Introduction
An important share of the world's food supply is produced on smallholder farmsReference Herrero, Thornton, Notenbaert, Wood, Msangu, Freeman, Bossio, Dixon, Peters, van de Steeg, Lynam, Parthasarathy Rao, Macmillan, Gerard, McDermott, Seré and Rosegrant 1 . Worldwide there are about 500 million farms with less than 2 ha of landReference Wiggins, Kirsten and Llambí 2 . Many people in rural areas depend directly on productivity on these holdings, which are often located in biophysically marginal production areas. Despite policies to support family-based agriculture in many Latin American countries, persistent lack of productivity leads to rural poverty, resulting in migration of young people to cities and, in the case of Mexico, to the USA.
Maize-based smallholder systems dominate the Southern Mexican states of Chiapas, Oaxaca and Guerrero. These states have a high degree of marginalization 3 , and are home to about 8 million poor people. In Mexico, farm sizes up to 3 ha account for 71% of farms, representing around 5 million ha. These smallholders are mainly maize producers; on 42% of maize land they produce 22% of the national volume of maizeReference Gómez 4 . Smallholder maize production systems resemble the traditional milpa systems that date back to ancient civilizations in Mesoamerica, by producing maize with beans and squash in the same fieldReference Zizumbo-Villareal and Colunga-García Marín 5 . However, multi-year fallowing after 2–3 years of production as was practiced in the milpa system has been abandoned due to land shortage. Besides, soil management is largely based on inorganic fertilizer input and crop residues, which, however, are also eaten by roaming cattleReference Kass and Somarriba 6 , Reference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 . In the state of Guerrero, subsidy schemes exist to financially support farmers in purchasing fertilizer for up to 2 ha of land. Since 1994 the schemes include a limited choice of artificial fertilizers that provide nitrogen (N) and phosphorus (P) but not potassium (K) or lime. In 2005, the subsidy schemes were reorganized by adjusting rates and types of fertilizers to soil pH and including bio-fertilizers, such as Azospirillum brasilense and Glomus intraradices but K still was not considered as part of the packages 8 . Recently, the State Government promoted the production and use of organic fertilizers 9 . To obtain the necessary amounts, the Ministry proposed the establishment of composting facilities in each municipality.
A farm survey in five communities of the Costa Chica, a region of Guerrero along the Pacific coast, outlined the constraints that smallholders in the region face: farm sizes of maximum 5 ha, fields that are often only accessible on foot or horseback, production on steep slopes, soils with low to extremely low levels of organic matter (OM) and fertilizer schemes that lead to low as well as imbalanced nutrient availability. Yields of maize grain varied from 750 to 3000 kg ha−1, with very little relation to N or P inputReference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 . Between-field variation in maize yield was significant and dominated variation between farms and communities. A series of on-farm experimentsReference Flores-Sanchez, Pastor, Lantinga, Rossing and Kropff 10 confirmed the lack of N and K and the relative abundance of P, and indicated that application of vermicompost together with inorganic fertilizers and intercropping with canavalia (Canavalia brasiliensis Mart. ex Benth) can contribute to enhance OM input and increases in soil OM (SOM) in the longer term. Increase in SOM is considered as essential for long-term productivity by enhancing water retention, erosion mitigation and increasing cycling of plant nutrients, all of which increase resilience to predicted higher temperatures and increased precipitation variability in MesoamericaReference Marengo, Liebmann, Grimm, Misra, Silva Dias, Cavalcanti, Carvalho, Berbery, Ambrizzi, Vera, Saulo, Nogues-Paegle, Zipser, Seth and Alves 11 .
Livestock farming is an important activity in the Costa Chica, characterized by extensive grazing systems. The largest number of heads (around 250,000) is concentrated in the municipalities Cuajinicuilapa and San Marcos on the coast and Ayutla. Livestock production faces problems, such as the poor facilities and equipment, low genetic quality of the herds, low levels of production and insufficient availability of credit 12 . Livestock are kept in the region as a means of transport, draught for soil cultivation, household consumption or as cash reserve. Cattle are usually confined during the growing season of maize and are released to graze on rangeland and crop residues on the mostly unfenced fields during the dry seasonReference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 , Reference López 13 . The animal unit in the region is 1.62 ha due to livestock pressure on communal land and farmer's fieldsReference Cervantes, Hernández and Jiménez 14 .
Cropping systems that seem promising at field level may be unfeasible when considered at farm level, where constraints emerge such as self-sufficiency in food production, labor availability or restrictive cash flow. Thus, proposals on alternative crop management need to be combined with assessments at higher levels of organization and considered within the context set by policiesReference Hyman, Fujisaka, Jones, Wood, de Vicente and Dixon 15 . Greater SOM input necessitates availability of OM sources, which may need to be found at farm or even regional levels.
This paper sets out to evaluate at farm level the feasibility of various maize production systems that are proposed at field level. We focused on systems as proposed by the state government, the national extension service and our own research, and evaluated the consequences of applying these alternatives at eight specific farms for which detailed data on resource endowment were available. The farm-level explorations were organized in four scenarios, the first two focusing on redressing current nutrient and OM input imbalances and the latter two on exploring high fertilizer input and animal husbandry trajectories.
The aim of this paper is therefore to present and apply a methodology for ex-ante evaluation of alternative maize production strategies to improve farm-level performance in terms of economic, environmental and social objectives. The method is applied to analyze the room for increasing family income and SOM levels and decreasing labor input for farmers in the Costa Chica based on existing technologies.
Materials and Methods
Approach
Using soil and management data from on-farm surveys, we parameterized all relevant farm-specific inputs and outputs of current and alternative maize-based cropping systems (MBCSs) for eight case study farms in the Costa Chica. The alternative systems were based on on-farm experimentsReference Flores-Sanchez, Pastor, Lantinga, Rossing and Kropff 10 and data from local government and extension sources, and focused on different rates and combinations of fertilizer and OM applications. Emphasis was put on maize production, ignoring production of intercropped species, such as bean (Phaseolus vulgaris L.) and squash (Cucurbita pepo L.) that are produced at low densities with average yields of 80 and 95 kg ha−1, respectively and for self-consumption or for local marketing. Roselle (Hibiscus sabdariffa L.), which is produced in more substantial amounts in the region, was included both as mono-culture and intercropped with maize. Canavalia (C. brasiliensis Mart. ex Benth) was included as a potential new cropping activity that could enhance soil fertility or provide feed for cattle. Crop response to inorganic fertilizers and resources was predicted using field-specific information as input for the QUEFTS modelReference Janssen, Guiking, Van der Eijk, Smaling, Wolf and van Reuler 16 , which has been tested on a local datasetReference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 . The field-level data were integrated at the farm level using the FarmDESIGN modelReference Groot, Oomen and Rossing 17 , which allowed exploration of alternative farm configurations that outperformed the current farms in terms of the objectives family income, labor requirements and SOM balance. Exploration of possible farm configurations was organized in four incremental scenarios, the first of which aimed at redressing imbalances in nutrient and SOM balances, and the latter investigated fertilizer and livestock-based intensification options. Details of the approach are described in the next sections.
Case-study farms
We selected eight farms from two communities in the municipality Tecoanapa (16°48′N, 99°09′W) for which data on resource endowment and productivity (Table 1) were available from previous workReference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 , Reference Flores-Sanchez, Pastor, Lantinga, Rossing and Kropff 10 . Mean annual rainfall in the municipality was 1300 mm concentrated between June and October. Minimum and maximum temperatures varied with altitude, from 12 to 27 °C in the higher areas (900 m a.s.l.) and 18–33 °C at less than 300 m a.s.l. 18 . Soils, like elsewhere in the Costa Chica region, were of volcanic origin, classified as regosols. In 2010, the population of Tecoanapa comprised 44,079 inhabitants, over 66% of whom were engaged in farming activities 19 . Total agricultural area was about 14,000 ha.
Table 1. Resource characteristics and actual production levels of maize and roselle on the eight case-study farms in the communities Las Animas (A1–A4) and Xalpathlahuac (X1–X4). Data fromReference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 .
1 A, Community Las Animas; X, Community Xalpathlahuac.
2 MR, maize–roselle intercrop; M, maize monocrop; R, roselle monocrop.
3 Field-specific variation; average value between parentheses.
4 Value between parentheses corresponds to estimated yield due to field being infested by Trichoplusia ni.
The case-study landholdings ranged from 1 to 4.2 ha spread over 1–4 fields (Table 1) and were managed by households with 4–12 family members. The main crops, maize and roselle, were grown during the rainy season (June–November). Most commonly, maize was intercropped with roselle as well as low densities of squash and beans, although there were fields in which maize and roselle were grown as monocrops. Farmers practiced no-till without mechanization and weeds were controlled by herbicides or by cutting with hand-held implements. Mineral N and P fertilizers were the main inputs for crop nutrition applied at sowing and before tasseling, and application rates varied widely among farms. Manure was used only when farmers owned animals, mainly as cattle manure at a rate of 0.6 t ha−1, on average, and applied before sowing. Average maize grain yield was 1.8 Mg DM ha−1; roselle calyx yield was 135 kg ha−1. Only the harvestable products (maize cobs and roselle calyxes) were removed from the field by the farmers. As most of the fields were not fenced, roaming cattle had free access to crop residues during the dry season.
Livestock resources fell into two broad categories: (1) equines (donkeys, horses and mules), pigs, poultry (chickens, turkeys); and (2) goats and cows. The first category was used for transport of materials, traction and/or home consumption, and was occasionally also sold. The second category served primarily as capital and was sold in case of immediate need of cash. Pigs and poultry were kept near the house and fed with household leftovers and grains. During the cropping season, equines, goats and cows were fed by means of cut-and-carry forage provided around the farmstead and by grazing in communal fields. In the dry season, animals were grazed in communal fields, own fields or in other farmers’ fields if unfenced.
The land characteristics (Table 2) show that farm fields were located on steep slopes, were prone to light to severe erosion, were slightly to considerably acid and had low levels of SOM and plant macro-nutrients 21 .
Table 2. Land characteristics of the eight case study farms. Ranges refer to variation among fields. Soil erosion was calculated using RUSLEReference Renard, Foster, Weesies, McCool and Yoder 20 . Data summarized fromReference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 .
Field-level re-design
Maize-based production systems
In this section, we describe the steps taken to re-design field-level maize cropping systems. Since the objective of the study focused on alternatives to current systems for the short term, we concentrated on existing technologies or technologies that could be mobilized without major research effort.
Design criteria
We used three sources of information to select design criteria, i.e., attributes distinguishing the different alternative production systemsReference Hengsdijk and van Ittersum 22 , the state government fertilizer subsidy scheme, the recommendations of the national extension service and results from our own experiments and surveys. We did not find other sources of information that were locally relevant. The resulting design criteria and the associated variants are listed in Table 3.
Table 3. Criteria used to design alternative maize-based cropping systems (MBCSs) for smallholders in the Costa Chica, Mexico.
The design criteria comprised the origin of fertilization strategies, sources of nutrients, use of canavalia and level of residue retention. Cropping systems were constructed by combining variants of the various criteria. Not all combinations resulted in cropping systems that were parameterized. In particular, information on the effect of combinations of organic and mineral sources of nitrogen is still limited, and information on canavalia only existed from the on-farm experiments. In total, 14 MBCSs were parameterized (Table 4). For each MBCS all relevant inputs and outputs were defined as described in the next section.
Table 4. Design criteria used to create maize-based cropping systems (MBCSs) for smallholder farming systems in the Costa Chica, the abbreviations of the variants, their use in scenarios during exploration and the information sources used in their quantification.
1 Scenario S4 comprises the MBCS of S3 and number of animals as decision variables.
Quantification of outputs and inputs of MBCS
Outputs and inputs of the cropping systems were quantified to be able to evaluate their performance at farm level in the next stage using the FarmDESIGN model. Marketable outputs comprised maize grain and roselle calyx yields. Non-marketable outputs comprised changes in the SOM balance resulting from application and decomposition of OM. Inputs included seeds of maize, roselle and canavalia, mineral and organic fertilizer, herbicide and labor. For each field, crop products were characterized in terms of biomass and yield (see below), and N, P, K and ash contents using on-farm measurementsReference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 , complemented with information from the literatureReference Mitra and Shanker 23 – Reference Harrington, Thatcher and Kemp 26 .
Using the design criteria outlined in Table 3, soil fertility strategies were created (Table 4). Farm-specific current fertilizer use (Cu) served as a reference. The simplest change comprised application of K at the same rate as N to compensate for the lack of K in current strategies and in the soilsReference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 . Also for the subsidized fertilizer package (69–30–00 kg ha−1 N–P–K) an alternative which included 25 kg ha−1 K was created based on INIFAP recommendationsReference Gómez, González, Manjarrez, Murillo and Cruzaley 28 . Finally, a system with a fertilizer rate of 135–39–83 (kg ha−1 N–P–K) was included corresponding to the agronomic recommendation (R) for maize–roselle systemsReference Navarro, Cruzaley, Reyes, Noriega and Miranda 29 . Other soil fertility strategies were based on experimental trials (E), which included mineral fertilizers at a rate of 55–5–46, vermicompost at a rate of 2.5 Mg DM ha−1, equivalent to 23–6–20 (kg ha−1 N–P–K), and a combination of both equivalent to 78–11–66 (kg ha−1 N–P–K).
Maize production levels for each cropping system and each farm were calculated in an input-oriented manner using the model QUEFTSReference Janssen, Guiking, Van der Eijk, Smaling, Wolf and van Reuler 16 , which were evaluated for the region by Flores-Sanchez et al.Reference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 . The model uses soil chemical properties as inputs, including organic carbon (g kg−1) assuming 58% C in SOM, total N (g kg−1), P-Bray-1 (mg kg−1) (B.H. Janssen; personal communication), K-exchangeable (cmol kg−1), pH (H2O) and cropping system-specific rates of fertilizer. The model first calculates crop uptake rates of N, P and K based on the potential supply by the soil, the applied amounts of fertilizer and an estimated nutrient recovery of applied nutrients. Next, three intermediate yield estimates are made, one for each of the nutrient pairs based on the uptake of N, P and K, taking into account for each nutrient values for maximum accumulation (i.e., the nutrient is not yield-limiting) and maximum dilution (i.e., the nutrient is yield-limiting). In the final step, yield is predicted based on the smaller of the three yield estimates.
Weed management was assumed to be conventional with herbicide applications at maize sowing and 3 weeks later, resulting in the same biomass of weeds as found for current practicesReference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 . In all land-use activities, roselle yield was assumed to be similar to that found currently on the farms, as roselle was found to show little response to different rates of fertilizer (Flores-Sanchez et al., unpublished data).
A number of fertilization strategies included canavalia as cover crop (Table 4), which was assumed to be sown 4 weeks after sowing maize. Experimental results did not reveal direct effects of canavalia on maize grain yield, but did demonstrate a substantial reduction of weed biomassReference Flores-Sanchez, Pastor, Lantinga, Rossing and Kropff 10 . Assuming similar conditions as in the experiments for those MBCS that included canavalia, weed biomass was reduced by 66% compared to current practices. Except for those strategies in which fields were assumed to be fenced, resulting in 100% residue retention, 70% of the biomass of crop residues, weeds and canavalia was assumed to be removed by roaming animals (on farms without animals) or fed to the own farm animalsReference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 . Input from N2 fixation by canavalia was set at 6 kg ha−1 yr−1 in a maize–roselle intercrop, and 16 kg ha−1 yr−1 in a maize monocrop, similar to field estimates (Flores-Sanchez et al., unpublished data).
Quantification of labor input was based on current labor use observed on each of the case-study farms. A fixed amount (2 labor-days=16 h) of additional labor was added to account for the time needed to cover the fertilizer and compost after application to the plant base. This technique was assumed to be a ‘best technical’ meansReference van Ittersum and Rabbinge 30 to maximize use efficiency by avoiding washing off. Labor was hired outside the farm to deal with labor peaks; the remainder was supplied by the farmer and his family (Table 5). We denote the former as casual labor and the latter as regular labor.
Table 5. Regular and casual labor (h ha−1) for each MBCS variant and for each farm.
1 Variants as described by their labels (Table 3).
For each cropping system production costs were estimated based on quantities and prices of inputs: mineral and organic fertilizers, herbicides, seeds and labor. Cost of inputs per MBCS and variation among farms are presented in Table 6. Prices of crop products (maize grains and roselle calyces) and animal products (meat) were obtained from databases 31 , 32 using data of 2003. Both family and hired labor were valued at 50 MX$ h−1.
Table 6. Average costs per ha (in Mexican pesos (MX$) for each maize-based cropping stem. Ranges are due to differences in amounts used per farm. For abbreviations of the MBCS see Table 4.
1 Cost of canavalia seed. Farmers used own or exchanged seeds of maize and roselle.
Animal production systems
The design of animal production activities concentrated on goats and cattle, as numbers of horses, donkeys and mules were limited to one or two animals per farm. No detailed information was available on management systems, but based on farmer interviews and experiences in the area we assumed on-farm feeding for 2–3 months, depending on available feed resources, and roaming outside the farm during the rest of the year. This had implications for nutrient cycles as roaming was assumed not to contribute to the on-farm nutrient or OM cycles, while on-farm feeding allowed manure to be collected.
We distinguished cows, heifers, calves and goats with body weights of 450, 300, 170 and 75 kg, respectively. Marketable outputs comprised meat of culled cows and goats. Calves stayed with the mother and used her milk. Non-marketable outputs comprised changes in the SOM balance resulting from application and decomposition of manure during the time the animals were on the farm. Animal feed stuffs comprised grain, straw and weeds which were produced within the farm. Feed was characterized by three feed value indicators: dry matter (g kg−1), metabolizable energy (Mcal kg−1) and crude protein content (g kg−1), which were based on published and regional sources 33 – Reference Douxchamps 35 (Cortez-Arriola, personal communication). The amount of labor needed for herd and stable management and for general farm management was estimated based on Cortez-Arriola et al. (unpublished data) who collected data on smallholders in West-Michoacán. Parameters such as carcass%, milk protein and fat content, and energy and protein requirements were taken from the literature 33 , 34 , Reference Martínez, Palacios, Bustamante, Ríos, Vega and Montaño 36 .
Farm-level analyses: performance of current and possible future farming systems
The performance of current and possible future farming systems was evaluated in terms of family income, SOM balance and input of regular (own) labor using the FarmDESIGN modelReference Groot, Oomen and Rossing 17 . The model calculates transfers of dry and OM, C, N, P and K between the farm compartments crops, animals, manure and soil, all based on production ecological relations. Imports, e.g., through fertilizers and exports, e.g., sales or losses are taken into account. The model allows characterization of the current farming system, as well as exploration of future farming systems that perform better in terms of the objectives by varying areas of current and alternative MBCS and numbers of farm animals under a set of user-specified constraints, e.g., related to maximum area and feed balance deviation. Results are expressed as trade-offs among objectives.
The crop component comprised the current and alternative land-use systems, i.e., maize and roselle as monocrop, and/or maize–roselle as intercrop, as well as the land-use system products, i.e., maize grains, maize residues, roselle calyces, roselle residues and weeds. Crop products have one or several destinations: application to the soil (crop residues and cut weeds left on the field), feed for animals (crop residues and weeds), home use by the farm family (grains) and selling on the market (grains and calyces).
The animal component included goats and cows. Feed balances and manure produced by animals were calculated for the part of the dry season that the animals were around the homestead. The duration of this on-farm feeding phase was estimated to be 100 days for farm A3, and 130 days for farms A2 and A4, based on information of the farmers and calculations with the model. In the explorations (Scenarios S1–S4, see below), it was assumed that manure produced within the farm was applied to the crops. The amount of N–P–K, in the manure of the farms was: 20–2–14 for A2, 29–2–20 for A3 and 12–1–8 for A4. These amounts were included in the calculation of grain yield per production system, using QUEFTS.
Family income (FI) represents the actual amount of money available to the farm family on an annual basis, calculated as the sum of the margins of crop and animal products minus costs of fertilizers, pesticides and casual labor. The amount of maize sold equaled the amount of maize produced minus the amount used for self-consumption. Daily per capita consumption of maize was assumed to be 0.5 kg. We compared family income with the ‘basic food basket’, a local indicator of the minimum amount of money required for food self-sufficiency, when basic needs are met through the market 37 .
The OM balance was calculated by combining four ‘sub-balances’: root residues, aboveground crop residues, manure and SOM. Balances were calculated as the difference between annual input and output. Of the maize and weed residues, 30% biomass was assumed to remain in the field where they were produced, the remainder being taken up by animals. In case of farm-owned animals, the resulting manure was assumed to stay on the farm. If the farm did not own animals, roaming animals were assumed to export the OM from the farm system. Roselle residues were assumed to be not suitable for animal consumption and remained in the field. Similarly, in MBCS with residue retention due to fencing of the fields no export, or 100% of residue retention, was assumed.
The net contribution of root and aboveground crop residues to the SOM balance was quantified as the amount of OM remaining 1 year after application in the fieldReference Groot, Oomen and Rossing 17 . Root biomass was estimated as 15% of the total crop biomassReference Rodríguez 38 . After calibration with litterbag experiments in farmers’ fields (unpublished data), a mono-component modelReference Yang and Janssen 39 was used to predict the amount of root OM remaining per field after 1 year.
Estimates of bulk density (1.3 Mg m−3) and annual rate of SOM decomposition (0.5% yr−1) were taken fromReference Grace, Jain and Harrington 40 for no-tillage conditions in long-term trials at CIMMYT, Central Mexico. This information was used together with field-specific estimates of soil depth to calculate annual SOM degradationReference Groot, Oomen and Rossing 17 .
Erosion was considered a cause of SOM and plant nutrient losses. Loss rates were calculated using RUSLE estimates of soil loss, multiplied by SOM, N, P or K fractions as established in an earlier farm diagnosisReference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 .
Balances for regular labor were calculated by subtracting farmer-provided own labor input for herd and animal management and for the various activities in each MBCS from the amount of labor available given the size of the farm family, assuming each person to provide 2190 h per annum.
Exploration of possible future farming systems started from the current farming systems (Scenario S0), which for three farms included animals (Table 1), and then proceeded in four incremental steps (Scenarios S1–S4). Scenario S1 included the MBCS Cu(Fcr), Cu+K(Fcr), S(Fcr), S+K(Fcr), E(Fcr), E(FCr) and E(FCR) in which plant nutrient provision was improved compared to the current system (Cu(Fcr), Table 4) by relying on imported fertilizers. Canavalia and crop residue retention were included in two MBCS: E(FCr) and E(FCR). In Scenario S2, the set of MBCS was further expanded by those designed to enhance SOM balances (E(Vcr), E(F-Vcr), E(FCr), E(VCr), E(F-VCr), E(FCR), E(VCR) and E(F-VCR); see Table 4). In this scenario, six MBCS included canavalia, in three of which residue was retained by fencing. Scenarios S1 and S2 thus assumed fine-tuning of current cropping system management by redressing nutrient and SOM imbalances without substantially changing production per unit area. Scenarios S3 and S4 assumed incremental intensification of production by allowing the FarmDESIGN model to select R(Fcr), the MBCS with the highest fertilizer inputs (in Scenario S3) and associated yield, and to select a herd of goats and/or cows that was fed on the farm during 120 days (in Scenario S4). In all scenarios, canavalia was included as a monocrop, to evaluate its potential to compete with other land-use activities to restore soil functions. Scenario S4 was not run for farms X2 and X4, as their size (1 ha) was assumed to be prohibitive for providing on-farm feed for animals during 120 days. On the farms that owned animals in Scenario S0, the number of animals was kept constant in Scenarios S1–S3, and optimized in S4.
The explorations were performed in FarmDESIGN using a genetic algorithm and a fitness function based on Pareto-based ranking and crowing metricsReference Groot, Oomen and Rossing 17 . Three objectives were addressed simultaneously: maximization of family income, maximization of the SOM balance and minimization of the regular labor balance. For Scenarios S1–S3, the decision variables included cultivated areas of MBCS. When a farm cultivated roselle either in mixture with maize or as monocrop, these variants were included as well, thus assessing the strength of roselle to compete with other land use activities. For Scenario S4 number of milking cows, their replacement rate, and number of goats were added as decision variables. Cultivated area was set to a maximum of the whole farm area and included as constraint in S1–S3. In S4, additional constraints were included on feed intake dry matter (g kg−1), metabolizable energy (Mcal kg−1) and crude protein (g kg−1) to ensure reasonable feeding patterns.
The results are described in two steps. In the first step, each MBCS is assumed to be deployed on the whole farm and consequences are assessed in terms of family income, SOM balance and regular labor balance. This analysis highlights the variation in performance among MBCS per farm and among farms. In the second step, the trade-offs among the objectives for the four scenarios are summarized by triangles which connect current system performance with best performances for each of the objectives. By projecting these in two dimensions, triangles are obtained, which show the amount of improvement possible compared to the current farming systems and the severity of the trade-offs among the objectives.
Results
Single MBCS deployed on the whole farm
Compared to the current farm-specific systems, the redesigned MBCS resulted in yields that were up to a factor 2 greater (Table 7). Largest yields were mostly associated with the agronomic recommendation R(Fcr), which resulted in yields at or above 3.5 Mg ha−1. Farms responded differently to a particular MBCS, reflecting the current differences in soil fertility status of their individual fields (Table 2). Applying the cropping systems to the entire farm area showed that six out of eight farms remained below the ‘basic food basket’ (Fig. 1) indicating that livelihoods would need to rely on barter, remittances or on income from hiring out labor. We define the family income gap as the difference between family income from a maize production technology and the basic food basket. The two farms that achieved family income close to or slightly above the basic food basket included animals (farms A2 and A4). These were also the farms with the largest per capita land area (PCLA). In contrast, farm A3, which had the largest number of animals (14 cows and 14 goats, Table 1) attained less family income and suffered a larger family income gap than the other two animal farms. The gap was even larger than that of farms X1 and X3, which did not include cows or goats while having a slightly smaller PCLA. This suggests that animal husbandry on the land-limited farm A3 did not provide advantages over a purely crop-based strategy.
Figure 1. Family income for eight smallholder farms in the Costa Chica, Mexico, for different maize-based cropping systems (MBCSs). Gray bars: family income from agriculture. White bars: family income from the PROCAMPO subsidy (MX$1030 ha−1). The horizontal line represents the basic food basket 33 for the particular farm family. For details on the MBCSs see Table 4.
Table 7. Range of field-specific maize grain yield (kg DM ha−1) for different maize-based cropping systems (MBCSs) on eight farms in the communities Las Animas (A1–A4) and Xalpathlahuac (X1–X4). Farms A2, A3 and A4 were assumed to also apply on-farm produced manure. Inclusion of canavalia or residue retention did not affect maize yield and is denoted as xx in the MBCS identifiers. For explanation of MBCS see Table 4.
1 Farms A1 and X2 have only one field.
The subsidies provided through the PROCAMPO program of the Mexican government are linked to farmed hectares and therefore differed among farms. The contribution from PROCAMPO to total family income varied from MX$ 1030 to 4300 (white bars in Fig. 1).
Assuming uptake of a particular MBCS on the entire area of each of the eight farms, the trade-offs among family income, SOM balance and labor input are shown in Fig. 2. Many systems are better than the current system (system A in Fig. 2) in both family income and SOM balance. Values of family income less than the current are obtained when vermicompost is included without additional fertilizer (systems G, J, M) as additional costs of vermicompost purchase and transportation are not fully compensated by yield increases. Cropping systems E, H, K and N are on the trade-off frontier (Figs. 2A and 2D) as they contribute most positively to SOM balance and family income. The four systems all rely on fertilizer input, albeit in different amounts (cf. Table 4). Cropping systems H, K and N are associated with 78–11–66 kg ha−1 N–P–K input, partly in vermicompost. Cropping system E is associated with 135–39–83 kg ha−1 N–P–K input, but does not include an organic source of nutrients. Residue retention and inclusion of canavalia contribute positively to the SOM balance; their effect is about 2/3 of the effect of vermicompost when compared to the fertilizer-only cropping system E (Figs. 2A and 2B). System E contributes more than the current system to the SOM balance due to its larger biomass production, which partly stays in the field. Improvements in SOM balance demand more labor (Figs. 2B and 2E). On farms without animals (farms A1, X1, X2, X3 and X4) increasing crop residue retention substantially increased SOM balances (systems L, M and N), but also demanded more labor. The alternative MBCS demanded relatively more labor; however, labor balance showed that there was a surplus of family labor, but this excess of labor can be linked to the fact that farming activities comprised only between 6 and 7 months, but family income was improved (Figs. 2C and 2F).
Figure 2. Trade-offs among family income, regular labor requirements and organic matter balance assuming that eight smallholder farms dedicate their entire cropping area to a single maize-based cropping system (MBCS). Results are expressed as averages across farms. Bars indicate standard errors. Panels on the left concern farms with animals, farms on the right without animals. MBCS codes: A: Cu(Fcr), B: Cu+K(Fcr), C: S(Fcr), D: S+K(Fcr), E: R(Fcr), F: E(Fcr), G: E(Vcr), H: E(F-Vcr), I: E(FCr), J: E(VCr), K: E(F-VCr), L: E(FCR); M: E(VCR), N: E(F-VCR). For details on the MBCSs described in the legend see Table 4.
Explorations based on the four scenarios
Explorations for the eight farms were conducted for Scenarios S1–S4. From the current system (S0) to Scenario S3, the number of possible cropping systems increased sequentially, and FarmDESIGN was used to find combinations of areas that optimized farm performance for the three objectives maximize family income, minimize own labor and maximize SOM balance, simultaneously. Scenario S4 comprised the set of cropping systems of S3 plus animal husbandry on those farms that did not have animals to start with. Decision variables comprised hectares of cropping activities and, for S4, number of goats and cows and their replacement rate. The results demonstrate that improvements are feasible for both for family income, SOM balance and required family labor as illustrated for farm A3 in Scenario 4 (Fig. 3). The relation between labor requirement and the other two objectives was similar for all farms: relatively small differences between best and worst values of labor requirement. As a result, in the rest of this section, we concentrate on the trade-off between the other two objectives, which was, however, calculated including the labor balance objective. The results are shown in Fig. 4 and Table 8 for each farm and for each of the four scenarios.
Figure 3. Illustration of the trade-offs among family income, regular labor input and organic matter balance on farm A3 for explorations based on Scenario 4. The open circle corresponds to current maize-based cropping system. Closed circles represent alternative farming systems. The triangle in the top panel illustrates the procedure used to arrive at Fig. 4.
Figure 4. Relationship between family income and organic matter balance obtained by exploration of four scenarios for eight farms. The horizontal dotted line represents the basic food basket 33 for the farm family.
Table 8. Major MBCS and numbers of animals selected to maximize family income or organic matter (SOM) balance under four intensification scenarios. For explanation of abbreviations of the MBCS see Table 4.
1 Value between parentheses refers to the number of ha.
2 Value between parentheses refers to the percentage of land that occupied each maize-based cropping system (MBCS).
For Scenario S1, maximum family income was mostly associated with the current fertilizer strategy plus K (Table 8). On two farms, MBCS based on subsidized rates of fertilizer were selected, and on the farm with the largest number of animals (A3) canavalia was grown on 40% of the area. To maximize SOM balance, on some farms MBCS with current or subsidized rates of fertilizer rates dominated, always with a supplementation of K. On other farms moderate fertilizer rates were combined with canavalia and residue retention (E(FCR)). For Scenario S2, the inclusion of vermicompost was selected among the options that maximized SOM balances. To maximize family income, the land-use systems selected were similar to those for S1. For Scenario S3 the R(Fcr) option that included the largest fertilizer rates, resulting in the largest maize biomass and yield, was selected on all farms to maximize family income. Large biomass and hence residue production by R(Fcr) also made it the option that maximized SOM balance on two of the eight farms. On the other farms maximizing SOM balance required MBCS relying on canavalia and/or vermicompost. Optimization of number of animals in Scenario S4 maintained R(Fcr) as the best MBCS to maximize family income. Only on half of the farms animal husbandry was a means to increase family income. However, animal husbandry together with the application of vermicompost with or without canavalia (E(F-VCr) and E(F-Vcr)) were important to maximize SOM balances (Table 8).
Trade-off triangles were constructed for the objectives family income and SOM balance by linearly connecting the current farming system with those that exhibited best performance in each of the objectives, as illustrated in Fig. 3. The trade-off triangles (Fig. 4) show that as the number and type of land-use options increases from Scenario S0 to S4, the trade-off frontier shifts outward. Both family income and SOM balance can be improved, although with large differences among the farms, as revealed by the size of the triangles. The triangles are not congruent, indicating that the trade-offs between the objectives change when progressing through the scenarios. For some scenarios and farms narrow triangles were found (e.g., farm A1, Scenario S3; farm X4, Scenario S3) indicating that trade-offs were replaced by (a few) optimal solution(s).
Discussion
This study set out to investigate options for improving socio-economic performance and resource use of smallholder livelihoods in the Costa Chica, by bringing together information on alternative MBCSs and animal husbandry in a context of actual farms.
The study addressed socio-economic performance in terms of family income and use of regular labor, and evaluated resource use in terms of changes in the SOM balance.
Opportunities for improving family income, SOM balance and labor balance
Results for the eight farms from two communities representative of the Costa ChicaReference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 showed the need to increase family income, which on most farms was found to be considerably smaller than the minimum needed to sustain the family members, i.e., the basic food basket. Results also showed that considerable improvements were possible by intensification of maize production (Figs. 2 and 4). Roselle is one of the main sources of income; however, its cultivation demands high labor input, making it an expensive crop to grow. Therefore, improvements in maize production appeared to be more attractive.
Whether increases to the level of the basic food basket were feasible was dependent on PCLA. For six out of eight farms the PCLA was too small even at the highest farm productivities calculated in Scenario S4. Only farms A2 and A4 (PCLAs 0.8 and 1.05 ha, respectively (Table 1)) reached family income above the livelihood threshold for the current farm systems. Calculations for Scenario S4 showed that farm X1 (PCLA 0.42 ha) also could attain the threshold and farm X3 (PCLA 0.44 ha) could come close to it, but with trade-offs with the OM balance. Worldwide, PCLA was 0.21 ha in 2007 41 .
Current maize systems on all farms were associated with negative annual SOM balances, whilst SOM levels in the area are already generally low. We found SOM levels of 5–28 g kg−1. Similarly, values from 3 to 52 g kg−1 were reported for the same regionReference Navarro, Cruzaley, Reyes, Noriega and Miranda 29 ; others found a range from 10 to 46 g kg−1 18 . This high level of soil degradation affects the potential supply of N, P and KReference Janssen, Guiking, Van der Eijk, Smaling, Wolf and van Reuler 16 , Reference Mulder 42 . As a result, the attainable yield predictions with QUEFTS varied greatly—up to 100%—across fields at similar rates of fertilizer input (Table 7). To increase long-term productivity and input use efficiency on these sandy soils, addition of OM sources should be an integral component of any restoration strategy or policyReference Mann, Tolbert and Cushman 43 – Reference Chivenge, Murwira, Giller, Mapfumo and Six 45 . In such strategies, the existing variation among fields should be taken into account, e.g., by allocating greater amounts of organic inputs to fields that are more degraded. In the explorations increases in annual SOM balances were achieved in various ways. Firstly, producing a larger amount of biomass by increasing inorganic fertilizer application resulted in greater yields as well as larger amounts of residues. The results of Scenario S3 compared to S2 indicated that the net effect on annual SOM balance was usually limited as 70% of the residues were assumed to be removed by roaming animals (Fig. 4). Secondly, use of vermicompost contributed directly to the SOM balance, but required important labor and monetary expenditure (Tables 5 and 6). Finally, residue retention by fencing fields and use of canavalia as cover crop contributed positively to SOM balances, while requiring less input than vermicompost. The higher costs of maintaining OM inputs caused the trade-off between annual SOM balance and family income (Fig. 4).
Calculations for Scenario S4 showed that increasing animal numbers always served to maintain greater balances of SOM. However, maximum family income strategies did not always include animals, indicating the relatively important direct costs for labor and feed associated with them.
Labor requirements increased roughly linearly with family income. Even though a large number of family members were available for working on the farm, according to the farmers casual labor was hired to deal with peaks in field work. A more in-depth analysis is needed to reveal whether this does not concern reciprocal labor, where families work on each other's farms and sow, weed and harvest together based on monetary remuneration. If that is the case, the fraction of family labor in the total labor requirement would be used to a much greater degree than the on average 31% we found.
Land-use scenarios for the short and the middle term
The land-use Scenarios S1 and S2 were set up to reveal the importance of redressing imbalances in current nutrient and OM application within the opportunities offered by government subsidy schemes for fertilizers (to a maximum of 2 ha) and by local vermicomposting facilities. For family income maximization in Scenarios S1 and S2 fertilizer rates always included K application, indicating the farm economic benefit of a more balanced but also more costly crop nutrition. According to these results, purchasing K at market rates in addition to purchasing subsidized N and P fertilizer would be beneficial to family income (Table 8).
Compared to Scenario S1 maximization of SOM balance in S2 relied strongly on vermicompost. Vermicompost production and transport costs were taken into account, assuming a maximum application rate of 2.5 ton ha−1. At this rate, total regional vermicomposting capacity may be insufficient to produce the amounts needed to provide for the entire 14,000 ha of agricultural area. For instance, the composting facility in Tecoanapa produced 35 ton in 2006, 1 per mil of the total regional requirement. Government support would thus need to address not only fertilizer purchasing subsidies but also local initiatives to recycle household waste and produce vermicompost.
For Scenario S1 and particularly for S2 the land-use activities for family income and SOM balance are quite distinct. To strike a balance between the two objectives thus requires a mix of fertilizer application that includes K and inputs of external OM, the costs of which fit within the financial constraints of individual farms.
Scenarios S3 and S4 represented more drastic changes, including high fertilizer inputs largely without subsidies and the option to have goats or cattle. Under the S3 scenario, family income was maximized for all farms by applying the high fertilizer MBCS on a substantial fraction of the farm area. However, this MBCS did not prove the best for increasing SOM balance. For that purpose MBCS were selected that included vermicompost, canavalia and residue retention through fencing (Table 8). Current subsidies for purchasing fertilizers provided by the State and municipal Governments amount to 600 million Mexican pesos and constitute the major agronomic support instrument. This policy has been criticized for its unilateral focus on fertilizers at the expense of development of human and other rural support resourcesReference Mendez 46 . A major challenge for policy will be to balance support for short-term gains in yields through fertilizer subsidies and support for long-term benefits from soil improvement.
In Scenario S4, animals were excluded from the maximum family income solutions on those farms where feeding the animals (partly with maize grains) and selling the meat was less profitable than selling maize. The fact that on some farms animals were selected, and on others, none, indicates the delicate farm-specific balance between costs and returns. Goats were never selected as part of the optimal systems, although farms in the region often have considerable numbers of goats. It is known that animals in smallholder systems constitute a source of savings for subsistence needs. Animals can be kept as an insurance against eventualities, and provide an instrument of liquidity and consumption smoothingReference McDermott, Randolph and Staal 47 – Reference Thornton 49 . On the farms that had animals to start with (A2, A3 and A4) the contribution of the animal component to family income was only 16%, emphasizing that it is not the immediate cash contribution that makes animal husbandry important for smallholders. Similar to our findings, a study on smallholders in different countries demonstrated that income from livestock was 12% on averageReference Pica-Ciamarra, Tasciotti, Otte and Zezza 50 . To avoid undesired side-effects of policies that support or not animal husbandry, more information is needed to understand the trade-offs that farmers strike between costs of maintaining the animals and benefits provided by animals.
On all farms, cattle became part of the optimal system when maximizing SOM balances, due to enhanced recycling of residues on the farm rather than exporting residues with roaming cattle. As a result, farms with cattle had positive SOM balances. Manure is a valuable resource for improving SOM balances and sustaining crop productionReference Randolph, Schelling, Grace, Nicholson, Leroy, Cole, Demment, Omore, Zinsstag and Ruel 48 , Reference Pica-Ciamarra, Tasciotti, Otte and Zezza 50 , Reference McDermott, Staal, Freeman, Herrero and van de Steeg 51 . We assumed that all manure produced during the on-farm period was applied on the farm fields assuming N losses during storage of approximately 30%Reference Groot, Oomen and Rossing 17 . Better loss estimates requires information on the ways in which farmers collect, store and apply the manureReference Rufino, Rowe, Delve and Giller 52 .
Policy implications
For subsistence farmers in Mexico, maize accounts for 70% of calories and 60% of proteinsReference Hellin, Groenewald and Keleman 53 . All eight farms were able to meet the level of food self-sufficiency, and family income was increased. However, family income from farming was insufficient to meet basic family needs, even after re-design of the farms, for five out of eight farms and the only option open to these families is to complement their farm income with off-farm employmentReference García-Barrios and García-Barrios 54 , Reference Hellin 55 . These results took into account the federal support by the programs PROCAMPO (Program for Assistance in Agriculture) and Oportunidades (a program to alleviate poverty), and the fertilizer subsidy program from the state of Guerrero. These results are consistent with reports about the persistence of poverty and the high degree of marginalization in the region 3 .
The results suggest that the technological options that are currently available may be insufficient to enable farm families to meet the basic food basket. Diversification of options both on- and off-farm is needed to allow farmers to select activities that are suitable to their constraints and objectives, and policies aimed at regional development are needed. A key element of such rural policies should be agricultural extension and training of smallholders to understand the agro-ecological processes they manage and to promote land-use activities that both provide increased returns in the short run as well as rendering the system resilient to changes in prices of inputs and to increased variability in weather as predicted for MesoamericaReference Pachauri and Reisinger 56 . Such ecological intensification can improve the resource base and the living standard of smallholdersReference McDermott, Staal, Freeman, Herrero and van de Steeg 51 , Reference Hellin, Groenewald and Keleman 53 . Recently, the Federal Government and CIMMYT announced the implementation of Modernización Sustentable de la Agricultura Tradicional (MASAGRO, the Sustainable Modernization of Traditional Agriculture) targeted at smallholders. The program is aimed at increasing maize production in rainfed areas through improving agronomic practices, and the use of improved maize varietiesReference Hellin 55 , Reference González-Rojas, García-Salazar, Matus-Gardea and Martínez-Saldaña 57 . It will be important to take into account that maize landraces, or criollos, are preferred over hybrids by the rural families due to better taste, ease of shelling cobs, time needed for cooking, better quality of tortillas, storage time and so onReference Hellin 55 . A key issue to be taken into account is the restoration of the resource base given the high degree of soil degradation in smallholder systemsReference Flores-Sanchez, Kleine Koerkamp-Rabelista, Navarro-Garza, Lantinga, Groot, Kropff and Rossing 7 , 58 .
Conclusions
Model-based explorations for eight real farms in the Costa Chica, Mexico, demonstrated that farm family income can be increased and SOM balances enhanced, without drastically changing labor input. Labor input will need to be increased proportionally with family income. Variation in response to fertilizer-based cropping strategies among the eight farms highlighted the need for farm- and field-specific nutrient management strategies. Notwithstanding the progress possible, most of the farms did not reach a minimum living standard as specified by the basic food basket, due to low productivity in combination with low per capita land availability. Policies to support smallholders should therefore take a multifaceted regional development approach that also develops off-farm economic options, rather than focusing on fertilizer subsidies as is currently the case. Policy support for the regeneration of the degraded soils through OM-based technologies will be necessary as short-term benefits favor purely fertilizer-based land-use systems. Such support should address logistics as well as development of farmer and scientific knowledge.
