Field Crops Research 236 (2019) 132–144
Contents lists available at ScienceDirect
Field Crops Research
journal homepage: www.elsevier.com/locate/fcr
Can yield variability be explained? Integrated assessment of maize yield T
gaps across smallholders in Ghana
Marloes P. van Loona,⁎, Samuel Adjei-Nsiahb, Katrien Descheemaekera,
Clement Akotsen-Mensahb, Michiel van Dijkc, Tom Morleyc, Martin K. van Ittersuma,
Pytrik Reidsmaa
a Plant Production Systems, Wageningen University, P.O. Box 430, 6700 AK, Wageningen, the Netherlands
b Forest and Horticultural Crops Research Centre, School of Agriculture, University of Ghana, Legon, Ghana
cWageningen Economic Research, Wageningen University & Research, Alexanderveld 5, 2585 DB, The Hague, the Netherlands
A R T I C L E I N F O A B S T R A C T
Keywords: Agricultural production in Ghana should more than double to fulfil the estimated food demand in 2050, but this
Yield gaps is a challenge as the productivity of food crops has been low, extremely variable and prone to stagnation. Yield
Yield potential gap estimations and explanations can help to identify the potential for intensification on existing agricultural
Integrated assessment land. However, to date most yield gap analyses had a disciplinary focus. The objective of this paper is to assess
Smallholder farms the impact of crop management, soil and household factors on maize (Zea mays) yields in two major maize
Crop modelling
Crop experiments growing regions in Ghana through an integrated approach.
Farm household survey We applied a variety of complementary methods to study sites in the Brong Ahafo and Northern region. Farm
household surveys, yield measurements and soil sampling were undertaken in 2015 and 2016. Water-limited
potential yield (Yw) was estimated with a crop growth simulation model, and two different on-farm demon-
stration experiments were carried out in 2016 and 2017.
There is great potential to increase maize yields across the study sites. Estimated yield gaps ranged between
3.8Mg ha−1 (67% of Yw) and 13.6Mg ha−1 (84% of Yw). However, there was no consistency in factors affecting
maize yield and yield gaps when using complementary methods. Demonstration experiments showed the po-
tential of improved varieties, fertilizers and improved planting densities, with yields up to 9Mg ha−1. This was
not confirmed in the analysis of the household surveys, as the large yield variation across years on the same
farms impeded the disclosure of effects of management, soil and household factors. The low-input nature of the
farming system and the incidence of fall armyworm led to relatively uniform and low yields across the entire
population. So, farmers’ yields were determined by interacting, and strongly varying, household, soil and
management factors.
We found that for highly variable and complex smallholder farming systems there is a danger in drawing
oversimplified conclusions based on results from a single methodological approach. Integrating household
surveys, crop growth simulation modelling and demonstration experiments can add value to yield gap analysis.
However, the challenge remains to improve upon this type of integrated assessment to be able to satisfactorily
disentangle the interacting factors that can be managed by farmers in order to increase crop yields.
1. Introduction example, it was estimated that on average, 20% of maize (Zea mays)
yield potential is achieved across Ghana (GYGA, 2018). In addition, due
Agricultural production in sub-Saharan Africa (SSA) should triple to to climate change, the frequency of drought in SSA is expected to in-
fulfil the estimated food demand in 2050 (Godfray et al., 2010; van crease (Hulme et al., 2001; IPCC, 2014) and as a result, Ghana’s de-
Ittersum et al., 2016). In Ghana, as in much of SSA, this is a challenge as pendence on rain fed agriculture for food production would become
yields of food crops are low and productivity is extremely variable and more precarious (De Pinto et al., 2012).
has been stagnating in many areas over recent years (MOFA, 2010). For Sustainable agricultural intensification is recognised as one of the
⁎ Corresponding author.
E-mail address: marloes.vanloon@wur.nl (M.P. van Loon).
https://doi.org/10.1016/j.fcr.2019.03.022
Received 3 July 2018; Received in revised form 29 March 2019; Accepted 29 March 2019
Available online 09 April 2019
0378-4290/ © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license 
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
M.P. van Loon, et al. Field Crops Research 236 (2019) 132–144
Fig. 1. a) Map of Ghana indicating the location of Nkoranza and Savelugu municipalities in the Brong Ahafo and Northern regions, respectively; b) Location of
surveyed villages in Savelugu municipality; c) Location of surveyed villages in Nkoranza municipality.
main strategies for increasing food production (van Ittersum et al., 2012; Silva et al., 2018). Moreover, a body of work analysing technical
2016), especially in densely populated areas where the potential to efficiency of maize producers in Ghana also points to numerous, and
expand agricultural land is limited. Moreover, as experienced in Ghana, varied factors explaining yield differences among farmers, including
available land for farming is becoming increasingly scarce due to po- seed input, agrochemicals, fertilization, labour, farm experience, age,
pulation increase and competition with other economic activities education and access to credit (Appendix A).
(Braimoh and Vlek, 2006). The diversity of smallholder farming systems and households in SSA
Yield gap estimations and explanations provide important in- (e.g. in terms of soils, risk perceptions, wealth) (Affholder et al., 2013;
formation on the potential for intensification on existing agricultural Berre et al., 2017; Tittonell et al., 2010, 2007; Tittonell and Giller,
land (Laborte et al., 2012; Lobell et al., 2009; Van Ittersum et al., 2013), 2013) compel one to dig deeper into yield gap analysis. However, until
and provide a starting point for ensuring that intensification happens in now, most yield gap analyses of Ghana are based on either surveys
a sustainable matter. The yield gap (Yg) is the difference between actual (Abdulai et al., 2013; Addai and Owusu, 2014; Kuwornu et al., 2013),
farmers’ yield (Ya) and yield potential under irrigated (Yp) or rainfed modelling (Mueller et al., 2012) or field experiments (Sileshi et al.,
(Yw) conditions. Yp is yield determined by growth-defining factors, i.e. 2010). Integrated assessments, using multiple methods at multiple le-
plant characteristics, temperature, CO2, and solar radiation (Evans, vels, might be able to unravel some of the complexity experienced by
1996; Van Ittersum and Rabbinge, 1997). Besides the growth-defining farmers. Causes of yield gaps have been analysed through models and
factors, Yw is also influenced by soil type, field topography and pre- surveys in different countries (Rattalino Edreira et al., 2017; Silva et al.,
cipitation (Evans, 1996; Van Ittersum and Rabbinge, 1997). 2017, 2018), but up to now these approaches were not combined with
A review of yield gap explanation factors indicated that for SSA, field experiments. Combining the approaches and different levels of
fertilization (timing and amount) often explained the yield gap for analysis (including field and farm level) is expected to result in relevant
various crops (Beza et al., 2017). Other analyses found that improving insight into the factors explaining yield gaps. Our main objective in this
fertilization alone could narrow the yield gap in SSA by 50% (Mueller study is therefore to assess the impact of crop management, soil and
et al., 2012). In Ghana, increased fertilizer use has been shown to en- household factors on maize yields and yield gaps in Ghana by in-
hance yields, but, a low soil fertility status can limit the yield response tegrating different methods and data sources. Subsequently, we aim to
(Chapoto et al., 2015). In practice, current fertilizer use in Ghana is low identify options that could be used to mitigate yield gaps. We focus on
(i.e., on average 5 kg N ha−1) (FAO, 2018) and this has been explained maize production by smallholder farmers because in Ghana agriculture
in terms of the risk associated with relatively high costs of fertilizer in is mainly carried out by smallholder farmers (Chamberlin, 2008) who
relation to the uncertainty in yield gains (Agyare et al., 2014; Sileshi widely produce maize, a major food crop in the country (Angelucci
et al., 2010). et al., 2013; Wood, 2013).
Besides fertilization, additional biophysical and socioeconomic
factors also contribute to the yield gap. This complexity has been
confirmed by studies showing that factors affecting the yield gap can
vary substantially between farms (Kihara et al., 2015; Mueller et al.,
133
M.P. van Loon, et al. Field Crops Research 236 (2019) 132–144
Fig. 2. Monthly precipitation (bars) and average monthly minimum (squares) and maximum (triangles) daily temperatures for 2015 and 2016 in a) Nkoranza
(Source: Sunyani weather station) and b) Savelugu (Source: Savelugu weather station). Grey dotted lines and arrows indicate the growing seasons, which last in
Nkoranza from September to January (minor season) and from April to August (major season), and in Savelugu from May to October.
2. Material and methods and Langa (9.57 °N, 0.89 °W), which are located six, eight and 11 km
away from Savelugu, the municipal capital, respectively (Fig. 1b). The
2.1. Study sites villages in both municipalities were selected in consultation with the
local agricultural extension agents based on the extent of maize culti-
The study was conducted in 2015, 2016, and 2017 growing seasons vation in the community, proximity to the main capital, i.e. the market,
in Nkoranza and Savelugu municipalities, located in two different and for representativeness.
agroecological zones (AEZ) that represent major maize growing regions
in Ghana and across large areas of West Africa (Fig. 1). Each AEZ has
markedly different agroecological and market conditions. 2.2. On-farm data
Nkoranza, located in the Brong Ahafo region of Ghana, is in the
forest/savanna transitional AEZ (Fig. 1a). The region experiences a 2.2.1. Household survey
major rainy season commencing in April and a minor rainy season Farm household surveys were conducted during the 2015 and 2016
commencing in September (Fig. 2a). This region accounts for the maize growing seasons. The same households were surveyed in both
highest proportion of maize production within the country (MOFA, years (as yields and management differ per year), 15 per village and 90
2015). Data collection in Nkoranza was carried out in three villages, households in total. The selected households represented different
namely, Bibiani (7.55 °N, 1.77 °W), Dandwa (7.52 °N, 1.73 °W), and wealth statuses within each of the villages. Wealth status was based on,
Broahohuo (7.48 °N, 1.73 °W), which are located eight, five and 12 km among others, household size, livestock herd size, and farm area.
away from Nkoranza, the municipal capital, respectively (Fig. 1c). Individuals within the village, with an overview of all households, as-
Savelugu, located in the Northern region, is in the southern Guinea sisted in categorising the households into low-, medium- and high-
savannah AEZ and has a single rainy season commencing in May and wealth categories. Depending on the proportion of households in the
ending in October (Fig. 2b). Data collection was carried out in three three wealth categories in each village, three to eight households within
villages, namely, Nyatua (9.63 °N, 0.77 °W), Kpendua (9.66 °N, 0.89 °W) each category were selected to constitute 15 farms per village.
In Nkoranza, each of the selected households were surveyed three
134
M.P. van Loon, et al. Field Crops Research 236 (2019) 132–144
times during consecutive maize harvests. Households were visited maize varieties with the local maize variety being cultivated by farmers.
during the 2015 minor season and both the major and minor seasons in The maize varieties used were the farmers’ local variety, a drought
2016. In Savelugu the survey was conducted during harvest in tolerant maize (open pollinated variety), Obatanpa (the recommended
November in both 2015 and 2016. open pollinated variety) and a hybrid (Pannar in 2016 and Proseed in
The survey included information on the household’s socio-economic 2017). In total there were four treatments, and a treatment plot mea-
characteristics, farm characteristics and field-level maize management sured 10 by 12m. The recommended planting density of 5.33 plants
(Appendices C, D). Questions related to the household included topics m−2 was used on all plots.
on household composition, schooling, farm experience, labour avail- In 2016, there was an outbreak of fall armyworm (Spodoptera fru-
ability, household assets, and off-farm income. Questions related to the giperda) in both Nkoranza and Savelugu with the incidence being more
farm included farm structure, livestock herd size, and crops grown on severe in Nkoranza than in Savelugu. The pest was controlled in
each of the plots. Furthermore, questions on management regarding the Nkoranza using K-Optima EC ® (Acetamiprid, 20 g l−1) and Lambda
maize plots included variety, sources of seed, labour, agrochemical cyhalothrin (16 g l−1), a systemic insecticide and in Savelugu with
inputs and use of maize residues. After the first round of the household Sunpyrifos 48 EC® (Chlorpyrifos ethyl, 480 g l−1). Two different che-
survey, the questions related to household characteristics were ex- micals were used based on their availability. In experiment 1 in both
cluded as we assumed these factors did not change between seasons. 2016 in Nkoranza and 2017 in Savelugu, yield data from only three
farms were obtained, because of the outbreak completely destroyed the
2.2.2. Maize yield measurement maize crop in one of the fields.
Maize yields were measured in each maize plot belonging to each of For both demonstration experiments maize yields were determined
the surveyed farmers. In each maize plot, three random areas mea- in the same way as during the household survey (Section 2.2.2).
suring five by five metres, were selected for sampling. Maize plants Farmers’ field days were held in each of the villages where the ex-
were counted to determine plant density at harvest and, cobs were periments were running at three different periods during crop growth:
removed, de-husked and marketable cobs counted and weighed. From first fertilizer application, tasselling, and harvest. During the field days
the marketable cobs, ten were randomly selected, weighed and oven- in 2016 there were on average 57 participants per field day, resulting in
dried at 70 °C for two days to determine the dry matter content. Maize a total attendance of 680. In 2017, there were on average 47 partici-
grain yield data are presented with a correction for moisture content of pants per field day, resulting in a total attendance of 713. During the
12%. field day at harvest, group discussions were held to rank the different
treatments based on farmers’ preferences, and a final ranking was ob-
2.2.3. Soil sampling and analysis tained based on a general consensus among the group.
Soil was sampled from every maize plot in the household survey.
Soil samples were randomly collected at four sites within each field.
Samples were taken from 0 to 20 cm depth in a y-shaped manner. These 2.3. Yield gap concepts and determination
were bulked and a composite sample was taken for analysis. The sam-
ples were air dried, sieved with 2mm mesh and analysed for both Since maize production by smallholder farmers in Ghana is mainly
physical and chemical properties at the Soil Research Institute of Ghana under rain fed conditions, we took the water-limited potential yield
in Kumasi. The chemical properties analysed were: organic matter (Yw) as a benchmark. The yield gap (Yg) in this study was thus defined
content (Walkely Black Procedure), total N (Kjeldahl method), plant as the difference between Yw and the farmers’ average actual yield (Ya)
available P (Bray-1 method), exchangeable cations (K, Mg, Ca, Na) (1M (Van Ittersum et al., 2013). We refer to the relative yield gap (%) as
NH4OAc method) and micro-nutrients (Fe, Mn, Cu and Zn; ammonium equal to [1-Ya/Yw] * 100. Next, we defined the highest farmer yield
acetate-EDTA extractable method). During the first round of the survey, (Yhf), which is the average yield above the 90th percentile (Silva et al.,
soil samples were taken on every maize plot. Given that maize plots 2017). Farmers’ average actual yield (Ya) and Yhf were based on the
were not necessarily the same in every season, soil samples were only measured maize yields from the household survey (section 2.2.2). To
taken in the next season if farmers cultivated maize on a different plot, estimate Yw for Nkoranza and Savelugu for the 2015 and 2016 seasons
with the assumption that the soil status was unchanged for plots which we used the crop growth simulation model Hybrid-Maize (Yang et al.,
were the same. 2004 2006), which has been successful in simulating maize yields under
various environmental conditions (Grassini et al., 2015; Meng et al.,
2.2.4. Demonstration experiments 2013), including Ghana (Van Ittersum et al., 2016). Details on model
In 2016 and 2017 two demonstration experiments were carried out calibration and testing can be found in Grassini et al. (2015); Meng
on four farmers’ fields in each of the selected villages, experiments and et al. (2013); Van Ittersum et al. (2016). The model simulates maize
years (Appendix B). Two villages from each Nkoranza and Savelugu development and growth using a daily time-step, based on temperature
were selected in 2016, and in 2017 again two from Savelugu, but three driven development, temperature-sensitive maintenance respiration,
villages from Nkoranza were selected. In Nkoranza, in 2016 the ex- vertical canopy integration of photosynthesis, and organ specific
periment was conducted in the minor season and in 2017 in the major growth respiration.
season. The experiments in 2016 included 16 farmers and those in 2017 Daily weather data for 2015 and 2016 were obtained from the
included 20 farmers (Appendix B); selection of the farmers was based Savelugu weather station and, in the case of Nkoranza, from the
on accessibility of the plot and willingness of the farmer to donate a Sunyani weather station.
plot. We used the same protocol for determination of the simulated
In experiment 1, we tested the effect of recommended and farmers’ sowing date and the planting density as that published in the Global
planting density, at three different fertilizer application rates, on maize Yield Gap Atlas (Grassini et al., 2015; GYGA, 2018; van Bussel et al.,
yield. The recommended planting density was 5.33 plants m−2, and 2015). For sowing, this means that it was done when cumulative
each of the participating farmers chose their own planting density. For rainfall was> 20mm within seven consecutive days in the specified
each of the planting densities an N:P:K (23:10:5) fertilizer was applied sowing window, as defined by the local agronomist from the Global
at rates 0, 250, 375 or 500 kg ha-1. These treatment combinations re- Yield Gap Atlas (Dobor et al., 2016; Wolf et al., 2015). Planting density
sulted in a total of eight treatments per farm. A treatment plot measured was obtained from the actual average water deficit in the region and the
10 by 12m. The hybrids Pannar (variety Pannar 53) and Proseed were relation between seasonal water deficit and planting density (Grassini
used in 2016 and 2017, respectively, on all plots. et al., 2009). A generic maize variety was used, with growing degree
In experiment 2, we compared the performance of three improved days specific for the area.
135
M.P. van Loon, et al. Field Crops Research 236 (2019) 132–144
Fig. 3. Actual maize yields of plots for a) 2015 minor season in Nkoranza; b) 2016 major season in Nkoranza; c) 2016 minor season in Nkoranza; d) 2015 main season
in Savelugu and; e) 2016 main season in Savelugu. The red continuous line is the average farmers’ yield (Ya), the blue dashed line is the highest farmers’ yield (Yhf) as
defined by the average yield above the 90th percentile, and the green dashed line is the simulated water-limited potential yield (Yw). The arrows indicate the yield
gaps (Yg). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).
2.4. Statistical analysis multiple linear regression (MLR) was performed using the stats package
in R. The MLR models had yield as the response variable and either the
Quality control of the data was done in consultation with the per- crop management, soil, or household factors as predictor variables. As
sons carrying out the household survey. Suspicious data or outliers were with the PCAs, the MLR analyses were performed for each season and
verified at households involved and rectified if needed. If deviations municipality combination and, variables with a large number of
could not be traced back, they were considered an outlier and removed missing values were excluded from the analyses. In addition, a model
when the value was outside the range of the mean +/- 3 times the was tested including both nutrient management and soil factors to-
standard deviation (McClelland, 2000). Different methods were used to gether. Including all factors and interaction was not feasible, because of
study variance in yield, and crop management, soil and household limited degrees of freedom. Factors which caused collinearity were not
variables. included in the MLR model (selection was based on Variance Inflation
First, to test for significant differences between seasons and muni- Factor, using vifstep from the usdm package in R). Both PCA and MLR
cipalities in crop management, soil, and household factors, Least analyses were performed for each season and municipality combination
Significant Difference (LSD) tests were performed using the agricolae separately, as the factors which affect maize yield were expected to
package in R. Correlation analysis was performed on the yield data to differ for each of the seasons and municipalities. We also performed
test for relationship per plot and per farm (based on average yield) pooled analyses using multi-level models, in which both season and
between the different seasons. municipality were included as random factors. These models did not
Second, to visualise patterns in the data and to explain variance in provide additional insights and results are therefore only presented in
yield and crop management and soil factors, a Principle Component Appendix I and not discussed in the paper.
Analysis (PCA) was performed using the Ade4 package in R. Five PCAs Finally, to test for significant differences between pairs of treat-
were performed: Nkoranza 2015 (minor season), Nkoranza 2016 (major ments in the demonstration experiments a LSD test was performed.
season), Nkoranza 2016 (minor season), Savelugu 2015 and Savelugu
2016.
Third, to test which factors significantly affected maize yield,
136
M.P. van Loon, et al. Field Crops Research 236 (2019) 132–144
3. Results correlated with nutrient inputs (Fig. 5a). In the 2016 major season in
Nkoranza, yield was negatively correlated with cation exchange capa-
3.1. Farming system characteristics city (Fig. 5b) and in the 2016 minor season, negatively correlated with
the amount of organic matter in the soil (Fig. 5c). In 2015 in Savelugu,
In Nkoranza, the households were characterised by their small fa- yield was positively correlated with nutrient inputs and negatively
mily size, numbers of livestock and total farm area. This contrasts to correlated with the percentage of clay in the soil (Fig. 5d), while in
Savelugu, where households had significantly more members, owned 2016 there was a positive correlation between yield and the percentage
greater numbers of livestock and had larger total farm areas (Appendix of clay in the soil (Fig. 5e).
C). In Nkoranza, rental of land for maize cultivation was fairly common,
while this was not the case in Savelugu (Appendix D). Both munici- 3.4. Factors affecting maize yield
palities also had similar characteristics: the use of inputs (fertilizer,
manure, seeds, and herbicides) was comparably low (Appendices C, D); Results from the MLR analysis were not consistent across seasons
next to herbicide no other agrochemicals were used; and the majority of and municipalities (Table 1 and Appendix H). This is not surprising,
the farmers used their own seeds saved from previous harvests and only because there was no correlation between yields from different seasons
a minority used certified improved seeds. on individual fields or farms (Fig. 4, Appendix E).
The soil texture of all maize plots in Nkoranza and Savelugu was The MLR models with the most explanatory power in general were
sandy or sandy loam. The soils in Nkoranza were generally more fertile those based on the combination of nutrient management and soil fac-
with more available soil P, and higher percentages of N and organic tors, where the R2 values ranged from 0.45 to 0.84 (Table 1). Individual
matter compared to Savelugu (Appendix C). factors that were statistically significant within the models pre-
dominantly related to soil variables and these varied between each
3.2. Maize yields and maize yield gaps model.
In general, there was not a strong and positive relationship between
The simulated Yw were generally high, reaching 16.2Mg ha−1 in N input and yield, except in 2015 in Savelugu (Fig. 6). Likewise, the
the 2016 major season in Nkoranza (Fig. 3b) and 14.1Mg ha−1 in 2015 other crop management variables included in the MLR analysis did not
in Savelugu (Fig. 3d). A comparison with a previous, long-term mod- help to explain the variation in yield. Plant density at harvest had a
elling study from the same regions confirmed that these values fall at significant effect on yield in three models. With respect to the house-
the top of the range of attainable yields (GYGA, 2018). In both muni- hold factors, the only significant, and consistent result from the MLR
cipalities, these high values are largely explained by the high amount of was that in Savelugu, for both seasons, there was a positive effect of
well-distributed rainfall during the season. In the 2016 major season in livestock herd size (TLU) on yields.
Nkoranza the total precipitation was 538mm while estimated reference
evapotranspiration was 663mm. In Savelugu, in both 2015 and 2016, 3.5. Demonstration experiments
total precipitation (608mm and 585mm, respectively) slightly ex-
ceeded evapotranspiration. Demonstration experiment 1, to test the effect of planting density at
The yield gap (Yg) was large in all season and municipality combi- different fertilizer rates, showed fertilizer application resulted in higher
nations, except for the 2016 minor season in Nkoranza (Fig. 3). Here maize yields compared to the treatment without fertilizer (Fig. 7;
the relative yield gap was 68%, while for the other season municipality P < 0.01). This was not the case for the 2016 minor season in Nkor-
combinations it ranged from 81% to 87%. The low relative yield gap in anza, where no significant differences were found between any of the
the 2016 minor season in Nkoranza (Fig. 3c) was the result of a lower treatments (Fig. 7a), which was probably due to the severe yield re-
Yw due to water limitation. The water limitation ([1 – Yw/Yp] * 100) duction across all plots due to the fall armyworm outbreak. Across all
was 60% for the 2016 minor season in Nkoranza, while it was less than sites, there was no significant difference in yield when comparing ap-
4% in the other season and municipality combinations presented here. plication rates of 500 kg ha−1 and 375 kg ha−1 (P= 0.65). During the
For each season municipality combination, the simulated and the actual group discussions of the field days the farmers indicated that 375 kg
lengths of the growing seasons were similar. However, in Savelugu the ha−1, i.e. equivalent to 86.25 kg N ha−1, was their preferred fertilizer
simulated sowing dates were substantially earlier than the farmers ac- application rate (Table 2), which is much higher than the farmers re-
tual sowing dates (day of year 152 versus 199 in 2015; day of year 155 ported average application rate of 27 kg N ha−1 (Appendix C).
versus 182 in 2016). This difference in sowing dates contributed to the No clear effect of planting density on yield was found in all muni-
large difference between Yw and Ya. cipality and season combinations (Fig. 7). Nevertheless, farmers in-
Overall, Ya was similar between the two municipalities (Fig. 3, dicated that they preferred the recommended planting density in
Appendix C). Within Nkoranza, Ya was significantly higher in the 2016 Nkoranza in both 2016 and 2017 and in Savelugu in 2016 (Table 2).
major season compared to the minor seasons in 2015 and 2016 This preference led to farmers’ planting at higher densities in the 2017
(P < 0.01). In Savelugu, Ya in 2016 was significantly higher than that round of the experiment, thus potentially impacting the comparison of
in 2015 (P < 0.01). results for that year. Looking at all factors together, farmers mentioned
For both municipalities, there was not a significant relationship they preferred the combination of 375 kg ha−1 of fertilizer, together
between yields from different seasons in the same plots (Fig. 4), and with the recommended planting density and the hybrid seed.
also not between the yield from different seasons when averaged per Yield was high (up to 9Mg ha−1) in the 2017 major season in
household (Appendix E). These unexpected results show that farmers Nkoranza across all treatments compared with the other season muni-
obtaining high yields in one season did not necessarily obtain high cipality combinations and when compared to yield levels found in the
yields in other seasons. survey. This is probably the result of the use of hybrid seeds in 2017.
The hybrid was also used in 2017 in Savelugu, but here, most likely due
3.3. Principle component analysis to drought at the end of July and beginning of August, yield levels were
similar to those in 2016.
Maize yield did not make substantial contributions to any of the Demonstration experiment 2 showed some differences in yield due
PCs. Whereas, relations between yield and other variables were de- to the maize varieties planted (Fig. 8). In the minor season in 2016 in
tected (Fig. 5, Appendix F, G), there was no consistency across seasons Nkoranza, there was no significant difference among the different
and sites. In the 2015 minor season in Nkoranza, yield was positively varieties (Fig. 8a), while in Savelugu in 2016, the hybrid Pannar had a
correlated with the amount of organic matter in the soil but negatively significantly higher yield than the farmers’ variety (Fig. 8b) (P= 0.04).
137
M.P. van Loon, et al. Field Crops Research 236 (2019) 132–144
Fig. 4. Comparison between seasonal maize yields per
plot in a) Nkoranza – the 2016 major season versus the
2015 minor season (Spearman's rank correlation=
0.27, P= 0.14); b) Nkoranza – the 2016 minor season
versus 2016 major season (Spearman’s rank correla-
tion = -0.16, P= 0.35); c) Nkoranza – the 2016 minor
season versus 2015 minor season (Spearman’s rank
correlation = 0.17, P= 0.39) and; d) Savelugu – the
2016 versus 2015 main season (Spearman’s rank cor-
relation = 0.20, P= 0.29). The continuous lines are
the 1:1 lines.
In 2017, in both Nkoranza and Savelugu, the hybrid Proseed resulted in 4.1. Growth-defining factors
significantly higher yields than the other three varieties in the experi-
ment (Fig. 8) (P < 0.01 for both municipalities). Despite the high yield The demonstration experiment showed higher yields could be ob-
of Pannar in 2016 the farmers indicated during the group discussion at tained when improved maize varieties are used (up to 9Mg ha−1), al-
harvest that they preferred the drought tolerant maize because those though this effect was not always observed (Fig. 8, Fig. 7b). Other
seeds are less expensive and can be recycled (Table 2). Proseed was studies confirm the potential of improved varieties compared to farmer-
preferred in 2017 due to the high yields, but farmers indicated that the saved seeds (Asiedu et al., 2008), under favourable production en-
price of the seeds is not within their reach (Table 2). vironments, including good management (Adu et al., 2014). Never-
theless, all farmers from the survey used open pollinated varieties and
saved their seeds for subsequent seasons. This is attributed to the lack of
4. Discussion cash at planting to obtain seeds in combination with the prohibitive
cost of hybrid seeds, as they explained during the group discussion.
The large yield gaps observed in this study in Nkoranza (i.e., 69% The demonstration experiments did not confirm that the re-
and 76% in the minor and major season) and Savelugu (Yg was 75%) commended planting densities out-performed farmers’ planting den-
municipalities show that there is substantial potential to increase maize sities in terms of maize yield. However, farmers perceived that in-
yields in the study sites and, more broadly across the forest/savannah creasing planting density did increase yield as they observed more
transitional and southern Guinea AEZs. This is consistent with what has plants per unit area (Table 2). Interestingly, those participating in the
been found for the whole of Ghana, for which the average yield gap is experiment changed their behavior and increased their planting density
80% (GYGA, 2018). However, despite the large potential to increase the in the second year of experiment 1. From farmers’ fields, we do not have
maize yield in Nkoranza and Savelugu municipalities in Ghana, we did precise data on planting densities, which hinders clear conclusions.
not find consistency in factors that have impact on maize yields and
yield gaps when combining data sources and methods of analyses across 4.2. Growth-limiting factors
sites, seasons and levels of analysis. In a review of yield gap analyses,
Beza et al. (2017) showed that, across studies, explanations of the yield According to model simulations, water supply was only sub-
gap often differ which can be a result of the factors considered in each stantially limiting crop growth during the 2016 minor season in
study. In discussing our results, we used the framework of van Ittersum Nkoranza. The favourable conditions are reflected in the high simulated
and Rabbinge (1997), which groups potential causes of the yield gap values for Yw (GYGA, 2018). It is possible that these simulations un-
into growth-defining, growth-limiting and growth-reducing factors. We derestimated water limitations on individual farmers’ fields as they
also discuss the challenging aspects of the results and the methodolo- were based on precipitation recorded at a single weather station for
gical, practical, and policy implications of this study. each municipality.
Analysis of the farm survey data did not clearly show that the use of
fertilizer increased maize yields for any of the municipality and season
138
M.P. van Loon, et al. Field Crops Research 236 (2019) 132–144
Fig. 5. Plots showing the relation between, and contribution of, different variables to the first two principle components (PC1 and PC2) for a) 2015 minor season in
Nkoranza; b) 2016 major season in Nkoranza; c) 2016 minor season in Nkoranza; d) 2015 main season in Savelugu and; e) 2016 main season in Savelugu. Variables
with latent vectors less than 0.5 (with the exception of yield) are not labelled (a variable with a latent vector with a value close to one indicates that the variable
makes a large contribution to the component axis). The average contribution to PC1 and 2 (%) = [(C1 * Eig1) + (C2 * Eig2)] / (Eig1 + Eig2), where C1 and C2 are
the latent vectors and Eig1 and Eig2 are the eigenvalues for PC1 and PC2, respectively. The correlation matrix of latent vectors is provided in Appendix G. The
percentage of variance explained by each principle component is presented on the axis labels in brackets (see also Appendix F).
combinations, while in the demonstration experiments fertilizer did environmental factors that interact with fertilizer input (Getnet et al.,
increase yields (Fig. 7 versus Fig. 8). A possible explanation for the 2016) were controlled for in the demonstration experiment, namely
differences in yield response to fertilizer between the household survey sowing density, time of fertilizer application, crop variety and pest
and the demonstration experiment could be the low fertilizer applica- control (Tittonell et al., 2008). Thus, the value of combining these re-
tion rates applied on farmers’ fields (i.e., on average 27 kg N ha−1; note sults lies in the implication that at least a minimum rate of fertilizer
this is substantially more than the 5 kg N ha-1 reported as average N should be supplied to positively impact yields.
input in Ghana by the FAO (FAO, 2018)) compared to the fertilization
rates used in the demonstration experiments (i.e., ranging between 57.5
and 115 kg N ha−1). It is thus not a great surprise to see such low rates 4.3. Growth-reducing factors
having minor effects on yields. Moreover, several management and
Our study lacked sufficient data to analyse the impact of pest
139
M.P. van Loon, et al. Field Crops Research 236 (2019) 132–144
Table 1
Significant factors from the multiple linear regression models of crop management, soil and household factors which affect maize grain yields in Nkoranza and
Savelugu for the different seasons. Estimates of each of the factors are given and the significance stated: *** P < 0.001, ** P < 0.01, * P < 0.05, + P < 0.1. R2 of
the models and the number of observations (n) are provided below the response variables. An x in the table indicates that the variable was excluded from the analysis
due to a large number of missing values. The full regression model can be found in Appendix H.
Nkoranza Savelugu
Crop management Minor season - 2015 Major season - 2016 Minor season - 2016 Main season - 2015 Main season - 2016
Plant density (harvest) −0.14 0.83*** 0.77** 0.07 0.54*
N input −0.08* 0.00 −0.04 0.08** 0.02
Model R2 0.30 0.36 0.37 0.40 0.38
n 36 46 34 42 40
Soil
Mn −0.39* −0.26+ −0.08 x x
Percentage silt 0.96 1.30* −0.95 2.06 −0.66
CEC K x exchange acidity 0.65 1.20 −0.36 0.50 2.54*
Soil P x Cu −1.36 −0.29 −1.17+ 2.44 0.90
CEC Mg 9.53 0.11 1.96 −6.15 0.04
CEC Ca x CEC Mg −6.28 −0.01 −1.40 3.29 0.45
Model R2 0.58 0.68 0.49 0.25 0.57
n 51 48 45 47 43
Nutrient management and soil
N input −0.04 0.01 −0.02 0.07* −0.01
Dummy variable no P input −0.02 0.05 −0.57* 0.11 −0.35
Soil P x Cu −3.55* 0.50 −0.94+ 2.21 1.70
CEC Mg 19.43* 0.55 1.54 −7.75 0.55
CEC Mg x CEC Ca −12.17* −0.43 −1.38 5.01 0.18
Mn −0.16 −0.31* −0.40 x x
Percentage silt 0.92 1.53* 0.60 2.34 −0.73+
Model R2 0.84 0.66 0.56 0.45 0.65
n 40 47 44 46 43
Household
TLU 0.00 0.01 −0.04 0.17** 0.14*
Family labour available −0.10 0.12 0.61 −1.66* −1.38
People in the household −0.22 −0.40+ 0.05 −2.97* −0.54
Family labour available x people in the household 0.35 0.26 −0.28 0.74* 0.55
Model R2 0.53 0.57 0.37 0.66 0.64
n 35 33 34 37 36
incidences and diseases. Fall armyworm reduced yields in both de- by the low-input nature of the systems (see relatively large difference
monstration experiments in 2016 and 2017, and probably also the between Yw and Yhf values in Fig. 3) and low soil quality in general (e.g.
surveyed farmers’ yields from the 2016 season, but no reporting was very low soil Zn). Finally, errors in data collection (e.g. arising from
done on actual pest infestations. The farm survey data indicated that no relying on farmer recall) complicate the identification of consistent
farmers applied pesticides during the 2016 seasons. Day et al. (2017) relationships (Kassie et al., 2013). Compared to the separate models per
indicate that maize yields in Ghana were reduced by 45% in 2017 due location and season, no additional insights were obtained by pooling
to fall armyworm attack. Other studies have indicated that agrochem- data in the multi-level regression model (Appendix I).
icals can be positive determinant of maize yield (Addai and Owusu,
2014; Awunyo-Vitor et al., 2016; Bempomaa et al., 2014; Oppong et al.,
2014) and integrated pest management should be part of sustainable 4.5. Methodological, practical and policy implications
intensification of maize production in our study sites as well.
The practical aim of this study was to identify options to mitigate
yield gaps. However, disentangling spatial and temporal factors in
4.4. Challenging results biophysical, social and economic domains that interact and impact crop
yields proved to be difficult. Nevertheless, our integrated approach
A first challenge of our results is that a relatively short term study combining different data sources and methods helped to reveal com-
like this one is not well equipped to capture factors that vary between plexity that must not be ignored, but instead be incorporated in yield
season, e.g. management decisions, pest and disease outbreaks and gap analysis. Employing different methods provided both confounding
weather patterns. This calls for caution in the interpretation of results, and complementary results. Therefore, we highlight the danger of
and for more long-term studies to better understand relationships and drawing over-simplified conclusions based on a single method that in
dynamics. In this study, we observed little relation between yield ob- turn may lead to potentially unjustified crop management re-
tained by individual farmers between seasons. This is likely due to the commendations.
temporal variability in climate, management, and household factors, Smallholder systems are complex and heterogeous and blanket re-
the impact of which becomes larger with sub-optimal management commendations to increase yields are often irrelevant, infeasible and
conditions (Fermont et al., 2009). Another study in Uganda reported risky (Berre et al., 2017; Giller et al., 2011). Our results confirm that, in
that farmers make significant changes to their crop management from terms of offering recommendations to mitigate yield gaps, a “basket of
season to season (Ronner et al., 2018). Even in long-term studies, cor- options” that includes technologies and practices that farmers can
relations between yields of farmers in different years are relatively low choose, try and adapt is more appropriate (Ronner et al., 2018). In this
(Silva et al., 2018). case, the farm-level “basket” may include a selection of adapted vari-
Second, all farmers in the study sites achieved yields far below Yw eties, fertilizer rates and pest control methods, i.e. options that all
with relatively small differences between the farms. The uniform and contribute to good agronomy. Further research that investigates how
low yields can be explained by the incidence of fall armyworm, but also such a “basket” is used over multiple seasons would be extremely
140
M.P. van Loon, et al. Field Crops Research 236 (2019) 132–144
Fig. 6. Yield versus N input for Nkoranza and Savelugu for the 2015 and 2016 seasons. Dashed lines represent the recommended N input.
valuable. Innovative data collection methods including crowdsourcing socioeconomic constraints faced by farmers, policies can contribute to
and sensor technology (Beza et al., 2017), and data science (e.g., closure of the maize yield gap in Ghana by improving the availability
Janssen et al., 2017) can contribute to integrated assessments to explain and access of farmers to improved seeds. Next, affordability of fertili-
yield gaps. We can nevertheless conclude that given the importance of zers is likely to contribute to yield improvements. Finally, we
Fig. 7. Average maize yield with standard error from demonstration experiment 1 for farmers’ chosen planting densities (F, dark colored bars) and recommended
planting density (5.33 plants m−2) (R, light colored bars) at fertilizer application rates (N:P:K 23:10:5) of 0 (grey bars), 250 (yellow bars), 375 (green bars) or 500
(blue bars) kg ha-1 for a) 2016 minor season in Nkoranza; b) 2017 major season in Nkoranza; c) 2016 in Savelugu and; d) 2017 in Savelugu. Bars labelled with
different letters within each graph indicate significant differences in yield, P < 0.05 based on the LSD analysis. The red continuous line is the average actual yield
(Ya) as measured during the household survey, the blue dot dashed line is the highest farmers yield (Yhf) as measured during the household survey, and the green
dashed line is the simulated water-limited potential yield (Yw) (see also Fig. 4). (For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article).
141
M.P. van Loon, et al. Field Crops Research 236 (2019) 132–144
Table 2
Ranking (1= best, 4=worst) of varieties, fertilizer application rates and planting densities based on general consensus of farmers’ preferences as indicated during
farmer field days at harvest of the demonstration experiments in the different seasons and municipalities. Comments regarding preferences are also included.
Nkoranza Savelugu
2016 2017 2016 2017 Comments
Fertilizer rate
(N:P:K, 23:10:5) (kg ha−1)
0 4 4 4 4 In 2016 and 2017 375 kg ha−1 of N:P:K fertilizer preferred, because: yield and growth performance was high and
250 2 2 2 3 was not much different from that of 500 kg ha−1, but was better compared to 0 and 250 kg ha−1; cob sizes were
375 1 1 1 1 uniform and their weights were close to that of 500 kg ha−1; 375 kg ha−1 fertilizer is similar to what is
500 3 3 3 2 recommended by research and extension; more cost effective to use 375 kg ha-1 than 500 kg ha-1; financial
constraints to purchase 500 kg ha−1 of fertilizer; in 2017 resulted in delayed drying of cobs thus delaying
harvesting for this treatment.
Planting density
Farmers 2 2 2 1 Recommended planting density preferred in Nkoranza 2016, and 2017 and Savelugu 2016 because: higher yields
Recommended 1 1 1 2 could be achieved with this higher planting density; in Nkoranza farmers noted that it gives more plants per unit
area, and they are already used to row planting so increasing the rate may not be too difficult.
Farmers planting density preferred in Savelugu 2017 because: recommended planting density involves row
planting, which requires more labour due to obstruction by trees in the field
Variety
Drought Tolerant Maize (DTM) 1 3 1 2 In 2016 DTM preferred because: uniform and with filled cobs; yield was high and consistent across all fields;
Farmer variety 4 2 4 4 recycling of seeds is possible unlike Pannar seeds; Pannar seeds are expensive.
Obatanpa 3 4 3 3 In 2017 Proseed preferred even though the price is not within reach of farmers because: high and consistent yield
Pannar 2 – 2 – across most fields; good and uniform germination; well-filled cobs; large grain size.
Proseed – 1 – 1
recommend improving integrated pest management by farmer capacity light of complementary methods that account for the multiple yield-
building and improved information access through extension services. determining factors acting at different spatial and temporal levels.
The combination of household survey, crop growth simulation
5. Conclusion modelling, and demonstration experiments with farmers’ feedback used
in this study adds value to yield gap analysis of maize in Ghana.
We conclude that there is great potential to increase maize yields Demonstration experiments showed that improved varieties and in-
within low-input smallholder farming systems in Ghana. However, creased fertilizer rates can lead to yields much closer to water-limited
identifying recommendations to achieve such yield improvements is far yields, whereas farmers’ yields differed among plots and seasons, and
less straightforward because farmers’ yields are determined by inter- explaining factors differed per season and site. Nevertheless, the ap-
acting, and often strongly varying, household, soil and management proach can be improved upon with additional data sources and
factors. Our study showed that it is often not possible to make sense of methods that better disentangle variation in space and time.
these factors on the basis of a single methodological approach. Instead,
results obtained through individual methods should be interpreted in
Fig. 8. Average maize yield with standard error for different varieties used in demonstration experiment 2 (note that in 2016 Pannar is used as a hybrid and in 2017
Proseed is used) in a) Nkoranza and b) Savelugu. The dark grey bars show results for 2016 and the light grey bars for 2017. Different letters indicate significant
differences in yield, P < 0.05, between the different treatments within each season and municipality combination.
142
M.P. van Loon, et al. Field Crops Research 236 (2019) 132–144
Acknowledgements resource use across farming zones in the Central Rift Valley of Ethiopia. Exp. Agric.
52, 493–517. https://doi.org/10.1017/S0014479715000216.
This research was part of the project Integrated assessment of the Giller, K.E., Tittonell, P., Rufino, M.C., van Wijk, M.T., Zingore, S., Mapfumo, P., Adjei-Nsiah, S., Herrero, M., Chikowo, R., Corbeels, M., Rowe, E.C., Baijukya, F., Mwijage,
determinants of the MAize yield Gap in Sub-Saharan Africa: towards A., Smith, J., Yeboah, E., van der Burg, W.J., Sanogo, O.M., Misiko, M., de Ridder, N.,
farm INnovation and Enabling policies (IMAGINE), funded by the Karanja, S., Kaizzi, C., K’ungu, J., Mwale, M., Nwaga, D., Pacini, C., Vanlauwe, B.,
DEGRP programme, the DFID-ESRC Growth Research Programme. We 2011. Communicating complexity: integrated assessment of trade-offs concerning soilfertility management within African farming systems to support innovation and de-
thank the Ministry of Food and Agriculture of Ghana and farmers in velopment. Agric. Syst. 104, 191–203. https://doi.org/10.1016/j.agsy.2010.07.002.
Savelugu and Nkoranza who participated in the project. We also thank Godfray, H.C.J., Beddington, J.R., Crute, I.R., Haddad, L., Lawrence, D., Muir, J.F., Pretty,
Wim Paas and Lenny van Bussel, from Wageningen University, for de- J., Robinson, S., Thomas, S.M., Toulmin, C., 2010. Food security: the challenge offeeding 9 billion people. Science 327, 812–818. https://doi.org/10.1126/science.
veloping the questionnaires for the household surveys and starting up 1185383.
the data collection, and Kwasi Gyan, from the University of Grassini, P., Yang, H.S., Cassman, K.G., 2009. Limits to maize productivity in Western
Development Studies at Nyankpala, who assisted in the data collection Corn-Belt: A simulation analysis for fully irrigated and rainfed conditions. Agric. For.
Meteorol. 149 (8), 1254–1265.
for both the surveys and the field experiments. In addition we thank Grassini, P., van Bussel, L.G.J., Van Wart, J., Wolf, J., Claessens, L., Yang, H., Boogaard,
Mink Zijlstra for assisting in the data analysis, and finally Madeleine H., de Groot, H., van Ittersum, M.K., Cassman, K.G., 2015. How good is good enough?
Florin for her great contribution to finalize the writing of the paper. Data requirements for reliable crop yield simulations and yield-gap analysis. Field
Crops Res. 177, 49–63. https://doi.org/10.1016/j.fcr.2015.03.004.
GYGA, 2018. Global Yield Gap and Water Productivity Atlas [WWW Document]. URL.
References (Accessed 9 June 2018). www.yieldgap.org.
Hulme, M., Doherty, R., Ngara, T., New, M., Lister, D., 2001. African climate change:
1900-2100. Clim. Res. 17, 145–168. https://doi.org/10.3354/cr017145.
Abdulai, S., Nkegbe, P.K., Donkoh, S.A., 2013. Technical efficiency of maize production in IPCC, 2014. Climate change 2014: impacts, adaptation, and vulnerability. In: Barros,
Northern Ghana. Afr. J. Agric. Res. 8, 5251–5259. https://doi.org/10.5897/ V.R., Field, C.B., Dokken, D.J., Mastrandrea, M.D., Mach, K.J., Bilir, T.E., M, C, Ebi,
ajar2013.7753. K.L., Estrada, Y.O., Genova, R.C., Girma, B., Kissel, E.S., Levy, A.N., MacCracken, S.,
Addai, K.N., Owusu, V., 2014. Technical efficiency of maize farmers across various agro Mastrandrea, P.R., L, W.L (Eds.), Part B: Regional Aspects. Contribution of Working
ecological zones of Ghana. J. Agric. Environ. Sci. 3, 149–172. Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate
Adu, G., Abdulai, M., Alidu, H., K Nustugah, S., Buah, S., Kombiok, J., Obeng-Antwi, K., Change. Cambridge University Press, Cambridge and New York, pp. 688.
Abudulai, M., M Etwire, P., 2014. Recommended Production Practices for Maize in Kassie, B.T., Hengsdijk, H., Rötter, R., Kahiluoto, H., Asseng, S., Van Ittersum, M., 2013.
Ghana. https://doi.org/10.13140/2.1.4376.3527. Adapting to climate variability and change: experiences from cereal-based farming in
Affholder, F., Poeydebat, C., Corbeels, M., Scopel, E., Tittonell, P., 2013. The yield gap of the Central Rift and Kobo Valleys, Ethiopia. Environ. Manage. 52, 1115–1131.
major food crops in family agriculture in the tropics: assessment and analysis through https://doi.org/10.1007/s00267-013-0145-2.
field surveys and modelling. Field Crops Res. 143, 106–118. https://doi.org/10. Kihara, J., Tamene, L.D., Massawe, P., Bekunda, M., 2015. Agronomic survey to assess
1016/j.fcr.2012.10.021. crop yield, controlling factors and management implications: a case-study of Babati
Agyare, W.A., Asare, I.K., Sogbedji, J., Clottey, V.A., 2014. Challenges to maize fertili- in northern Tanzania. Nutr. Cycl. Agroecosyst. 102, 5–16. https://doi.org/10.1007/
zation in the forest and transition zones of Ghana. Afr. J. Agric. Res. 9, 593–602. s10705-014-9648-3.
https://doi.org/10.5897/ajar12.1698. Kuwornu, J.K.M., Amoah, E., Seini, W., 2013. Technical efficiency analysis of maize
Angelucci, F., Asante-Poku, A., Anaadumba, P., 2013. Analysis of Incentives and farmers in the Eastern Region of Ghana. J. Soc. Dev. Sci. 4, 84–99.
Disincentives for Rice in Ghana. Technical notes series, MAFAP, FAO, Rome. Laborte, A.G., de Bie, K., C. A.J.M, Smaling, E.M.A., Moya, P.F., Boling, A.A., Van
Asiedu, E.A., Asante, R., Sallah, P.Y., Baduon, A., Avah, E., 2008. Effects of farmers’ seed Ittersum, M.K., 2012. Rice yields and yield gaps in Southeast Asia: past trends and
source on maize seed quality and crop productivity. Ghana J. Agric. Sci. 40, 105–111. future outlook. Eur. J. Agron. 36, 9–20. https://doi.org/10.1016/j.eja.2011.08.005.
https://doi.org/10.4314/gjas.v40i1.2159. Lobell, D.B., Cassman, K.G., Field, C., 2009. Crop yield gaps: their importance, magni-
Awunyo-Vitor, D., Wongnaa, C.A., Aidoo, R., 2016. Resource use efficiency among maize tudes, and causes. Annu. Rev. Environ. Resour. 34, 179–204.
farmers in Ghana. Agric. Food Secur. 5 (28). https://doi.org/10.1186/s40066-016- Mcclelland, G.H., 2000. Nasty Data. Handbook of Research Methods in Social Psychology.
0076-2. pp. 393–411.
Bempomaa, B., Henry, De-, Acquah, 2014. Technical efficiency analysis of maize pro- Meng, Q., Hou, P., Wu, L., Chen, X., Cui, Z., Zhang, F., 2013. Understanding production
duction: evidence from Ghana. Appl. Stud. Agribus. Commer. 8, 73–79. potentials and yield gaps in intensive maize production in China. Field Crops Res.
Berre, D., Corbeels, M., Rusinamhodzi, L., Mutenje, M., Thierfelder, C., Lopez-Ridaura, S., 143, 91–97. https://doi.org/10.1016/j.fcr.2012.09.023.
2017. Thinking beyond agronomic yield gap: smallholder farm efficiency under MOFA, 2010. Agriculture in. Facts and figures., Ghana.
contrasted livelihood strategies in Malawi. Field Crops Res. 214, 113–122. https:// MOFA, 2015. Agriculture in. Facts and figures., Ghana.
doi.org/10.1016/j.fcr.2017.08.026. Mueller, N.D., Gerber, J.S., Johnston, M., Ray, D.K., Ramankutty, N., Foley, J.A., 2012.
Beza, E., Silva, J.V., Kooistra, L., Reidsma, P., 2017. Review of yield gap explaining Closing yield gaps through nutrient and water management. Nature 490, 254–257.
factors and opportunities for alternative data collection approaches. Eur. J. Agron. https://doi.org/10.1038/nature11420.
82, 206–222. https://doi.org/10.1016/j.eja.2016.06.016. Oppong, B.A., Onumah, E.E., Asuming-Brempong, S., 2014. Stochastic frontier modeling
Braimoh, A.K., Vlek, P.L.G., 2006. Soil quality and other factors influencing maize yield in of maize production in Brong-Ahafo Region of Ghana. Agris On-line Pap. Econ.
northern Ghana. Soil Use Manag. 22, 165–171. https://doi.org/10.1111/j.1475- Inform. 6, 2.
2743.2006.00032.x. Ronner, E., Descheemaeker, K., Almekinders, C.J.M., Ebanyat, P., Giller, K.E., 2018.
Chamberlin, J., 2008. It’s a Small World After All: Defining Smallholder Agriculture in Farmers’ use and adaptation of improved climbing bean production practices in the
Ghana. highlands of Uganda. Agric. Ecosyst. Environ. 261, 186–200. https://doi.org/10.
Chapoto, A., Sabasi, D., Asante-Addo, C., Ghana, N., 2015. Fertilizer Intensification and 1016/j.agee.2017.09.004.
Soil Fertility Impact on Maize Yield Response in Northern Ghana. Agricultural & Sileshi, G., Akinnifesi, F.K., Debusho, L.K., Beedy, T., Ajayi, O.C., Mong’omba, S., 2010.
Applied Economics Association’s 2015 AAEA Annual Meeting, San Francisco, Variation in maize yield gaps with plant nutrient inputs, soil type and climate across
California. sub-Saharan Africa. Field Crops Res. 116, 1–13. https://doi.org/10.1016/j.fcr.2009.
Day, R., Abrahams, P., Bateman, M., Beale, T., Clottey, V., Cock, M., Colmenarez, Y., 11.014.
Corniani, N., Early, R., Godwin, J., Gomez, J., Moreno, P.G., Murphy, S.T., Oppong- Silva, J.V., Laborte, A.G., van Ittersum, M.K., 2017. Explaining rice yields and yield gaps
Mensah, B., Phiri, N., Pratt, C., Silvestri, S., Witt, A., 2017. Fall armyworm: impacts in Central Luzon, Philippines: an application of stochastic frontier analysis and crop
and implications for Africa. Outlooks Pest Manage. 28, 196–201. https://doi.org/10. modelling. Eur. J. Agron. 82, 223–241. https://doi.org/10.1016/j.eja.2016.06.017.
1564/v28_oct_02. Silva, J.V., Reidsma, P., Lourdes Velasco, M., Laborte, A.G., van Ittersum, M.K., 2018.
De Pinto, A., Demirag, U., Haruna, A., JawooAsamoah, M., 2012. Climate change, agri- Intensification of rice-based farming systems in Central Luzon, Philippines: con-
culture, and food crop production in Ghana. GSSP – IFPRI Policy note 3. . straints at field, farm and regional levels. Agric. Syst. 165, 55–70. https://doi.org/10.
Dobor, L., Barcza, Z., Hlasny, T., Arendas, T., Spitko, T., Fodor, N., 2016. Crop planting 1016/j.agsy.2018.05.008.
date matters: estimation methods and effect on future yields. Agric. For. Meteorol. Tittonell, P., Giller, K.E., 2013. When yield gaps are poverty traps: the paradigm of
223, 103–115. ecological intensification in African smallholder agriculture. Field Crops Res. 143,
Edreira, J.I.R., Mourtzinis, S., Conley, S.P., Roth, A.C., Ciampitti, I.A., Licht, M.A., Kandel, 76–90. https://doi.org/10.1016/j.fcr.2012.10.007.
H., Kyveryga, P.M., Lindsey, L.E., Mueller, D.S., Naeve, S.L., Nafziger, E., Specht, J.E., Tittonell, P., Vanlauwe, B., de Ridder, N., Giller, K.E., 2007. Heterogeneity of crop pro-
Stanley, J., Staton, M.J., Grassini, P., 2017. Assessing causes of yield gaps in agri- ductivity and resource use efficiency within smallholder Kenyan farms: soil fertility
cultural areas with diversity in climate and soils. Agric. For. Meteorol. 247, 170–180. gradients or management intensity gradients? Agric. Syst. 94, 376–390. https://doi.
Evans, L.T., 1996. Crop Evolution, Adaptation and Yield. Cambridge University Press. org/10.1016/j.agsy.2006.10.012.
FAO, 2018. FAOSTAT 2018: FAO Statistical Databases. URL (accessed 6.19.18). http:// Tittonell, P., Vanlauwe, B., Corbeels, M., Giller, K.E., 2008. Yield gaps, nutrient use ef-
www.fao.org/faostat/en/#country/81. ficiencies and response to fertilisers by maize across heterogeneous smallholder farms
Fermont, A.M., van Asten, P.J.A., Tittonell, P., van Wijk, M.T., Giller, K.E., 2009. Closing of western Kenya. Plant Soil 313, 19–37. https://doi.org/10.1007/s11104-008-
the cassava yield gap: an analysis from smallholder farms in East Africa. Field Crops 9676-3.
Res. 112, 24–36. https://doi.org/10.1016/j.fcr.2009.01.009. Tittonell, P., Muriuki, A., Shepherd, K.D., Mugendi, D., Kaizzi, K.C., Okeyo, J., Verchot,
Getnet, M., Van Ittersum, M., Hengsdijk, H., Descheemaeker, K., 2016. Yield gaps and L., Coe, R., Vanlauwe, B., 2010. The diversity of rural livelihoods and their influence
143
M.P. van Loon, et al. Field Crops Research 236 (2019) 132–144
on soil fertility in agricultural systems of East Africa – a typology of smallholder A. 113, 14964–14969. https://doi.org/10.1073/pnas.1610359113.
farms. Agric. Syst. 103, 83–97. https://doi.org/10.1016/j.agsy.2009.10.001. Van Ittersum, M.K., Cassman, K.G., Grassini, P., Wolf, J., Tittonell, P., Hochman, Z., 2013.
van Bussel, L.G.J., Grassini, P., Van Wart, J., Wolf, J., Claessens, L., Yang, H., Boogaard, Yield gap analysis with local to global relevance-A review. Field Crops Res. 143,
H., de Groot, H., Saito, K., Cassman, K.G., van Ittersum, M.K., 2015. From field to 4–17. https://doi.org/10.1016/j.fcr.2012.09.009.
atlas: upscaling of location-specific yield gap estimates. Field Crops Res. 177, 98–108. Wolf, J., Ouattara, K., Supit, I., 2015. Sowing rules for estimating rainfed yield potential
https://doi.org/10.1016/j.fcr.2015.03.005. of sorghum and maize in Burkina Faso. Agric. For. Meteorol. 214, 208–218.
van Ittersum, M.K., Rabbinge, R., 1997. Concepts in production ecology for analysis and Wood, T.N., 2013. Agricultural Development in the Northern Savannah of Ghana.
quantification of agricultural input-output combinations. Field Crops Res. 52, Doctoral Documents From Doctor of Plant Health Program. pp. 1.
197–208. https://doi.org/10.1016/S0378-4290(97)00037-3. Yang, H., Dobermann, A., Lindquist, J., Walters, D., Arkebauer, T., Cassman, K., 2004.
van Ittersum, M.K., van Bussel, L.G.J., Wolf, J., Grassini, P., van Wart, J., Guilpart, N., Hybrid-maize—a maize simulation model that combines two crop modeling ap-
Claessens, L., de Groot, H., Wiebe, K., Mason-D’Croz, D., Yang, H., Boogaard, H., van proaches. Field Crops Res. 87, 131–154. https://doi.org/10.1016/j.fcr.2003.10.003.
Oort, P.A.J., van Loon, M.P., Saito, K., Adimo, O., Adjei-Nsiah, S., Agali, A., Bala, A., Yang, H., Dobermann, A., Cassman, K.G., Walters, D.T., 2006. Features, applications, and
Chikowo, R., Kaizzi, K., Kouressy, M., Makoi, J.H.J.R., Ouattara, K., Tesfaye, K., limitations of the Hybrid-Maize simulation model. Agron. J. 98, 737. https://doi.org/
Cassman, K.G., 2016. Can sub-saharan Africa feed itself? Proc. Natl. Acad. Sci. U. S. 10.2134/agronj2005.0162.
144