Our approach to explore the cost effectiveness of mitigation measures combines models and participatory methods as summarised in Fig. 1. The methodology includes five sequential steps: first, surveys to 120 farmers detailing farming systems were used to characterize the baseline scenario. Second, focus groups that include experts in agricultural, environmental and economic sciences were used to define the possible choices of mitigation measures adapted to the reality of the farming systems. These two steps represent the methodology`s participatory approach. Third, the Cool Farm Tool model was used to define the mitigation potential of the selected measures (Haverkort & Hillier, 2011). Fourth, an optimisation model was used to estimate changes in the farmers´ net margin as a proxy for the cost of implementing the measures. Last, MACC were used to calculate the cost-effectiveness of the mitigation measures in order to guide policy choices.

The study was carried out in 2013 in three sites in Ecuador that exemplify subsistence farming systems in the Central Andes. These farming systems are located between 2,200 and 3,800 meters above the sea level; the temperature ranges from 4 to 21°C and precipitation ranges from 500 to 2000 mm yr^-1. These characteristics produce particular agro-ecological environments. The potato-milk system is the key farming system in this region, with 96% of land dedicated to potato cultivation and 4% of land dedicated to livestock pasture (INIAP, 2013). This system consists of growing potatoes during 2-3 cycles on soils previously occupied by pastures for approximately 2 years, completing a cycle of 4-5 years (Proaño & Paladines, 1998; Paladines & Jacome, 1999; Barrera et al., 2004). Pasture is used for livestock feeding and therefore milk production. After potato production, other crops can be planted for one cycle only to take advantage of residual fertilizers; this third crop is not considered in the study. About 75% of the cultivated area is non-irrigated (INEC, 2016). Soils in this region are Andosols which are primarily characterized by having a high content of organic matter (IUSS Working Group WRB, 2014).

The Andes is topographically variable and has high agro-ecological heterogeneity, with different production zones that exemplify the production range in high altitude zones. Location selection was done according to a bio-socio-economic characterisation which distinguishes three different zones (Table 1).

We selected three zones and representative locations of the mixed production systems (potato and milk) located in the central Andes of Ecuador (Table 1 and Fig. 2). This selection was established based on the following criteria: a) potato and milk production in the area is important for family income and nutrition, b) technology is essential for production sustainability, and c) selected locations are representative for the region.

As such, each of the three locations represents a different producer and production systems: Carchi represents large producers and intensive systems. Chimborazo represents medium producers of mixed systems, and Cañar represents small producers, with extensive systems and family farming.

Sample size was estimated based on the number of potato and milk farmers registered in the Ministry of Agriculture and Livestock: 112 in Carchi, 98 in Chimborazo, and 85 in Cañar. Therefore, 40 surveys in each study area were done for this research.

The surveys gathered information related to crop management such as yield, area, and amount of fertilizer, labour and fuel used. Additional information related to the representativeness of the statistical data for the local realities was obtained from face-to-face interviews with a group of experts from the INIAP and CIP-Ecuador. Socioeconomic information of the population and agro-meteorological characteristics (soil-water-plant-animal) from the selected locations were obtained from the Ministry of Agriculture of Ecuador (MAGAP), the National Institute of Meteorology and Hydrology (INAMHI), and the Military Geographical Institute (IGM).

Based on INIAP (2013) the study areas can be characterized from a sociodemographic perspective. The inhabitants are 48% male and 52% female. An illiteracy rate of approximately 20% is observed. Household average size is 5. The level of education of 70% of heads of household reaches primary education and 18% reaches secondary education. Both men and women have access to secondary education, but usually, neither of them shows interest in continuing education because young people are used to managing their own money and they know that the opportunity for higher education is not possible because it is very expensive. Instead, they decide to invest their capital in the livestock business, selling coal, buying and selling food, making handicrafts or simply growing potatoes. They are aware that education is important, but they prefer not to study in universities. It should be noted that in recent years, the creation of extensions of several universities, as well as the availability of distance university studies, has increased the number of people who have access to university. About 79% of producers have their own land with a property title and 6% of producers without a title. Moreover, 13% of producers use loaned land, and 2% of producers produce on leased land. Finally, 80% of the producers own their own home, while the rest have their homes leased and/or mortgaged. The majority of agricultural activities are under the responsibility of the heads of household or of the male children, although women also participate in these as well as taking care of housework-related activities and the production of handicrafts, mainly textiles. Everyone in the household works; the children are in charge of collecting grass to the guinea pigs and feeding the chickens. After fulfilling their school tasks; the younger children accompany their parents to milk the cows and change the cattle from one paddock to another.

We made an initial list of GHG mitigation measures in agricultural soils based on a published literature review and the datasets available at INIAP. The potential list of measures and their feasibility was discussed and evaluated by scientists on pastures, crops and soil science in the above-mentioned focus groups. Focus groups participants were selected based on the following criteria: i) prior involvement in research; ii) experience related to agricultural GHG mitigation for at least five years; iii) regular contact with farmers; and iv) crop specific knowledge of pastures and potatoes. The measures which showed less implementation barriers were selected, resulting in a list of measures with a higher abatement potential rate chosen for this region and specific crop system.

Using the data obtained in the farmer survey GHG emissions in the baseline scenario and mitigation scenarios were calculated using the open-source software Cool Farm Tool (Haverkort & Hillier, 2011). This site-specific software has been tested in potato farm systems in various countries: The Netherlands (Haverkort & Hillier, 2011), United Kingdom (Hillier et al., 2011), Spain and Peru (Cayambe et al., 2015). Estimates emissions from potato and livestock were analysed separately. Recognising that the mitigation potential varies greatly in different agro-climatic regions and agricultural practices; Cool Farm Tool software defines the mitigation potential for all study sites. In order to calculate the mitigation potential of the measures, emissions data in baseline scenarios were required.

The input information required by the software was: general information (location, year, product, production area, and weather), crop management (agricultural operations, crop protection, fertilizer use, and management of crop residues), carbon sequestration (soil use and soil management, plant biomass), livestock (feed options, enteric fermentation, nitrogen excretion, and manure management), and energy use (irrigation, agricultural machinery). All these data were obtained from the surveys, except weather data which was obtained from weather stations (INAMHI, 2015).

The output information generated by the program consisted of CO₂e emissions from the time of sowing to harvest and can be represented - for each agricultural management choice - per unit area, or per tonne of product produced, of nitrous oxide in the soil, emissions generated in the field by applied fertilizers, emissions generated for pesticide use, energy use, and crop residue management (Haverkort & Hillier, 2011).

The Cool Farm Tool program calculates GHG emissions from different sources using conversion rates that are internationally agreed and used by the IPCC. Table 2 summarises the conversion rates and the sources of data.

Production costs in the baseline scenario are derived from data published in the National Information System of the MAGAP (2016a,b), in which the yearly input and output prices are recorded. The costs of implementing alternative measures were estimated through an optimisation model (i.e. net margin maximization). This method consists of maximizing the net margin of producers when a GHG mitigation practice is implemented in the system. For this calculation we relied on a linear programming model valid for potato-milk systems in Ecuador; which considers different production alternatives imposing restrictions related to the practices and available resources (Annexes A and B). The model defines the production plan that optimizes the use of resources to achieve the maximization of net margin. Net margin was obtained from the difference between the profit and the input cost (fertilizer, seed, crop protection, labour costs, and domestic consumption).

Baseline and mitigation scenarios were optimized under the assumption of maximizing net margin using the linear programming software LINDO (LINDO Systems, 2003).

The cost-effectiveness of a mitigation measure (CEp) is the ratio between the change in the net margin related to practice (ΔNMp) and the change in GHG emissions associated with practice (ΔGHG p) (Eq. 1) (Cayambe et al., 2015; Dequiedt & Moran, 2015). ‘ΔGHGp’ was calculated as the difference between emissions generated in the baseline scenario less emissions from mitigation scenarios.

‘ΔNMp’ was calculated as the difference between the net margin in mitigation scenarios (Eq. 2).

The characterisation of current potato-milk farming systems is defined by the baseline. This is characterised with the information obtained from surveys (Table 3). The results showed that productivity and production costs vary greatly among the three locations. According to the surveys, farmers employ mostly household labour but sometimes hired labour as well. Soil tillage is mostly done manually with 38 people ha^-1 for the production of potatoes in Cañar to 103 people ha^-1 in Carchi. In some cases, the tillage of the soil is done with agricultural machinery using 60 L of fuel ha^-1. High rates of fertilization and frequent use of pesticides are common in the potato-milk system. In general, potato production in Carchi is more mechanized, uses more inputs, and also attains higher productivity. The baseline productions in Chimborazo and Cañar were very similar. In all cases the year to year variability is very high (from 80% to 110% of average production). Milk production varied among the three locations and was also more mechanized in Carchi, especially with high labour inputs. This results in higher milk productivity. Cañar is the location that favours milk production over potato production in terms of pasture area occupied.

Seven potential mitigation measures were selected by scientists at the focus group (Table 4). The selection also included a qualitative analysis of the barriers (climate, agronomic and social constraints) and incentives for implementing the measures. Some of the measures identified are already implemented by some farmers in the region. Additionally, some references for the mitigation measures implementation in countries with similar agricultural and climatic conditions were included in Table 4. The mitigation measures were: M1, management of organic soils; M2, multipurpose fodder trees; M3, manure management; M4, grazing management, pasture improvement and fertilization; M5, reduced tillage + use of crop residues; M6, cropland management agronomy; and M7, nutrient management.

These measures have been selected based on the results obtained in other investigations. Organic soils in the Andes contain high carbon densities. Emissions in these soils can be reduced by avoiding row crops and deep ploughing and keeping the water table close to the surface.

The measure M2 was included as tree species contribute to the soil`s organic carbon through biomass and create shade for livestock. There are species of trees in the legume family that can increase the nitrogen content in soils by fixing them through nitrifying bacteria.

The “huacho rozado” practice is a pre-Columbian system of reduced tillage and cover that is applied in potato cultivation. It is practiced mainly by farmers in the province of Carchi. This system is applied to convert an old pasture into a potato crop, with yields greater than or equal to those of conventional tillage. Being a manual system prevents soil erosion and compacting. In addition, rotting of the plant cover (commonly called chamba) allows microbial activity, creating an antagonistic environment for the development of the white worm (Premnotrypes vorax) and the potato blight (Phytophthora infestans). The requirements to establish a plot with grazed huacho are: pasture with Kikuyo (Pennisetum clandestinum) of more than 3 years, soil slope between 15% and 45% and precipitation of at least 1000 mm yr^-1.

Most of these measures have been the most promising of this analysis, since some of these measures have already been implemented by farmers in the region with similar agricultural and climatic conditions, except those options very difficult to adopt because of the high implementation costs. In addition, the opinion of farmers regarding the barriers to implementation was considered to ensure the adoption of mitigation options raised in consensus with farmers and promote equity and social justice.

This research also showed that there are implications in terms of well-being. That is, if low-input farms are also those managed/owned by poorer households, the costs would be higher for the most economically dis-advantaged.

The barriers to the implementation of measures can be directly related to the socioeconomic data of the farmers.

Using the Cool Farm Tool model, we estimated the emissions in the baseline scenario. Potato and milk production account for an average of 10 ton CO₂e ha^-1 yr^-1. This value includes emissions from potato and milk in the three locations (Fig. 3). Milk production emissions were calculated as the sum of dairy livestock and pasture emissions. In relative terms, crop residue management on the pasture generates the greatest contribution to the total, 40-50% of emissions generated by the mixed systems of the three zones. Emissions from livestock enteric fermentation account for an average of 2 ton CO₂e ha^-1 yr^-1, which are equivalent to 18% of emissions generated in these systems. Emissions attributed to fertilizer use in pastures were not significant, accounting for an average of 0.4 ton CO₂e ha^-1 yr^-1; however, direct and indirect N₂O from excessive fertilizer application in the potato crop reached 20% of emissions generated by the mixed systems of the three zones.

The costs and benefits for the farm (effects on yield, input costs, labour costs and machinery) are used to calculate the impact of each measure in farmers´ net margins when they apply GHG mitigation measures.

The estimation of the costs and benefits of the small measures has been published based on the assumptions of costs and benefits detailed in the published literature (Table 5). Although there is no local empirical data to prove them, these assumptions were analysed and approved with the members of the focus group and with INIAP researchers.

Costs and benefits were used as input data to calculate the net margin per farm and hectare in each case study by using the optimization model (net margin maximization). The change in net margin was defined as the cost or benefit of implementing a particular mitigation measure. Negative changes in net margin mean cost savings and positive changes mean more expensive measures (Table 6).

In Fig. 4, the rate of change in total costs depending on changes in GHG reduction and the relative rate of increase of marginal cost observed. In all scenarios of our study, MACC have negative slopes. Each point represents a mitigation measure, differentiated by the implementation cost per tonne of reduced CO₂e (y-axis) and the quantity of reduced CO₂e emissions (x-axis). Measures below the x-axis are cost-effective, for example removing emissions and representing positive changes in net margin.

Regarding the cost-effectiveness of the mitigation measures to guide policy choices, our results showed that the MAAC of Cañar (the blue curve in Fig. 4) was steeper because of the use of few agricultural inputs. MACC analysis showed that the potential for reducing emissions in farms with high inputs consumption exceeds the abatement potential of farms with low consumption. Thus, in the scenario with high inputs consumption (Carchi, mainly) the cost to reduce 2 ton CO₂e ha^-1 would be USD 300 ton^-1 CO₂e, while in the case of farms with low inputs consumption (Cañar) the effort to reduce 0.7 ton CO₂e ha^-1 would be, equally, about USD 300 ton^-1 CO₂e.

The abatement potential between different geographical areas can be represented for the same MACC. An alternative representation of the cost heterogeneity is presented in Fig. 5, which contains the results of the three locations in regional MACC. The results showed considerable geographical variability. Management of organic soils, reduced tillage and crop residues management in Cañar is more profitable than in Carchi and Chimborazo. Manure management is profitable only in Carchi (USD -7.7 ton^-1 CO₂e. In Chimborazo and Cañar, even for low levels of mitigation, the marginal abatement cost is positive (respectively USD 8.9 ton^-1 CO₂e and USD 4.2 ton^-1 CO₂e). Therefore, in the Andes, it would be expected that the marginal abatement cost range will be USD 4-8 ton^-1 CO₂e.

The most efficient measures in terms of potential and cost reduction were the management of organic soils (M1) and reduced tillage (M5), which had the lowest negative costs and the greatest reduction in CO₂ emissions. Costs and benefits of reduced tillage and crop residue use were derived from studies conducted in Ecuador specifically under the “Huacho rozado” system (INIAP, 2013).

Additionally, there may be mitigation options with mitigation potential, but they cannot be adopted by the farmers, due to the high costs or due to socioeconomic factors of the farmers in each region. This would be considered as an implementation barrier, which will generate inequality among farmers of the three regions. An analysis of implementation barriers could be relevant for a better understanding of the reasons why farmers do not adopt these mitigation practices.

According to information from the Ministry of Environment of Ecuador (MAE, 2011), 210,000 ton CO₂e yr^-1 were generated by the agriculture sector in 2011. The GHG mitigation analysis without considering adoption barriers or costs, and assuming total adoption of practices by farmers, shows that approximately 97,234 ton CO₂e yr^-1 would be reduced in the Central Andes (specifically in regard to Ecuador), that is equivalent to 46% reduction in agricultural emissions generated in 2011 (Table 7). The analysis in relative terms (Table 8) indicates that, the application of measures M4 (grazing management) and M5 (reduced tillage), would reduce agricultural emissions by 12.5% and 8.28%, respectively, of the total emissions generated by agriculture in the year 2011.