Agricultural Systems 201 (2022) 103473
Contents lists available at ScienceDirect 
Agricultural Systems 
journal homepage: www.elsevier.com/locate/agsy 
The cocoa yield gap in Ghana: A quantification and an analysis of factors 
that could narrow the gap 
Paulina A. Asante a,b,f,*, Eric Rahn d, Pieter A. Zuidema b, Danaё M.A. Rozendaal a,c, 
Maris E.G. van der Baan b, Peter Läderach d,e, Richard Asare f, Nicholas C. Cryer g, 
Niels P.R. Anten a 
a Centre for Crop Systems Analysis, Wageningen University, P.O. Box 430, 6700 AK Wageningen, the Netherlands 
b Forest Ecology and Management Group, Wageningen University, P.O. Box 47, 6700 AA Wageningen, The Netherlands 
c Plant Production Systems Group, Wageningen University, P.O. Box 430, 6700 AK Wageningen, The Netherlands 
d International Center for Tropical Agriculture (CIAT), km 17 recta Cali-Palmira, 763537 Cali, Colombia 
e CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS), Alliance of Bioversity International and CIAT, Headquarters – Rome, Via dei Tre 
Denari 472/a, 00054 Maccarese (Fiumicino), Rome, Italy 
f International Institute of Tropical Agriculture (IITA), PMB L56 Legon-Accra, Ghana 
g Mondelēz UK R&D Ltd., Bournville Lane, Bournville, Birmingham B30 2LU, UK   
H I G H L I G H T S  G R A P H I C A L  A B S T R A C T  
• Increasing cocoa yields per unit area is a 
means to meet growing demand, secure 
food security& reduce pressure on 
forest. 
• We quantified cocoa yield gap by 
comparing water-limited, attainable 
yield in high- & low-input systems with 
farmer yields. 
• Considerable yield gaps on all cocoa 
farms but water-limited yield gaps were 
much larger than in high- and low-input 
systems. 
• Relative yield gaps are substantial, and 
driven mostly by management practices, 
cocoa tree density & black pod control. 
• Improved agronomic practices offer op-
portunities to substantially increase 
production of present-day cocoa 
plantations.  
A R T I C L E  I N F O   A B S T R A C T   
Editor: Kairsty Topp  CONTEXT: Global cocoa production is largely concentrated in West Africa where over 70% of cocoa is produced. 
Here, cocoa farming is largely a rain-fed, low-input system with low average yields, which are expected to 
Keywords: decline with climate change. With increasing demand, there is a need to evaluate opportunities to increase 
Cocoa (Theobroma cacao L.) production whilst avoiding deforestation and expansion to croplands. Thus, it is important to know how much 
Crop model 
* Corresponding author at: Wageningen University & Research, Centre for Crop Systems Analysis (CSA), P.O. Box 430, 6700 AK Wageningen, the Netherlands. 
E-mail addresses: paulinaansaa.asante@wur.nl (P.A. Asante), e.rahn@cgiar.org (E. Rahn), pieter.zuidema@wur.nl (P.A. Zuidema), danae.rozendaal@wur.nl 
(D.M.A. Rozendaal), p.laderach@cgiar.org (P. Läderach), r.asare@cgiar.org (R. Asare), niels.anten@wur.nl (N.P.R. Anten).  
https://doi.org/10.1016/j.agsy.2022.103473 
Received 29 January 2022; Received in revised form 21 July 2022; Accepted 22 July 2022   
Available online 28 July 2022
0308-521X/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
P.A. Asante et al.                                                                                                                                                                                                  A   g r i c  u l t u  r a  l  S  y  s t e m   s 201 (2022) 103473
Water-limited yield additional cocoa can be produced on existing farmland, and what factors determine this potential for increased 
Yield gap yield. 
Cocoa planting density OBJECTIVE: The objective was to quantify the cocoa yield gap in Ghana and identify the factors that can 
Black pod control contribute to narrowing the gap. 
METHODS: We calculated the cocoa yield gap as the difference between potential yield (i. water-limited potential 
(Yw) quantified using a crop model, ii. attainable yield in high-input systems(YE), iii. attainable yield in low- 
input systems(YF)) and actual farmer yield. Both absolute and relative yield gaps were calculated. We then 
related each yield gap (absolute & relative) as a function of environment and management variables using mixed- 
effects models. 
RESULTS AND CONCLUSIONS: There were considerable yield gaps on all cocoa farms. Maximum water-limited 
yield gaps (YGW) were very large with a mean absolute gap of 4577 kg/ha representing 86% of Yw. Attainable 
yield gap in high-input (YGE) was lower with mean absolute gap of 1930 kg/ha representing 73% of YE. The yield 
gap in low-input (YGF) was even lower with mean absolute gap of 469 kg/ha representing 42% of YF. Mixed- 
effects models showed that, absolute YGW were larger at sites with higher precipitation in the minor wet and 
minimum temperature in the minor dry season explaining 22% of the variability in YGW. These same factors and 
cocoa planting density explained 28% of variability in absolute YGE. Regardless of climate, absolute YGF and 
relative YGW, YGE and YGF were reduced by increasing cocoa planting density and application of fungicide 
against black pod. The models explained 25% of the variability in absolute YGF, and 33%, 33% and 25% in 
relative YGW, YGE and YGF respectively. 
In conclusion, climate determined absolute YGW in Ghana whilst absolute YGE were determined by both climate 
and management. In contrast, absolute YGF and relative YGW, YGE and YGF can be reduced by agronomic 
management practices. 
SIGNIFICANCE: Our study is one of the first to quantify cocoa yield gaps in West Africa and shows that these can 
be closed by improved agronomic practices.   
1. Introduction available room to increase yield requires robust estimates of potential 
yield, which is the maximum yield a crop can achieve in a specific 
Global cocoa production is largely concentrated in West Africa where environment with no limitation of water and nutrients nor reductions 
77.4% (of the total 5,175,000 tons) of cocoa beans are produced on an from pests and diseases (van Ittersum et al., 2013). Under rain-fed 
estimated six million ha of land by nearly two million smallholder cropping systems, which is the norm for cocoa farming in West Africa, 
farmers (ICCO, 2021; Wessel and Quist-Wessel, 2015). Ghana is the potential yield is limited by plant available water and therefore, water- 
second largest producer after Côte d'Ivoire and globally these two limited potential yield (Yw) is a more relevant benchmark. 
countries supply about 64% of cocoa beans. While these countries lead Dynamic simulation models are commonly used to estimate potential 
in total cocoa production, their yield per hectare in smallholder farms – yield, which are developed on the basis of current understanding of 
typically 300–600 kg/ha – is amongst the lowest in the world (Asante ecophysiological crop processes in response to environmental and 
et al., 2021; Wessel and Quist-Wessel, 2015). In addition, climate suit- management factors (Monzon et al., 2021; Rahn et al., 2018; Zuidema 
ability is expected to decrease in response to climate change with po- et al., 2005). For cocoa, only one such model, namely Sucros-cocoa/ 
tential negative effects on yields (Anim-Kwapong and Frimpong, 2004; Cacao Simulation Engine 2 (CASE2), has been developed and tested 
Gateau-Rey et al., 2018; Läderach et al., 2013; Schroth et al., 2016). for simulating cocoa growth and yield under irrigated and rain-fed 
Over the past three decades, increases in production have been driven by conditions (Zuidema et al., 2005). 
a sharp increase in plantation area with only marginal increases in yield Another means to estimate potential yield is based on direct mea-
(van Vliet and Giller, 2017; Wessel and Quist-Wessel, 2015). Expansion surements from long-term field experiments which utilize crop man-
of the land area under cocoa cultivation is driving deforestation as cocoa agement practices designed to eliminate all yield-reducing factors (e.g., 
is grown mainly in regions that used to be covered with highly diverse nutrient deficiencies, incidence of pests and diseases) (Lobell et al., 
moist tropical forests (Abu et al., 2021; Ruf et al., 2015). Another 2009; van Ittersum et al., 2013). Attained yields from experimental trials 
challenge is that cocoa is also replacing food croplands, threatening food are expected to come close to model-based potential values, however, it 
security in the cocoa growing belt, as exemplified for Ghana (Ajagun is generally impossible to exclude all yield limiting and reducing factors 
et al., 2021). In the coming decades, increased demand for cocoa under field conditions (Aggarwal et al., 2008; Lobell et al., 2009; van 
(growing at approximately 3% per year (Beg et al., 2017)), and the Ittersum et al., 2013). Location-specific yield limiting and reducing 
projected potential loss of about 50% of the current cocoa growing area factors such as year-to-year climate variation can be large for some lo-
due to decreasing climatic suitability (Läderach et al., 2013; Schroth cations, which means required optimal management practices can vary 
et al., 2016) could drive producers to new areas, resulting in additional substantially from one year to another (Aggarwal et al., 2008; Daymond 
deforestation (Ruf et al., 2015) and food insecurity (Ajagun et al., 2021). et al., 2020; Lobell et al., 2009). These location-specific yield-reducing 
To avoid further deforestation and expansion of cocoa fields into other factors can lower the experimental yields by up to two-thirds of model- 
sensitive areas, there is a need to evaluate opportunities to increase based potential yields (Chapman et al., 2021; Hoffmann et al., 2020). In 
yields per unit area on existing lands to meet the growing demand for West Africa, experimental trials are unavailable for most cocoa growing 
cocoa. Whilst increasing productivity may not necessarily lead to a areas. Thus, even though model-based potential yields may probably be 
reduction in deforestation without supporting governmental policies an overestimation of what can be achieved in experimental trials, it does 
that contribute to forest protection (e.g., The Cocoa Forest REDD+, The provide a reference of what can be obtained theoretically in optimally 
Cocoa & Forests Initiative) and a social safety net that ensures strong managed fields (best agronomic practices in place) with no nutrient 
farmer livelihood through improved negotiation skills, it can be a limitation (fertilized fields) and no incidence of pests and diseases. In 
necessary step to reduce pressure on areas designated for forests and Ghana, a few studies have reported experimentally-based potential 
other land uses. yields including 1891.3 kg/ha (Ofori-Frimpong et al., 2006 in Aneani 
Yield gap analysis provides a means for evaluating the scope to in- and Ofori-Frimpong, 2013), 3500 kg/ha (Ahenkorah et al., 1974), 2000 
crease production on existing lands as it can provide information on the kg/ha (Ahenkorah et al., 1987) and 3245.97 kg/ha (Appiah et al., 2000), 
factors that limit current yields (van Ittersum et al., 2013). Evaluating but the estimated national-level experimental-based cocoa yield gap was 
2
P.A. Asante et al.                                                                                                                                                                                                  A   g r i c  u l t u  r a  l  S  y  s t e m   s 201 (2022) 103473
obtained using only one experimental yield (1891.3 kg/ha) benchmark management factors as effects of variation in potential/attainable yields 
obtained from one location (Aneani and Ofori-Frimpong, 2013). on yield gaps is normalized and variation in actual farm-based yields in 
The use of maximum farmer yields based on surveys represents Ghana was shown to be driven more by management than by climate or 
another way to estimate potential yield. This is most suitable in inten- soil factors (Asante et al., 2021). We expect agronomic management 
sively managed cropping systems where it is reasonable to assume that practices like pest and disease control, cocoa planting density, and fer-
at least some farmers apply management practices capable of tilizer use to reduce relative yield gaps whilst high shade levels, tree age 
approaching the potential yield (Lobell et al., 2009). In Ghana, two and farm size are expected to increase relative yield gaps. 
studies have quantified and explained yield gaps for cocoa using 
maximum farmer yields as benchmark (Abdulai et al., 2020; Aneani and 2. Materials & method 
Ofori-Frimpong, 2013). However, considering that cocoa cropping sys-
tems in West Africa are largely low-input, it is likely that even maximum 2.1. Study area 
farmer yields are well below the potential under rainfed conditions and 
using them as benchmark would not allow to assess the potential yield The study was conducted at 93 different cocoa farm locations 
gain that could be achieved under high input. Also, from a previous spanning the cocoa growing areas of Ghana, to represent the range of 
study it appears that actual cocoa yields in Ghana are not very sensitive environmental conditions and production systems in the cocoa belt 
to climate as they are strongly limited by low level of agronomic man- (Fig. 1). Cocoa is grown in southern Ghana within three agroecological 
agement, yet strong climatic influence is expected with good agronomic zones; i.e., evergreen rainforest, deciduous forest and forest/savanna 
management (Asante et al., 2021). Hence, we believe that using both transition zones. The pattern of rainfall distribution within this region is 
model-based and maximum farmer yield-based benchmarks will give a bimodal, with two wet (main wet season from April to June/July, and 
comprehensive indication of the potential yield gains that could be minor wet season from September to November) and two dry seasons 
achieved at the different levels of intensification. To our knowledge this (main dry season from December to February/March and a short dry 
has not previously be done for cocoa. period from July/August during which relative humidity is still high). 
The difference between the benchmark (i.e., either model-simulated, Mean rainfall is highest in the south-west and decreases gradually to-
experimental attained or based on farmer maximum) and actual farmer wards the North (Fig. 1). Temperature is less variable across the cocoa 
yields (Ya), which is the yield achieved in a farmer's field is the absolute belt with mean monthly values of about 25 ◦C and a diurnal range of 
yield gap, a measure which provides relevant information on the scope 5–9 ◦C. The dominant soil types within the region are the strongly 
for production increase in kg per ha (Lobell et al., 2009; van Ittersum weathered Acrisols (Ochrosols - Ghana Great Soils Group) found in the 
et al., 2013). Defining this in relative terms (relative yield gap), which deciduous forest and parts of the forest/savanna transition agro- 
expresses the absolute yield gap as a percent of the potential yield ecological zones and the highly leached, strongly weathered Ferrasols 
calculated as; Yg = benchmark− actualrel benchmark *100%, has the methodological (Oxysols - Ghana Great Soils Group) with low soil pH (strong acidity) 
advantage of allowing comparison of the absolute yield gaps between occurring in areas with high rainfall such as in the south west 
different locations and with different crops (Van Oort et al., 2017). Also, (Adjei–Gyapong & Adjei Gyapong and Asiamah, 2002; Appiah et al., 
in the case of the model-simulated benchmark normalization of the 1997). The high acidity, and low amounts of nutrients make Ferralsols 
absolute yield gap reduces the dominant effect of Yw on yield gap when unfavourable for cocoa growth (Appiah et al., 1997). 
this is mainly driven by variation in Yw. 
The objective of this study was to quantify the cocoa yield gap for 2.2. Quantifying the water-limited potential cocoa yield 
Ghana and to identify the factors that contribute to narrowing the gap. 
We provide three different yield gap estimates: (1) a yield gap estimate Simulation of water-limited potential cocoa yield was done using the 
where we obtain Yw as upper limit that can be achieved on existing land CASE2 model (Zuidema et al., 2005). This is a dynamic crop simulation 
in a rain-fed system using the crop simulation model Sucros Cocoa/ model for cocoa that simulates all major processes of crop growth and 
CASE2 (Zuidema et al., 2005) and field-level Ya data obtained on farmer production, including light interception, photosynthesis, maintenance 
fields (maximum water-limited yield gap; YGW), (2) a yield gap estimate respiration, evapotranspiration, biomass production and associated 
based on attainable yield from experimental trials and Ya (attainable growth respiration and biomass allocation. Resulting bean yield of cocoa 
yield gap in high-input systems; YGE) and (3) a yield gap estimate based trees can be simulated for conditions with or without shade from asso-
on maximum farmer yield and Ya (attainable yield gap in low-input ciated trees and with or without water-limitation. CASE2 is originally 
systems; YGF). YGW, YGE and YGF were calculated in both absolute implemented in FORTRAN using the Fortran Simulation Environment 
and relative terms for 93 (84 in the case of YGF) cocoa farms spanning (FSE) (van Kraalingen, 1995) which makes it difficult to automate 
the cocoa growing belt of Ghana. We then analysed the association of simulations for different inputs. To address this, RCASE2, a wrapper 
yield gaps (absolute and relative) with variation in a set of environ- around CASE2 has been developed by Wageningen University and 
mental conditions (climate, soil) and agronomic management factors. Research, which allows CASE2 to be run with R statistical software (R 
This is important for identifying potential causes of yield gaps and op- Core Team, 2018). 
portunities and entry points for sustainable intensification. We CASE2 has been parameterised based on existing information of 
addressed the following questions: (1) What are the current cocoa yield cocoa physiology and morphology with values obtained from literature 
gaps on farms across cocoa growing areas of Ghana? (2) To what extent (Zuidema et al., 2003). It uses information on weather, soil and cropping 
and how do environmental and management factors explain these yield system as inputs for growth and yield simulations at a daily time step. 
gaps? For weather, the CASE2 model requires input data on daily minimum 
We expect that variation in absolute yield gaps will be mostly driven and maximum temperature, precipitation, solar radiation, and early 
by climatic factors as potential yields tend to be very sensitive to climate morning vapour pressure for at least an eight year period (Zuidema 
(Zuidema et al., 2005). Absolute yield gaps are expected to become et al., 2005). Assumed climatic limitations for growth and yield in 
smaller in drier areas as Yw and attainable yields will be lower due to CASE2 include: average temperature between 10 and 40 ◦C and an 
negative impacts of low water availability and high temperature. The annual precipitation of at least 1250 mm. Soil data required in CASE2 
climate effect on absolute yield gaps will be smaller for YGF than for the includes information on thickness; number and depth of soil layers, the 
others because low-input attainable yield is expected to be less climate- sum of which should add up to 1.5 m, and soil physical characteristics 
sensitive than high-input attainable yield and Yw yields. On the other including, the water content at saturation, field capacity, wilting point 
hand, relative yields gaps are expected to be driven more by and for air-dried soil with standard values defined based on the Driessen 
soil types (Driessen, 1986). With regard to data on cropping systems, 
3
P.A. Asante et al.                                                                                                                                                                                                  A   g r i c  u l t u  r a  l  S  y  s t e m   s 201 (2022) 103473
Fig. 1. (a)Simulated cocoa water-limited yields (Yw, circles) and b) actual mean cocoa yield (Ya, circles) for 93 farm locations and annual precipitation (background 
colour) in southern Ghana. Rainfall and cocoa yields are averages of the 2012/2013 and 2013/2014 cocoa crop years on a 11-km resolution. The size of the circle is 
proportional to the average Yw and Ya for that location. 
CASE2 requires information on cocoa tree age, planting density and for bean yield, standing biomass, leaf area and size-age relations (Zui-
shade levels. Simulations can be carried out for cocoa trees (assuming dema et al., 2005). Yield estimates from the model were not far off es-
planting material is uniform) between the age of 3 to 40 years (i.e., timates of experimental yields in some countries and the represented 
18.5–70 kg dry weight per plant; CASE2 does not include the juvenile processes represent our current understanding of cocoa growth and 
phase), with planting density ranging from 700 to 2500 trees/ha. Hor- yield formation (Zuidema et al., 2005). 
izontally homogeneous shading is assumed and the shade level is In simulations of Yw, the model assumes non-limited nutrient supply 
calculated as a function of shade tree leaf area index (SLAI) and light while yield losses caused by pests and diseases are considered absent. 
extinction coefficient (k) which varies between 0.4 and 0.8 (Zuidema Most climate variation (e.g. temperature, radiation and precipitation) is 
et al., 2005). Simulations can be carried out for shade levels between considered with the exception of flooding. Simulations of Yw were 
0 and 3 SLAI (i.e., with 0 representing no shading to 3 representing carried out for a period of 8 years (from 2007 to 2014), using weather, 
heavy shading). Here, we calculated the relative light intensity reaching soil and cropping system information observed at 93 cocoa farm loca-
the cocoa canopy using the modified Lambert-Beer equation (Monsi and tions within the cocoa growing areas of Ghana. Simulations were carried 
Saeki, 2005); PARb/PARi = e ^ (− k * SLAI),where PARb refers to the out for cocoa trees with initial average tree age of 14 years (based on the 
Photosynthetically Active Radiation below the shade tree canopy (but average, observed cocoa tree age), a planting density of 1246 trees per 
above the cocoa tree canopy), and PARi the incident Photosynthetically hectare (based on the average observed across the cocoa farms) and 
Active Radiation above the shade tree canopy (i.e., unobstructed day under a shade tree canopy of 10% (based on average SLAI calculated for 
light) and k is the light extinction coefficient. PARb values were the cocoa farms). Fixing these factors in our calculation of Yw allows us 
measured with hemispherical photographs in cocoa farms from which to compare how yield gap affecting factors vary across farms. 
yield data was obtained (Daymond et al., 2017). The value of k was 
taken as 0.6, the standard setting in CASE2 (Zuidema et al., 2005). 2.3. Weather and soil data 
Although validating the CASE2 model is difficult due to limited avail-
ability of yield data that approach potential or water-limited yield, a Daily minimum and maximum temperature (◦C), precipitation 
validation study comparing model output with available cocoa planta- (mm), and solar radiation (MJ m− 2 d− 1) at a spatial resolution of 0.1◦
tion outputs from locations where empirical data (regularly reported (approximately 11 km) for the period of 2007 to 2014 were obtained 
values) was available, showed that the model produces realistic outputs from the Copernicus AgERA5 database (Boogaard and van der Grijn, 
4
P.A. Asante et al.                                                                                                                                                                                                  A   g r i c  u l t u  r a  l  S  y  s t e m   s 201 (2022) 103473
2020). Early morning vapour pressure was estimated following the Table 1 
calculation procedure by FAO (Allen et al., 1998). In the FAO procedure, Definitions and descriptive statistics for yield gap estimates.  
actual vapour pressure per day was estimated from relative humidity Abbreviation Variable Unit Definition Mean 
and air temperature using the following equation, e = (std. a
[ ] dev.) 
RH 0 0mean e (Tmax)+e (Tmin)
100 2 where ea. is the actual vapour pressure [kPa], and Ya Actual yield kg/ Yield achieved in a 717 
RHmean is the mean relative humidity, whilst e◦(Tmin) and e◦(Tmax) is ha farmer's field (343.7) 
the saturation vapour pressure at daily minimum temperature [kPa] and Yw Simulated water- kg/ Theoretical maximum 5294 limited potential ha yield limited by water, (553.7) 
at daily maximum temperature [kPa], respectively. This saturation yield temperature and light as 
vapour pressure at minimum and maximum air temperature is calcu- simulated with a crop 
[ ]
lated as, e0(T) = 0.6108exp 17.27 TT 237 3 where T is the minimum or model + . Y Attainable yield kg/ 50% of Yw, determined 2647 
maximum temperature (◦C), respectively. We included the saturated E in high-input ha based on reported (276.8) 
vapour pressure derived from minimum temperature e◦(Tmin) as early systems average yields from 
morning vapour pressure values, as the lowest temperature is registered experimental trials in 
in the early morning and e◦(Tmin) is often lower than actual vapour Ghana 
pressure (ea) but when relative humidity is below ~70%, e is lower than 
YF Attainable yield kg/ Average yield from the 1109 
a in low-input ha 10% best performing 
e◦(Tmin). systems farmers across the 93 
Soil texture data, classified based on the USDA system at six standard cocoa farms 
depths (0–5, 5–15, 15–30, 30–60, 60–100 & 100–200 cm) at a spatial Absolute Absolute kg/ Difference between Yw 4, 577 
resolution of 250 m were obtained from the ISRIC database (Hengl et al., YGW maximum water- ha and Ya expressed in kg/ (641.7) 
limited yield gap ha 
2017). Since the sum of the depth of all soil layers (thickness) should not Relative Relative % The maximum water- 86(6.8) 
exceed 1.5 m, we took the mean of the 100-200 cm standard depth layer YGW maximum water- limited absolute yield 
in addition to the first five layers of the soil data from ISRIC. For in- limited yield gap gap as a percentage of 
formation on physical characteristics (i.e., standard values of soil water Yw 
content at saturation, field capacity, wilting point and for air-dried soil), Absolute Absolute kg/ Difference between 1930 YG attainable yield ha attainable yield in high- (433.9) 
we compared the soil texture classification of the soil classes of the USDA E gap in high-input input systems and Ya 
system to the soil texture properties of the Driessen soil types, to be able systems expressed in kg/ha 
to include the soil type in the simulations with CASE2 (Table S1, Relative YGE Relative % The Ya as a percentage of 73(13.5) 
Driessen, 1986; Zuidema et al., 2003). attainable yield attainable yield in high- 
gap in high-input input systems. 
systems 
2.4. Actual cocoa yield Absolute YGF Absolute kg/ Difference between 469 
attainable yield ha attainable yield in low- (248.9) 
Actual cocoa yield data from farmer fields with information on gap in low-input input systems and Ya 
a 
management (cocoa planting density, cocoa tree age, radiation inter- systems expressed in kg/ha Relative YG Relative % The Ya as a percentage of 42(22.4)  
ception by shade trees, fungicide application against black pod (Phy- F attainable yield attainable yield in low- 
tophthora palmivora and megakarya), insecticide application against gaps in low-input input 
capsid (Sahlbergella singularis and Distanfiella theobroma) and fertilizer systemsa 
use) and soil (field measured pH, carbon (%), nitrogen (%), available a Yield gap was calculated for only the 90% lowest performing farmers (84 
phosphorus (μg/g), potassium (meq/100 g), and magnesium (meq/100 cocoa farms). 
g)) for 93 farms with georeferenced locations across the cocoa belt of 
Ghana were obtained from Mondelez International ‘Mapping Cocoa 
Productivity’ project data (Daymond et al., 2017). Yield data was Ybench − YaYGrel = *100% (2) 
available for a period of two years (2012/2013 and 2013/2014 cocoa Ybench
cropping season). We defined cocoa yield as the amount of dried beans These yield gaps (eq. 1 and 2) were calculated for every farm in our 
(pod to kilogram conversion based on field measured mean pod value of sample. The attainable yield in high-input systems was defined as 50% 
24.2 (±3.6) to 1 kg) harvested per year (cocoa crop year is defined as of Yw based on the average of the maximum experimental potential 
March of a given year – February of the next year), per unit of cocoa yields (2500 kg/ha) from four experimental trial studies in Ghana 
plantation area (ha). Production data was collected using pod counts (Ahenkorah et al., 1987; Ahenkorah et al., 1974; Aneani and Ofori- 
and field size determined using GPS measurements. Frimpong, 2013; Appiah et al., 2000). On the other hand, attainable 
yield in low-input systems was defined as the average yield from the 
2.5. Yield gap definition and statistical analysis 10% best performing farmers across the 93 cocoa farms. Thus, the YGF 
was calculated for only the 90% lowest performing farmers (84 cocoa 
With reference to Table 1, we defined the absolute yield gap for YGW, farms). 
YGE, YGF as the difference between Yw (YGW) or attainable yield in high- We examined the drivers of the absolute and relative yield gaps for 
input (YGE) or attainable yield in low-input systems (YGF) and actual YGW, YGE, and YGF by modelling the absolute (or relative) yield gap as a 
farmers' yield (Ya). Hence the absolute yield gap is given as: function of climate, soil and management variables using mixed-effects 
YG Ybench Ya (1)  models (MEMs) (Zuur et al., 2009). For management, we considered abs = −
farm size, fertilizer use, application of fungicide against black pod, 
where Ybenchis the benchmark yield: the water-limited potential yield application of insecticide against capsid, cocoa planting density, tree age 
(Yw), the high-input attainable yield (YE) or the low-input attainable and radiation interception by shade trees. As soil variables, we consid-
yield (YF) in the cases of YGW, YGE and YGF, respectively. The relative ered measured soil properties including soil pH, carbon, nitrogen, 
yield gap (for YGW, YGE, YGF) was calculated as a percentage of the available phosphorus, potassium and magnesium. For climate, we 
benchmark yield using the following equation considered seasonal variables (i.e. all four seasons; the main wet season 
(March–June), the minor dry season (July–August), the minor wet sea-
son (September–November), and the main dry season 
5
P.A. Asante et al.                                                                                                                                                                                                  A   g r i c  u l t u  r a  l  S  y  s t e m   s 201 (2022) 103473
(December–February) Fig. 4). Thus, daily weather data was aggregated performed in R (R Core Team, 2018). 
to seasonal climate variables. We performed MEM between the absolute 
(or relative) yield gap and the seasons of each climate variable sepa- 3. Results 
rately. This was done to select the season for which the climate variables 
most strongly influenced the yield gap. We included for each climate 3.1. Magnitude of actual yield (Ya), water-limited yield (Yw), and the 
variable the season that was included in the best model (i.e., lowest yield gap for cocoa farms in Ghana 
Bayesian Information Criterion; BIC) (Table 2). We excluded solar ra-
diation as an explanatory variable for YGF as MEM between the seasons The Ya across the 93 cocoa farms of the 2012/2013 and 2013/2014 
(of solar radiation) and YGF did not converge. cropping seasons was generally low with a mean of 717 kg/ha. Ya for 
To obtain the most parsimonious MEM that explains most of the some farms was as low as 78 kg/ha whilst other farms achieved yields as 
variation in the absolute or relative yield gap, we used a two-step high as 2331 kg/ha depending on the year. Relatively small differences 
approach; correlation analyses and stepwise regression. We first con- in Ya were observed between wet and dry areas within the study area 
ducted correlation analyses for all explanatory variables (which (Fig. 1(b)). Yw values, on the other hand, were generally high with a 
included all selected climate, soil and management variables) to identify mean of 5294 kg/ha. Average maximum Yw yields of 6567 kg/ha and 
and remove one variable out of variable pairs that were strongly average minimum of 4178 kg/ha were observed across farms and 
correlated (i.e., having r > 0.7) in order to avoid collinearity. Based on cropping seasons. Lowest Yw were observed in dry areas and highest Yw 
this procedure, none of the variables was excluded from the list of in wet areas (Fig. 1(a)). Across all cocoa farms, Ya was lower than Yw 
explanatory variables for the absolute yield gap and for the relative yield (Fig. 1). 
gap of YGW, YGE and YGF as we found no case of explanatory variables The resulting estimated YGW was accordingly very large with a mean 
having r > 0.7 (Figs. S3, S4 and S5). Next, we included all explanatory absolute yield gap of 4577 kg/ha, representing a relative yield gap of 
variables after the correlation analyses (Table 2) in the MEM as fixed 86% (Fig. 2.). Across farms, absolute YGW ranged between 2223 kg/ha 
effects and farm ID as random intercept to account for non- and 6072 kg/ha which represents a range of 49–98% for relative YGW 
independence of data points (more than one year yield data) from the over the two-year period. Absolute YGW was largely driven by Yw. The 
same farm. We tested including year as random intercept but this did not spatial pattern of the distribution of absolute YGW across the study area 
improve the model, hence only farm ID was used as random intercept. was similar to Yw, with larger absolute YGW observed in wet areas and 
To allow comparison of the relative importance of explanatory vari- low absolute YGW in dry areas (Fig. S1(a)). Yet, relatively small differ-
ables, we standardized all continuous variables by subtracting the mean ences in relative YGW were observed across dry and wet areas (Fig. S1 
value of the variable and dividing it by the standard deviation (Gelman (b)). 
and Hill, 2006). A backward stepwise elimination of MEM models was The YGE, was obviously lower than YGW with mean absolute YGE of 
conducted using the “buildglmmTMB” function from R package 1930 kg/ha (representing 73% of the relative yield gap). For some 
“buildmer” to identify the most parsimonious model. The final model farms, YGE was negative, − 53.9 kg/ha (i.e. relative yield gap of − 2%), 
was selected based on BIC. Conditional and marginal R2 for the models thus achieved yields were beyond the reference attainable yield, whilst 
were estimated to evaluate variation explained by only the fixed effects others had YGE as high as 2873 kg/ha (i.e., relative yield gap of 97%) 
(i.e. the explanatory variables) and both the fixed effects and random (Fig. 2). The YGF was generally lower with mean absolute YGF of 469 kg/ 
effects, respectively (Nakagawa and Schielzeth, 2010). All analyses were ha which represents 42% of the relative yield gap. Across farms, YGF 
Fig. 4. Monthly data of precipitation (bars) and minimum temperature (red line) of Ghana (Tafo) and annual cocoa cropping cycle. Adapted from van Vliet and 
Giller, 2017 (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 
6
P.A. Asante et al.                                                                                                                                                                                                  A   g r i c  u l t u  r a  l  S  y  s t e m   s 201 (2022) 103473
Table 2 
Descriptive statistics of selected climate, soil and management (explanatory) variables based on model selection using the Bayesian Information Criterion and cor-
relation analyses for each of the dependent variables in the first step of the analysis.  
Explanatory variables Unit min max mean std.dev Dependent variables 
Climate variables       
Precipitation (minor wet season) mm 189 476 287 71 All yield gaps 
Solar radiation (minor dry season) MJ 12,756 14,899 13,481 509.8 All yield gaps 
Maximum temperature (main ◦C 28.5 30.6 29.8 0.4 Absolute YGW 
wet season) 
Maximum temperature (minor ◦C 28.0 29.6 28.8 0.3 Relative YGW, 
wet season) Absolute YGE, 
& Relative YGE 
Maximum temperature (minor dry season) ◦C 26.8 28.4 27.6 0.3 Absolute YGF, 
& Relative YGF 
Minimum temperature (minor ◦C 20.8 22.1 21.6 0.2 All yield gaps 
dry season) 
Management variables       
Cocoa planting density trees ha 276 3626 1221 531 All yield gaps 
Radiation interception by shade trees % 0 0.31 0.07 0.08 All yield gaps) 
Tree age years 6 57 22.2 11.0 All yield gaps 
Farm size hectares 0.26 7.7 1.7 1.4 All yield gaps 
Application of insecticides against capsid yes/no     All yield gaps 
Application of fungicides against Black pod yes/no     All yield gaps 
Fertilizer use yes/no     All yield gaps 
Soil variables       
Soil pH – 4.3 7.5 5.8 0.7 All yield gaps 
Soil carbon content (C) % 0.7 2.8 1.5 0.4 All yield gaps 
Soil nitrogen (N) % 0.07 0.28 0.1 0.0 All yield gaps 
Available Phosphorus in soil (P) μg/g 3.7 58.6 17.6 12.1 All yield gaps 
Soil potassium content (K) meq/100 g 0.2 0.8 0.4 0.1 All yield gaps 
Soil magnesium content (Mg) meq/100 g 0.01 5.8 1.5 1.1 All yield gaps 
The selected variables were subsequently included in the mixed-effects models for relative (or absolute) YGW, YGE, YGF (dependent variables) in the second step of the 
analysis, to select the final best model. 
Fig. 2. Variation in (a) the absolute yield gap (difference between potential and actual yield) for maximum water-limited (YGW), high-input attainable (YGE) and 
low-input attainable (YGF) yield, and (b) the relative values for YGW, YGE and YGF, across 93 (84 in the case of YGF) cocoa farms in Ghana for the 2012/2013 and 
2013/2014 cocoa crop years. Yield refers to dry bean yield and cocoa crop year is March of a given year to February of the next year of 2012/2013 and 2013/2014, 
respectively. 
ranged from 4 kg/ha (relative yield gap of 0.3%) to 1031 kg/ha (relative 3.2. Determining factors of the absolute cocoa yield gap 
yield gap of 93%) (Fig. 2). Similarly to actual yields, relatively small 
differences in both absolute and relative YGF were observed between Results of initial correlation analyses between the absolute YGW, YGE 
wet and dry areas within the study area (Fig. S2). and YGF and explanatory variables showed that absolute YGW was 
significantly and positively correlated with precipitation of the minor 
wet season, solar radiation of the minor dry season, minimum temper-
ature of the minor dry season and radiation interception by shade trees 
7
P.A. Asante et al.                                                                                                                                                                                                  A   g r i c  u l t u  r a  l  S  y  s t e m   s 201 (2022) 103473
(Fig. S3). Significant negative correlations with absolute YGW were effects (farm-to-farm variation) were included (conditional R2 = 0.70). 
found for soil magnesium content (Mg), soil pH and available phos- Thus, variation in the absolute YGW was largely driven by other vari-
phorus (P) (Table 2, Fig. S3). Absolute YGE was also significantly and ables than those tested as fixed effects. Absolute YGE on the other hand, 
positively correlated with precipitation of the minor wet season and was driven by both climatic and management variables. Amongst cli-
minimum temperature of the minor dry season (Fig. S4). Significant matic factors, precipitation of the minor wet season (Fig. S7a) and 
negative correlations with absolute YGE were found for cocoa planting minimum temperature of the minor dry season (Fig. S7b) significantly 
density, soil pH, P, and Mg. On the other hand, correlations between increased this gap (Table 3(iii). Amongst management factors, only 
absolute YGF and explanatory variables differed from YGW and YGE. In cocoa planting density (Fig. S7c) was influential and significantly 
this case only cocoa planting density showed a significant negative reduced the absolute YGE. The fixed effects of the final model for YGE 
correlation with absolute YGF (Spearman's rank correlation (r) of 0.47) explained 28% (marginal R2 of 0.28) of the variation whilst 66% of the 
(Fig. S5). variation in absolute YGE is explained when including random effects 
The mixed-effects models indicated that the absolute YGW was (conditional R2 = 0.66) (Table 3(iii)). 
driven by only climatic factors, with precipitation of the minor wet The final mixed-effects model for absolute YGF revealed that only 
season (Fig. 3a) having the strongest influence followed by minimum management variables explained absolute YGF. Cocoa planting density 
temperature of the minor dry season (Fig. 3b). Precipitation of the minor (Fig. S8a), which showed a significant correlation with absolute YGF, 
wet season and minimum temperature of the minor dry season showed a and application of fungicides for controlling black pod disease (Fig. S8b) 
relatively strong positive correlation (r of 0.44 and 0.34 respectively) were the most important variables (Table 3(v)). These two factors (fixed 
with absolute YGW (Fig. S3), and significantly increased this gap effects) explained 25% (marginal R2 of 0.25) of the variation in absolute 
(Table 3(i)). These two factors (fixed effects) explained 22% (marginal YGF whilst 61% (conditional R2 of 0.61) of the variation was explained 
R2 of 0.22) of the variation in the absolute YGW and 70% when random by fixed and random effects together. 
Fig. 3. Relationship between absolute YGW and (a) precipitation of minor wet season and (b) minimum temperature of minor dry season and between YGW relative 
yield gap and (c) cocoa planting density and (d) application of fungicide against black pod (use vs. no use) based on 93 cocoa farms from 2012 to 2014. Lines are 
predicted relations from the mixed-effects model, other predictors were kept constant at mean values. 
8
P.A. Asante et al.                                                                                                                                                                                                  A   g r i c  u l t u  r a  l  S  y  s t e m   s 201 (2022) 103473
Table 3 4. Discussion 
Results of the mixed-effects models for the YGW, YGE, YGF absolute yield gap and 
relative yield gaps as a function of environmental and management factors.   4.1. Magnitude of the cocoa yield gap in Ghana 
Estimates std. Confidence Marginal R2 / 
Error Interval Conditional R2 The YGW of the 93 farms in Ghana was very large. Actual cocoa yields 
(i) Absolute YGW predictors 0.22 / 0.70 per annum ranged between 78 and 2331 kg/ha (mean = 717 kg/ha) and 
Precipitation 219.81 were considerably lower than simulated water-limited yields (range 48.26 124.49–315.14  
(minor wet season) *** between 4178 and 6567 kg/ha with mean = 5294 kg/ha) at all locations 
Minimum 
168.95 over the two-year period (2012–2014). The absolute YG ranged from temperature (minor 46.68 76.75–261.14  W 
dry season) *** 2223 to 6071 kg/ha (mean = 4577 kg/ha) representing a relative yield 
(ii) Relative YGW predictors 0.33 / 0.65 gap of 49 to 98% (mean = 86%). These yield gap values are amongst the 
Cocoa planting − 2.89 
0.48 3.84 to 1.94  highest documented globally for perennial tree crops grown under − −
density *** rainfed conditions by smallholder farmers. For instance, YGW for oil 
Application of 
fungicide against 3.38 ** 1.16 5.67 to 1.10  palm was 63% on average in smallholder farms in Indonesia (Monzon − − −
black pod (yes) et al., 2021). Euler et al. (2016) also found average oil palm yield gaps 
(iii) Absolute YGE predictors 0.28 / 0.66 ranging from 43% to 55% for smallholder oil palm producers in Jambi 
Precipitation 119.47 
31.07 58.10–180.84  (Sumatra, Indonesia) under irrigated conditions. Besides these studies, (minor wet season) *** other yield gap studies for tropical tree crops including cocoa (Aneani 
Minimum 
temperature (minor 93.05 ** 30.24 33.32–152.77  and Ofori-Frimpong, 2013), coffee (Bhattarai et al., 2017; Wang et al., 
dry season) 2015), banana (Wairegi et al., 2010) and oil palm (Rhebergen et al., 
Cocoa planting − 126.45 33.74 − 193.10 to 2018) used empirical approaches. Thus, to the best of our knowledge, 
density *** − 59.81  our study is the first to quantify yield gaps at field level for cocoa using a 
(iv) Relative YGE predictors 0.33 / 0.65 
Cocoa planting 5.79 crop modelling approach. The YGW of cocoa we found is slightly com-−
density *** 0.96 − 7.69 to − 3.88  parable but still higher than yield gaps reported for some annual crops 
Application of (e.g. rainfed maize =80%, rainfed rice =81.8%, millet =75% etc.) 
− 11.33 to 
fungicide against − 6.76 ** 2.31 
− 2.19  produced by smallholder farmers in Ghana (Global Yield Gap Atlas, 
black pod (yes) 2022). This shows that cocoa farmers are producing far below what is 
(v) Absolute YGF predictors 0.25 / 0.61 
Cocoa planting 94.95 138.91 to theoretically achievable under ideal management in a rain-fed system (i. − −
density *** 22.24 − 50.98  e., where only water availability limits yields), and that this at least to 
Application of 
160.19 263.18 to some extent is comparable to large yield gaps in other crops. This large fungicide against − 52.09 −
** − 57.21  gap also reveals an enormous potential for yield improvement as means 
black pod (yes) 
(vi) Relative YG predictors 0.25 / 0.61 to increase cocoa production without the need to further expand the F 
Cocoa planting − 8.56 area planted. 2 − 12.52 to 
density *** − 4.60  The cocoa yield gap calculated as the difference between attainable 
Application of yield in high-input systems (estimated as 50% of Yw) where improved or 
fungicide against − 14.44 4.7 − 23.72 to 
** − 5.16  recommended management practices are applied and actual yields were black pod (yes) relatively larger but comparable to other experiment based yield gap 
Only variables retained in the final model are shown. Significance levels are estimates for cocoa in Ghana (Aneani and Ofori-Frimpong, 2013). The 
indicated (* p < 0.05 ** p < 0.01 *** p < 0.001). mean absolute YGE we found was 1930 kg/ha (relative yield gap of 73%) 
which is slightly larger than the national experimental yield gap esti-
3.3. Determining factors of the relative cocoa yield gap mate of 1553.4 kg/ha (relative yield gap of 82.1%) for cocoa in Ghana 
(Aneani and Ofori-Frimpong, 2013). In relative terms however, our YGE 
The drivers of the relative YGW and YGE differed from the drivers of value 73% was lower than the national experimental-based relative 
the absolute YGW and YGE but drivers of absolute and relative YGF were yield gap of 82.1% indicating that relying only on a relative yield gap 
the same. Results of initial correlation analysis between relative YGW, can lead to low or high prioritization of impact if not compared with the 
YGE and YGF and explanatory variables showed that only cocoa planting absolute yield gap (Van Oort et al., 2017). The attainable, relative yield 
density had a significant negative correlation (i.e., r of 0.54, 0.54, 0.47 gap values for cocoa are again amongst the highest documented globally 
for relative YGW, YGE and YGF respectively) with relative YGW, YGE and for perennial tree crops. In oil-palm, a mean attainable yield gap of 47% 
YGF (Figs. S3, S4, S5). was found for small-holder farmers in Indonesia when attainable yield 
The final mixed-effects model for relative YGW, YGE and YGF all was defined as 70% of simulated water-limited yields (Monzon et al., 
revealed that management variables primarily drove relative YGW, YGE 2021). With a relatively lower attainable yield benchmark (50% of 
and YGF. Cocoa planting density (Fig. 3c, Figs. S7d, S8c), which was simulated water-limited yields) for cocoa, our YGE of 73% still remains 
strongly correlated with the relative YGW, YGE and YGF and application higher than the yield gap estimate for oil palm in that study. Euler et al. 
of fungicides for controlling black pod disease (Fig. 3d, Figs. S7e, S8d) (2016) also found attainable oil palm yield gaps of between 46% to 50% 
were the most important variables which significantly reduced the for smallholder oil palm producers in Jambi (Sumatra, Indonesia), 
relative YGW, YGE and YGF (Table 3(ii, iv, vi)). These two factors (fixed where attainable yield was defined as 85% of the potential yield (irri-
effects) explained 33% (marginal R2 of 0.33) of the variation in relative gated crops). These large attainable cocoa yield gaps results suggest 
YGW whilst 65% (conditional R2 of 0.65) of the variation was explained large opportunities for further increases in cocoa yields beyond current 
by fixed and random effects together. Similarly, the two factors levels. 
explained 33% (marginal R2 of 0.33) of the variation in relative YGE and Yield gap estimates based on maximum farmer yields in Ghana (YGF) 
65% (conditional R2 of 0.65) when including random effects. For rela- where cocoa farming is dominated by low-input systems were consistent 
tive YGF 25% (marginal R2 of 0.25) of the variation was explained by the with findings of other yield gap studies for cocoa in Ghana (Abdulai 
two factors and 61% (conditional R2 of 0.61) when including random et al., 2020; Aneani and Ofori-Frimpong, 2013). Across the dry, mid and 
effects. wet cocoa growing areas in Ghana, Abdulai et al. (2020) reported ab-
solute YGF of 434 kg/ha, 697 kg/ha, and 1126 kg/ha which represent a 
relative yield gap of 67%, 59% and 53%, respectively. Thus, in their 
9
P.A. Asante et al.                                                                                                                                                                                                  A   g r i c  u l t u  r a  l  S  y  s t e m   s 201 (2022) 103473
study absolute yield gaps increased significantly along a rainfall decline in photosynthesis (Balasimha et al., 1991). Increasing minimum 
gradient but relative yield gaps between dry and mid zones were not temperature is expected to increase respiration (increases exponentially 
significantly different, although the wet zone was significantly different with increasing temperature) and pod development (increases linearly 
from the dry zone. While we found similarly low YGF values (i.e. from 4 from 20 ◦C to 28 ◦C) (Zuidema et al., 2003). Higher respiration sup-
to 1031 kg/ha with a mean of 469 kg/ha representing a relative yield presses net assimilation rates and tends to result in lower yields. More 
gap range of 0.3 to 93% and mean of 42%) for the 84 cocoa farms in our rapid pod development on the other hand tends to allow pods to pass 
study, we did not observe this spatial pattern of absolute YGF increasing more quickly to maturing developmental stages with higher sink 
along a rainfall gradient (Fig. S2). Instead, the spatial pattern of absolute strength, which would thus positively affect yields. The minor dry sea-
and relative YGF differed less across the rainfall gradient, indicating that son in Ghana coincides with the early/mid stage of pod development as 
YGF was relatively insensitive to climate variation (Asante et al., 2021). the bulk of pods initiate development in the main wet season (April to 
Also, our study differs from the study of Abdulai et al. (2020), as we do June) and pods take approximately 5–6 months after pollination to 
not analyse data separately for the different climatic zones but for the reach maturity (Fig. 4) (Gerritsma, 1995; Toxopeus, 1985; Wessel, 
entire cocoa growing region. We did this because the analysis of a huge 1971). The net positive effect of temperature on yield suggests that 
(~3800 cocoa farms) dataset on cocoa yields in Ghana found climate did temperature-driven stimulation of pod development had a stronger ef-
not show strong effects on actual yields, as yield variability was mainly fect than the negative effects of higher temperature on net assimilation. 
driven by management (Asante et al., 2021). At the national level, Thus, in our simulations increasing minimum temperature increased 
Aneani and Ofori-Frimpong (2013) found YGF of 1537.2 kg/ha (relative simulated yields and thereby the absolute yield gap. 
yield gap of 82%) which is somewhat larger than our value and the value 
obtained by Abdulai et al. (2020). 4.3. Cocoa planting density and application of fungicide against black pod 
reduces cocoa yield gaps in Ghana 
4.2. Climate drives absolute maximum water-limited and attainable yield 
gaps in high-input systems, but not in low-input systems Agronomic management factors reduced both absolute YGF and YGE 
and the relative yield gaps (YGW, YGE and YGF) highlighting the 
Climate factors were identified as the main determinants of absolute importance of improved management practices for closing the cocoa 
YGW and YGE but not absolute YGF, which supports our hypothesis. yield gap and confirms our hypothesis. Absolute yield gap for YGF, was 
Climate variables explained 22% of the variation in absolute YGW but determined by only agronomic management factors and explained 25% 
when both climate and farm-to-farm variation are considered 70% of the of the variation and 61% when farm-to-farm variation was considered. 
variation is explained. This suggests that, other factors, including other Whilst absolute YGE, was driven by agronomic management in addition 
climate, soil and management factors not tested as fixed effects, drive to climate factors. In Ghana, Asante et al. (2021) found strong climatic 
the absolute YGW. The strong effect of climate on absolute YGW was influence for farms with best agronomic management but farms with 
mainly due to strong effects of climate on simulated water-limited yields average yields were less sensitive to climate. 
(Fig. S6) (Zuidema et al., 2005). Water-limited yields are more climate On the other hand, quantifying not only the absolute, but also the 
sensitive than the actual yields because all non-climatic factors, other relative yield gap, helps to quantify the relative importance of specific 
than crop traits, are, by definition, assumed to be non-limiting (Asante controllable measures for closing the yield gap, as the climatic effects 
et al., 2021; Zuidema et al., 2005). For YGE, climate together with that drive the water-limited yield predominate as drivers of the absolute 
agronomic management drove the absolute yield gap and explained YGW. Agronomic management factors were identified as the main de-
28% of the variation and 65% when farm-to-farm variation is considered terminants of relative YGW, which explained a large part (33%) of the 
thus also suggesting that factors not tested played a large role. variation in relative YGW. Similar to relative YGW, agronomic manage-
Absolute YGW and YGE were significantly and positively related to ment factors were the main determinants of relative YGE and relative 
precipitation of the minor wet season and minimum temperature of the YGF, also explaining a large part, namely 33% in the case of relative YGE 
minor dry season, (Table 3(i, iii)). The positive effects of precipitation of and 25% of the variation in relative YGF. 
the minor wet season on the absolute YGW and YGE may relate to pos- Increasing cocoa planting density significantly reduced the absolute 
itive effects of water availability on simulated water-limited cocoa yields YGE and YGF and relative values of YGW, YGE and YGF. Planting density 
(Fig. S6a) (Zuidema et al., 2005). In CASE2, bean yield is determined has consistently been identified as an important yield-limiting factor for 
largely by water-availability to cocoa trees and water limitation reduces cocoa (Abdulai et al., 2020; Asante et al., 2021; Daymond et al., 2017; 
yields (Gateau-Rey et al., 2018; Zuidema et al., 2005). The minor wet Efron et al., 2005; Sonwa et al., 2018; Souza et al., 2009), as well as for 
season (i.e. September to November) coincides with the period when the other crops (Duvick and Cassman, 1999) including tree crops like coffee 
major cocoa harvest starts in Ghana (Fig. 4), hence, when cocoa trees (Bhattarai et al., 2017; Wang et al., 2015). The simulations of water- 
have many maturing pods. Assimilate demand for pod growth in this limited yield with CASE2 were based on a standardized planting den-
period is therefore high. Water-limitation induced reductions in photo- sity of 1246 trees per hectare. This was based on the assumption that 
synthesis at this time will thus have a relatively large negative effect on density can be controlled and changed by the farmer to reduce the yield 
pod yield, whilst increasing precipitation has positive effects on pod gap. However, increasing densities also tend to increase disease inci-
yield hence on the absolute YGW and YGE. These results support our dence (e.g. due to microclimate effects and greater ease of transmission) 
hypothesis. but also greater competition between trees especially in mature stands 
The positive effect of minimum temperature of the minor dry season (Sonwa et al., 2018; Souza et al., 2009). The latter can be controlled by 
(July/August) on absolute YGW and YGE may be related to the temper- thinning (Lachenaud and Oliver, 1998) and pruning (Tosto et al., 2022). 
ature effects on pod development. In CASE2, minimum temperature Breeding for high yielding cocoa genotypes, that are smaller but also 
affects photosynthesis, respiration and pod development. Minimum have a higher allocation to pods, as a means to suppress competition and 
temperature values observed within the minor dry season in Ghana stimulate the positive effect of planting density on yields is recom-
range from 20.8 to 22.1 ◦C (Table 2) and are expected to drive average mended (Lockwood and Pang, 1996). 
temperature (23.9 to 25.1 ◦C) within this period as relative humidity is Application of fungicides against black pod reduces absolute YGF and 
still high with overcast weather conditions (Anim-Kwapong and Frim- relative values of YGW, YGE and YGF. Black pod disease which occurs in 
pong, 2004). For photosynthesis, average daytime temperature of 30 to all cocoa growing areas is considered as one of the most destructive 
32.1 ◦C are considered optimal for obtaining maximum photosynthesis diseases that prevents pod development and ripening and reduces yields 
rates (Balasimha et al., 1991; Zuidema et al., 2003). Higher tempera- (Andrews Yaw Akrofi et al., 2015; Anim-Kwapong and Frimpong, 2004; 
tures beyond 34 ◦C and temperatures below 24 ◦C result in a rapid Daymond et al., 2017; Opoku et al., 2000). This disease has been found 
10
P.A. Asante et al.                                                                                                                                                                                                  A   g r i c  u l t u  r a  l  S  y  s t e m   s 201 (2022) 103473
to be more prevalent under damp conditions (wet and humid conditions optimal density, other management practices that would help increase 
and shaded systems), particularly in the minor dry season (Anim-Kwa- yields need to be evaluated. For instance, high density may increase the 
pong and Frimpong, 2004) and can cause mean annual pod losses of need for adequate pruning (Tosto et al., 2022). A stepwise management 
about 40% and higher (Idachaba and Olayide, 1976 in Aneani and approach has been recommended, which targets yield limiting practices 
Ofori-Frimpong, 2013; Opoku et al., 2000; Wessel and Quist-Wessel, step-by-step. Only after implementing good agricultural practices (e.g. 
2015). Cocoa farmers who do not apply fungicide against black pod planting improved material, weeding, pruning, pest and disease control) 
suffer yield losses whilst application increases yields (Akrofi et al., 2003) nutrient management is considered (Wessel and Quist-Wessel, 2015) to 
and therefore reduces the yield gap. Adequate knowledge of techniques ensure that nutrient addition actually results in increased yields. Also, 
of fungicide application, the use of more black pod disease resistant monitoring and better surveys (improved data quality and additional 
genotypes and management practices that improves air circulation and management variables) are needed to evaluate the effect of management 
reduce humidity (e.g. pruning, regular harvesting of infected pods, factors on the yield gap. 
removal of pod husk heaps) have been recommended for controlling 
black pod disease (Adejumo, 2005; Akrofi et al., 2003; Cilas et al., 2018; 5. Conclusion 
Opoku et al., 2000). The reduction in relative yield gaps for YGW, YGE, 
and YGF due to cocoa planting density and application of fungicides We quantified three cocoa yield gap estimates based on model-based 
against black pod supports our hypothesis. However, application of in- maximum water-limited yield, and attainable yield in high- and low- 
secticides against capsid, fertilizer use, shade level, tree age and farm input systems both in absolute and relative terms. A considerable 
size had no effects, contrary to our expectations. model-based, mean absolute yield gap of 4577 kg/ha representing a 
relative yield gap of 86%, was found for the cocoa growing areas in 
4.4. Limitations and future steps Ghana. The attainable yield gap in high-input systems where improved 
or recommended management practices are applied was relatively lower 
This study had several limitations. First, it should be noted that there (mean absolute yield gap of 1930 kg/ha representing a relative yield gap 
are still important knowledge gaps regarding to how cocoa responds to of 73%) than the maximum water-limited estimate but larger than yield 
water limitation and hence modelled Yw estimates based on a physio- gap estimates in low-input systems (where the mean absolute yield gap 
logical model such as CASE2 need to be treated with some care. The was 469 kg/ha, representing a relative yield gap of 42%). These yield 
extent to which seasonal fluctuations in water supply affect growth and gaps suggest large opportunities for increasing cocoa yield beyond 
productivity under field conditions, is not well understood and probably current levels. Climate factors including precipitation and minimum 
not fully captured by CASE2. For instance, how the dynamics in leaf temperature were found to primarily drive absolute maximum water- 
flushing and cherelle wilt are mediated by seasonal fluctuation in limited and attainable yield gaps in high-input systems. The absolute 
assimilate supply is not well understood. There are also insufficient field and relative attainable yield gap in low-input systems and the relative 
data of these dynamics to validate model simulations. Second, we only yield gaps based on maximum water-limited yield and attainable yield 
analysed data for two years, and may have failed to capture the negative in high-input systems were reduced by increased cocoa planting density 
effects of extreme climatic conditions on yields (Abdulai et al., 2018; and control of black pod disease. This suggests that irrespective of cur-
Gateau-Rey et al., 2018). There was no case of extreme climatic condi- rent climate conditions, investments in good management practices, 
tions during the period for which data was available; hence, we could such as cocoa planting density and improved access to pest and disease 
not evaluate this. Furthermore, regarding the effect of planting density, control by smallholder farmers, offer opportunities to substantially in-
it is important to note that there is a huge variability in planting den- crease production in present-day cocoa farms. 
sities across cocoa farms. Even though we have planting density as a co- 
variate in the regression analysis, it is difficult to assess how much of the Declaration of Competing Interest 
climate sensitivity is actually captured in the regression as compared to 
a data set with more homogeneous planting densities (effects could be The authors declare that they have no known competing financial 
stronger in this case) along a climate gradient. Finally, even if planting interests or personal relationships that could have appeared to influence 
density is similar, farms can differ in the number of unproductive trees the work reported in this paper. 
(Jagoret et al., 2017; Wibaux et al., 2018), which we did not have any 
information on. Data availability 
What are the options to close the yield gap? We recommend 
considering variability in the absolute yield gap for cocoa across Ghana. The authors do not have permission to share data. 
Areas with large absolute yield gaps such as the wetter areas indicate 
potential for larger yield gains, whilst farmers in areas with low absolute 
yield gaps maybe more vulnerable due to climate change. Progressive Acknowledgement 
climate change may alter simulated water-limited yields (upper limit of 
yields in rain-fed system) through direct changes in temperature and This research was conducted within the framework of the CocoaSoils 
water availability (Bunn et al., 2019; Läderach et al., 2013; Schroth program, funded by the Norwegian Agency for Development Coopera-
et al., 2016). Thus, it is important for climate change impact studies to tion (NORAD), Grant number RAF-17/0009-Cocoasoils. We thank 
carefully evaluate projected changes in climate and potential responses Mondelez International for data sharing and Alejandro S. Morales for 
of cocoa growth and yield. Even though yield gaps are lower in the dry support in developing RCASE2, a wrapper around CASE2. 
area, there is still a significant potential for yield increase following best 
management practices. Furthermore, using irrigation (Carr and Lock- Appendix A. Supplementary data 
wood, 2011), mulching (Acheampong et al., 2021), shading (but with 
careful consideration of compatible shade tree species selection) Supplementary data to this article can be found online at https://doi. 
(Abdulai et al., 2018) and planting drought-resistant cocoa varieties org/10.1016/j.agsy.2022.103473. 
(Dzandu et al., 2021) are often specific recommended practices to in-
crease yields under dry conditions. Based on the relative yield gap, References 
management aspects like increasing planting density and application of 
Abdulai, I., Vaast, P., Hoffmann, M.P., Asare, R., Graefe, S., Jassogne, L., Graefe, S., 2018. 
fungicide against black pod are highlighted to be important for closing Cocoa agroforestry is less resilient to sub-optimal and extreme climate than cocoa in 
the yield gap regardless of climatic conditions. However, after achieving full sun. Glob. Chang. Biol. 24 (1), 273–286. https://doi.org/10.1111/gcb.13885. 
11
P.A. Asante et al.                                                                                                                                                                                                  A   g r i c  u l t u  r a  l  S  y  s t e m   s 201 (2022) 103473
Abdulai, I., Hoffmann, M.P., Jassogne, L., Asare, R., Graefe, S., Tao, H.-H., Rötter, R.P., Daymond, A.J., Prawoto, A., Abdoellah, S., Susilo, A.W., Cryer, N.C., Lahive, F., 
2020. Variations in yield gaps of smallholder cocoa systems and the main Hadley, P., 2020. Variation in Indonesian cocoa farm productivity in relation to 
determining factors along a climate gradient in Ghana. Agric. Syst. 181 (October management, environmental and edaphic factors. Exp. Agric. 56 (5), 738–751. 
2019), 102812. https://doi.org/10.1016/j.agsy.2020.102812. https://doi.org/10.1017/S0014479720000289. 
Abu, I.O., Szantoi, Z., Brink, A., Robuchon, M., Thiel, M., 2021. Detecting cocoa Driessen, P.M., 1986. The water balance of the soil. In: Van Keulen, H., Wolf, J. (Eds.), 
plantations in Côte d’Ivoire and Ghana and their implications on protected areas. Modelling of Agricultural Production: Weather, Soils and Crops. PUDOC, 
Ecol. Indic. 129 (February), 107863 https://doi.org/10.1016/j. Wageningen, pp. 76–116. 
ecolind.2021.107863. Duvick, D.N., Cassman, K.G., 1999. Post–green revolution trends in yield potential of 
Acheampong, K., Daymond, A.J., Hadley, P., 2021. Improving Field Establishment of temperate maize in the North-Central United States. Crop Sci. 39, 1622–1630. 
Cacao (Theobroma cacao) Through Mulching, Irrigation and Shading, 2019, Dzandu, E., Enu-kwesi, L., Markwei, C.M., Ayeh, K.O., 2021. Heliyon Screening for 
pp. 898–912. https://doi.org/10.1017/S0014479718000479. drought tolerance potential of nine cocoa (Theobroma cacao L.) genotypes from 
Adejumo, T.O., 2005. Crop protection strategies for major diseases of cocoa, coffee and Ghana. Heliyon 7 (August), e08389. https://doi.org/10.1016/j.heliyon.2021. 
cashew in Nigeria. Afr. J. Biotechnol. 4 (2), 143–150. e08389. 
Adjei Gyapong, T., Asiamah, R.D., 2002. The Interim Ghana Soil Classification System Efron, Y., Epaina, P., Tade, E., Marfu, J., 2005. The relationship between vigour, yield 
and its Relation with the World Reference Base for Soil Resources. Kumasi. and yield efficiency of cocoa clones planted at different densities. In: Proceedings of 
Aggarwal, P.K., Hebbar, K.B., Venugopal, M., Rani, S., Biswal, A., Wani, S.P., 2008. the International Workshop on Cocoa Breeding for Improved Production Systems, 
Global Theme on Agroecosystems Quantifi cation of Yield Gaps in Rain-fed Rice, Accra, Ghana, 19th-21st October 2003, pp. 92–102. 
Wheat, Cotton and Mustard in India. International Crops Research Institute for the Euler, M., Hoffmann, M.P., Fathoni, Z., Schwarze, S., 2016. Exploring yield gaps in 
Semi-Arid Tropics, 36 Pages, 43, p. 36. smallholder oil palm production systems in eastern Sumatra, Indonesia. Agric. Syst. 
Ahenkorah, Yaw, Akrofi, G.S., Adri, A.K., 1974. The end of the first cocoa shade and 146, 111–119. https://doi.org/10.1016/j.agsy.2016.04.007. 
manurial experiment at the Cocoa Research Institute of Ghana. J. Hortic. Sci. 49 (1), Gateau-Rey, L., Tanner, E.V.J., Rapidel, B., Marelli, J.-P., Royaert, S., 2018. Climate 
43–51. https://doi.org/10.1080/00221589.1974.11514550. change could threaten cocoa production: effects of 2015-16 El Niño-related drought 
Ahenkorah, Y., Halm, B.J., Appiah, M.R., Akrofi, G.S., Yirenkyi, J.E.K., 1987. Twenty on cocoa agroforests in Bahia, Brazil. PLoS One 13 (7), e0200454. https://doi.org/ 
years’ results from a shade and fertilizer trial on Amazon cocoa (Theobroma cacao) 10.1371/journal.pone.0200454. 
in Ghana. Exp. Agric. 23 (1), 31–39. https://doi.org/10.1017/S0014479700001101. Gelman, A., Hill, J., 2006. Data Analysis Using Regression and Multilevel/Hierarchical 
Ajagun, E.O., Ashiagbor, G., Asante, W.A., Gyampoh, B.A., Obirikorang, K.A., Models. Data Analysis Using Regression and Multilevel/Hierarchical Models. 
Acheampong, E., 2021. Cocoa Eats the Food: Expansion of Cocoa into Food Cambridge University Press, Cambridge.  
Croplands in the Juabeso District. Food Security, Ghana. https://doi.org/10.1007/ Gerritsma, W., 1995. Physiological Aspects of Cocoa Agronomy and Its Modelling. 
s12571-021-01227-y.  Retrieved from. https://agris.fao.org/agris-search/search.do?recordID=NL2012 
Akrofi, A.Y., Appiah, A.A., Opoku, I.Y., 2003. Management of Phytophthora pod rot 090394. 
disease on cocoa farms in Ghana. Crop Prot. 22 (3), 469–477. https://doi.org/ Global Yield Gap Atlas, 2022. gygaviewer. Retrieved January 26, 2022, from. https:// 
10.1016/S0261-2194(02)00193-X. www.yieldgap.org/gygaviewer/index.html. 
Akrofi, Andrews Yaw, Amoako-Atta, I., Assuah, M., Asare, E.K., 2015. Black pod disease Hengl, T., Mendes de Jesus, J., Heuvelink, G.B.M., Ruiperez Gonzalez, M., Kilibarda, M., 
on cacao (Theobroma cacao, L) in Ghana: spread of Phytophthora megakarya and Blagotić, A., Kempen, B., 2017. SoilGrids250m: global gridded soil information 
role of economic plants in the disease epidemiology. Crop Prot. 72, 66–75. https:// based on machine learning. PLoS One 12 (2), e0169748. https://doi.org/10.1371/ 
doi.org/10.1016/j.cropro.2015.01.015. journal.pone.0169748. 
Allen, R.G., Pereira, L.S., Dirk, R., Smith, M., 1998. Crop Evapotranspiration-Guidelines Hoffmann, M.P., Cock, J., Samson, M., Janetski, N., Janetski, K., Rötter, R.P., 
for Computing Crop Water Requirements- FAO Irrigation and Drainage Paper 56. Oberthür, T., 2020. Fertilizer management in smallholder cocoa farms of Indonesia 
FAO, Rome.  under variable climate and market prices. Agric. Syst. 178 (November 2019) https:// 
Aneani, F., Ofori-Frimpong, K., 2013. An analysis of yield gap and some factors of cocoa doi.org/10.1016/j.agsy.2019.102759. 
(Theobroma cacao) yields in Ghana. Sustain. Agric. Res. 2 (4), 117. https://doi.org/ ICCO, 2021. ICCO Quarterly Bulletin of Cocoa Statistics, Vol XLVII, No. 4, Cocoa Year 
10.5539/sar.v2n4p117. 2020/2021. Retrieved from. https://www.icco.org/wp-content/uploads/Product 
Anim-Kwapong, G.J., Frimpong, E.B., 2004. Vulnerability and Adaptation Assessment ion_QBCS-XLVII-No.-4.pdf. 
under the Netherlands Climate Change Studies Assistance Programme Phase 2 Idachaba, F.S., Olayide, S.O., 1976. The Economics of Pesticides Use in Nigerian 
(NCCSAP2). Cocoa Research Institute of Ghana, New Tafo Akim. Retrieved from. htt Agriculture. Federal Department of Agriculture, Lagos, Nigeria.  
p://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.494.4508&rep=rep1 Jagoret, P., Michel, I., Ngnogué, H.T., Lachenaud, P., Snoeck, D., Malézieux, E., 2017. 
&type=pdf.  Structural characteristics determine productivity in complex cocoa agroforestry 
Appiah, M.R., Sackey, S.T., Ofori-Frimpong, K., Afrifa, A.A., 1997. The Consequence of systems. Agron. Sustain. Dev. https://doi.org/10.1007/s13593-017-0468-0. 
Cocoa Production on Soil Fertility in Ghana: A Review. New Akim-Tafo. Retrieved Lachenaud, P., Oliver, G., 1998. Effect of thinning on cocoa hybrid yields. Plant. Rech. 
from. https://www.ajol.info/index.php/gjas/article/view/1970/10719. Dévelop. 1, 34–40. 
Appiah, M.R., Ofori-Frimpong, K., Afrifa, A.A., 2000. Evaluation of fertilizer application Läderach, P., Martinez-Valle, A., Schroth, G., Castro, N., 2013. Predicting the future 
on some peasant cocoa farms in Ghana. Ghana J. Agric. Sci. 33, 183–190. https:// climatic suitability for cocoa farming of the world’s leading producer countries, 
doi.org/10.4314/gjas.v33i2.1869. Ghana and Côte d’Ivoire. Clim. Chang. 119 (3–4), 841–854. https://doi.org/ 
Asante, P.A., Rozendaal, D.M.A., Rahn, E., Zuidema, P.A., Quaye, A.K., Asare, R., 10.1007/s10584-013-0774-8. 
Anten, N.P.R., 2021. Unravelling drivers of high variability of on-farm cocoa yields Lobell, D.B., Cassman, K.G., Field, C.B., 2009. Crop yield gaps: their importance, 
across environmental gradients in Ghana. Agric. Syst. 193, 103214 https://doi.org/ magnitudes, and causes. Annu. Rev. Environ. Resour. 34, 179–204. https://doi.org/ 
10.1016/J.AGSY.2021.103214. 10.1146/annurev.environ.041008.093740. 
Balasimha, D., Daniel, E.V., Bhat, P.G., 1991. Influence of environmental factors on Lockwood, G., Pang, J.T.Y., 1996. Yields of cocoa clones in response to planting density 
photosynthesis in cocoa trees. Agric. For. Meteorol. 55 (1–2), 15–21. https://doi. in Malaysia. Exp. Agric. 32 (1), 41–47. https://doi.org/10.1017/ 
org/10.1016/0168-1923(91)90019-M. s0014479700025837. 
Beg, M.S., Ahmad, S., Jan, K., Bashir, K., 2017. Status, supply chain and processing of Monsi, M., Saeki, T., 2005. On the factor light in plant communities and its importance 
cocoa - A review. Trends Food Sci. Technol. 66, 108–116. https://doi.org/10.1016/j. for matter production. Ann. Bot. 95 (1907), 549–567. https://doi.org/10.1093/aob/ 
tifs.2017.06.007. mci052. 
Bhattarai, S., Alvarez, S., Gary, C., Rossing, W., Tittonell, P., Rapidel, B., 2017. Monzon, J.P., Slingerland, M.A., Rahutomo, S., Agus, F., Oberthür, T., Andrade, J.F., 
Combining farm typology and yield gap analysis to identify major variables limiting Grassini, P., 2021. Fostering a climate-smart intensification for oil palm. Nat. 
yields in the highland coffee systems of Llano Bonito, Costa Rica. Agric. Ecosyst. Sustain. 4 (7), 595–601. https://doi.org/10.1038/s41893-021-00700-y. 
Environ. 243 (April), 132–142. https://doi.org/10.1016/j.agee.2017.04.016. Nakagawa, S., Schielzeth, H., 2010. Repeatability for Gaussian and non-Gaussian data: A 
Boogaard, H., van der Grijn, G., 2020. Agrometeorological Indicators from 1979 to practical guide for biologists. Biol. Rev. https://doi.org/10.1111/j.1469- 
Present Derived from Reanalysis. https://doi.org/10.24381/cds.6c68c9bb. 185X.2010.00141.x. John Wiley & Sons, Ltd. November 1.  
Bunn, C., Läderach, P., Quaye, A., Muilerman, S., Noponen, M.R.A., Lundy, M., 2019. Ofori-Frimpong, K., Afrifa, A.A., Appiah, M.R., 2006. The response of new cocoa hybrids 
Recommendation domains to scale out climate change adaptation in cocoa to phosphate and potassium fertilizers, report 2005/2006. Cocoa Research Institute, 
production in Ghana. Clim. Serv. 16 https://doi.org/10.1016/j.cliser.2019.100123. Ghana, pp. 38–43. 
Carr, M.K.V., Lockwood, G., 2011. The water relations and irrigation requirements of Opoku, I., Appiah, A., Akrofi, A., Owusu, G., 2000. Phytophthora megakarya: A potential 
cocoa (Theobroma cacao L.): A review. Exp. Agric. 47 (04), 653–676. https://doi. threat to the cocoa industry in Ghana. Ghana J. Agric. Sci. 33 (2) https://doi.org/ 
org/10.1017/S0014479711000421. 10.4314/gjas.v33i2.1876. 
Chapman, R., Cock, J., Samson, M., Janetski, N., Janetski, K., Gusyana, D., Oberthür, T., R Core Team, 2018. A Language and Environment for Statistical Computing. R 
2021. Crop response to El Niño-southern oscillation related weather variation to help Foundation for Statistical Computing, Vienna.  
farmers manage their crops. Sci. Rep. 11 (1), 1–8. https://doi.org/10.1038/s41598- Rahn, E., Vaast, P., Läderach, P., van Asten, P., Jassogne, L., Ghazoul, J., 2018. Exploring 
021-87520-4. adaptation strategies of coffee production to climate change using a process-based 
Cilas, C., Sounigo, O., Efombagn, B., Nyassé, S., Tahi, M., Bharath, S.M., 2018. Advances model. Ecol. Model. 371 (July 2017), 76–89. https://doi.org/10.1016/j. 
in Pest- and Disease-Resistant Cocoa Varieties. October, pp. 345–364. https://doi. ecolmodel.2018.01.009. 
org/10.19103/as.2017.0021.22. Rhebergen, T., Fairhurst, T., Whitbread, A., Giller, K.E., 2018. Yield gap analysis and 
Daymond, A.J., Acheampong, K., Prawoto, A., Abdoellah, S., Addo, G., Adu-Yeboah, P., entry points for improving productivity on large oil palm plantations and 
Hadley, P., 2017. Mapping Cocoa productivity in Ghana, Indonesia and Côte smallholder farms in Ghana. Agric. Syst. J. 165, 14–25. https://doi.org/10.1016/j. 
d'Ivoire. In: 2017 International Symposium on Cocoa Research (ISCR), Lima, Peru, agsy.2018.05.012. 
13–17 November 2017. Peru, Lima, pp. 13–17. 
12
P.A. Asante et al.                                                                                                                                                                                                  A   g r i c  u l t u  r a  l  S  y  s t e m   s 201 (2022) 103473
Ruf, F., Schroth, G., Doffangui, K., 2015. Climate change, cocoa migrations and van Vliet, J.A., Giller, K.E., 2017. Mineral nutrition of cocoa: A review. In: Advances in 
deforestation in West Africa: what does the past tell us about the future? Sustain. Sci. Agronomy, vol. 141. Academic Press Inc., pp. 185–270. https://doi.org/10.1016/bs. 
10 (1), 101–111. https://doi.org/10.1007/s11625-014-0282-4. agron.2016.10.017 
Schroth, G., Läderach, P., Martinez-Valle, A.I., Bunn, C., Jassogne, L., 2016. Vulnerability Wairegi, L.W.I., van Asten, P.J.A., Tenywa, M.M., Bekunda, M.A., 2010. Abiotic 
to climate change of cocoa in West Africa: patterns, opportunities and limits to constraints override biotic constraints in east African highland banana systems. Field 
adaptation. Sci. Total Environ. 556, 231–241. https://doi.org/10.1016/j. Crop Res. 117 (1), 146–153. https://doi.org/10.1016/j.fcr.2010.02.010. 
scitotenv.2016.03.024. Wang, N., Jassogne, L., Van Asten, P.J.A., Mukasa, D., Wanyama, I., Kagezi, G., Giller, K. 
Sonwa, D.J., Weise, S.F., Schroth, G., Janssens, M.J.J.J., Shapiro, H.-Y.Y., 2018. E., 2015. Evaluating coffee yield gaps and important biotic, abiotic, and 
Structure of cocoa farming systems in west and Central Africa: a review. Agrofor. management factors limiting coffee production in Uganda. Eur. J. Agron. 63, 1–11. 
Syst. 93 (5), 2009–2025. https://doi.org/10.1007/s10457-018-0306-7. https://doi.org/10.1016/j.eja.2014.11.003. 
Souza, C.A.S., dos Dias, L.A.S., Aguilar, M.A.G., Sonegheti, S., Oliveira, J., Costa, J.L.A., Wessel, M., 1971. Fertiliser requirements of cacao (Theobroma cacao L.) in South Western 
2009. Cacao yield in different planting densities. Braz. Arch. Biol. Technol. 52 (6), Nigeria. In: Communication 61. Koninklijk Instituut voor de Tropen. Amsterdam. 
1313–1320. Wessel, Marius, Quist-Wessel, P.M.F., 2015. Cocoa production in West Africa, a review 
Tosto, A., Zuidema, P.A., Goudsmit, E., Evers, J.B., Anten, N.P.R., 2022. The Effect of and analysis of recent developments. NJAS - Wageningen J. Life Sci. 74–75, 1–7. 
Pruning on Yield of Cocoa Trees Is Mediated by Tree Size and Tree Competition. https://doi.org/10.1016/j.njas.2015.09.001. 
Manuscript Submitted for Publication. Wibaux, T., Konan, D.C., Snoeck, D., Jagoret, P., Bastide, P., 2018. Study of tree-to-tree 
Toxopeus, H., 1985. Botany, types and populations. In: Wood, G.A.R., Lass, R.A. (Eds.), yield variability among seedling-based cacao populations in an industrial plantation 
Cocoa (Chapter 2, pp. 17–18). Blackwell Science, Oxford. https://doi.org/10.1002/ in Côte D’lvoire. Exp. Agric. 54 (5), 719–730. https://doi.org/10.1017/ 
9780470698983.ch2.  S0014479717000345. 
van Ittersum, M.K., Cassman, K.G., Grassini, P., Wolf, J., Tittonell, P., Hochman, Z., Zuidema, P.A., Gerritsma, W., Mommer, L., Leffelaar, P., a., 2003. A Physiological 
2013. Yield gap analysis with local to global relevance—A review. Field Crop Res. Production Model for Cacao: Model Description and Technical Program Manual of 
143, 4–17. https://doi.org/10.1016/j.fcr.2012.09.009. CASE2 Version 2.2, pp. 5–135. 
van Kraalingen, D., 1995. The FSE System for Crop Simulation, Version 2.1. Wageningen. Zuidema, P.A., Leffelaar, P.A., Gerritsma, W., Mommer, L., Anten, N.P.R., 2005. 
Retrieved from. https://research.wur.nl/en/publications/the-fse-system-for-crop-si A physiological production model for cocoa (Theobroma cacao): model presentation, 
mulation-version-21. validation and application. Agric. Syst. 84 (2), 195–225. https://doi.org/10.1016/j. 
Van Oort, P.A.J., Saito, K., Dieng, I., Grassini, P., Cassman, K.G., Van Ittersum, M.K., agsy.2004.06.015. 
2017. Can yield gap analysis be used to inform R & D prioritisation? Global Food Zuur, A.F., Ieno, E.N., Walker, N., Saveliev, A.A., Smith, G.M., 2009. Mixed Effects 
Secur. 12 (September 2016), 109–118. https://doi.org/10.1016/j.gfs.2016.09.005. Models and Extensions in Ecology with R. Springer New York, New York, NY.  
13