- Email: [email protected]
Identifying secure and low carbon food production practices: A case study in Kenya and Ethiopia
Agriculture, Ecosystems and Environment 197 (2014) 137–146
Contents lists available at ScienceDirect
Agriculture, Ecosystems and Environment journal homepage: www.elsevier.com/locate/agee
Identifying secure and low carbon food production practices: A case study in Kenya and Ethiopia Jessica Bellarby a,b, *, Clare Stirling c , Sylvia Helga Vetter a , Menale Kassie d, Fred Kanampiu d, Kai Sonder e, Pete Smith a , Jon Hillier a a
Scottish Food Security Alliance-Crops & Institute of Biological and Environmental Sciences, University of Aberdeen, Aberdeen AB24 3UU, UK Lancaster Environment Centre, Lancaster University, Bailrigg, Lancaster LA1 4YQ, UK c International Maize and Wheat Improvement Center (CIMMYT), Ynys Mon, Wales, UK d International Maize and Wheat Improvement Center (CIMMYT), Nairobi,Kenya e International Maize and Wheat Improvement Center (CIMMYT), Mexico, D.F., Mexico b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 11 October 2013 Received in revised form 24 July 2014 Accepted 25 July 2014 Available online xxx
The world population is projected to increase to 9–10 billion by 2050, during which time it will be necessary to reduce anthropogenic greenhouse gas emissions to mitigate climate change. The particular challenge this places on agriculture is to identify practices which ensure stable and productive food supply that also have a low greenhouse gas (GHG) intensity. Maize is the principle staple crop in many parts of Africa with low and variable yields, averaging only 1.6 t/ha in sub-Saharan Africa (SSA). Food security and increasing crop yields are considered priorities in SSA over impacts of food production on GHG emissions. Here we describe an approach that can be used to inform a decision support tree for optimal interventions to obtain sufficient food production with low GHG intensity, and we demonstrate its applicability to SSA. We employed a derivative of the farm greenhouse gas calculator ‘Cool Farm Tool’ (CFT) on a large survey of Kenyan and Ethiopian smallholder maize-based systems in an assessment of GHG intensity. It was observed that GHG emissions are strongly correlated with nitrogen (N) input. Based on the relationship between yield and GHG emissions established in this study, a yield of 0.7 t/ha incurs the same emissions as those incurred for maize from newly exploited land for maize in the region. Thus, yields of at least 0.7 t/ha should be ensured to achieve GHG intensities lower than those for exploiting new land for production. Depending on family size, the maize yield required to support the average consumption of maize per household in these regions was determined to be between 0.3 and 2.0 t/ha, so that the desirable yield can be even higher from a food security perspective. Based on the response of the observed yield to increasing N application levels, average optimum N input levels were determined as 60 and 120 kg N/ha for Kenya and Ethiopia, respectively. Nitrogen balance calculations could be applied to other countries or scaled down to districts to quantify the trade-offs, and to optimise crop productivity and GHG emissions. Crown Copyright ã 2014 Published by Elsevier B.V. All rights reserved.
Keywords: Greenhouse gas emissions Sub-Saharan Africa Smallholder farming system Maize Food security
1. Introduction The world population is projected to increase to 9–10 billion by 2050 (Godfray et al., 2010), during which time it will be necessary to reduce anthropogenic greenhouse gas emissions to mitigate climate change (Smith et al., 2008). The particular challenge this places on agriculture is to identify agricultural
* Corresponding author at: Lancaster Environment Centre, Lancaster University, Bailrigg, Lancaster LA1 4YQ, UK. Tel.: +44 1224 273810. E-mail address: [email protected] (J. Bellarby). http://dx.doi.org/10.1016/j.agee.2014.07.015 0167-8809/ Crown Copyright ã 2014 Published by Elsevier B.V. All rights reserved.
practices which ensure stable and productive food supply combined with food of low greenhouse gas intensity (Garnett et al., 2013; Godfray et al., 2010; Mueller et al., 2012; Smith et al., 2013; Tilman et al., 2011). In the developed world, access to water, nutrient resources (e.g. fertilisers), and crop protection chemicals have raised yields since 1965 (Foley et al., 2011). In contrast, yield increases in the developing world have been limited, so that there is substantial potential to increase food production (Mueller et al., 2012). Mueller et al. (2012) identify sub-Saharan Africa as one of the world regions showing considerable ‘low-hanging’ intensification opportunities for major cereals, stating also that closing maize
138
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
yield gaps to 50% of attainable yields (defined as the area-weighted 95th percentile of observed yield within zones of a similar climate) primarily requires addressing nutrient deficiencies, but that additional gains require both an increase in irrigated area and nutrient application over most of the region (Mueller et al., 2012). Tittonell and Giller (2012) discuss the need to increase soil fertility in sub-Saharan Africa and develop options to achieve this. The potential for yield improvement has also been demonstrated by the national subsidy programme for fertiliser and improved seeds in Malawi, which resulted in a doubling of national yields (Denning et al., 2009; Hickman et al., 2011). However, this comes at a cost by also increasing GHG emissions. During the 2000s, agriculture contributed between 10 and 12% to global anthropogenic greenhouse gas (GHG) emissions, consisting mainly of nitrous oxide and methane (Smith et al., 2008). These are projected to increase further, with one of the highest growth rates in emissions predicted in sub-Saharan Africa, due to increased fertiliser input and livestock numbers (Reay et al., 2012; Smith et al., 2007). Therefore, food supply needs to be improved whilst minimising the environmental impact (Garnett et al., 2013), which is a strategy that has been termed ‘sustainable intensification’. There has been much debate about the precise definition of this term (Tilman et al., 2011; Smith et al., 2013). In order to make this term practically usable, quantifiable measures – such as GHG emissions per unit of production – are needed for evidence-based assessment of practices. Although in many developed countries, improvement in this measure can be achieved with modest increases in production in conjunction with reduced emissions – through for example precision agriculture – in developing countries the focus and benefits will be more on increasing the efficiency of production through good agronomic practices (e.g. Mueller et al., 2012; Van Groenigen et al., 2010). Total GHG emissions from croplands are a sum of background emissions from the field (e.g. soil gaseous emissions from the breakdown of residual soil organic matter), and emissions from the production and application of pesticides, synthetic and organic fertilisers, and from the management of residues. Field measurements of GHG emissions are often expressed on a per hectare basis and may not report observed yield (Linquist et al., 2012). GHG emissions per hectare have a strong relationship to the amount of fertiliser N supplied, and are usually linked to yield (Hillier et al., 2009; Van Groenigen et al., 2010). The relationship between GHG emissions and agronomic productivity can be illustrated by calculating the yield-scaled emissions (Linquist et al., 2012; Mapanda et al., 2011; Mosier et al., 2006; Van Groenigen et al., 2010). This enables the identification of best management options in terms of GHG emissions as a function of N input (Van Groenigen et al., 2010) in addition to other variables. Current estimates of GHG emissions for African regions are mainly based on the IPCC Tier 1 emission factor approaches, which are often considered to be too simplistic for informative site-level assessments (Hickman et al., 2011). In contrast, the use of more complex models is hampered by a lack of data. These include soil and weather data used as inputs, but also the dearth of GHG emissions measurements available for model validation (Hickman et al., 2011). The capacity to assess the impact of practices on GHG emissions on working farms is relatively recent. Although for example, Smith et al. (2008) list many mitigation options for agriculture together with an estimate of the per hectare GHG impact, until relatively recently decision support capacity has been lacking at the site level (Hillier et al., 2011a). This has led to the development of several “GHG calculators”, some examples of which were recently reviewed (Whittaker et al., 2013). These models enable the analysis of typically collated farm survey data and so can capture mitigation options at a scale, and with a degree
of representativeness, not possible with models that are more data demanding. Such tools are generally not sensitive to seasonal climatic variation or subtle variation in soil properties. However, some allow an assessment of GHG emissions as a function of management practice, and enable the user to examine and optimise different management options (Hillier et al., 2011b) at the desired scale (e.g. Hillier et al., 2012). In this study we used a derivative of one such model – the Cool Farm Tool (CFT; Hillier et al., 2011b) – to estimate GHG emissions of smallholder maize in Kenya and Ethiopia to assess best practice. Kenya and Ethiopia represent countries in SSA where maize is the dominant staple crop in both countries with smallholder farmers occupying about 75% of the total maize land area (CSA, 2013; Kang’ethe, 2011). Kenyan smallholders usually have complex intercropping regimes with two or three different crops grown in a single plot, whereas maize is mainly planted on its own as a sole crop in rotation with other crops by smallholders in Ethiopia. In this study, we have restricted the analysis to mono-cropped maize to avoid issues of allocation of emissions to different crops on the same plot. Both countries are representative of maize-based smallholder farming systems in SSA. According to FAOSTAT, Ethiopia is a country almost self-sufficient in maize (FAOSTAT, 2014). An integrated livestock system is part of Ethiopian culture, reflected in it having the highest livestock population in Africa. In contrast, Kenya produces only 60% of the maize required for its own needs (FAOSTAT, 2014). Both countries are characterized by erratic rainfall and hence are prone to drought (Mulwa et al., 2013; Muricho et al., 2012). The CFT was chosen partly because it does not exclusively adopt a Tier 1 emissions factor approach. As such it can produce site and technology sensitive estimates of GHG emissions as a function of management practice, but at the same time, it is relatively datalight in comparison to many process-based models. We hypothesise that farm survey data and a version of the CFT are sufficient to identify optimum levels of N input (the key driver of GHG emissions in croplands) based on currently observed attainable yields (defined as the 95th percentile of surveyed observed yield), for sustainable food production in sub-Saharan Africa. Datasets from Kenya and Ethiopia are used as representative of smallholder maize growing systems in the region. We further show that the method is generalizable, and scalable for screening approaches to identify best practice to guide sustainable intensification on a regional basis. 2. Materials and methods 2.1. Data Data were derived from household surveys conducted by the Kenyan Agricultural Research Institute (KARI) and the Ethiopian Institute of Agriculture Research (EIAR) in collaboration with the International Maize and Wheat Improvement Centre (CIMMYT) as part of the ‘Sustainable Intensification of Maize–Legume Systems for Food Security in Eastern and Southern Africa’ (SIMLESA) programme. In this survey, 613 and 896 households were interviewed on a multitude of farm and general social aspects in 5 Kenyan, and 9 Ethiopian districts, respectively (Mulwa et al., 2013; Muricho et al., 2012) (Fig. 1). The survey also included farmer experiences of stresses such as “waterlogging”. This study makes use of 376 mono-cropped plots from 187 households in Kenya, and 1433 mono-cropped plots from 765 households in Ethiopia. 2.1.1. Calculation of GHG emissions GHG emissions on smallholder farms were calculated using a derivative of the CFT. Particular features of the tool are that it has a
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
global scope, is transparent, and open source, and has been piloted on numerous systems in many different regions and products (Cool Farm Tool Institute, 2012; Hillier et al., 2011b; Whittaker et al., 2013). In this study, the data availability only allowed the use of components related to site characteristics and crop management. These have previously been scripted in Matlab (version 7.14, http://www.mathworks.com) and were available to allow the throughput of high sample numbers. The GHG emissions derived from soil and crop management include:
139
2.2. Data input requirement and matching with the smallholder database The inputs as supplied from the survey data, including the assumptions and generalizations that had to be made to match the available information to the inputs required by the CFT, are described here. There are two types of input: 1) Categorical: data fall within a certain range and are assigned to
one group (Table 1). 2) Continuous: amounts are provided by the database as they
(i) emissions of fertiliser production and distribution given by
emission factors from the Ecoinvent database (Ecoinvent Centre, 2007), (ii) soil (fertiliser induced and background) emissions calculated using the multivariate empirical model of Bouwman et al. (2002) for nitrous oxide (N2O) and nitric oxide (NO) emissions, and the model of FAO/IFA (2001) for ammonia (NH3) emissions; residue emissions are included only when they are returned to the field using IPCC N2O Tier 1 emission factors, (iii) emissions from other agro-chemicals using figures from Audsley (1997). Emissions of CO2 from soil resulting from urea application or liming are also accounted for using IPCC emission factors (IPCC, 2006). All GHG emissions are reported as CO2-equivalents (CO2e) using the global warming potential over a 100 year horizon, which is 298 times higher for N2O, and 25 times for methane, than for CO2 (IPCC, 2007).
occur (amount of fertiliser and farmyard manure, proportion of residue returned to the field, area, amount harvested product and number of pesticide applications). Most continuous data were available from the SIMLESA database with the exception of the amount of residue returned to the field, which instead was calculated as a proportion of the yield. Local experience indicated that the default method applied in the CFT for estimating the residue for maize was inaccurate and was therefore estimated to be twice the harvested yield (Mapanda et al., 2011). In the categorical data, all soil information was estimated as shown in Table 1, with details of residue management provided by the SIMLESA database. 2.3. Analysis The analysis was carried out in R2.15.2 (R Core Team, 2012) using the additional package sqldf for database queries. For
Fig. 1. Map of Ethiopia and Kenya showing districts and ‘maize megaenvironments’ according to Bellon et al. (2005). Numbers in brackets are the percentage of maize cropped on its own in relation to total maize cropped as sole and inter-cropped maize. Plot data are derived from the SIMLESA (Sustainable Intensification of Maize–Legume Systems for Food Security in Eastern and Southern Africa) survey data.
140
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
analysis, outliers of total nitrogen (N) application and yield were removed. These were: – Maize: yield greater than 10 t/ha and lower than 0.02 t/ha. – N application rate greater than 200 kg N/ha.
The N application rate is the sum of total N from the synthetic fertilisers di-ammonium-phosphate (DAP) and urea, as well as the organic fertiliser, farm yard manure (FYM). The maximum yield and N application rates were decided upon by CIMMYT experts working in the region. The minimum yield threshold was determined by the average seed-rate of Kenya, which had the lower average of the two. As a result of these rules, we omitted 5% and 0.9% of the dataset as outliers in Kenya and Ethiopia, respectively. The rate of FYM application was considered high (“FYM_N high”) when more than 50% of the N applied was applied as FYM. Accordingly, those points that are not designated as such applied less that 50% of the N as FYM. The N content of aboveground and belowground maize was calculated using IPCC emission factors (IPCC, 2006), as is done in the CFT. This factor was used only for the amount of residue that was actually applied to the soil. This does not mean that all of the residue N is available to the crop. “Residue high” refers to those farmers that returned more than 50% of the residue available to the soil. The absolute amount of residue depended on the yield. The functions and significance of the different regression analyses to estimate the relationship between variables has been determined by performing multiple linear regression analysis in R as described by Townend (2002). Total GHG emissions calculated in the CFT were compared to total GHG emissions (including emissions from land use change) resulting from extensification based on Palm et al. (2010) in order to determine the yield threshold below which GHG emissions per tonne maize grown on existing agricultural lands are greater than those associated with land use conversion. Palm et al. (2010) reported GHG emissions of 2600 kg CO2e/ha associated with maize production following land use change from a “dense woody cover” for which a yield of 1.7 t/ha was assumed with a crop area expansion of 75% in the ‘Millenium Village’ Sauri, Siaya district in Kenya. The corresponding GHG emissions per tonne of maize were 1529 kg CO2e/t maize grain yield, which was reported for
some yields within the SIMLESA dataset. The maximum of these observed yields with GHG emissions greater than 1529 kg CO2e/t maize grain yield was used to provide the threshold yield that needs to be achieved in order to ensure that GHG emissions are lower than extensifying production, hence triggering land use change. This threshold maize grain yield was calculated to be 0.7 and 0.6 t/ha for Kenya and Ethiopia, respectively. In 2009, average maize consumption has been reported to be 77.2 kg/year/capita in Kenya and 44.2 kg/year/capita in Ethiopia. The average overall food calorie intake was almost 2100 kcal/capita/day for both countries in the same year (FAOSTAT, 2014), which is below the recommended daily calorie intake of 2400 and 2800 kcal/capita/day for women and men with a height of 1.7 m and a weight of 60 kg (FAO/WHO/UNU, 2001). A household’s ability to meet current average maize consumption was calculated by summing up the total maize harvested in a year from all mono- and inter-cropped plots available to a household. The harvested amount was then checked to determine whether it provides the average amount of maize consumed for the respective size of the family. It was found that yields to meet current average maize consumption needed to be between 0.8 and 2.0 t/ha in Kenya and 0.3 and 0.9 t/ha in Ethiopia, depending on family size (ranges of average adult equivalents in districts: Kenya 4–5.9; Ethiopia 3.8–6.3) and area available (ranges of average available area in districts: Kenya 0.23–0.41 ha; Ethiopia 0.26–0.74 ha). To calculate the theoretical yield for a given amount of available N, the N requirement of maize was derived by applying the figures given by Belfield and Brown (2008) where each tonne of maize requires 40 kg N/ha. This N requirement includes the fact that the plant can only access about 50% of available nitrateN in the soil and that an average of 20% of fertiliser N is lost. This would be equivalent to a nitrogen use efficiency of 40% (Belfield and Brown, 2008). The N application rate has been binned into 20 kg N/ha input ranges with exception of the first and last bin of effectively only 10 kg N/kg (0–10, 10–30, 30–50, . . . , 190–200 kg N/ha). The theoretical GHG emissions per tonne maize have been derived by calculating GHG per hectare if the amount of N, using the midpoint of N bins, was applied as urea. This has then been divided by the theoretical yield at that N bin. Here, the zero N bin has an average N application of 5 kg/ha corresponding with a yield of 0.125 t/ha.
Table 1 Cool Farm Tool (CFT) categories, which are used in this study (for a description of all categories (Hillier et al., 2011b)) and match with data. Label (number of total categories used by CFT)
CFT category used in analysis
Comments
Soil texture (3)
Medium
Soil organic matter (4)
0–1.72
Soil moisture (2)
Moist Dry
Soil drainage (3)
Good Poor
Soil pH (4)
1–5.5 5.5–7.3
Most districts in Kenya were estimated to have medium textured soil with the exception of Siaya with coarsemedium. However, as the texture was still not overall considered to be coarse, medium was still used. For Ethiopia medium conditions were also used throughout All districts fall within both categories as SOM of data ranges between 0.9 and 3.4%, this study uses 0– 1.72 only for simplification All plots are considered to be “moist” Unless: variable “stress” is given as drought at plot level and plots are not irrigated Percentage of dry plots are 55% and 18% for Kenya and Ethiopia, respectively All plots are considered to be of “good” drainage Unless: variable “stress” is given as “waterlogged” Percentage of waterlogged soils are 0.8% and 3.4% for Kenya and Ethiopia, respectively Districts: Bungoma (4.4–5.9), Imenti South (5.2) Districts: Siaya (5.0–6.6), Embu (5.6), Meru South (5.8) All districts in Ethiopia were considered to be of soil pH (5.5–7.3)
Croptype (5) Climate (2) Fertiliser application method (5) Residue management (6)
Other crop Tropical Broadcast Incorporated into field or “Left on field for soil fertility” left as mulch “Burnt” Burned Exported Used for “construction”, “sold”, “fed to livestock”, “firewood”, ”other uses”
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
141
Fig. 2. The relationship between greenhouse gas (GHG) emissions per hectare and nitrogen (N) application; N application is the sum of synthetic fertilisers and farmyard manure (FYM), (A) for Kenya and (B) for Ethiopia show the CFT result next to calculations based on IPCC guidelines and field measurements as presented in Linquist et al. (2012), bars give the range of the 95% confidence interval; ‘CFT total’ are total emissions as calculated by the Cool Farm Tool (CFT), ‘CFT field’ are emissions from the field only, i.e. excluding emission from the production of fertiliser and agrochemicals, ‘IPCC default’ are GHG emissions as calculated by IPCC guideline methodology considering field N2O emissions from fertiliser, CO2 emissions from urea and N2O emissions from residue management.
Fig. 3. The relationship between greenhouse gas (GHG) emissions per hectare and nitrogen (N) application given as open circles unless otherwise described in legend; N application is the sum of synthetic fertilisers and farmyard manure (FYM), (A) for Kenya and (B) for Ethiopia show the impact of different management options and soil conditions on GHG emissions. The legend refers to the N application: ‘FYM_N high’ is where more than 50% of the N applied is from FYM, ‘residue high’ refers to plots where the proportion of residue returned to the soil by the farmer is more than 50% of the residue available and ‘residue high +N’ refers to the same plots but where the N content of the residues is added to the total N application.
The local attainable yield was determined based on the upper 95th percentile of data from a farmer survey following the approach described by Van Ittersum et al. (2013). The amount of soil organic carbon (SOC) depleted with each 20 kg N/ha supplied by the soil was derived by using the C:N ratio and average bulk density of Kenyan soils using the SOTER database (ISRIC, 2013). These data were not available for Ethiopia, so the same values were used as for Kenya. Considering the high uncertainty of soil data in general and the use of only one value for different soil types, it was considered an appropriate approximation. The average proportion of N in relation to C was determined to be 11.5%, and the bulk density in the top 20 cm horizon was 1.25 g/cm3. Therefore 20 kg N/ha is equivalent to about 0.001% N in the top horizon, hence corresponding to about 0.01% SOC. 3. Results 3.1. Greenhouse gas emissions per hectare The most important factor in determining GHG emissions per hectare is the amount of N applied as synthetic and organic fertiliser (Figs. 2 and 3), with a significant regression line for both
countries (Table 2). The GHG emissions per hectare with and without inclusion of emissions from production of fertiliser and agrochemicals are within the averages reported for global field measurements as collated by a meta-analysis by Linquist et al. (2012) (Fig. 2). In comparison, results from IPCC guideline Table 2 Linear regressions for greenhouse gas emissions (GHG) per hectare and nitrogen (N) applications (GHG per hectare = Intercept + Slope N application) for Kenya and Ethiopia. N application is the sum of synthetic fertilisers and farmyard manure; N from maize residue is added to this where stated. Intercept Kenya N application N application + N from residue N application for waterlogged soils only N application for ‘FYM high’ only Ethiopia N application N application + N from residue N application for waterlogged soils only N application for ‘FYM high’ only a
Slope
351 7.51 328a 7.45 Not available as only 3 cases 318a 7.41
326 319a 524a 290a
7.95 7.98a 7.95 7.85
Significantly different (p < 0.05) to top function of respective country group.
142
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
Fig. 4. Relationship between greenhouse gas (GHG) emissions per tonne of maize grain and yield for Kenya (A) and Ethiopia (B) given as open circles unless otherwise described in legend. The legend refers to the N application, which is the sum of synthetic fertilisers and farmyard manure (FYM) as described in Fig. 3; ‘FYM_N high’ is where more than 50% of the N applied is from FYM, ‘residue high’ refers to plots where the proportion of residue returned to the soil by the farmer is more than 50% of the residue available and ‘residue high +N’ refers to the same plots but where the N content of the residues is added to the total N application. The solid black horizontal line marks the GHG emissions of extensifying production at the cost of deforestation based on Palm et al. (2010). The solid black vertical line marks the maximum yield observed at the equivalent GHG emissions in the data. The dashed vertical line mark the yield that is required for the household in the different districts to meet the average maize consumption depending on the average family size and maize area of the household (see Section 2 for full details), The dotted horizontal lines mark the range of GHG emissions per tonne maize reported by Linquist et al. (2012) of their meta analysis of real field measurements.
procedures would have a smaller slope though they are also still within the lower confidence interval of the global field measurement dataset. Some of the variation in Fig. 3 can be explained by the drainage of the soil, type of fertiliser applied and amount of residue returned to the soil. The size of the intercept is the result of background emissions (due to breakdown of residual soil organic matter (SOM), and use of residual nutrients), which can be observed on cropland without any amendments of N. Residual emissions are mainly dependent on soil type and level of water saturation. The impact of waterlogged soils on GHG emissions per hectare is demonstrated clearly in Ethiopia, where emissions on waterlogged plots are distinctly above the average regression line, with an intercept that is about 40% higher (Fig. 3B, Table 2). Kenya has an overall higher intercept than Ethiopia, as several factors contribute to background GHG emissions in the absence of fertiliser. In this case, the main driver is the proportion of residue returned to the field, which increases GHG emissions due to mineralisation of the N in the residue (round symbols in Fig. 3). The actual amount of residue will depend on the reported yield. When residue N is considered as part of the total N application the intercept reduces significantly (p < 0.05; triangular symbols in Fig. 3), and the difference is only about 9 kg CO2e/ha. This difference can be explained by emissions from agrochemicals, the
use of which is more widespread in Kenya compared to Ethiopia (data not shown). Fig. 3 also demonstrates that residue addition can contribute significant amounts of N to the soil. It leads to lower emissions than when N is supplied as synthetic fertiliser only, however, N from residue will not usually be fully available to the plant. It also makes apparent that smallholders in Kenya return, on average, far more residue to the soil directly in comparison to Ethiopia (Fig. 3). Generally, crop residues are very valuable for smallholders, and the main alternative uses of residues are feed for livestock and for burning for fuel (data not shown). Like residue application to the field, FYM results in lower GHG emissions compared to equal amounts of fertiliser in synthetic form, which can also be observed in Kenya to some extent. Lower emissions from FYM are partly due to the main emissions from livestock not being assigned to cropland within the CFT whereas fertiliser production is. 3.2. Greenhouse gas emissions per tonne of maize A strong relationship between GHG emissions per tonne of maize and yield was found (Fig. 4). A relationship between N application and yield is not very strong due to various other stresses such as, e.g. water or phosphorus limitation and pests (Fig. 5). The minimum GHG
Fig. 5. Relationship between greenhouse gas yield and nitrogen (N) application in tonne per hectare for Kenya (A) and Ethiopia (B).
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
143
Table 3 Mean values of Kenyan and Ethiopian districts for sole cropped maize from the SIMLESA (Sustainable Intensification of Maize–Legume Systems for Food Security in Eastern and Southern Africa) dataset. Values in brackets are the amount of nitrogen (N) applied as farmyard manure (kilogram per hectare) in the ‘N applied’ column and the 95th percentile of observed yields to represent observed attainable yield in ‘yield’ column. Values of attainable yield corresponding to FAOSTAT 2009 country statistics have been taken from Mueller et al. (2012). District
Kenya Bungoma Embu Imenti South Meru South Siaya Mean of districts Kenya: FAOSTAT 2009 Ethiopia Adami Tulu Bako Tibe Dugda Gubuesyo Hawasa Zurya Meskan Mesrak Badawacho Pawe Shalla Mean of districts Ethiopia: FAOSTAT 2009
Number of plots (n)
N applied (kg/ha)
Residue applied (%)
39 99 72 136 30
38 (6) 36 (5) 45 (3) 47 (6) 17 (5)
35 12 9 11 76
1.4 1.4 1.5 1.7 1.2
(1.8) (1.9) (2.0) (2.4) (1.7)
646 615 695 691 517
1292 911 856 659 1183
376
40
18
1.5 (2.0) 1.3 (3.7)
653
870
203 277 109 124 106 136 133 112 233
7 (1) 60 (2) 16 (4) 56 (2) 70 (3) 25 (3) 29 (1) 31 (8) 18 (1)
6 1 14 2 4 1 0 9 9
2.0 2.4 1.8 2.8 2.7 2.0 2.0 2.1 2.4
(2.8) (3.2) (2.4) (4.0) (3.2) (2.7) (2.4) (3.0) (3.2)
389 796 453 793 888 503 536 565 480
335 519 650 475 430 346 382 426 282
1433
34
5
2.3 (3.0) 2.2 (5.7)
598
418
emissions per tonne maize achieved are 82 and 67 kg CO2e/t maize grain for Kenya and Ethiopia, respectively. The strong relationship between emissions and yield demonstrates that it is of utmost importance to ensure that maize yields are above a level where extensification (associated with land use change), would be more efficient. This would correspond to a maize grain yield of 0.7 and 0.6 t/ha for Kenya and Ethiopia, respectively. In Kenya, this threshold is even higher from a food security perspective, increasing the required maize grain yield to between 0.8 and 2.0 t/ha for Kenya, depending on the average family size and area cropped with maize in a given household of the respective region. This range for maize grain yield is between 0.3 and 0.9 t/ha for Ethiopia. Generally, maize yields are much lower in Kenya than in Ethiopia, and the best yielding district in Kenya is equivalent to the worst yielding district in Ethiopia with respect to yield and GHG emissions per tonne maize (Table 3). Districts in Kenya with high
Yield (t/ha)
GHG emissions in (kg CO2e/ha) (kg CO2e/t maize)
Food insecure (%)
25 12 10 6 34
1 5 11 2 2 13 1 28 3
GHG emissions per tonne maize tend to be those with a high percentage of food insecure households (i.e. Bungoma and Siaya in the western Kenyan highlands). Generally, Kenyan smallholders struggle far more to produce enough for their own consumption (Table 3), and yields are much lower, with some exceptionally low yields resulting in very high GHG emissions per tonne of maize. 3.3. N balance, yield gap and the identification of “climate smart” practices Fig. 6 depicts a box plot for yield and GHG emissions of each N bin, displaying a general trend of higher yields with a higher N input. Furthermore, the yield gap is demonstrated by a comparison between the actual yield and the theoretical yield for a given amount of available N based on the N requirement of maize (black triangles in Fig. 6). Lastly the theoretical GHG emissions for the
Fig. 6. Boxplots of greenhouse gas (GHG) emissions per tonne maize (dark grey) and corresponding observed yield (light grey) for each nitrogen (N) bin for Kenya (A) and Ethiopia (B). N bins are ranges of N application 10 kg N/ha. The number of plots represented within each boxplot is shown at the bottom of the graph. The region shaded grey contains those boxplots with less than 10 observations. Theoretic values for GHG emissions per tonne of maize (black squares) and yield (black triangles) have been provided to enable judgement of adequacy of N supply versus demand and its associated GHG emissions (see Section 2 for full details).
144
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
expected associated theoretical yield are plotted as black squares. Each 20 kg N/ha available to the plant will increase the theoretical yield by 0.5 t/ha. The theoretical GHG emissions per tonne maize demonstrate that for an N input above about 60 kg/ha the impact of the associated higher yields with higher N input starts levelling off though a real stagnation is only observed at levels of 160 kg N/ha. A comparison of the data to the theoretical values shows that, for Kenya, yields are actually higher than expected below an N input of 60 kg/ha and mainly less than expected above this amount. These higher than expected yields from lower N rates require the plant to “mine” N from the soils unless N is applied in a different form. This may well be the case to some extent in Siaya, reporting the lowest N rates and the highest proportion of residue returned to the soil (Table 3). Insufficient N supply would slowly decrease soil organic nitrogen (and soil organic carbon) content and reduce fertility and yields. For each 20 kg/ha N that is supplied by the soil, 0.01% SOC has to be mineralised. In contrast, if the yields are not achieved with higher N input, the N is vulnerable to be leached unless it contributes to an increase of SOC and may be available for the following season. Yields are likely to be limited by other factors in addition to N such as other nutrients, water or soil conditions. The data from Ethiopia look very different in that much higher yields are achieved and an N input of up to 120 kg N/ha results in yields greater than would be expected at that N addition rate (Fig. 6). Higher N rates may again be prone to losses. However, it should also be noted that fewer plots are represented at high N levels, and so these values are not so reliable, which is the reason they have been highlighted in grey. Generally, despite the higher maize grain yields achieved in Ethiopia (2.3 t/ha), overall, the average N rates (34 kg N/ha) are smaller than in Kenya (Table 3) resulting in a potentially higher N deficit. However, for a true analysis of an N balance, further factors would need to be considered such as, e.g. residual N from legumes grown in the previous year and atmospheric N deposition. With the information available, an amount of 60 kg N/ha and 120 kg N/ha can be recommended on average for Kenya and Ethiopia, respectively. However, if higher yields are achieved than would be expected according to the N input, this level would have to be higher. Generally, the location of the 95th percentile indicates that most districts should have the potential to increase their yield by at least 0.5 t/ha. 4. Discussion The results from Ethiopia and Kenya of GHG from maize production illustrate the dependence of GHG emissions on fertiliser usage. Background GHG emissions (320–524 kg CO2e/ha, range of intercept in Fig. 3, Table 2) were comparable to the measurements reported by Mapanda et al. (2011) (294–404 kg CO2e/ha) and Linquist et al. (2012) (Fig. 2). In comparison to Mapanda et al. (2011), there are considerable differences in the composition of GHGs, and N2O was reported to be much lower in their study. These are likely to be at least partially a result of technical issues in the field when measuring N2O. A major issue with non-continuous sampling is that it can miss spatial and temporal hotspots such as the “birch effect” (Dick et al., 2008; Mapanda et al., 2011). It is clear though, that the CFT does not capture differences in emissions with, for example, rainfall pattern and crop performance, which have been observed by Mapanda et al. (2011). It should be recognised that the CFT estimates of GHG emissions from N application based on algorithms developed by Bouwman et al. (2002), who stated that their “estimates cannot be compared to individual fields” as their results summarize a diversity of studies in regard to methodology and management. This is reflected in the match to globally reported averages as reported by Linquist et al. (2012). Generally, any positive or negative bias is of less importance when comparing samples from the same region with each other.
A clear outcome of this study supports the findings of Palm et al. (2010), that there are synergies between climate change mitigation and food security. Even though this study did not look in detail at scenarios of mitigation, it is clear that the achievement of a minimum yield, which will be specific for each country or region, is the main driver to reduce emission intensity, i.e. the GHG emission on a per-tonne-product basis. This has also been suggested by other researchers, but for temperate systems (Brentrup et al., 2004; Ma et al., 2012). Although mitigation on a per-hectare basis might simplistically appear to favour lower input production systems, this clearly overlooks the primary role of agriculture to provide food and materials, and also risks creating drivers for indirect land use change (Carlton et al., 2010). This emphasizes the importance of increased yields and supporting smallholder food security, in order to avoid opening up new land for cultivation. The reduction of GHG emissions on a per-tonne basis is compatible with the aim of closing the yield gap, as discussed in several recent papers (Mueller et al., 2012; Smith et al., 2013; Tittonell and Giller, 2013; Van Ittersum et al., 2013). The approach in this study also addresses the need for strong, locally relevant, bottom-up approaches compared to strictly global analysis performed at coarse spatial scales (Van Ittersum et al., 2013). Sufficient nutrient input is usually mentioned as a key to achieving and maintaining yields (Tittonell and Giller, 2012). The recommended average amount of N of 60 and 120 kg N/ha for Kenya and Ethiopia, respectively, is close to the amount of 80 kg N/ha already recommended by local agricultural extension services (Palm et al., 2010). However, our recommendation is based on actual achieved yields, here specific for two countries though it could be transferred to different areas and scaled to different levels. The respective nutrient input is selected to avoid nutrient depletion when supplying too little N or SOC loss due to soil organic matter mineralisation, as well as N leaching at superoptimal applications. Organic sources of N are usually not available in sufficient quantity to meet the needs of high yielding crops. In addition, the maize residues and FYM often immobilize available N initially (Gentile et al., 2011; Zingore et al., 2008). Organic fertilisers are often, therefore, applied with synthetic fertilisers (Bationo et al., 2004; Gentile et al., 2011; Nyamangara et al., 2004). Organic forms of N do have the advantage that they will contribute to the general fertility and water retention of the soil in the long-term. However, it should also be noted that especially in countries with erratic rainfall, the overall water balance and specific timing of rainfall also limits yields (Van Ittersum et al., 2013). The effect of water limitation has not been considered here as all the yields are water-limited. A change from current application levels to an N application of 60 and 120 kg N/ha for Kenya and Ethiopia, respectively, would increase nutrient input for most plots (Fig. 6). This would have a considerable impact on GHG emissions per hectare and regional total emissions, as is predicted for SSA (Reay et al., 2012; Smith et al., 2007). In this study it would double and triple the total GHG emissions for the area that maize is grown on for Kenya and Ethiopia, respectively. However, at the same time yield would be expected to increase accordingly. Van der Velde et al. (2013), e.g. estimated an average yield increase of 36% with addition of 10 kg/ha of N and P and up to a maximum of 190% with additions of 80 kg/ha N and 20 kg/ha P. There is a clear need to raise production to feed the population, particularly in Kenya, where currently almost 40% of the consumed maize is imported. This approach could be used to identify the optimum for increasing production with the lowest GHG costs. The modelling approach here, based on simple regional survey data, allowed the identification of optimal N rates, aiming for the most GHG efficient production systems. This is particularly valuable for low input systems as GHG emission per tonne product
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
are very sensitive to changes in yields at low levels. Expressing results in terms of a GHG intensity metric leads to analogous conclusions as those based on N balance and nutrient use efficiency alone. Indeed, the GHG optimum is achieved for both countries over the ranges at which the balance of N inputs and crop N uptake is achieved. As such, the identification of practices to minimise GHG intensity are compatible with the objectives of optimizing productivity and ensuring food security. Generally, the approach described here can be used to support policy decisions for optimal interventions to obtain realistically achievable food production with low GHG intensity in different regions. The analysis of N balance should be combined with other important factors such as water availability, soil properties (especially fertility and drainage characteristics), and the use of crop protection chemicals. This concerted approach can quantify the trade-offs between, and optimise, crop productivity and GHG emissions, essential for addressing the joint global challenges of food security and climate change mitigation. Acknowledgements We thank two anonymous reviewers for comments which considerably improved the manuscript. This research was conducted under the CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS) with additional funding from the Australian Centre for International Agricultural Research (ACIAR) for generation of the survey data under the ‘Sustainable Intensification of Maize–Legume Systems for Food Security in Eastern and Southern Africa (SIMLESA)’ project. The work contributes to the University of Aberdeen Environment and Food Security theme and the Scottish Food Security Alliance-Crops. The scripting of the CFT tool into Matlab was funded by the Social and Environmental Economic Research (SEER) into Multi-Objective Land Use Decision Making project (which in turn is funded by the Economic and Social Research Council (ESRC); Funder Ref: RES-060-25-0063). References Audsley, E., 1997. Harmonisation of environmental life cycle assessment for agriculture (No. Final Report. Concerted Action AIR3-CT94-2028. (Coordinator). European Commission DG VI Agriculture. Bationo, A., Nandwa, S.M., Kimetu, J.M., Kinyangi, J.M., Bado, B.V., Lompo, F., Kimani, S.K., Kihanda, F., Koala, S., 2004. Sustainable intensification of crop-livestock systems through manure management in eastern and western Africa: lessons learned and emerging research opportunities. Sustainable Crop-Livestock Production for Improved Livelihoods and Natural Resource Management in West Africa. Wageningen, The Netherlands. Belfield, S., Brown, C., 2008. Field Crop Manual: Maize – A Guide to Upland Production in Cambodia. NSW Department of Primary Industries, New South Wales. Bellon, M.R., Hodson, D., Bergvinson, D., Beck, D., Martinez-Romero, E., Montoya, Y., 2005. Targeting agricultural research to benefit poor farmers: relating poverty mapping to maize environments in Mexico. Food Policy 30, 476–492. Bouwman, A.F., Boumans, L.J.M., Batjes, N.H., 2002. Modeling global annual N2O and NO emissions from fertilized fields. Global Biogeochem. Cycles 16, 28-1–28-9. Brentrup, F., Küsters, J., Lammel, J., Barraclough, P., Kuhlmann, H., 2004. Environmental impact assessment of agricultural production systems using the life cycle assessment (LCA) methodology II. The application to N fertilizer use in winter wheat production systems. Eur. J. Agron. 20, 265–279. Carlton, R., Berry, P., Smith, P., 2010. Impact of crop yield reduction on greenhouse gas emissions from compensatory cultivation of pasture and forested land. Int. J. Agric. Sustain. 8, 164–175. Cool Farm Tool Institute, 2012. URL http://www.coolfarmtool.org/ (accessed 10.06.13). CSA (Central Statistical Agency), 2013. Agricultural Sample Survey 2012/2013 (2005 E.C.). Report on area and production of major crops (Private Peasant Holdings, Meher Season). Statistical Bulletin 532. Addis Ababa, Ethiopia. Denning, G., Kabambe, P., Sanchez, P., Malik, A., Flor, R., Harawa, R., Nkhoma, P., Zamba, C., Banda, C., Magombo, C., Keating, M., Wangila, J., Sachs, J., 2009. Input subsidies to improve smallholder maize productivity in Malawi: toward an African green revolution. PLoS Biol. 7, e1000023. Dick, J., Kaya, B., Soutoura, M., Skiba, U., Smith, R., Niang, A., Tabo, R., 2008. The contribution of agricultural practices to nitrous oxide emissions in semi-arid Mali. Soil Use Manage. 24, 292–301.
145
Ecoinvent Centre, 2007. Ecoinvent data v2.0 (No. Ecoinvent reports No. 1e25). Swiss Centre for Live Cycle Inventories, Dübendorf. FAO/IFA, 2001. Global estimates of gaseous emissions of NH3, NO and N2O from agricultural land. FAO and IFA, Rome. FAOSTAT, 2014. FAOSTAT Agricultural Data. Available at: http://faostat.fao.org/FAO/ WHO/UNU, 2001. Human Energy Requirements. FAO Food and Nutrition Technical Report Series, Rome. FAO/WHO/UNU, 2001. Human Energy Requirements. FAO Food and Nutrition Technical Report Series, Rome. Foley, J.A., Ramankutty, N., Brauman, K.A., Cassidy, E.S., Gerber, J.S., Johnston, M., Mueller, N.D., O'Connell, C., Ray, D.K., West, P.C., Balzer, C., Bennett, E.M., Carpenter, S.R., Hill, J., Monfreda, C., Polasky, S., Rockström, J., Sheehan, J., Siebert, S., Tilman, D., Zaks, D.P.M., 2011. Solutions for a cultivated planet. Nature 478, 337–342. Garnett, T., Appleby, M.C., Balmford, A., Bateman, I.J., Benton, T.G., Bloomer, P., Burlingame, B., Dawkins, M., Dolan, L., Fraser, D., Herrero, M., Hoffmann, I., Smith, P., Thornton, P.K., Toulmin, C., Vermeulen, S.J., Godfray, H.C.J., 2013. Sustainable intensification in agriculture: premises and policies. Science 341, 33–34. Gentile, R., Vanlauwe, B., Chivenge, P., Six, J., 2011. Trade-offs between the short- and long-term effects of residue quality on soil C and N dynamics. Plant Soil 338, 159–169. Godfray, H.C.J., Beddington, J.R., Crute, I.R., Haddad, L., Lawrence, D., Muir, J.F., Pretty, J., Robinson, S., Thomas, S.M., Toulmin, C., 2010. Food security: the challenge of feeding 9 billion people. Science 327, 812–818. Hickman, J.E., Havlikova, M., Kroeze, C., Palm, C.A., 2011. Current and future nitrous oxide emissions from African agriculture. Curr. Opin. Environ. Sustain. 3, 370–378. Hillier, J., Hawes, C., Squire, G., Hilton, A., Wale, S., Smith, P., 2009. The carbon footprints of food crop production. Int. J. Agric. Sustain. 7, 107–118. Hillier, J., Smith, P., Bandel, T., Daniels, S., Malin, D., Hamilton, H., Walter, C., 2011a. Farm-scale greenhouse gas emissions using the Cool Farm Tool: application of a generic farming emissions calculator in developing countries. In: Wollenberg, E., Nihart, A., Tapio-Biström, M.L., Grieg-Gran, M. (Eds.), Climate change mitigation and agriculture. Routledge, London, UK. Hillier, J., Walter, C., Malin, D., Garcia-Suarez, T., Mila-i-Canals, L., Smith, P., 2011b. A farm-focused calculator for emissions from crop and livestock production. Environ. Model. Softw. 26, 1070–1078. Hillier, J., Brentrup, F., Wattenbach, M., Walter, C., Garcia-Suarez, T., Mila-i-Canals, L., Smith, P., 2012. Which cropland greenhouse gas mitigation options give the greatest benefits in different world regions? Climate and soil-specific predictions from integrated empirical models Global Change Biol. 18, 1880–1894. IPCC, 2006. In: The National Greenhouse Gas Inventories Programme, Eggleton, H.S., Buendia, L., Miwa, K., Ngara, T., Tanabe, K. (Eds.), 2006 IPCC Guidelines for National Greenhouse Gas Inventories. IPCC, Institute for Global Environmental Strategies, Tokyo, Japan. IPCC, 2007. Summary for policymakers. In: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M., Miller, H.L. (Eds.), The physical science basis. Contribution of working group I to the fourth assessment report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. ISRIC, 2013. Development of a Soil and Terrain Database for Kenya (KENSOTER). http://www.isric.org/projects/soter-kenya-kensoter. Kang’ethe, E., 2011. Situation Analysis: Improving Food Safety in the Maize Value Chain in Kenya. Report prepared for FAO. College of Agriculture and Veterinary Science, University of Nairobi, Nairobi. Mulwa, C., Kassie, M., Odendo, M., Ouma, J., 2013. Socioeconomic Profile of Maizelegume Based Farming System in Ethiopia: Characterization of Livelihood Strategies, Production Conditions and Markets. Mimeo, CIMMYT, Nairobi. Linquist, B., van Groenigen, K.J., Adviento-Borbe, M.A., Pittelkow, C., van Kessel, C., 2012. An agronomic assessment of greenhouse gas emissions from major cereal crops. Global Change Biol. 18, 194–209. Ma, B.L., Liang, B.C., Biswas, D.K., Morrison, M.J., McLaughlin, N.B., 2012. The carbon footprint of maize production as affected by nitrogen fertilizer and maize– legume rotations. Nutr. Cycl. Agroecosyst. 94, 15–31. Mapanda, F., Wuta, M., Nyamangara, J., Rees, R.M., 2011. Effects of organic and mineral fertilizer nitrogen on greenhouse gas emissions and plant-captured carbon under maize cropping in Zimbabwe. Plant Soil 343, 67–81. Mosier, A.R., Halvorson, A.D., Reule, C.A., Liu, X.J., 2006. Net global warming potential and greenhouse gas intensity in irrigated cropping systems in Northeastern Colorado. J. Environ. Qual. 35, 1584–1598. Mueller, N.D., Gerber, J.S., Johnston, M., Ray, D.K., Ramankutty, N., Foley, J.A., 2012. Closing yield gaps through nutrient and water management. Nature 490, 254–257. Muricho, G., Kassie, M., Odendo, M., Ouma, J., 2012. Characterization of Maize– Legume Farming Systems and Farm Households in Kenya: Analysis of Technology Choice, Resource Use, Gender, Risk Management, Food Security and Poverty Profiles. Mimeo, CIMMYT, Nairobi. Nyamangara, J., Piha, M.I., Giller, K.E., 2004. Effect of combined cattle manure and mineral nitrogen on maize N uptake and grain yield. Afr. Crop Sci. J. 11, 289–300. Palm, C.A., Smukler, S.M., Sullivan, C.C., Mutuo, P.K., Nyadzi, G.I., Walsh, M.G., 2010. Identifying potential synergies and trade-offs for meeting food security and climate change objectives in sub-Saharan Africa. PNAS 107, 19661–19666. R Core Team, 2012. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.
146
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
Reay, D.S., Davidson, E.A., Smith, K.A., Smith, P., Melillo, J.M., Dentener, F., Crutzen, P.J., 2012. Global agriculture and nitrous oxide emissions. Nat. Clim. Change 2, 410–416. Smith, P., Martino, D., Cai, Z., Gwary, D., Janzen, H.H., Kumar, P., McCarl, B., Ogle, S., O'Mara, F., Rice, C., Scholes, R.J., Sirotenko, O., Howden, M., McAllister, T., Pan, G. X., Romanenkov, V., Rose, S., Schneider, U., Towprayoon, S., 2007, 2007. Agriculture, in: Mitigation. Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 497–540. Smith, P., Martino, D., Gwary, D., Janzen, H., Kumar, P., McCarl, B., Ogle, S., O’Mara, F., Rice, C., Scholes, B., Sirotenko, O., Howden, M., McAllister, T., Romanenkov, V., Schneider, U., Towprayoon, S., Wattenbach, M., Smith, J., 2008. Greenhouse gas mitigation in agriculture. Philos. Trans. R. Soc. B Biol. Sci. 363, 789–813. Smith, P., Haberl, H., Popp, A., Erb, K., Lauk, C., Harper, R., Tubiello, F.N., de Siqueira Pinto, A., Jafari, M., Sohi, S., Masera, O., Böttcher, H., Berndes, G., Bustamante, M., Ahammad, H., Clark, H., Dong, H., Elsiddig, E.A., Mbow, C., Ravindranath, N.H., Rice, C.W., Robledo Abad, C., Romanovskaya, A., Sperling, F., Herrero, M., House, J.I., Rose, S., 2013. How much land-based greenhouse gas mitigation can be achieved without compromising food security and environmental goals? Global Change Biol. 19, 2285–2302.
Tilman, D., Balzer, C., Hill, J., Befort, B.L., 2011. Global food demand and the sustainable intensification of agriculture. PNAS 108, 20260–20264. Tittonell, P., Giller, K.E., 2012. When yield gaps are poverty traps: the paradigm of ecological intensification in African smallholder agriculture. Field Crops Res. 143, 76–90. Townend, J., 2002. Practical Statistics for Environmental and Biological Sciences. John Wiley & Sons Ltd., Chichester, pp. UK. Van der Velde, M., See You, L.L., Balkovic, J., Fritz, S., Khabarov, N., Obersteiner, M., Wood, S., 2013. Affordable nutrient solutions for improved food security as evidenced by crop trials. PLoS One 8. Van Groenigen, J.W., Velthof, G.L., Oenema, O., Van Groenigen, K.J., Van Kessel, C., 2010. Towards an agronomic assessment of N2O emissions: a case study for arable crops. Eur. J. Soil Sci. 61, 903–913. Van Ittersum, M.K., Cassman, K.G., Grassini, P., Wolf, J., Tittonell, P., Hochman, Z., 2013. Yield gap analysis with local to global relevance–a review. Field Crops Res. 143, 4–17. Whittaker, C., McManus, M.C., Smith, P., 2013. A comparison of carbon accounting tools for arable crops in the United Kingdom. Environ. Model. Softw. 46, 228–239. Zingore, S., Murwira, H.K., Delve, R.J., Giller, K.E., 2008. Variable grain legume yields, responses to phosphorus and rotational effects on maize across soil fertility gradients on African smallholder farms. Nutr. Cycl. Agroecosyst. 80, 1–18.
Contents lists available at ScienceDirect
Agriculture, Ecosystems and Environment journal homepage: www.elsevier.com/locate/agee
Identifying secure and low carbon food production practices: A case study in Kenya and Ethiopia Jessica Bellarby a,b, *, Clare Stirling c , Sylvia Helga Vetter a , Menale Kassie d, Fred Kanampiu d, Kai Sonder e, Pete Smith a , Jon Hillier a a
Scottish Food Security Alliance-Crops & Institute of Biological and Environmental Sciences, University of Aberdeen, Aberdeen AB24 3UU, UK Lancaster Environment Centre, Lancaster University, Bailrigg, Lancaster LA1 4YQ, UK c International Maize and Wheat Improvement Center (CIMMYT), Ynys Mon, Wales, UK d International Maize and Wheat Improvement Center (CIMMYT), Nairobi,Kenya e International Maize and Wheat Improvement Center (CIMMYT), Mexico, D.F., Mexico b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 11 October 2013 Received in revised form 24 July 2014 Accepted 25 July 2014 Available online xxx
The world population is projected to increase to 9–10 billion by 2050, during which time it will be necessary to reduce anthropogenic greenhouse gas emissions to mitigate climate change. The particular challenge this places on agriculture is to identify practices which ensure stable and productive food supply that also have a low greenhouse gas (GHG) intensity. Maize is the principle staple crop in many parts of Africa with low and variable yields, averaging only 1.6 t/ha in sub-Saharan Africa (SSA). Food security and increasing crop yields are considered priorities in SSA over impacts of food production on GHG emissions. Here we describe an approach that can be used to inform a decision support tree for optimal interventions to obtain sufficient food production with low GHG intensity, and we demonstrate its applicability to SSA. We employed a derivative of the farm greenhouse gas calculator ‘Cool Farm Tool’ (CFT) on a large survey of Kenyan and Ethiopian smallholder maize-based systems in an assessment of GHG intensity. It was observed that GHG emissions are strongly correlated with nitrogen (N) input. Based on the relationship between yield and GHG emissions established in this study, a yield of 0.7 t/ha incurs the same emissions as those incurred for maize from newly exploited land for maize in the region. Thus, yields of at least 0.7 t/ha should be ensured to achieve GHG intensities lower than those for exploiting new land for production. Depending on family size, the maize yield required to support the average consumption of maize per household in these regions was determined to be between 0.3 and 2.0 t/ha, so that the desirable yield can be even higher from a food security perspective. Based on the response of the observed yield to increasing N application levels, average optimum N input levels were determined as 60 and 120 kg N/ha for Kenya and Ethiopia, respectively. Nitrogen balance calculations could be applied to other countries or scaled down to districts to quantify the trade-offs, and to optimise crop productivity and GHG emissions. Crown Copyright ã 2014 Published by Elsevier B.V. All rights reserved.
Keywords: Greenhouse gas emissions Sub-Saharan Africa Smallholder farming system Maize Food security
1. Introduction The world population is projected to increase to 9–10 billion by 2050 (Godfray et al., 2010), during which time it will be necessary to reduce anthropogenic greenhouse gas emissions to mitigate climate change (Smith et al., 2008). The particular challenge this places on agriculture is to identify agricultural
* Corresponding author at: Lancaster Environment Centre, Lancaster University, Bailrigg, Lancaster LA1 4YQ, UK. Tel.: +44 1224 273810. E-mail address: [email protected] (J. Bellarby). http://dx.doi.org/10.1016/j.agee.2014.07.015 0167-8809/ Crown Copyright ã 2014 Published by Elsevier B.V. All rights reserved.
practices which ensure stable and productive food supply combined with food of low greenhouse gas intensity (Garnett et al., 2013; Godfray et al., 2010; Mueller et al., 2012; Smith et al., 2013; Tilman et al., 2011). In the developed world, access to water, nutrient resources (e.g. fertilisers), and crop protection chemicals have raised yields since 1965 (Foley et al., 2011). In contrast, yield increases in the developing world have been limited, so that there is substantial potential to increase food production (Mueller et al., 2012). Mueller et al. (2012) identify sub-Saharan Africa as one of the world regions showing considerable ‘low-hanging’ intensification opportunities for major cereals, stating also that closing maize
138
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
yield gaps to 50% of attainable yields (defined as the area-weighted 95th percentile of observed yield within zones of a similar climate) primarily requires addressing nutrient deficiencies, but that additional gains require both an increase in irrigated area and nutrient application over most of the region (Mueller et al., 2012). Tittonell and Giller (2012) discuss the need to increase soil fertility in sub-Saharan Africa and develop options to achieve this. The potential for yield improvement has also been demonstrated by the national subsidy programme for fertiliser and improved seeds in Malawi, which resulted in a doubling of national yields (Denning et al., 2009; Hickman et al., 2011). However, this comes at a cost by also increasing GHG emissions. During the 2000s, agriculture contributed between 10 and 12% to global anthropogenic greenhouse gas (GHG) emissions, consisting mainly of nitrous oxide and methane (Smith et al., 2008). These are projected to increase further, with one of the highest growth rates in emissions predicted in sub-Saharan Africa, due to increased fertiliser input and livestock numbers (Reay et al., 2012; Smith et al., 2007). Therefore, food supply needs to be improved whilst minimising the environmental impact (Garnett et al., 2013), which is a strategy that has been termed ‘sustainable intensification’. There has been much debate about the precise definition of this term (Tilman et al., 2011; Smith et al., 2013). In order to make this term practically usable, quantifiable measures – such as GHG emissions per unit of production – are needed for evidence-based assessment of practices. Although in many developed countries, improvement in this measure can be achieved with modest increases in production in conjunction with reduced emissions – through for example precision agriculture – in developing countries the focus and benefits will be more on increasing the efficiency of production through good agronomic practices (e.g. Mueller et al., 2012; Van Groenigen et al., 2010). Total GHG emissions from croplands are a sum of background emissions from the field (e.g. soil gaseous emissions from the breakdown of residual soil organic matter), and emissions from the production and application of pesticides, synthetic and organic fertilisers, and from the management of residues. Field measurements of GHG emissions are often expressed on a per hectare basis and may not report observed yield (Linquist et al., 2012). GHG emissions per hectare have a strong relationship to the amount of fertiliser N supplied, and are usually linked to yield (Hillier et al., 2009; Van Groenigen et al., 2010). The relationship between GHG emissions and agronomic productivity can be illustrated by calculating the yield-scaled emissions (Linquist et al., 2012; Mapanda et al., 2011; Mosier et al., 2006; Van Groenigen et al., 2010). This enables the identification of best management options in terms of GHG emissions as a function of N input (Van Groenigen et al., 2010) in addition to other variables. Current estimates of GHG emissions for African regions are mainly based on the IPCC Tier 1 emission factor approaches, which are often considered to be too simplistic for informative site-level assessments (Hickman et al., 2011). In contrast, the use of more complex models is hampered by a lack of data. These include soil and weather data used as inputs, but also the dearth of GHG emissions measurements available for model validation (Hickman et al., 2011). The capacity to assess the impact of practices on GHG emissions on working farms is relatively recent. Although for example, Smith et al. (2008) list many mitigation options for agriculture together with an estimate of the per hectare GHG impact, until relatively recently decision support capacity has been lacking at the site level (Hillier et al., 2011a). This has led to the development of several “GHG calculators”, some examples of which were recently reviewed (Whittaker et al., 2013). These models enable the analysis of typically collated farm survey data and so can capture mitigation options at a scale, and with a degree
of representativeness, not possible with models that are more data demanding. Such tools are generally not sensitive to seasonal climatic variation or subtle variation in soil properties. However, some allow an assessment of GHG emissions as a function of management practice, and enable the user to examine and optimise different management options (Hillier et al., 2011b) at the desired scale (e.g. Hillier et al., 2012). In this study we used a derivative of one such model – the Cool Farm Tool (CFT; Hillier et al., 2011b) – to estimate GHG emissions of smallholder maize in Kenya and Ethiopia to assess best practice. Kenya and Ethiopia represent countries in SSA where maize is the dominant staple crop in both countries with smallholder farmers occupying about 75% of the total maize land area (CSA, 2013; Kang’ethe, 2011). Kenyan smallholders usually have complex intercropping regimes with two or three different crops grown in a single plot, whereas maize is mainly planted on its own as a sole crop in rotation with other crops by smallholders in Ethiopia. In this study, we have restricted the analysis to mono-cropped maize to avoid issues of allocation of emissions to different crops on the same plot. Both countries are representative of maize-based smallholder farming systems in SSA. According to FAOSTAT, Ethiopia is a country almost self-sufficient in maize (FAOSTAT, 2014). An integrated livestock system is part of Ethiopian culture, reflected in it having the highest livestock population in Africa. In contrast, Kenya produces only 60% of the maize required for its own needs (FAOSTAT, 2014). Both countries are characterized by erratic rainfall and hence are prone to drought (Mulwa et al., 2013; Muricho et al., 2012). The CFT was chosen partly because it does not exclusively adopt a Tier 1 emissions factor approach. As such it can produce site and technology sensitive estimates of GHG emissions as a function of management practice, but at the same time, it is relatively datalight in comparison to many process-based models. We hypothesise that farm survey data and a version of the CFT are sufficient to identify optimum levels of N input (the key driver of GHG emissions in croplands) based on currently observed attainable yields (defined as the 95th percentile of surveyed observed yield), for sustainable food production in sub-Saharan Africa. Datasets from Kenya and Ethiopia are used as representative of smallholder maize growing systems in the region. We further show that the method is generalizable, and scalable for screening approaches to identify best practice to guide sustainable intensification on a regional basis. 2. Materials and methods 2.1. Data Data were derived from household surveys conducted by the Kenyan Agricultural Research Institute (KARI) and the Ethiopian Institute of Agriculture Research (EIAR) in collaboration with the International Maize and Wheat Improvement Centre (CIMMYT) as part of the ‘Sustainable Intensification of Maize–Legume Systems for Food Security in Eastern and Southern Africa’ (SIMLESA) programme. In this survey, 613 and 896 households were interviewed on a multitude of farm and general social aspects in 5 Kenyan, and 9 Ethiopian districts, respectively (Mulwa et al., 2013; Muricho et al., 2012) (Fig. 1). The survey also included farmer experiences of stresses such as “waterlogging”. This study makes use of 376 mono-cropped plots from 187 households in Kenya, and 1433 mono-cropped plots from 765 households in Ethiopia. 2.1.1. Calculation of GHG emissions GHG emissions on smallholder farms were calculated using a derivative of the CFT. Particular features of the tool are that it has a
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
global scope, is transparent, and open source, and has been piloted on numerous systems in many different regions and products (Cool Farm Tool Institute, 2012; Hillier et al., 2011b; Whittaker et al., 2013). In this study, the data availability only allowed the use of components related to site characteristics and crop management. These have previously been scripted in Matlab (version 7.14, http://www.mathworks.com) and were available to allow the throughput of high sample numbers. The GHG emissions derived from soil and crop management include:
139
2.2. Data input requirement and matching with the smallholder database The inputs as supplied from the survey data, including the assumptions and generalizations that had to be made to match the available information to the inputs required by the CFT, are described here. There are two types of input: 1) Categorical: data fall within a certain range and are assigned to
one group (Table 1). 2) Continuous: amounts are provided by the database as they
(i) emissions of fertiliser production and distribution given by
emission factors from the Ecoinvent database (Ecoinvent Centre, 2007), (ii) soil (fertiliser induced and background) emissions calculated using the multivariate empirical model of Bouwman et al. (2002) for nitrous oxide (N2O) and nitric oxide (NO) emissions, and the model of FAO/IFA (2001) for ammonia (NH3) emissions; residue emissions are included only when they are returned to the field using IPCC N2O Tier 1 emission factors, (iii) emissions from other agro-chemicals using figures from Audsley (1997). Emissions of CO2 from soil resulting from urea application or liming are also accounted for using IPCC emission factors (IPCC, 2006). All GHG emissions are reported as CO2-equivalents (CO2e) using the global warming potential over a 100 year horizon, which is 298 times higher for N2O, and 25 times for methane, than for CO2 (IPCC, 2007).
occur (amount of fertiliser and farmyard manure, proportion of residue returned to the field, area, amount harvested product and number of pesticide applications). Most continuous data were available from the SIMLESA database with the exception of the amount of residue returned to the field, which instead was calculated as a proportion of the yield. Local experience indicated that the default method applied in the CFT for estimating the residue for maize was inaccurate and was therefore estimated to be twice the harvested yield (Mapanda et al., 2011). In the categorical data, all soil information was estimated as shown in Table 1, with details of residue management provided by the SIMLESA database. 2.3. Analysis The analysis was carried out in R2.15.2 (R Core Team, 2012) using the additional package sqldf for database queries. For
Fig. 1. Map of Ethiopia and Kenya showing districts and ‘maize megaenvironments’ according to Bellon et al. (2005). Numbers in brackets are the percentage of maize cropped on its own in relation to total maize cropped as sole and inter-cropped maize. Plot data are derived from the SIMLESA (Sustainable Intensification of Maize–Legume Systems for Food Security in Eastern and Southern Africa) survey data.
140
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
analysis, outliers of total nitrogen (N) application and yield were removed. These were: – Maize: yield greater than 10 t/ha and lower than 0.02 t/ha. – N application rate greater than 200 kg N/ha.
The N application rate is the sum of total N from the synthetic fertilisers di-ammonium-phosphate (DAP) and urea, as well as the organic fertiliser, farm yard manure (FYM). The maximum yield and N application rates were decided upon by CIMMYT experts working in the region. The minimum yield threshold was determined by the average seed-rate of Kenya, which had the lower average of the two. As a result of these rules, we omitted 5% and 0.9% of the dataset as outliers in Kenya and Ethiopia, respectively. The rate of FYM application was considered high (“FYM_N high”) when more than 50% of the N applied was applied as FYM. Accordingly, those points that are not designated as such applied less that 50% of the N as FYM. The N content of aboveground and belowground maize was calculated using IPCC emission factors (IPCC, 2006), as is done in the CFT. This factor was used only for the amount of residue that was actually applied to the soil. This does not mean that all of the residue N is available to the crop. “Residue high” refers to those farmers that returned more than 50% of the residue available to the soil. The absolute amount of residue depended on the yield. The functions and significance of the different regression analyses to estimate the relationship between variables has been determined by performing multiple linear regression analysis in R as described by Townend (2002). Total GHG emissions calculated in the CFT were compared to total GHG emissions (including emissions from land use change) resulting from extensification based on Palm et al. (2010) in order to determine the yield threshold below which GHG emissions per tonne maize grown on existing agricultural lands are greater than those associated with land use conversion. Palm et al. (2010) reported GHG emissions of 2600 kg CO2e/ha associated with maize production following land use change from a “dense woody cover” for which a yield of 1.7 t/ha was assumed with a crop area expansion of 75% in the ‘Millenium Village’ Sauri, Siaya district in Kenya. The corresponding GHG emissions per tonne of maize were 1529 kg CO2e/t maize grain yield, which was reported for
some yields within the SIMLESA dataset. The maximum of these observed yields with GHG emissions greater than 1529 kg CO2e/t maize grain yield was used to provide the threshold yield that needs to be achieved in order to ensure that GHG emissions are lower than extensifying production, hence triggering land use change. This threshold maize grain yield was calculated to be 0.7 and 0.6 t/ha for Kenya and Ethiopia, respectively. In 2009, average maize consumption has been reported to be 77.2 kg/year/capita in Kenya and 44.2 kg/year/capita in Ethiopia. The average overall food calorie intake was almost 2100 kcal/capita/day for both countries in the same year (FAOSTAT, 2014), which is below the recommended daily calorie intake of 2400 and 2800 kcal/capita/day for women and men with a height of 1.7 m and a weight of 60 kg (FAO/WHO/UNU, 2001). A household’s ability to meet current average maize consumption was calculated by summing up the total maize harvested in a year from all mono- and inter-cropped plots available to a household. The harvested amount was then checked to determine whether it provides the average amount of maize consumed for the respective size of the family. It was found that yields to meet current average maize consumption needed to be between 0.8 and 2.0 t/ha in Kenya and 0.3 and 0.9 t/ha in Ethiopia, depending on family size (ranges of average adult equivalents in districts: Kenya 4–5.9; Ethiopia 3.8–6.3) and area available (ranges of average available area in districts: Kenya 0.23–0.41 ha; Ethiopia 0.26–0.74 ha). To calculate the theoretical yield for a given amount of available N, the N requirement of maize was derived by applying the figures given by Belfield and Brown (2008) where each tonne of maize requires 40 kg N/ha. This N requirement includes the fact that the plant can only access about 50% of available nitrateN in the soil and that an average of 20% of fertiliser N is lost. This would be equivalent to a nitrogen use efficiency of 40% (Belfield and Brown, 2008). The N application rate has been binned into 20 kg N/ha input ranges with exception of the first and last bin of effectively only 10 kg N/kg (0–10, 10–30, 30–50, . . . , 190–200 kg N/ha). The theoretical GHG emissions per tonne maize have been derived by calculating GHG per hectare if the amount of N, using the midpoint of N bins, was applied as urea. This has then been divided by the theoretical yield at that N bin. Here, the zero N bin has an average N application of 5 kg/ha corresponding with a yield of 0.125 t/ha.
Table 1 Cool Farm Tool (CFT) categories, which are used in this study (for a description of all categories (Hillier et al., 2011b)) and match with data. Label (number of total categories used by CFT)
CFT category used in analysis
Comments
Soil texture (3)
Medium
Soil organic matter (4)
0–1.72
Soil moisture (2)
Moist Dry
Soil drainage (3)
Good Poor
Soil pH (4)
1–5.5 5.5–7.3
Most districts in Kenya were estimated to have medium textured soil with the exception of Siaya with coarsemedium. However, as the texture was still not overall considered to be coarse, medium was still used. For Ethiopia medium conditions were also used throughout All districts fall within both categories as SOM of data ranges between 0.9 and 3.4%, this study uses 0– 1.72 only for simplification All plots are considered to be “moist” Unless: variable “stress” is given as drought at plot level and plots are not irrigated Percentage of dry plots are 55% and 18% for Kenya and Ethiopia, respectively All plots are considered to be of “good” drainage Unless: variable “stress” is given as “waterlogged” Percentage of waterlogged soils are 0.8% and 3.4% for Kenya and Ethiopia, respectively Districts: Bungoma (4.4–5.9), Imenti South (5.2) Districts: Siaya (5.0–6.6), Embu (5.6), Meru South (5.8) All districts in Ethiopia were considered to be of soil pH (5.5–7.3)
Croptype (5) Climate (2) Fertiliser application method (5) Residue management (6)
Other crop Tropical Broadcast Incorporated into field or “Left on field for soil fertility” left as mulch “Burnt” Burned Exported Used for “construction”, “sold”, “fed to livestock”, “firewood”, ”other uses”
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
141
Fig. 2. The relationship between greenhouse gas (GHG) emissions per hectare and nitrogen (N) application; N application is the sum of synthetic fertilisers and farmyard manure (FYM), (A) for Kenya and (B) for Ethiopia show the CFT result next to calculations based on IPCC guidelines and field measurements as presented in Linquist et al. (2012), bars give the range of the 95% confidence interval; ‘CFT total’ are total emissions as calculated by the Cool Farm Tool (CFT), ‘CFT field’ are emissions from the field only, i.e. excluding emission from the production of fertiliser and agrochemicals, ‘IPCC default’ are GHG emissions as calculated by IPCC guideline methodology considering field N2O emissions from fertiliser, CO2 emissions from urea and N2O emissions from residue management.
Fig. 3. The relationship between greenhouse gas (GHG) emissions per hectare and nitrogen (N) application given as open circles unless otherwise described in legend; N application is the sum of synthetic fertilisers and farmyard manure (FYM), (A) for Kenya and (B) for Ethiopia show the impact of different management options and soil conditions on GHG emissions. The legend refers to the N application: ‘FYM_N high’ is where more than 50% of the N applied is from FYM, ‘residue high’ refers to plots where the proportion of residue returned to the soil by the farmer is more than 50% of the residue available and ‘residue high +N’ refers to the same plots but where the N content of the residues is added to the total N application.
The local attainable yield was determined based on the upper 95th percentile of data from a farmer survey following the approach described by Van Ittersum et al. (2013). The amount of soil organic carbon (SOC) depleted with each 20 kg N/ha supplied by the soil was derived by using the C:N ratio and average bulk density of Kenyan soils using the SOTER database (ISRIC, 2013). These data were not available for Ethiopia, so the same values were used as for Kenya. Considering the high uncertainty of soil data in general and the use of only one value for different soil types, it was considered an appropriate approximation. The average proportion of N in relation to C was determined to be 11.5%, and the bulk density in the top 20 cm horizon was 1.25 g/cm3. Therefore 20 kg N/ha is equivalent to about 0.001% N in the top horizon, hence corresponding to about 0.01% SOC. 3. Results 3.1. Greenhouse gas emissions per hectare The most important factor in determining GHG emissions per hectare is the amount of N applied as synthetic and organic fertiliser (Figs. 2 and 3), with a significant regression line for both
countries (Table 2). The GHG emissions per hectare with and without inclusion of emissions from production of fertiliser and agrochemicals are within the averages reported for global field measurements as collated by a meta-analysis by Linquist et al. (2012) (Fig. 2). In comparison, results from IPCC guideline Table 2 Linear regressions for greenhouse gas emissions (GHG) per hectare and nitrogen (N) applications (GHG per hectare = Intercept + Slope N application) for Kenya and Ethiopia. N application is the sum of synthetic fertilisers and farmyard manure; N from maize residue is added to this where stated. Intercept Kenya N application N application + N from residue N application for waterlogged soils only N application for ‘FYM high’ only Ethiopia N application N application + N from residue N application for waterlogged soils only N application for ‘FYM high’ only a
Slope
351 7.51 328a 7.45 Not available as only 3 cases 318a 7.41
326 319a 524a 290a
7.95 7.98a 7.95 7.85
Significantly different (p < 0.05) to top function of respective country group.
142
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
Fig. 4. Relationship between greenhouse gas (GHG) emissions per tonne of maize grain and yield for Kenya (A) and Ethiopia (B) given as open circles unless otherwise described in legend. The legend refers to the N application, which is the sum of synthetic fertilisers and farmyard manure (FYM) as described in Fig. 3; ‘FYM_N high’ is where more than 50% of the N applied is from FYM, ‘residue high’ refers to plots where the proportion of residue returned to the soil by the farmer is more than 50% of the residue available and ‘residue high +N’ refers to the same plots but where the N content of the residues is added to the total N application. The solid black horizontal line marks the GHG emissions of extensifying production at the cost of deforestation based on Palm et al. (2010). The solid black vertical line marks the maximum yield observed at the equivalent GHG emissions in the data. The dashed vertical line mark the yield that is required for the household in the different districts to meet the average maize consumption depending on the average family size and maize area of the household (see Section 2 for full details), The dotted horizontal lines mark the range of GHG emissions per tonne maize reported by Linquist et al. (2012) of their meta analysis of real field measurements.
procedures would have a smaller slope though they are also still within the lower confidence interval of the global field measurement dataset. Some of the variation in Fig. 3 can be explained by the drainage of the soil, type of fertiliser applied and amount of residue returned to the soil. The size of the intercept is the result of background emissions (due to breakdown of residual soil organic matter (SOM), and use of residual nutrients), which can be observed on cropland without any amendments of N. Residual emissions are mainly dependent on soil type and level of water saturation. The impact of waterlogged soils on GHG emissions per hectare is demonstrated clearly in Ethiopia, where emissions on waterlogged plots are distinctly above the average regression line, with an intercept that is about 40% higher (Fig. 3B, Table 2). Kenya has an overall higher intercept than Ethiopia, as several factors contribute to background GHG emissions in the absence of fertiliser. In this case, the main driver is the proportion of residue returned to the field, which increases GHG emissions due to mineralisation of the N in the residue (round symbols in Fig. 3). The actual amount of residue will depend on the reported yield. When residue N is considered as part of the total N application the intercept reduces significantly (p < 0.05; triangular symbols in Fig. 3), and the difference is only about 9 kg CO2e/ha. This difference can be explained by emissions from agrochemicals, the
use of which is more widespread in Kenya compared to Ethiopia (data not shown). Fig. 3 also demonstrates that residue addition can contribute significant amounts of N to the soil. It leads to lower emissions than when N is supplied as synthetic fertiliser only, however, N from residue will not usually be fully available to the plant. It also makes apparent that smallholders in Kenya return, on average, far more residue to the soil directly in comparison to Ethiopia (Fig. 3). Generally, crop residues are very valuable for smallholders, and the main alternative uses of residues are feed for livestock and for burning for fuel (data not shown). Like residue application to the field, FYM results in lower GHG emissions compared to equal amounts of fertiliser in synthetic form, which can also be observed in Kenya to some extent. Lower emissions from FYM are partly due to the main emissions from livestock not being assigned to cropland within the CFT whereas fertiliser production is. 3.2. Greenhouse gas emissions per tonne of maize A strong relationship between GHG emissions per tonne of maize and yield was found (Fig. 4). A relationship between N application and yield is not very strong due to various other stresses such as, e.g. water or phosphorus limitation and pests (Fig. 5). The minimum GHG
Fig. 5. Relationship between greenhouse gas yield and nitrogen (N) application in tonne per hectare for Kenya (A) and Ethiopia (B).
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
143
Table 3 Mean values of Kenyan and Ethiopian districts for sole cropped maize from the SIMLESA (Sustainable Intensification of Maize–Legume Systems for Food Security in Eastern and Southern Africa) dataset. Values in brackets are the amount of nitrogen (N) applied as farmyard manure (kilogram per hectare) in the ‘N applied’ column and the 95th percentile of observed yields to represent observed attainable yield in ‘yield’ column. Values of attainable yield corresponding to FAOSTAT 2009 country statistics have been taken from Mueller et al. (2012). District
Kenya Bungoma Embu Imenti South Meru South Siaya Mean of districts Kenya: FAOSTAT 2009 Ethiopia Adami Tulu Bako Tibe Dugda Gubuesyo Hawasa Zurya Meskan Mesrak Badawacho Pawe Shalla Mean of districts Ethiopia: FAOSTAT 2009
Number of plots (n)
N applied (kg/ha)
Residue applied (%)
39 99 72 136 30
38 (6) 36 (5) 45 (3) 47 (6) 17 (5)
35 12 9 11 76
1.4 1.4 1.5 1.7 1.2
(1.8) (1.9) (2.0) (2.4) (1.7)
646 615 695 691 517
1292 911 856 659 1183
376
40
18
1.5 (2.0) 1.3 (3.7)
653
870
203 277 109 124 106 136 133 112 233
7 (1) 60 (2) 16 (4) 56 (2) 70 (3) 25 (3) 29 (1) 31 (8) 18 (1)
6 1 14 2 4 1 0 9 9
2.0 2.4 1.8 2.8 2.7 2.0 2.0 2.1 2.4
(2.8) (3.2) (2.4) (4.0) (3.2) (2.7) (2.4) (3.0) (3.2)
389 796 453 793 888 503 536 565 480
335 519 650 475 430 346 382 426 282
1433
34
5
2.3 (3.0) 2.2 (5.7)
598
418
emissions per tonne maize achieved are 82 and 67 kg CO2e/t maize grain for Kenya and Ethiopia, respectively. The strong relationship between emissions and yield demonstrates that it is of utmost importance to ensure that maize yields are above a level where extensification (associated with land use change), would be more efficient. This would correspond to a maize grain yield of 0.7 and 0.6 t/ha for Kenya and Ethiopia, respectively. In Kenya, this threshold is even higher from a food security perspective, increasing the required maize grain yield to between 0.8 and 2.0 t/ha for Kenya, depending on the average family size and area cropped with maize in a given household of the respective region. This range for maize grain yield is between 0.3 and 0.9 t/ha for Ethiopia. Generally, maize yields are much lower in Kenya than in Ethiopia, and the best yielding district in Kenya is equivalent to the worst yielding district in Ethiopia with respect to yield and GHG emissions per tonne maize (Table 3). Districts in Kenya with high
Yield (t/ha)
GHG emissions in (kg CO2e/ha) (kg CO2e/t maize)
Food insecure (%)
25 12 10 6 34
1 5 11 2 2 13 1 28 3
GHG emissions per tonne maize tend to be those with a high percentage of food insecure households (i.e. Bungoma and Siaya in the western Kenyan highlands). Generally, Kenyan smallholders struggle far more to produce enough for their own consumption (Table 3), and yields are much lower, with some exceptionally low yields resulting in very high GHG emissions per tonne of maize. 3.3. N balance, yield gap and the identification of “climate smart” practices Fig. 6 depicts a box plot for yield and GHG emissions of each N bin, displaying a general trend of higher yields with a higher N input. Furthermore, the yield gap is demonstrated by a comparison between the actual yield and the theoretical yield for a given amount of available N based on the N requirement of maize (black triangles in Fig. 6). Lastly the theoretical GHG emissions for the
Fig. 6. Boxplots of greenhouse gas (GHG) emissions per tonne maize (dark grey) and corresponding observed yield (light grey) for each nitrogen (N) bin for Kenya (A) and Ethiopia (B). N bins are ranges of N application 10 kg N/ha. The number of plots represented within each boxplot is shown at the bottom of the graph. The region shaded grey contains those boxplots with less than 10 observations. Theoretic values for GHG emissions per tonne of maize (black squares) and yield (black triangles) have been provided to enable judgement of adequacy of N supply versus demand and its associated GHG emissions (see Section 2 for full details).
144
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
expected associated theoretical yield are plotted as black squares. Each 20 kg N/ha available to the plant will increase the theoretical yield by 0.5 t/ha. The theoretical GHG emissions per tonne maize demonstrate that for an N input above about 60 kg/ha the impact of the associated higher yields with higher N input starts levelling off though a real stagnation is only observed at levels of 160 kg N/ha. A comparison of the data to the theoretical values shows that, for Kenya, yields are actually higher than expected below an N input of 60 kg/ha and mainly less than expected above this amount. These higher than expected yields from lower N rates require the plant to “mine” N from the soils unless N is applied in a different form. This may well be the case to some extent in Siaya, reporting the lowest N rates and the highest proportion of residue returned to the soil (Table 3). Insufficient N supply would slowly decrease soil organic nitrogen (and soil organic carbon) content and reduce fertility and yields. For each 20 kg/ha N that is supplied by the soil, 0.01% SOC has to be mineralised. In contrast, if the yields are not achieved with higher N input, the N is vulnerable to be leached unless it contributes to an increase of SOC and may be available for the following season. Yields are likely to be limited by other factors in addition to N such as other nutrients, water or soil conditions. The data from Ethiopia look very different in that much higher yields are achieved and an N input of up to 120 kg N/ha results in yields greater than would be expected at that N addition rate (Fig. 6). Higher N rates may again be prone to losses. However, it should also be noted that fewer plots are represented at high N levels, and so these values are not so reliable, which is the reason they have been highlighted in grey. Generally, despite the higher maize grain yields achieved in Ethiopia (2.3 t/ha), overall, the average N rates (34 kg N/ha) are smaller than in Kenya (Table 3) resulting in a potentially higher N deficit. However, for a true analysis of an N balance, further factors would need to be considered such as, e.g. residual N from legumes grown in the previous year and atmospheric N deposition. With the information available, an amount of 60 kg N/ha and 120 kg N/ha can be recommended on average for Kenya and Ethiopia, respectively. However, if higher yields are achieved than would be expected according to the N input, this level would have to be higher. Generally, the location of the 95th percentile indicates that most districts should have the potential to increase their yield by at least 0.5 t/ha. 4. Discussion The results from Ethiopia and Kenya of GHG from maize production illustrate the dependence of GHG emissions on fertiliser usage. Background GHG emissions (320–524 kg CO2e/ha, range of intercept in Fig. 3, Table 2) were comparable to the measurements reported by Mapanda et al. (2011) (294–404 kg CO2e/ha) and Linquist et al. (2012) (Fig. 2). In comparison to Mapanda et al. (2011), there are considerable differences in the composition of GHGs, and N2O was reported to be much lower in their study. These are likely to be at least partially a result of technical issues in the field when measuring N2O. A major issue with non-continuous sampling is that it can miss spatial and temporal hotspots such as the “birch effect” (Dick et al., 2008; Mapanda et al., 2011). It is clear though, that the CFT does not capture differences in emissions with, for example, rainfall pattern and crop performance, which have been observed by Mapanda et al. (2011). It should be recognised that the CFT estimates of GHG emissions from N application based on algorithms developed by Bouwman et al. (2002), who stated that their “estimates cannot be compared to individual fields” as their results summarize a diversity of studies in regard to methodology and management. This is reflected in the match to globally reported averages as reported by Linquist et al. (2012). Generally, any positive or negative bias is of less importance when comparing samples from the same region with each other.
A clear outcome of this study supports the findings of Palm et al. (2010), that there are synergies between climate change mitigation and food security. Even though this study did not look in detail at scenarios of mitigation, it is clear that the achievement of a minimum yield, which will be specific for each country or region, is the main driver to reduce emission intensity, i.e. the GHG emission on a per-tonne-product basis. This has also been suggested by other researchers, but for temperate systems (Brentrup et al., 2004; Ma et al., 2012). Although mitigation on a per-hectare basis might simplistically appear to favour lower input production systems, this clearly overlooks the primary role of agriculture to provide food and materials, and also risks creating drivers for indirect land use change (Carlton et al., 2010). This emphasizes the importance of increased yields and supporting smallholder food security, in order to avoid opening up new land for cultivation. The reduction of GHG emissions on a per-tonne basis is compatible with the aim of closing the yield gap, as discussed in several recent papers (Mueller et al., 2012; Smith et al., 2013; Tittonell and Giller, 2013; Van Ittersum et al., 2013). The approach in this study also addresses the need for strong, locally relevant, bottom-up approaches compared to strictly global analysis performed at coarse spatial scales (Van Ittersum et al., 2013). Sufficient nutrient input is usually mentioned as a key to achieving and maintaining yields (Tittonell and Giller, 2012). The recommended average amount of N of 60 and 120 kg N/ha for Kenya and Ethiopia, respectively, is close to the amount of 80 kg N/ha already recommended by local agricultural extension services (Palm et al., 2010). However, our recommendation is based on actual achieved yields, here specific for two countries though it could be transferred to different areas and scaled to different levels. The respective nutrient input is selected to avoid nutrient depletion when supplying too little N or SOC loss due to soil organic matter mineralisation, as well as N leaching at superoptimal applications. Organic sources of N are usually not available in sufficient quantity to meet the needs of high yielding crops. In addition, the maize residues and FYM often immobilize available N initially (Gentile et al., 2011; Zingore et al., 2008). Organic fertilisers are often, therefore, applied with synthetic fertilisers (Bationo et al., 2004; Gentile et al., 2011; Nyamangara et al., 2004). Organic forms of N do have the advantage that they will contribute to the general fertility and water retention of the soil in the long-term. However, it should also be noted that especially in countries with erratic rainfall, the overall water balance and specific timing of rainfall also limits yields (Van Ittersum et al., 2013). The effect of water limitation has not been considered here as all the yields are water-limited. A change from current application levels to an N application of 60 and 120 kg N/ha for Kenya and Ethiopia, respectively, would increase nutrient input for most plots (Fig. 6). This would have a considerable impact on GHG emissions per hectare and regional total emissions, as is predicted for SSA (Reay et al., 2012; Smith et al., 2007). In this study it would double and triple the total GHG emissions for the area that maize is grown on for Kenya and Ethiopia, respectively. However, at the same time yield would be expected to increase accordingly. Van der Velde et al. (2013), e.g. estimated an average yield increase of 36% with addition of 10 kg/ha of N and P and up to a maximum of 190% with additions of 80 kg/ha N and 20 kg/ha P. There is a clear need to raise production to feed the population, particularly in Kenya, where currently almost 40% of the consumed maize is imported. This approach could be used to identify the optimum for increasing production with the lowest GHG costs. The modelling approach here, based on simple regional survey data, allowed the identification of optimal N rates, aiming for the most GHG efficient production systems. This is particularly valuable for low input systems as GHG emission per tonne product
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
are very sensitive to changes in yields at low levels. Expressing results in terms of a GHG intensity metric leads to analogous conclusions as those based on N balance and nutrient use efficiency alone. Indeed, the GHG optimum is achieved for both countries over the ranges at which the balance of N inputs and crop N uptake is achieved. As such, the identification of practices to minimise GHG intensity are compatible with the objectives of optimizing productivity and ensuring food security. Generally, the approach described here can be used to support policy decisions for optimal interventions to obtain realistically achievable food production with low GHG intensity in different regions. The analysis of N balance should be combined with other important factors such as water availability, soil properties (especially fertility and drainage characteristics), and the use of crop protection chemicals. This concerted approach can quantify the trade-offs between, and optimise, crop productivity and GHG emissions, essential for addressing the joint global challenges of food security and climate change mitigation. Acknowledgements We thank two anonymous reviewers for comments which considerably improved the manuscript. This research was conducted under the CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS) with additional funding from the Australian Centre for International Agricultural Research (ACIAR) for generation of the survey data under the ‘Sustainable Intensification of Maize–Legume Systems for Food Security in Eastern and Southern Africa (SIMLESA)’ project. The work contributes to the University of Aberdeen Environment and Food Security theme and the Scottish Food Security Alliance-Crops. The scripting of the CFT tool into Matlab was funded by the Social and Environmental Economic Research (SEER) into Multi-Objective Land Use Decision Making project (which in turn is funded by the Economic and Social Research Council (ESRC); Funder Ref: RES-060-25-0063). References Audsley, E., 1997. Harmonisation of environmental life cycle assessment for agriculture (No. Final Report. Concerted Action AIR3-CT94-2028. (Coordinator). European Commission DG VI Agriculture. Bationo, A., Nandwa, S.M., Kimetu, J.M., Kinyangi, J.M., Bado, B.V., Lompo, F., Kimani, S.K., Kihanda, F., Koala, S., 2004. Sustainable intensification of crop-livestock systems through manure management in eastern and western Africa: lessons learned and emerging research opportunities. Sustainable Crop-Livestock Production for Improved Livelihoods and Natural Resource Management in West Africa. Wageningen, The Netherlands. Belfield, S., Brown, C., 2008. Field Crop Manual: Maize – A Guide to Upland Production in Cambodia. NSW Department of Primary Industries, New South Wales. Bellon, M.R., Hodson, D., Bergvinson, D., Beck, D., Martinez-Romero, E., Montoya, Y., 2005. Targeting agricultural research to benefit poor farmers: relating poverty mapping to maize environments in Mexico. Food Policy 30, 476–492. Bouwman, A.F., Boumans, L.J.M., Batjes, N.H., 2002. Modeling global annual N2O and NO emissions from fertilized fields. Global Biogeochem. Cycles 16, 28-1–28-9. Brentrup, F., Küsters, J., Lammel, J., Barraclough, P., Kuhlmann, H., 2004. Environmental impact assessment of agricultural production systems using the life cycle assessment (LCA) methodology II. The application to N fertilizer use in winter wheat production systems. Eur. J. Agron. 20, 265–279. Carlton, R., Berry, P., Smith, P., 2010. Impact of crop yield reduction on greenhouse gas emissions from compensatory cultivation of pasture and forested land. Int. J. Agric. Sustain. 8, 164–175. Cool Farm Tool Institute, 2012. URL http://www.coolfarmtool.org/ (accessed 10.06.13). CSA (Central Statistical Agency), 2013. Agricultural Sample Survey 2012/2013 (2005 E.C.). Report on area and production of major crops (Private Peasant Holdings, Meher Season). Statistical Bulletin 532. Addis Ababa, Ethiopia. Denning, G., Kabambe, P., Sanchez, P., Malik, A., Flor, R., Harawa, R., Nkhoma, P., Zamba, C., Banda, C., Magombo, C., Keating, M., Wangila, J., Sachs, J., 2009. Input subsidies to improve smallholder maize productivity in Malawi: toward an African green revolution. PLoS Biol. 7, e1000023. Dick, J., Kaya, B., Soutoura, M., Skiba, U., Smith, R., Niang, A., Tabo, R., 2008. The contribution of agricultural practices to nitrous oxide emissions in semi-arid Mali. Soil Use Manage. 24, 292–301.
145
Ecoinvent Centre, 2007. Ecoinvent data v2.0 (No. Ecoinvent reports No. 1e25). Swiss Centre for Live Cycle Inventories, Dübendorf. FAO/IFA, 2001. Global estimates of gaseous emissions of NH3, NO and N2O from agricultural land. FAO and IFA, Rome. FAOSTAT, 2014. FAOSTAT Agricultural Data. Available at: http://faostat.fao.org/FAO/ WHO/UNU, 2001. Human Energy Requirements. FAO Food and Nutrition Technical Report Series, Rome. FAO/WHO/UNU, 2001. Human Energy Requirements. FAO Food and Nutrition Technical Report Series, Rome. Foley, J.A., Ramankutty, N., Brauman, K.A., Cassidy, E.S., Gerber, J.S., Johnston, M., Mueller, N.D., O'Connell, C., Ray, D.K., West, P.C., Balzer, C., Bennett, E.M., Carpenter, S.R., Hill, J., Monfreda, C., Polasky, S., Rockström, J., Sheehan, J., Siebert, S., Tilman, D., Zaks, D.P.M., 2011. Solutions for a cultivated planet. Nature 478, 337–342. Garnett, T., Appleby, M.C., Balmford, A., Bateman, I.J., Benton, T.G., Bloomer, P., Burlingame, B., Dawkins, M., Dolan, L., Fraser, D., Herrero, M., Hoffmann, I., Smith, P., Thornton, P.K., Toulmin, C., Vermeulen, S.J., Godfray, H.C.J., 2013. Sustainable intensification in agriculture: premises and policies. Science 341, 33–34. Gentile, R., Vanlauwe, B., Chivenge, P., Six, J., 2011. Trade-offs between the short- and long-term effects of residue quality on soil C and N dynamics. Plant Soil 338, 159–169. Godfray, H.C.J., Beddington, J.R., Crute, I.R., Haddad, L., Lawrence, D., Muir, J.F., Pretty, J., Robinson, S., Thomas, S.M., Toulmin, C., 2010. Food security: the challenge of feeding 9 billion people. Science 327, 812–818. Hickman, J.E., Havlikova, M., Kroeze, C., Palm, C.A., 2011. Current and future nitrous oxide emissions from African agriculture. Curr. Opin. Environ. Sustain. 3, 370–378. Hillier, J., Hawes, C., Squire, G., Hilton, A., Wale, S., Smith, P., 2009. The carbon footprints of food crop production. Int. J. Agric. Sustain. 7, 107–118. Hillier, J., Smith, P., Bandel, T., Daniels, S., Malin, D., Hamilton, H., Walter, C., 2011a. Farm-scale greenhouse gas emissions using the Cool Farm Tool: application of a generic farming emissions calculator in developing countries. In: Wollenberg, E., Nihart, A., Tapio-Biström, M.L., Grieg-Gran, M. (Eds.), Climate change mitigation and agriculture. Routledge, London, UK. Hillier, J., Walter, C., Malin, D., Garcia-Suarez, T., Mila-i-Canals, L., Smith, P., 2011b. A farm-focused calculator for emissions from crop and livestock production. Environ. Model. Softw. 26, 1070–1078. Hillier, J., Brentrup, F., Wattenbach, M., Walter, C., Garcia-Suarez, T., Mila-i-Canals, L., Smith, P., 2012. Which cropland greenhouse gas mitigation options give the greatest benefits in different world regions? Climate and soil-specific predictions from integrated empirical models Global Change Biol. 18, 1880–1894. IPCC, 2006. In: The National Greenhouse Gas Inventories Programme, Eggleton, H.S., Buendia, L., Miwa, K., Ngara, T., Tanabe, K. (Eds.), 2006 IPCC Guidelines for National Greenhouse Gas Inventories. IPCC, Institute for Global Environmental Strategies, Tokyo, Japan. IPCC, 2007. Summary for policymakers. In: Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M., Miller, H.L. (Eds.), The physical science basis. Contribution of working group I to the fourth assessment report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. ISRIC, 2013. Development of a Soil and Terrain Database for Kenya (KENSOTER). http://www.isric.org/projects/soter-kenya-kensoter. Kang’ethe, E., 2011. Situation Analysis: Improving Food Safety in the Maize Value Chain in Kenya. Report prepared for FAO. College of Agriculture and Veterinary Science, University of Nairobi, Nairobi. Mulwa, C., Kassie, M., Odendo, M., Ouma, J., 2013. Socioeconomic Profile of Maizelegume Based Farming System in Ethiopia: Characterization of Livelihood Strategies, Production Conditions and Markets. Mimeo, CIMMYT, Nairobi. Linquist, B., van Groenigen, K.J., Adviento-Borbe, M.A., Pittelkow, C., van Kessel, C., 2012. An agronomic assessment of greenhouse gas emissions from major cereal crops. Global Change Biol. 18, 194–209. Ma, B.L., Liang, B.C., Biswas, D.K., Morrison, M.J., McLaughlin, N.B., 2012. The carbon footprint of maize production as affected by nitrogen fertilizer and maize– legume rotations. Nutr. Cycl. Agroecosyst. 94, 15–31. Mapanda, F., Wuta, M., Nyamangara, J., Rees, R.M., 2011. Effects of organic and mineral fertilizer nitrogen on greenhouse gas emissions and plant-captured carbon under maize cropping in Zimbabwe. Plant Soil 343, 67–81. Mosier, A.R., Halvorson, A.D., Reule, C.A., Liu, X.J., 2006. Net global warming potential and greenhouse gas intensity in irrigated cropping systems in Northeastern Colorado. J. Environ. Qual. 35, 1584–1598. Mueller, N.D., Gerber, J.S., Johnston, M., Ray, D.K., Ramankutty, N., Foley, J.A., 2012. Closing yield gaps through nutrient and water management. Nature 490, 254–257. Muricho, G., Kassie, M., Odendo, M., Ouma, J., 2012. Characterization of Maize– Legume Farming Systems and Farm Households in Kenya: Analysis of Technology Choice, Resource Use, Gender, Risk Management, Food Security and Poverty Profiles. Mimeo, CIMMYT, Nairobi. Nyamangara, J., Piha, M.I., Giller, K.E., 2004. Effect of combined cattle manure and mineral nitrogen on maize N uptake and grain yield. Afr. Crop Sci. J. 11, 289–300. Palm, C.A., Smukler, S.M., Sullivan, C.C., Mutuo, P.K., Nyadzi, G.I., Walsh, M.G., 2010. Identifying potential synergies and trade-offs for meeting food security and climate change objectives in sub-Saharan Africa. PNAS 107, 19661–19666. R Core Team, 2012. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.
146
J. Bellarby et al. / Agriculture, Ecosystems and Environment 197 (2014) 137–146
Reay, D.S., Davidson, E.A., Smith, K.A., Smith, P., Melillo, J.M., Dentener, F., Crutzen, P.J., 2012. Global agriculture and nitrous oxide emissions. Nat. Clim. Change 2, 410–416. Smith, P., Martino, D., Cai, Z., Gwary, D., Janzen, H.H., Kumar, P., McCarl, B., Ogle, S., O'Mara, F., Rice, C., Scholes, R.J., Sirotenko, O., Howden, M., McAllister, T., Pan, G. X., Romanenkov, V., Rose, S., Schneider, U., Towprayoon, S., 2007, 2007. Agriculture, in: Mitigation. Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 497–540. Smith, P., Martino, D., Gwary, D., Janzen, H., Kumar, P., McCarl, B., Ogle, S., O’Mara, F., Rice, C., Scholes, B., Sirotenko, O., Howden, M., McAllister, T., Romanenkov, V., Schneider, U., Towprayoon, S., Wattenbach, M., Smith, J., 2008. Greenhouse gas mitigation in agriculture. Philos. Trans. R. Soc. B Biol. Sci. 363, 789–813. Smith, P., Haberl, H., Popp, A., Erb, K., Lauk, C., Harper, R., Tubiello, F.N., de Siqueira Pinto, A., Jafari, M., Sohi, S., Masera, O., Böttcher, H., Berndes, G., Bustamante, M., Ahammad, H., Clark, H., Dong, H., Elsiddig, E.A., Mbow, C., Ravindranath, N.H., Rice, C.W., Robledo Abad, C., Romanovskaya, A., Sperling, F., Herrero, M., House, J.I., Rose, S., 2013. How much land-based greenhouse gas mitigation can be achieved without compromising food security and environmental goals? Global Change Biol. 19, 2285–2302.
Tilman, D., Balzer, C., Hill, J., Befort, B.L., 2011. Global food demand and the sustainable intensification of agriculture. PNAS 108, 20260–20264. Tittonell, P., Giller, K.E., 2012. When yield gaps are poverty traps: the paradigm of ecological intensification in African smallholder agriculture. Field Crops Res. 143, 76–90. Townend, J., 2002. Practical Statistics for Environmental and Biological Sciences. John Wiley & Sons Ltd., Chichester, pp. UK. Van der Velde, M., See You, L.L., Balkovic, J., Fritz, S., Khabarov, N., Obersteiner, M., Wood, S., 2013. Affordable nutrient solutions for improved food security as evidenced by crop trials. PLoS One 8. Van Groenigen, J.W., Velthof, G.L., Oenema, O., Van Groenigen, K.J., Van Kessel, C., 2010. Towards an agronomic assessment of N2O emissions: a case study for arable crops. Eur. J. Soil Sci. 61, 903–913. Van Ittersum, M.K., Cassman, K.G., Grassini, P., Wolf, J., Tittonell, P., Hochman, Z., 2013. Yield gap analysis with local to global relevance–a review. Field Crops Res. 143, 4–17. Whittaker, C., McManus, M.C., Smith, P., 2013. A comparison of carbon accounting tools for arable crops in the United Kingdom. Environ. Model. Softw. 46, 228–239. Zingore, S., Murwira, H.K., Delve, R.J., Giller, K.E., 2008. Variable grain legume yields, responses to phosphorus and rotational effects on maize across soil fertility gradients on African smallholder farms. Nutr. Cycl. Agroecosyst. 80, 1–18.