± 0.04 kg CO2e/kg FPCM), land use (0.11 ± 0.03 kg CO2e kg/FPCM), and energy/transport (0.04 ± 0.03 kg CO2e/kg FPCM) (Table 4). Numerical differences were found between farmers, obtaining lower CF values (1.12 vs 2.38 kg CO2e/kg FPCM) on farms with improved food quality (high digestibility and protein in the diet), larger production levels, and a higher percentage of lactating animals compared to dry cows.

Table 4 Greenhouse gas emissions per unit of product (FPCM and meat) according to emission sources for each farm evaluated in the San Martín region
Full size table

Discussion

Greenhouse gas emissions

The proportion of enteric CH4 emissions was similar to Rivera et al. (2016), who reported that CH4 represented up to 84% of the GHG for an intensive SPS (SPSi) of milk production in Colombia. However, it is greater than reported by Morais et al. (2018) in pastoral dairy farms in Portugal, with an emission range of 33 to 52%. Similarly, it is larger than emitted by pastoral dairy systems in New Zealand (62%) (Flysjö et al. 2011). In the same way, comparing with non-pastoral systems, the results of this study were higher to that reported by Flysjö et al. (2011) in Sweden for semi-stable systems, where the enteric CH4 emission represented 46%.

Regarding N2O emissions, this study estimated higher emissions (16%) than reported by Rivera et al. (2016) in Colombia, who estimated that 12% of total emissions inside farms with SPSi were for N2O (related to chemical and organic fertilizers, N excretions via manure and urine). Nonetheless, they are lower than those found in a Brazil study, where manure emissions and excrement deposited into the field represented between 20 and 33% (including the use of fertilizers, lime, and pesticides) (Cerri et al. 2015).

Finally, regarding CO2 emissions from energy use and transportation, this study obtained lower values (2%) than Rivera et al. (2016), who found an emission of 4% for CO2 in farms with SPSi, although the results observed on this study are inside the range observed by Cerri et al. (2015), who calculated those emissions from agricultural inputs, fossil fuels, and electricity that ranged from 1 to 11%.

There are many factors to determine these differences in the proportion of emissions. Firstly, “off-farm” emissions; therefore, when a study considers these emissions, N2O and CO2 quantity increases due to external inputs use; consequently, enteric methane emissions decrease.

Additionally, other factors are the annual milk yield and the number of cows (Knapp et al. 2014). This study found that on average 23.7% of total animals were cows in production, a nether percentage than reported by Morais et al. (2018) for pastoral dairy farms in Portugal, which had a greater average (49.2%). Hence, if the total environmental burden from CF is divided exclusively on milking cow’s category instead of total animals, CF will be lower.

Furthermore, both animal genetics and feed digestibility are crucial factors in CH4 emission (NRC 2001; Lassen and Løvendahl 2016). Other factors such as region (6.38 kg/animal/day; FAO 2018), heat stress (Polsky and von Keyserlingk 2017), the type of tropical pasture, and management practices (Pezo 2017), feed consumption rate, type of carbohydrate, quality and forage species, physical processing, forage conservation, and feeding frequency (Knapp et al. 2014) are important.

Concerning CO2 emissions, Cool Effect (2021) indicates that the carbon footprint of the solar panel is about 20 times less than the carbon output of coal-fired electricity sources. Thereby, the use of renewable energy, as well as reduced use of fuel compared to intensive systems, supports the small emissions in the Peruvian tropical zone.

Greenhouse gas allocation

The CF found in this study (1.68 kg CO2e/kg FPCM) is lower than reported by Rivera et al. (2016) for intensive silvopastoral systems (SSPi) and conventional systems (2.05 and 2.35 kg CO2e kg/FPCM, respectively) on dairy farms in Colombia. Therefore, it is worth highlighting the viability of SPS to avoid importing balanced feed in large quantities, favoring self-sufficiency with protein banks (e.g., Leucaena leucocephala) (Pezo et al. 2019), helping to reduce CF. Likewise, the results of this work were within the range reported by Rivera et al. (2014) in Colombia to pastoral and intensive systems (1.61 and 1.76 kg CO2e/kg FPCM, respectively).

Compared to pastoral farms, the average emission estimated in this study is greater than those reported by Del Prado et al. (2013) in the UK on farms with 195 days of grazing and with extended grazing (1.07 and 1.77 kg CO2e/kg FPCM, respectively). In the same way, the results of this study were greater to those reported by Flysjö et al. (2011) for pastoral systems in New Zealand (0.6–1.52 kg CO2e/kg FPCM), and by Laca et al. (2019) for Spain in pasture-based systems (0.69 kg CO2e/kg FPCM).

However, dairy production systems in Eastern and Western European countries (19.6 kg/cow/day) or New Zealand (15.23 L/cow/day) have milk production levels higher than this study. Thereby, CF decreases while productivity per animal increase, and other factors (Salas 2020; FAO and GDP 2018).

Compared to other Latin American studies, the average CF in this study was higher to reported by Wattiaux et al. (2016) in Costa Rica on pastoral dairy farms, where a range of partial CF (methane and nitrous oxide) from 0.38 to 1.02 kg CO2e/kg FPCM were found. Besides, the results of this work were larger than those reported by Lizarralde et al. (2014) in Uruguay on grazing dairy farms (0.99 ± 0.10 kg CO2e/kg FPCM).

Although, it should be emphasized that both milking cows’ numbers as average milk production were higher in these studies (Costa Rica and Uruguay) compared to this work. It highlights the importance of intensification of these systems.

Nevertheless, the CF calculated in this work (1.68 ± 0.41 kg CO2e kg/FPCM) is lower than that reported by Gaitán et al. (2016) in Nicaragua on small and medium pastoral farms (3.1 and 2.4 kg CO2e kg/FPCM, respectively). Since Nicaragua study evaluated emissions “on” and “off-farm”; while this research only carried out emissions “on-farm.” Moreover, it is minor then the CF on dairy farms in South Asia, Sub-Saharan Africa, West Asia, and North Africa, where emission intensities were larger (between 4.1 and 6.7 kg of CO2e kg/FPCM) (FAO and GDP 2018).

Compared to intensive dairy farms, the results of this study were larger than those reported by Del Prado et al. (2013) in the UK on farms in confinement (1.14 kg CO2e kg/FPCM) and by Vergé et al. (2013) in Canada on intensive dairy production systems (0.93–1.12 kg CO2e kg/FPCM). This study agrees with Wattiaux et al. (2016), who indicate that methane emissions were mainly reduced with the high up consumption of feed with respect to the amount of pasture provided to the animal. Although the GHG emissions found in this research were smaller than those estimated by Mazzetto et al. (2020), in Costa Rica on intensive specialized dairy farms in the highlands (3.86 kg CO2e kg/FPCM).

In Latin American (LA) and subtropical regions, the results of this study were higher to those reported by Ribeiro-Filho et al. (2020), who calculated emission of 0.88 to 1.04 kg CO2e/kg ECM on pastoral dairy systems in subtropical regions. Similarly, Zhu et al. (2016) reported lower values for LA (1.45 kg CO2e/kg FPCM). Differences between studies can be partly explained by various assumptions (e.g., emission factors, allocation of co-products, estimation of methane emissions, and CO2-C sequestered), in addition, by the herd productivity and manure management (Ribeiro-Filho et al. 2020).

Nationally, there are few studies carried out on GHG emissions. For example, Bartl et al. (2011) estimated a CF of 1.74 and 5.42 kg CO2e/kg ECM on farms in the coast and highlands of Peru, respectively. In the same way, Alvarado-Bolovich et al. (2021) estimated enteric methane production by cows, during dry and rainy seasons using IPCC Tier 3 in the highlands.

In conclusion, farms with improved livestock management (higher percentage of cows in production, and higher production levels) and better feed quality had a lower CF. Moreover, it was found that the principal source of emission was enteric fermentation. Consequently, dairy farms in the Peruvian Amazon region could reduce their emissions by improving their current management and feeding practices. Additionally, the CF can be decreased if sustainable management practices are incorporated such as solar panels use, and small use of chemical fertilizers. However, more studies are also needed that consider estimating carbon sequestration from trees, soil, and crops, and CF from off-farm feed to determine the true carbon balance of these systems.