Agriculture, Ecosystems and Environment 221 (2016) 50–59

Contents lists available at ScienceDirect

Agriculture, Ecosystems and Environment

journal homepage: www.elsevier.com/locate/agee

A diachronic study of greenhouse gas emissions of French dairy farms

according to adaptation pathways
G. Martina,* , M. Willaumea,b
a
INRA, UMR 1248 AGIR, F-31326 Castanet Tolosan, France
b
Université de Toulouse, INP, ENSAT, UMR 1248 AGIR, F-31326 Castanet Tolosan, France

A R T I C L E I N F O A B S T R A C T

Article history: Mitigation of greenhouse gas (GHG) emissions of dairy farms may conflict with farmers’ needs to adapt to
Received 4 August 2015 complex, dynamic and interrelated changes in production. We performed a diachronic study that
Received in revised form 10 December 2015 considers change in GHG emissions at the farm level following implementation of adaptations by dairy
Accepted 16 January 2016
farmers. We hypothesized (i) that farm adaptations follow three pathways discussed in the scientific
Available online xxx
literature: specialization and intensification, eco-efficient intensification, agroecological transition, and
(ii) that these three adaptation pathways drive the evolution of GHG emissions of dairy farms,
Keywords:
particularly by determining farms’ reliance on purchased nitrogen (N) fertilizers and N-rich concentrates.
Dairy farming
Mitigation
We identified 10 dairy farms in the Aveyron department, France, that capture the diversity of local
Adaptation farming systems on a gradient of intensified land use, dairy herd, or both. We considered two situations
Farm model for each farm: (i) the current situation and (ii) a former situation, at most 10 years before, suggested by
Eco-efficiency the farmer during the survey and different from the current situation. We estimated product-related (per
Agroecology kg of milk) and area-related (per ha) GHG emissions for each farm with a farm model that includes a GHG
emission calculator consistent with the International Panel on Climate Change methodology. Two farms
followed the trend of specialization and intensification by increasing their stocking rate and maintaining
or increasing their milk production per unit of area and livestock while simplifying and standardizing
farm management, such as feeding cows with larger amounts of maize silage and N-rich concentrates.
Three farms followed a trend of eco-efficient intensification by increasing their milk production per unit
of area and livestock while maintaining (or slightly changing) their consumption of N fertilizers and N-
rich concentrates. Three farms adopted agroecological practices by maintaining or decreasing their
stocking rate and farm production per unit of area and livestock and overall increasing their agricultural
diversity and substituting inputs (e.g., nitrogen fertilizers) with ecosystem services (e.g., nitrogen fixation
by legumes). Unlike our first hypothesis, there was a fourth trend, i.e., status quo, operated by two farms.
Results showed that agroecological transition reduced GHG emissions the most (by 2% to 40% per kg of
milk and by 18% to 29% per ha). The dairy farms following this transition had the lowest mean
product-related (0.89 kg CO2 eq./kg milk) and area-related (5306 kg CO2 eq./ha) GHG emissions. Eco-
efficient intensification reduced mean product-related GHG emissions (by 10%, to 0.93 kg CO2 eq./kg
milk), but increased mean area-related GHG emissions (by 18%, to 6239 kg CO2 eq./ha). In contrast,
specialization and intensification increased mean GHG emissions (by 6% per kg of milk, to 1.20 kg CO2 eq./
kg milk, and by 28% per ha, to 8741 kg CO2 eq./ha), making it appear unsuitable as a pathway towards
mitigating climate change at the farm level.
ã 2016 Elsevier B.V. All rights reserved.

1. Introduction et al., 2012) through enteric fermentation from animals, land-use

change, manure management, fertilizer use, feed processing and
Dairy farming is considered responsible for 3% of worldwide transport, etc. Mitigating these emissions may conflict with
greenhouse gas (GHG) emissions (Gerber et al., 2013; Hagemann farmers’ needs to adapt to complex, dynamic and interrelated
changes in production related to climate change, scarcity of natural
resources, rising energy costs, volatile input and output prices,
increasing food demand, agricultural policy changes, rising labour
* Corresponding author. Fax: +33 5 61 73 55 37.
productivity, etc. Consequently, despite the widely acknowledged
E-mail address: [email protected] (G. Martin).

http://dx.doi.org/10.1016/j.agee.2016.01.027
0167-8809/ã 2016 Elsevier B.V. All rights reserved.
 G. Martin, M. Willaume / Agriculture, Ecosystems and Environment 221 (2016) 50–59 51

urgency among the scientific and policy communities, mitigating levels and beyond (e.g., Gerber et al. (2013) for mitigation; Reidsma
GHG emissions is seldom among farmers’ main priorities when et al. (2010) for adaptation).
adapting their agricultural systems (Fleming and Vanclay, 2009). Two studies conducted in the Netherlands assessed impacts on
In the scientific field, despite calls for synergies (Rosenzweig GHG emissions of partly substituting grass and grass silage with
and Tubiello, 2007), mitigation and adaptation studies applied maize silage and grain (Vellinga and Hoving, 2010; van Middelaar
to agricultural systems are almost two distinct fields of research et al., 2013). They showed that emissions linked to land-use
(Füssel, 2007) leading to scarce examples of coupled studies change and losses of soil carbon can offset a decrease in methane
(e.g., Del Prado et al., 2013a). Mitigation studies estimate GHG emissions due to dietary changes. Tradeoffs between adaptation
emissions of agricultural systems while adaptation studies and mitigation are complex and location-specific. Moreover,
analyze their sustainability (e.g., maintenance of production, while the study by van Middelaar et al., 2013 was based on virtual
income variation) when confronted with unfavorable events dairy farms, the one by Vellinga and Hoving (2010) was based on a
such as drought and extreme temperatures. There is scope for real farm for which explorative adaptation scenarios were built
and a need to combine mitigation and adaptation analysis. and assessed. Proceeding this way, contextual changes that might
Adaptation options can have either synergetic or antagonistic influence adaptation decisions are necessarily limited and
effects on GHG emissions. For instance, up to a certain simplified. The need exists to further understand tradeoffs
threshold, intensification of grassland productivity by increasing between adaptation and mitigation in a real (and consequently
production or stocking density leads to positive effects through multifactorial—weather events, volatility of prices, family issues,
soil carbon sequestration; after this threshold, however, it leads etc.—and not fully predictable) agricultural adaptation context.
to negative effects due to excess of nitrogen (N) (Soussana and The objective of this study was to conduct a diachronic study of
Lemaire, 2014) the evolution of GHG emissions at the farm level along with
At the field level, estimates of GHG emissions due to soil-, crop- implementation of adaptations by farmers on dairy farms in
or grassland-management practices (e.g., tillage, fertilization) Aveyron, France. We hypothesized (i) that farm adaptations follow
(Datta et al., 2013; Louro et al., 2013) would benefit greatly from three pathways discussed in the scientific literature, and (ii) that
diachronic studies addressing the overall sustainability of these these three adaptation pathways drive the evolution of GHG
practices. Similarly, estimates of GHG emissions could enrich emissions of dairy farms. The three considered agricultural
sustainability assessment of an adaptation, such as adjusting adaptation pathways are distinguished according to evolution of
sowing dates (Cammarano et al., 2012). This combination is three types of indicators: farm reliance on input (external physical
valuable at other, higher organizational levels, i.e., animal ressources brought into farms to achieve output, i.e. milk and meat
(Grainger and Beauchemin (2011) for mitigation; Mejía et al. production) use assessed on a nitrogen basis, farm stocking rate
(2010) for adaptation), farm (e.g., Del Prado et al. (2013b) for and farm production per unit of area and livestock. These three
mitigation; Martin and Magne (2015) for adaptation), and regional adaptation pathways are:

Table 1
Characteristics of the 10 farms surveyed for the two scenarios considered, representing current and former situations. Types of animal housing are straw-bedded pen (SBP),
straw-bedded cubicles (SBC), straw bedding and scraped alley (SBSA), scraped cubicles (SC) and slatted cubicle (SlC). UAA = usable agricultural area. AU = animal units; a
Table detailing the correspondence between animal types and animal units is provided in Appendix A.

Farm Situation Usable % grassland Herd size (AU) Cow number Replacement Type of housing
agricultural rate (%)
for cows for heifers
area (ha)

Farm 1 Former 70 50 77 45 44 SBSA SBP

Current 60 47 89 55 36 SBSA SBP
% change 14% 6% +16% +22% 18%
Farm 2 Former 28 73 44 30 27 SBSA SBSA
Current 42 55 70 46 26 SC SC
% change +50% 25% +59% +53% 4%
Farm 3 Former 74 65 55 35 29 SC SBP
Current 74 65 78 45 44 SBC SBP
% change 0% 0% +42% +29% +52%
Farm 4 Former 190 79 109 45 56 SBSA SBP
Current 200 80 152 70 50 SlC SBP
% change +5% +1% +39% +56% 11%
Farm 5 Former 34 52 51 35 23 SBSA SBP
Current 34 52 60 40 20 SBSA SBP
% change 0% 0% +18% +14% 13%
Farm 6 Former 45 66 58 40 28 SBP SBP
Current 64 83 88 57 26 SBC SBP
% change +42% +26% +52% +43% 7%
Farm 7 Former 92 70 124 60 33 SC SBSA
Current 110 75 131 65 31 SC SBSA
% change +20% +7% +6% +8% 6%
Farm 8 Former 60 71 60 35 40 SBP SBP
Current 60 86 54 35 26 SBP SBP
% change 0% +21% 10% +0% 35%
Farm 9 Former 46 78 58 40 25 SBC SBSA
Current 46 78 53 35 29 SBC SBSA
% change 0% 0% 9% 13% +16%
Farm 10 Former 80 100 101 65 23 SlC SBP
Current 100 100 118 75 24 SlC SBP
% change +25% 0% +17% +15% +4%
52 G. Martin, M. Willaume / Agriculture, Ecosystems and Environment 221 (2016) 50–59

 specialization and intensification (Lemaire et al., 2013; Peyraud systems in lowland plains to less-productive grass-based systems
et al., 2014) consists of intensifying both land use and herds by in mountains. We identified the case study farms based on the
increasing stocking rate and maintaining or increasing produc- expertise of an advisor from the local chamber of agriculture. We
tion per unit of area and livestock. It goes along with a asked him to restrict the selection to farms specialized in milk
simplification and standardization of management practices. It is production (i.e., all the usable agricultural land is dedicated to
driven by an increased use of technologies, high-performance producing fodder and grain for the farm animals) displaying
equipment, and overall of inputs in the form of N-rich consistency in their technical functioning like achievement of the
concentrates and N fertilizers; farmers’ production objectives in most years, animal diets in
 eco-efficient intensification (Keating et al., 2010) is similar to accordance with targeted production levels, no major animal
sustainable intensification and ecological intensification (Doré health issues or financial troubles interfering with farmers’
et al., 2011). It consists of increasing stocking rate and farm technical choices, etc. The 10 selected dairy farms (Table 1) were
production per unit of area and livestock while maintaining the expected to capture the diversity of local dairy systems. These
consumption of inputs in the form of N fertilizers and N-rich 10 farms were individually surveyed. Data collection occurred in
concentrates. It results in an increase of farm input/output ratio mid-2014 and was driven by a template developed for the purpose
over time; of the study. Data collection relied as much as possible on
 agroecological transition (Altieri, 1989) consists of substituting measurements e.g., invoices for inputs (feedstuffs, fertilizers, fuel,
inputs (e.g., synthetic fertilizers, pesticides) with enhanced power, gas, etc.), and when unavailable was reported from farmers’
ecosystem services (e.g., nitrogen fixation by legumes, biological sayings, e.g., for animal diets. We considered two situations for
regulation of crop pests) (Power, 2010; Dumont et al., 2013). It is each farm: the current situation (year 2013) and a former situation
characterized by a decrease of input (fertilizers and N-rich (Table 1), at most 10 years before, suggested by the farmer during
concentrates) use, and a maintenance or a decrease of stocking the survey and different from the current situation.
rate and farm production per unit of area and livestock.
2.2. Farm modeling and estimation of GHG emissions
Section 2 describes the protocol for data collection through
farm surveys and data analysis through modelling of greenhouse Mechanistic approaches to estimating GHG emissions tend to
gas emissions. Results are presented and discussed in Section 3. be too complex or difficult to parameterize at the farm level (Del
Prado et al., 2013b). Therefore, a wide range of empirical GHG-
2. Material and methods emission calculators exists. In reviewing 18 such calculators,
Colomb et al. (2013) noted that all are based on International Panel
2.1. Case study systems on Climate Change (IPCC) guidelines, follow the IPCC tier
classification, and use IPCC methodology, which consists of
Dairy farm selection was comparable to the case-study research multiplying activity data from an emission source (e.g., land area,
approach (Eisenhardt, 1989). With this approach, qualitative number of animals) by an emission factor for a given gas. Major
surveys do not rely on statistically significant samples but rather differences in these approaches include inclusion/exclusion of
on samples that correspond to the diversity of studied objects (e.g., land-use change effects and estimation of N2O emissions according
herd, land use) to grasp the diversity of situations. For this reason, to the sources of N considered, which can be related to the
we chose to work in the French department of Aveyron. It has a definition of system components and boundaries. The farm model
wide diversity of dairy farms, from highly productive maize-based of Forage Rummy (Martin et al., 2011a,b; Martin, 2015) is well-

Fig. 1. Overview of the farm model of Forage Rummy (Martin et al., 2011) integrating a greenhouse-gas (GHG) emission calculator, and its main inputs, data flows and outputs.
 G. Martin, M. Willaume / Agriculture, Ecosystems and Environment 221 (2016) 50–59 53

suited to on-farm data collection and simulation. It focuses on It estimates carbon dioxide (CO2), methane (CH4) and nitrous oxide
interactions between land use (crop and grassland distribution and (N2O) emissions (Fig. 2), which are summed to estimate GHG
management) and herd management (batching, reproduction, emissions according to their characterization factors compared to
milk and meat production, and feeding). We chose to add a GHG- CO2-equivalence at 100 years (Global Warming Potential): 1 for
emission calculator to the farm model of Forage Rummy (Fig. 1). CO2, 25 for CH4 and 298 for N2O (IPCC, 2006). To allocate all
emissions among co-products, allocation factors used in France
2.2.1. Forage Rummy farm model (Beguin et al., 2012) and abroad (Basset-Mens et al., 2009) were
The farm model of Forage Rummy is a simulation model used: 85% allocated to milk and 15% allocated to meat.
dedicated to location-specific use (regarding soil–climate con- Fig. 2 synthesizes the farm components and GHG flows
ditions and management practices) to address local adaptation of included in our study and Appendix B gives a detailed list of
livestock systems (i.e., dairy and beef cattle, dairy and meat sheep, equations and emission factors used. Following the IPCC Tier
dairy goats). The simulation model is a stock and flow model with a 2 simplified method (IPCC, 2006), CH4 emissions related to enteric
4-week time step that enables multi-year simulations to be run. It fermentation of animals—dairy cows and young stocks—are
is composed of three components (Fig. 1): estimated from animal dry-matter and energy intake and from a
methane-conversion factor correlated with diet digestibility (Opio
1. A database of grazeable grass (kg/ha) per four-week period et al., 2013). Animal dry-matter and energy intake as well as diet
throughout the calendar year or of the yield (grain or grass) digestibility are provided by an empirical animal nutrition model
available from mechanized harvest once to several times a year (INRA, 2007) integrated into the farm model. Mean gross energy
for combinations of a particular climatic year, soil type, crop/ content of feed is assumed to be 18.45 MJ.kg DM 1 (IPCC, 2006).
grassland and its year-round management. This database was Methane emissions as well as direct and indirect N2O emissions
based on the expertise of researchers, farm advisors and/or from manure and slurry at stable, storage and spreading (Fig. 2) are
farmers, or the use of crop/grassland models (e.g., Duru et al., calculated based on the IPCC Tier 2 simplified method (IPCC, 2006)
2009). using emission factors parameterized for France (ADEME, 2013).
2. A database of daily dry-matter, energy and protein requirements Methane emissions from manure management depend on the type
for a type of animal characterized by morphological parameters, of manure produced (manure–dung, urine and bedding material—
a production level and a reproduction period. This database was or slurry–dung and urine-, management and treatment), which
based on an animal production and feeding model (INRA, 2007), depends on the type of animal housing, duration of animal stay
as well as on-farm measurements. inside the stable, and animal–dairy cows and young stocks—
3. A simulation engine enabling farm parameters to be specified excretion, which is estimated from animal dry-matter intake
(e.g., area allocated to each type of grassland and crop, type of corrected from its digestibility, urinary energy and ash-content
animals in each animal group, diet of each animal group losses. Nitrous oxide emissions are also estimated from the type of
throughout the year) and indicators to be calculated and plotted. manure produced, the type of animal housing, duration of animal
For instance, graphs illustrate the extent to which animal stay inside the stable and N excretion by animals. Nitrous oxide
feeding requirements (i.e., dry matter (DM), energy, protein) are emissions from dung and urine during grazing are estimated from
covered during each four-week period with the livestock system the duration of dairy cows and young stocks grazing and N
simulated. It also predicts, among other things, dynamics of excretion by these animals. Following Gac et al. (2010), N2O
stocks of forage, concentrates, and straw; and self-sufficiency in emissions from fertilizer applications are estimated from the
forage, concentrates, straw, energy and protein (Fig. 1). amount of N applied and the risk of leaching. Carbon dioxide
emissions related to farm inputs (e.g., purchased animals,
feedstuff, power, fertilizers; Fig. 2) are based on emission factors
2.2.2. GHG emission calculator from GES'TIM (Gac et al., 2010). Emissions related to land-use
The GHG emission calculator combined equations and emission change are not included. We considered that the carbon
factors from IPCC methodology (IPCC, 2006) and from the two sequestration potential of grasslands was yet uncertain (Jones
most widely used GHG-emission assessment methods calibrated and Donnelly, 2004) and highly variable depending on manage-
for France: Dia’Terre (ADEME, 2013) and GES’TIM (Gac et al., 2010). ment practices or climate (Jandl et al., 2014). Furthermore,

Fig. 2. Conceptual model of the farm components and flow. The greenhouse gas emissions included are circled in black (N2O), dark gray (CH4) and light gray (CO2) on the flows
and inputs.
54 G. Martin, M. Willaume / Agriculture, Ecosystems and Environment 221 (2016) 50–59

Fig. 3. Tradeoffs among intensification options among farms for former (light-gray line) and current (black line) situations. Indicator values are expressed as percentages of
the maximum value among all farms and situations. AU = animal units.
 G. Martin, M. Willaume / Agriculture, Ecosystems and Environment 221 (2016) 50–59 55

calculating emissions in that way enabled to address the decrease remaining farms (6, 7, 8, 9 and 10) are less intensively managed and
of total GHG emissions rather than their mitigation through have indicator values among the lowest for all farms, except for
increased carbon sequestration in soils. Emissions related to the percentage of organic N spread as slurry (farm 10), stocking rate
construction of farm buildings are not taken into account, and milk production per cow (Fig. 3).
considering that the required data was hard to collect, and the These results confirm that the sample of farms selected displays
resulting differences in GHG emissions among farms were a diversity of intensification patterns. Diversity is also found when
negligible in most cases. analyzing the farms' input-use efficiency and feed efficiency.
Efficiency is determined by comparing inputs and outputs (Van
2.3. Data analysis Ittersum and Rabbinge, 1997) and is key to the sustainability of
agricultural systems (De Koeijer et al., 2002). Farms 2, 3 and 4 are
Data analysis began by using Forage Rummy to verify the among the top 5 farms for amount of synthetic N fertilizer applied
consistency of data collected from documents and farmers’ per kg of milk produced (data not shown) and of N as concentrate
interviews. For instance, animal diets were simulated to verify purchased per kg of milk produced (Fig. 3). These three farms are
that herd concentrate consumption matched the sum of the probably among the least efficient of those surveyed. In contrast,
amounts of concentrates that farmers purchased and produced on farms 1, 7, 8, 9 and 10 appear the most efficient. More efficient
farms, or that feeding requirements were fulfilled in accordance systems require less input to produce a given amount of output and
with production levels. If necessary, farmers were contacted to are associated with lower GHG emissions (Smith et al., 2008). For
correct erroneous or questionable data. instance, more efficient synthetic N fertilizer applications (with
To characterize the diversity of local farming systems, their respect to crop demand) can decrease N2O emissions (Bouwman
adaptation pathways and analyze tradeoffs among adaptation et al., 2001). Similarly, high feed efficiency is associated with
options among farms and situations (former vs. current), we greater milk yield and improved feed digestibility (Britt et al.,
selected indicators that characterize changes in the intensity of 2003), which reduces CH4 emissions from enteric fermentation per
land use and animal breeding, i.e., farm stocking rate (animal kg of milk produced (Clemens and Ahlgrimm, 2001).
units (AU) per ha UAA to account for dairy cows and young stocks)
and farm production per unit of land (kg of milk produced/ha 3.2. Adaptation patterns among surveyed farms
UAA) and animal (kg of milk produced/cow/yr). As a consequence
of changes in the intensity of land use and animal breeding, When focusing on farm-adaptation pathways, comparison of
farmers often have to adapt farm reliance on input (mainly former and current situations reveals that the extent of the
fertilizers and concentrates) that we assessed on a nitrogen basis: adaptations implemented differs widely between farms (Fig. 3).
kilograms of N from fertilizer purchased per ha usable agricul- Farm 1 has intensified land use by increasing the stocking rate
tural area (UAA) and grams of N from cow concentrates purchased (from 1.10 AU/ha to 1.48 AU/ha) and milk production per ha (from
per kg of milk produced. Indeed, fertilizers and concentrates are 5428 kg of milk/ha to 7500 kg of milk/ha). Following these changes,
the two main source of N inputs in French dairy farms. The farm 1 has highly increased N inputs through concentrate feeding
percentage of organic N fertilizer that is applied as slurry (vs. (by 199% i.e., from 25 gN/kg of milk to 76 gN/kg of milk) between
manure) was also used as it reflects farmers' willingness to spread the two situations. A similar trend was observed on farm 2 that
quickly available N to plants. displayed a 112% increase, from 53 gN/kg of milk to 112 gN/kg of
All six indicators were used to differentiate adaptation path- milk even if neither stocking rate nor production per unit of area
ways corresponding to specialization and intensification, eco- and livestock significantly changed between the two situations. On
efficient intensification and agroecological transition. Indicator both farms, this increase in N-rich concentrate use corresponded to
values of farms and situations were compared by expressing them a switch in the form of N input resulting in a decrease in the
as percentages of the maximum value of each indicator among all amount of N fertilizer purchased. In addition to this change, farm
farms and situations. For each of the 10 dairy farms studied, we 2 has significantly increased the percentage of cropland (by 69%;
analyzed changes between the former and current situations in the Table 1). Connected to this decision, the farmer also changed the
relation between product-related (kg CO2 eq./kg of milk produced) manure collection and storage system from straw bedding and
and area-related (kg CO2 eq./ha UAA) GHG emissions, and in the scraped alley to scraped cubicles (Table 1), which changes all
distribution of GHG emissions per source of emissions (enteric organic N fertilizer spread on the fields to slurry providing quickly
fermentation, manure management, excreta during grazing, soil available N to crops and grasslands.
fertilization, input use and energy use). Farms 3, 4 and 5 have increased their stocking rates (by 11–42%)
and their milk production per ha (by 32–82%) and per cow (by 3–
3. Results and discussion 27%). All three farms have not significantly changed N inputs
through concentrate feeding. Like farm 2, farm 4 changed the cow
3.1. Diversity among surveyed farms manure collection and storage system to a slurry system and
operated a small reduction (by 5%) in amounts of N fertilizers
Surveyed farms in the current situation show a wide diversity of purchased. In contrast, farm 3 returned to a mixed system,
intensification patterns (Fig. 3). Some farms, especially farms 1 and including straw bedding, and compensated for the decreased N
5, have intensive land use and herd management, as illustrated by availability from slurry (by 42%) by purchasing 15% more N
high indicator values for stocking rate (1.48 AU/ha and 1.76 AU/ha fertilizer. Farm 3 was also the single farm that significantly
for farms 1 and 5 respectively) and milk production per ha increased replacement rate within the herd (Table 1) leading to a
(7500 kg/ha and 9412 kg/ha for farms 1 and 5 respectively) and per high proportion of non-productive animals (i.e., young stocks)
cow (Fig. 3). Along with farms 1 and 5, farm 3 and 4 share high milk similar to farm 4.
production per cow. Farms 2, 3, and 5 also have a high amount of Farms 7 and 8 have slightly reduced their stocking rates and
purchased N fertilizer. Farm 2 differs from farms 1, 3, 4 and milk production per ha and per cow. In contrast, farm 6 had a slight
5 because it has relatively low milk production per cow and per ha increase in these three indicators. Yet all three farms (6, 7 and 8)
(5119 kg/ha), yet has high amounts of N as concentrates in the cow have significantly reduced N inputs in the form of concentrates
diet. It is also characterized by a high stocking rate (1.67 AU/ha) and (from 164 and 69 gN/kg of milk to 85 and 36 gN/kg of milk) and N
by the highest percentage of organic N spread as slurry. The fertilizers (from 25 kg N/ha and 61 kg N/ha to none, for farms 7 and
56 G. Martin, M. Willaume / Agriculture, Ecosystems and Environment 221 (2016) 50–59

8 respectively), the latter being related to a decrease in the Table 2

Change in the relation between product-related greenhouse gas (GHG) emissions
percentage of cropland in UAA (by 17–52%, in favor of grasslands).
(kg CO2 eq./kg milk) and area-related GHG emissions (kg CO2 eq./ha usable
Farms 9 and 10 show little change between the two situations, agricultural area (UAA)) between the former and current situations for the 10 dairy
which can be considered a form of status quo. For these two farms, farms studied. = is for no change, + for an increase below 20%, ++ for an increase
the main adaptation is observed on farm 9, which decreased N above 20%, for a decrease below 20%, – for a decrease above 20%.
inputs through concentrate feeding by 12%, which in turn led to a Product-related GHG emissions Area-related GHG emissions
15% decrease in its milk production per ha (Fig. 3). Farm 9 also
(kg CO2 eq./kg milk) (kg CO2 eq./ha UAA)
increased herd replacement rate while decreasing herd size
(Table 1) leading to an increase in the proportion of non- Situation Former Current Former Current
Farm 1 1.04 1.09 + 6664 9587 ++
productive animals (i.e., young stocks).
Farm 2 1.21 1.31 + 7104 7896 +
These differences revealed that while being confronted to Farm 3 0.85 0.92 + 4818 5398 +
comparable adverse events (weather conditions, volatility of Farm 4 1.32 1.02 – 3338 3930 +
prices, changes in regulations, etc.), farms followed four rather Farm 5 1.00 0.85 7629 9388 +
than three types of adaptation pathways. Farms 1 and 2 had Farm 6 1.60 0.96 – 10219 7387 –
Farm 7 0.90 0.88 5927 4876 –
intensified both land use and herds by increasing herd size and
Farm 8 0.97 0.82 5139 3655 –
stocking rate, and maintaining or increasing production per unit of Farm 9 0.96 1.05 + 5655 5219
area and livestock. These farms followed the trend of specialization Farm 10 1.20 1.21 = 6021 5621
and intensification to improve labour productivity (Lemaire et al.,
2014; Peyraud et al., 2014). Farmers simplified management
practices and standardized cow feeding, which increased reliance Western Europe (Olesen et al., 2006; Vellinga et al., 2011; Del Prado
on maize silage and concentrates resulting in an increase of N-rich et al., 2013b) indicating that calculated emissions are not very
feed input use per kg of milk produced (Fig. 3). Management sensitive to emission factors specific for France used in our study.
practices were driven by technologies such as synthetic inputs Further comparison is difficult given that the results are highly
(e.g., pesticides and mineral fertilizers) and high-performance sensitive to the methodology and emission factors used (Del Prado
equipment: both farms had a milking robot and an automated et al., 2013b; van Middelaar et al., 2013).
concentrate feeder. Several types of changes in GHG emission patterns can be
In contrast, farms 6, 7 and 8 followed the trend of agroecology identified when addressing the transition between the former and
(Altieri, 1989). While maintaining (farm 6) or slightly decreasing current situations (Table 2). Farms 1 and 2 followed a trend of
(farms 7 and 8) stocking rate and farm production per unit of area specialization and intensification and significantly increased both
and livestock, all three farms succeeded in substituting inputs with product-related and area-related GHG emissions due to a large
ecosystem services (e.g., N fertilizers with legumes-rich grass- increase in total emissions related to manure management (+36%
lands, Fig. 3; Power, 2010; Dumont et al., 2013) to limit their and +120% in farms 1 and 2 respectively) and input use (+23% and
exposition to the volatility of input prices. When comparing +13% in farms 1 and 2 respectively) (Fig. 4).
current and former situations, farms 6, 7 and 8, increased their Eco-efficient farms displayed more diverse pathways. Eco-
levels of agricultural diversity (Altieri, 1989) as encouraged by the efficient intensification implemented by farm 3 led to an increase
greening of the European common agricultural policy. For instance, in product-related and area-related GHG emissions due to
farmer 8 replaced 10 ha of maize, 12 ha of triticale and 33 ha of increased emissions related to enteric fermentation (+7% consis-
perennial rye-grass with 5 ha of maize, 7 ha of a mix composed of tent with the increase in herd size, stocking rate and proportion of
wheat, triticale, oat and pea, and 43 ha of grasslands composed of non-productive animals) and input use (+17%). In contrast, farms
10 different species. 4 and 5 decreased product-related GHG emissions but increased
In between these two categories, farms 3, 4 and 5 followed a area-related GHG emissions as both herd sizes and stocking rates
trend of eco-efficient intensification (Keating et al., 2010) by increased as well. The former decreased emissions related to
increasing their stocking rate and production per unit of area and enteric fermentation ( 14%), soil fertilization ( 40%), and input
livestock while maintaining (or slightly changing) their consump- use ( 52%), while the latter decreased emissions related to enteric
tion of inputs in the form of fertilizers and N-rich concentrates. fermentation ( 20%) manure management ( 10%) and soil
Compared to other farms in the sample, farms 3, 4 and 5 have not fertilization ( 15%) (Fig. 4).
experienced significant changes in farm size. Thus farmers Following an agroecological transition, farms 6, 7 and 8 signifi-
invested most of their efforts (and succeeded) in increasing their cantly decreased both product-related and area-related GHG
overall efficiency, i.e., their farm N input/output ratio over time. emissions. These changes reflect changes in total GHG emissions
Finally, the status quo operated by farms 9 and 10 emerged as a last due to manure management ( 71% and 38% in farms 6 and 8
type of adaptation pathway neglected in our hypothesis. respectively), soil fertilization ( 38%, 19% and 53%, in farms 6, 7
and 8 respectively) and input use ( 57%, 27% and 50%,
3.3. GHG emissions of surveyed farms respectively) (Fig. 4). These changes compensated for slight
increases of emissions due to enteric fermentation in farms
In the current situation, surveyed farms have product-related 7 and 8 (+6%, and +7%, respectively). Farms 9 and 10, characterized
GHG emissions of 0.82–1.31 kg CO2 eq./kg milk and area-related by the status quo, slightly increased product-related GHG
GHG emissions of 3655–9587 kg CO2 eq./ha UAA (Table 2). Mean emissions, but also slightly decreased area-related GHG emissions.
impacts of all farms are 1.01 kg CO2 eq./kg milk and 6296 kg CO2 The main change was that farm 9 increased GHG emissions related
eq./ha UAA, respectively. In comparison, (Dollé et al., 2011) to input use (+18%) and to enteric fermentation (+6%) consistent
estimated a mean product-related GHG emission of 1.27 kg CO2 with the increase in the proportion of non-productive animal.
eq./kg milk for 153 French dairy farms (allocating 88% and 12% of The extent of changes in product-related and area-related GHG
impacts to milk and meat, respectively, and including GHG emissions varies widely among farms (Table 2). For instance, farm
emissions related to construction of farm buildings and machin- 6 has decreased product-related GHG emissions by 50% and area-
ery). Thus, the sample of farms surveyed has product-related related GHG emissions by 39%, while farm 9 has increased product-
emissions below the sample studied by Dollé et al., yet they lie related GHG emissions by only 6% and decreased area-related GHG
within the range of values of previous studies conducted in emissions by only 7%.
 G. Martin, M. Willaume / Agriculture, Ecosystems and Environment 221 (2016) 50–59 57

Fig. 4. Distribution of product-related greenhouse gas (GHG) emissions (kg CO2 eq./kg milk) for former and current situations of each farm per source of emissions (enteric
fermentation, manure management, excreta during grazing, soil fertilization, input use and energy use).

From these results, independently of size (usable agricultural area 2011; Van der Werf et al., 2009). Moreover, part of the data used in
and herd) effects, there is a large scope for reducing GHG emissions our study was based on farmers’ sayings. Despite the use of Forage
by better adjusting input use to farm outputs. Indeed, an Rummy to verify its consistency, it is subject to uncertainty.
agroecological transition (as performed by farms 6, 7 and 8) Another source of uncertainty is related to our methodology for
decreases both product-related and area-related GHG emissions GHG emissions accounting. Our methodology combined equations
the most. This transition is characterized by a decrease in stocking and emission factors that are regularly improved together with
rate and milk production per ha and by partial, if not complete, knowledge development on the processes underlying GHG
substitution of input use with ecosystem services (e.g., synthetic N emissions (Del Prado et al., 2013b; van Middelaar et al., 2013).
fertilizers with N fixation by legumes-rich grasslands), associated This is the reason why emission factors are regularly revised.
with an overall decrease in cropland area and diversification of crop Sensitivity analysis of farm GHG emissions to changes in emission
rotations and grassland types, as for the above-described case of farm factors could help to characterize the uncertainty of our results
8. In two out of three cases (farms 4 and 5, but not 3), eco-efficient inherent to the methodology.
intensification reduced product-related GHG emissions, as sug- The above-discussed sources of uncertainty indicate that
gested by Smith et al. (2008). In contrast, specialization and additional research should be conducted to evaluate the possibility
intensification (as performed by farms 1 and 2) appear an unsuitable for extrapolating our results in other temperate areas. Another
pathway towards mitigating climate change, as suggested by interesting perspective is to include emissions related to land-use
previous studies (Bonesmo et al., 2013; Vellinga and Hoving, 2011). change. Indeed, land use change effect upon farm GHG emissions
Veilinga et al. (2011) have shown that some mitigation options can be significant when taking into account the potential increase
(e.g., lowering replacement rate, replacing concentrates by single in soil carbon sequestration offered by an increase in grassland
products like farm-grown corn) can be cost effective in dairy area (Vellinga and Hoving, 2011).
systems of the Netherlands. We did not go into detailed analysis of
farm economic performances but farmers who had followed an 4. Conclusion
agroecological transition indicated that their incomes had not been
negatively affected suggesting that the cost of associated mitiga- We conducted a diachronic study of the change in GHG
tion options was affordable for them. emissions following implementation of adaptations by farmers on
10 dairy farms in Aveyron, France. Unlike our first hypothesis,
3.4. Limitations of our study surveyed farms follow four types of farm adaptation pathways:
specialization and intensification, eco-efficient intensification,
We conducted a case-study research work (Eisenhardt, 1989) agroecological transition and status quo. In accordance with our
and tried to define a sample of dairy farms corresponding to the second hypothesis, these adaptation pathways drive changes in
diversity of dairy farms in Aveyron. After classifying the GHG emissions on dairy farms, particularly by determining farms’
10 surveyed farms according to the adaptation pathway they reliance on input use. Our results show that an agroecological
had followed (status quo, specialization and intensification, eco- transition reduces GHG emissions the most while farmers’
efficient intensification, agroecological transition), the sample size incomes were not negatively affected indicating that the cost of
per pathway is small (2–3 farms for each pathway). It has been associated mitigation options is affordable for farmers. The two
shown that the contribution of a production mode to environ- dairy farms following this transition had the lowest product-
mental impacts of farms can be lower than the contribution of related and area-related GHG emissions. Additional research
differences among farms within a production mode (Eady et al., should be conducted for (i) evaluating the possibility for
58 G. Martin, M. Willaume / Agriculture, Ecosystems and Environment 221 (2016) 50–59

extrapolating our results obtained on a small sample of farms in Eisenhardt, K.M., 1989. Building theories from case study research. Acad. Manag.
other temperate areas; (ii) integrating the impacts of land-use Rev. 14, 532–550.
Fleming, A., Vanclay, F., 2009. Farmer responses to climate change and sustainable
changes on mitigation of GHG emissions. agriculture. A review. Agron. Sustain. Dev. 30, 11–19.
Füssel, H.M., 2007. Adaptation planning for climate change: concepts, assessment
approaches, and key lessons. Sustain. Sci. 2, 265–275.
Acknowledgements Gac, A. Le Gall, A. Deltour, L. Cariolle, M. Espagnol, S. Lagadec, S. Flenet, F. Lellahi, A.
Malaval, C. Tailleur, A. Ponchant, P., 2010. GES'TIM: Guide méthodologique
d'évaluation de l'impact des activités agricoles sur l'effet de serre. Available at:
This work was funded by INRA in the ACCAF metaprogram as http://idele.fr/?eID= cmis_download&oID =workspace://SpacesStore/
part of the FARMATCH project and by the French ANR Agrobio- ec8397f7-7e23-414f-82fd-41ee77b1533e.
sphère program as part of the TATABOX project (ANR-13-AGRO- Gerber, P.J., Steinfeld, H., Henderson, B., Mottet, A., Opio, C., Dijkman, J., Falcucci, A.,
Tempio, G., 2013. Tackling Climate Change Through Livestock—a Global
0006). The authors would like to thank Vladimir Segard for data
Assessment of Emissions and Mitigation Opportunities. Food and Agriculture
collection and Michel Weber from the Aveyron chamber of Organization of the United Nations (FAO), Rome.
agriculture for his support in selecting the farms and providing Grainger, C., Beauchemin, K.A., 2011. Can enteric methane emissions from
data from these farms. We also thank all the surveyed farmers for ruminants be lowered without lowering their production? Anim. Feed Sci.
Technol. 166-167, 308–320.
their time and conviviality. Hagemann, M., Ndambi, A., Hemme, T., Latacz-Lohmann, U., 2012. Contribution of
milk production to global greenhouse gas emissions. An estimation based on
typical farms. Environ. Sci. Pollut. Res Int. 19 (2), 390–402.
Appendix A. Supplementary data INRA (Ed.), 2007. Alimentation des bovins, ovins et caprins. Besoins des animaux—
Valeur des aliments. Tables INRA 2007. Quae Editions, Paris, France
IPCC, 2006. IPCC Guidelines for National Greenhouse Gas Inventories. Agriculture,
Supplementary data associated with this article can be
Forestry and Other Land Use, vol. 4. Available at: https://www.ipcc.ch/
found, in the online version, at http://dx.doi.org/10.1016/j. meetings/session25/doc4a4b/vol4.pdf.
agee.2016.01.027. Jandl, R., Rodeghiero, M., Martinez, C., Cotrufo, M.F., Bampa, F., van Wesemael, B.,
Harrison, R.B., Guerrini, I.A., Richter, D., deB. Rustad, L., Lorenz, K., Chabbi, A.,
Miglietta, F., 2014. Current status, uncertainty and future needs in soil organic
References carbon monitoring. Sci. Total Environ. 468-469, 376–383.
Jones, M.B., Donnelly, A., 2004. Carbon sequestration in temperate grassland
ADEME, 2013. Dia’Terre: Synthèse du guide de la méthode. 71 p. Available at: http:// ecosystems and the influence of management climate and elevated CO2. New
www.ademe.fr/sites/default/files/assets/documents/ Phytol. 164, 423–439.
diaterre_guide_methode_synthese_v3.pdf. Keating, B.A., Carberry, P.S., Bindraban, P.S., Asseng, S., Meinke, H., Dixon, J., 2010.
Altieri, M.A., 1989. Agroecology: a new research and development paradigm for Eco-efficient agriculture concepts, challenges, and opportunities. Crop Sci. 50,
world agriculture. Agric. Ecosyst. Environ. 27, 37–46. 109–119.
Basset-Mens, C., Ledgard, S., Boyes, M., 2009. Eco-efficiency of intensification Lemaire, G., Franzluebbers, A., Carvalho de, P.C.F., Dedieu, B., 2014. Integrated crop-
scenarios for milk production in New Zealand. Ecol. Econ. 68, 1615–1625. livestock systems: strategies to achieve synergy between agricultural
Beguin, E., Lorinquer, E., Pavie, J., Ferrand, M., 2012. Analyse environnementale production and environmental quality. Agric. Ecosyst. Environ. 190, 4–8.
multicritères et voies d'atténuation des émissions de gaz à effet de serre (GES) Louro, A., Sawamoto, T., Chadwick, D., Pezzolla, D., Bol, R., Báez, D., Cardenas, L.,
des systèmes d'exploitation bovins lait. Actes des 19èmes Rencontres Recherches 2013. Effect of slurry and ammonium nitrate application on greenhouse gas
Ruminants, p. 25–28. fluxes of a grassland soil under atypical South West England weather
Bonesmo, H., Beauchemin, K.A., Harstad, O.M., Skjelvåg, A.O., 2013. Greenhouse gas conditions. Agr. Ecosyst. Environ. 181, 1–11.
emission intensities of grass silage based dairy and beef production: a systems Martin, G., 2015. A conceptual framework to support adaptation of farming
analysis of Norwegian farms. Livest. Sci. 152, 239–252. systems—development and application with Forage Rummy. Agric. Syst. 132,
Bouwman, A., Boumans, L.J.M., Batjes, N.H., 2001. Global Estimates of Gaseous 52–61.
Emissions of Nh3, No and N2o from Agricultural Land. FAO and IFA, Rome, Italy, Martin, G., Felten, B., Duru, M., 2011a. Forage rummy: a game to support the
pp. 106. participatory design of adapted livestock systems. Environ. Model. Softw. 26,
Britt, J.S., Thomas, R.C., Speer, N.C., Hall, M.B., 2003. Efficiency of converting nutrient 1442–1453.
dry matter to milk in Holstein herds. J. Dairy Sci. 86, 3796–3801. Martin, G., Magne, M.A., 2015. Agricultural diversity to increase adaptive capacity
Cammarano, D., Payero, J., Basso, B., Stefanova, L., Grace, P., 2012. Adapting wheat and reduce vulnerability of livestock systems against weather variability—a
sowing dates to projected climate change in the Australian subtropics: analysis farm-scale simulation study. Agric. Ecosyst. Environ. 199, 301–311.
of crop water use and yield. Crop Pasture Sci. 63, 974–986. Martin, G., Martin-Clouaire, R., Rellier, J.P., Duru, M., 2011b. A simulation framework
Clemens, J., Ahlgrimm, H.J., 2001. Greenhouse gases from animal husbandry: for the design of grassland-based beef-cattle farms. Environ. Model. Softw. 26,
mitigation options. Nutr. Cycl. Agroecosyst. 60, 287–300. 371–385.
Colomb, V., Touchemoulin, O., Bockel, L., Chotte, J.-L., Martin, S., Tinlot, M., Bernoux, Mejía, R.M., Ortuño, J.A., Lascano, G.J., Vélez, M., 2010. Effects on lactation
M., 2013. Selection of appropriate calculators for landscape-scale greenhouse performance when slick hair gene is simulated in dairy cattle in the tropics. J.
gas assessment for agriculture and forestry. Environ. Res. Lett. 8, 015029. Dairy Sci. 93 (E-Suppl. 1), W184 Abstr..
Datta, A., Smith, P., Lal, R., 2013. Effects of long-term tillage and drainage treatments Olesen, J.E., Schelde, K., Weiske, A., Weisbjerg, M.R., Asman, W.A.H., Djurhuus, J.,
on greenhouse gas fluxes from a corn field during the fallow period. Agr. 2006. Modelling greenhouse gas emissions from European conventional and
Ecosyst. Environ. 171, 112–123. organic dairy farms. Agric. Ecosyst. Environ. 112, 207–220.
De Koeijer, T.J., Wossink, G.A.A., Struik, P.C., Renkema, J.A., 2002. Measuring Opio, C., Gerber, P., Mottet, A., Falcucci, A., Tempio, G., MacLeod, M., Vellinga, T.,
agricultural sustainability in terms of efficiency: the case of Dutch sugar beet Henderson, B., Steinfeld, H., 2013. Greenhouse Gas Emissions from Ruminant
growers. J. Environ. Manage. 66, 9–17. Supply Chains—A Global Life Cycle Assessment. Food and Agriculture
Del Prado, A., Crosson, P., Olesen, J.E., Rotz, C.A., 2013a. Whole-farm models to Organization of the United Nations (FAO), Rome.
quantify greenhouse gas emissions and their potential use for linking climate Peyraud, J.-L., Taboada, M., Delaby, L., 2014. Integrated crop and livestock systems in
change mitigation and adaptation in temperate grassland ruminant-based Western Europe and South America. A review. Eur. J. Agron. 57, 31–42.
farming systems. Animal 7:s2, 373–385. Power, A.G., 2010. Ecosystem services and agriculture: tradeoffs and synergies.
Del Prado, A., Mas, K., Pardo, G., Gallejones, P., 2013b. Modelling the interactions Philos. Trans. R. Soc. Lond. B. Biol. Sci. 365, 2959–2971.
between C and N farm balances and GHG emissions from confinement dairy Reidsma, P., Ewert, F., Lansink, A.O., Leemans, R., 2010. Adaptation to climate change
farms in northern Spain. Sci. Total Environ. 465, 156–165. and climate variability in European agriculture: the importance of farm level
Dollé, J.-B., Agabriel, J., Peyraud, J.-L., Faverdin, P., Manneville, V., Raison, C., Gac, A., responses. Eur. J. Agron. 32, 91–102.
Le Gall, A., 2011. Les gaz é effet de serre en àlevage bovin : àvaluation et leviers Rosenzweig, C., Tubiello, F.N., 2007. Adaptation and mitigation strategies in
d’action. INRA Prod. Anim. 24, 415–432. agriculture: an analysis of potential synergies. Mitig. Adapt. Strategies Glob.
Doré, T., Makowski, D., Malézieux, E., Munier-Jolain, N., Tchamitchian, M., Tittonell, Change 12, 855–873.
P., 2011. Facing up to the paradigm of ecological intensification in agronomy: Smith, P., Martino, D., Cai, Z., Gwary, D., Janzen, H., Kumar, P., McCarl, B., Ogle, S.,
revisiting methods, concepts and knowledge. Eur. J. Agron. 34, 197–210. O'Mara, F., Rice, C., Scholes, B., Sirotenko, O., Howden, M., McAllister, T., Pan, G.,
Dumont, B., Fortun-Lamothe, L., Jouven, M., Thomas, M., Tichit, M., 2013. Prospects Romanenkov, V., Schneider, U., Towprayoon, S., Wattenbach, M., Smith, J., 2008.
from agroecology and industrial ecology for animal production in the 21st Greenhouse gas mitigation in agriculture. Philos. Trans. R. Soc. Lond. B Biol. Sci.
century. Animal 7, 1028–1043. 363, 789–813.
Duru, M., Adam, M., Cruz, P., Martin, G., Ansquer, P., Ducourtieux, C., Jouany, C., Soussana, J.-F., Lemaire, G., 2014. Coupling carbon and nitrogen cycles for
Theau, J.P., Viegas, J., 2009. Modelling above-ground herbage mass for a wide environmentally sustainable intensification of grasslands and crop-livestock
range of grassland community types. Ecol. Modell. 220, 209–225. systems. Agric. Ecosyst. Environ. 190, 9–17.
Eady, S., Viner, J.B., MacDonnell, J.C., 2011. On-farm greenhouse gas emissions and Van der Werf, H.M.G., Kanyarushoki, C., Corson, M.S., 2009. An operational method
water use: case studies in the Queensland beef industry. Animal Prod. Sci. 51, for the evaluation of resource use and environmental impacts of dairy farms by
667–681. life cycle assessment. J. Environ. Manage. 90, 3643–3652.
 G. Martin, M. Willaume / Agriculture, Ecosystems and Environment 221 (2016) 50–59 59

Van Ittersum, M.K., Rabbinge, R., 1997. Concepts in production ecology for analysis Vellinga, V. Th., de Haan, M.H.A., Schils, R.L.M., Evers, A., van den Pol-van Dasselaar,
and quantification of agricultural input–output combinations. F. Crop. Res. 52, A., 2011. Implementation of GHG mitigation on intensive dairy farms: farmers’
197–208. preferences and variation in cost effectiveness. Livest Sci. 137, 185–195.
Van Middelaar, C.E., Berentsen, P.B.M., Dijkstra, J., De Boer, I.J.M., 2013. Evaluation of Vellinga, T.V., Hoving, I.E., 2010. Maize silage for dairy cows: mitigation of methane
a feeding strategy to reduce greenhouse gas emissions from dairy farming: the emissions can be offset by land use change. Nutr. Cycl. Agroecosyst. 89, 413–426.
level of analysis matters. Agric. Syst. 121, 9–22.