INFLUENCES OF SHADING AND FERTILIZATION ON ON-FARM YIELDS OF COCOA IN GHANA

Summary

Most cocoa farms in Ghana are cultivated in complex agroforest systems, with plant growth and cocoa productivity being affected. The objective of this study was to investigate how shade trees affect cocoa yield, temperature and soil nutrients in low-input cocoa systems. Establishing plots on 24 farms in four locations (districts) in Ghana, we assessed the influence of varying canopy cover and fertilization on cocoa yields. Results showed no relationship between canopy cover and cocoa yields in the light crop season (February to August). For the main crop season (September to January), there was an interaction between shade and yields: Yields were higher on no-shade plots than on shaded plots in two districts, whilst there were no differences at the two other districts possibly due to differences in precipitation and soil nutrient status. On the other hand, there was a positive effect of increased canopy cover on yields within the shaded plots. Soil nutrient analyses revealed no significant differences between shaded and no-shade plots and adequate levels of N, K⁺, Fe²⁺, Cu²⁺ and Zn²⁺ were recorded. However, soil contents of P, C, Mg²⁺ and Ca²⁺ were below recommended values. Peak temperatures recorded in the cocoa canopies were above the recommended range for this species. Although shade trees had a slight modifying effect on peak temperatures, the magnitude appeared too small to have any practical effects.

INTRODUCTION

Cocoa (Theobroma cacao L.) in Ghana and West Africa is grown in complex environments, in which multiple ecological, climatic and agronomic management factors can either limit or enhance growth and productivity (Cunningham and Arnold, 1962; de Almeida and Valle, 2007; Zuidema et al., 2005). In Ghana, traditional or extensively managed smallholder plantations are established by sowing cocoa seeds and/or planting seedlings at random on completely cleared forest (primary or secondary) or fallow lands. In these systems, fertilizers and agrochemical inputs are usually not applied or used in insufficient quantities (Baah et al., 2011). Some cocoa farms, however, are managed more intensively, as evidenced by cocoa seedlings planted in rows, the regular application of fertilizer and pesticides, and/or the complete removal of shade trees (Gockowski et al., 2013).

In its natural environment in the Amazon, cocoa evolved as an understory or sub-canopy tree species (Greenberg et al., 2000), and is widely recognized as having the capacity to grow in a low light environment. However, the shade requirements of mature cocoa have been questioned (Cunningham and Arnold, 1962). Results of long-term trials on relationship between shade, fertilizer application and yield conducted in a forest environment in Ghana showed very high yields after shade removal in well-established cocoa plantations on fertilized soils (Ahenkorah et al. 1974; 1987; Cunningham and Arnold, 1962). Cunningham and Arnold (1962) argued that ‘heavy shade is one of the greatest growth-limiting factors and the most vigorous growth and highest yields are only possible without shade’ (p. 220). Later, research by Ahenkorah et al., (1987) found that mean yield of heavily shaded cocoa (Amelonado) was approximately half that of no-shade cocoa, whilst the yield from a medium shade system was intermediate. Despite the economic gains that can be derived from increased yields, results from the fertilizer-shade trials also document numerous deleterious effects that offset the positive gains when eliminating shade over time. For example, the productive lifespan of no-shade, intensively cropped cocoa did not extend beyond 10 years; the point at which yields started to decline (Ahenkorah et al., 1974). The most prominent negative effect was the increase in pest and disease damage (Ahenkorah et al., 1987; Campbell, 1984). Faster weed growth and greater nutrient demands from the cocoa tree have also been observed (Ahenkorah et al., 1974). Whilst eliminating shade may boost yield in the short term, a non-shaded cocoa system is not necessarily economically justified considering the negative effects associated with the lack of shade and the increased demand for agro-chemical inputs in order to maintain productivity. Besides, shade in cocoa plantation can provide considerable ecological and economic benefits, especially in cases of low input agriculture, where sustainability rather than maximization of productivity is of major interest (Beer, 1987).

Cocoa is exceptionally demanding in its soil requirements (Smith, 1975). For good growth, forest areas with deep, well-drained soils that vary from loamy sands to friable clays, red or reddish brown in colour, are preferred (Charter, 1953). Ahenkorah (1981) has documented the critical thresholds for some selected macro and micro nutrients for cocoa cultivation in Ghana. According to Afrifa et al., (2009), the content of phosphorus (P) is very low compared to nitrogen (N) in most soils in Ghana. Especially the organic P content is important for the nutrition of cocoa (Appiah, 1975) as deficiency of soil P was reported as the main cause of decline in yields (Ahenkorah et al., 1974). In recent times, however, the notion that P is the most limiting factor, and that N is not critical for cocoa yield in Ghana, has been challenged. Trials conducted by the Cocoa Research Institute of Ghana (CRIG) indicated high productivity when N is applied; hence, revising fertilizer recommendations to also incorporate N (Afrifa et al., 2009). The effect of potassium (K) on cocoa yield in most experiments in Ghana was found to be negligible (Ahenkorah et al., 1982). Ahenkorah and company attributed this to high K buffering capacity of most soils in Ghana.

In addition to rainfall, temperature has been identified as one of the most critical climatic factors for the growth and development of cocoa (Daymond and Hadley, 2004). In their projections, Anim-Kwapong and Frimpong (2008) predict an increase between 0.6 and 5.4 °C in mean annual temperature over the next 70 years in the moist evergreen (ME) and moist semi-deciduous forest zones of Ghana, which encompasses the cocoa growing belt. These authors contend that longer periods of drought and higher temperatures are causing fluctuations in productivity, and that farmers are already experiencing losses by tree desiccation and death. Hence, they recommend that farmers develop resilient shade grown cocoa systems (Anim-Kwapong and Frimpong, 2008; Nellemann, 2009) as it is believed that shade trees play a key role in regulating humidity and temperature fluctuations (Beer et al., 1998). In cocoa plantations, the difference in temperature measured outside and inside canopy layers has been noted to be around 2 ± 0.5 °C, although higher differences were registered during summer weeks (de Almeida and Valle, 2009). In the Ghanaian context, the temperature in the cocoa growing system has been described mostly using standard weather data (Anim-Kwapong and Frimpong, 2008), which may represent the general scenario rather than the actual conditions in the cocoa environment (Bonaparte and Ampofo, 1977; de Almeida and Valle, 2007).

Despite the perceived role of shade trees in cocoa cultivation, there is limited – and to some extent contradictory – information on the ameliorating effect of canopy cover (CC) by shade trees on on-farm soil nutrients and temperature and how these may influence yields in smallholder plantations. Whilst a range of experiments has been conducted on-station, few studies aiming to measure key growing conditions and to understand the variation of cocoa yield on-farm have been developed (Isaac et al., 2007; Koko et al., 2013; Wade et al., 2010) across West Africa. The objective of this paper therefore was to measure the effect of CC by shade trees on cocoa yields, soil nutrient contents and temperature in two main cocoa growing agro-ecological zones in Ghana in an effort to better understand the conditions affecting production at the farm level. Specifically, the research sought to answer the questions: (i) What is the variation in temperature and the status of soil nutrients in shaded and non-shaded cocoa farms?; and (ii) What is the relationship between CC of shade trees, fertilizer use and cocoa yields?

MATERIALS AND METHODS

Study area

The study was conducted in four cocoa growing communities in four administrative districts located in the Ashanti and Western regions of Ghana (Figure S1). Sites for the Ashanti Region (Amansie West and Atwima Nwabiagya) fall under the Moist Semi-Deciduous Southeast subtype (MSSE) whilst sites for the Western Region (Sefwi Wiawso and Wassa Amenfi West) fall under the ME forest zones (Hall and Swaine, 1981). The MSSE forest zone is characterized by moderate annual rainfall (1250–1500 mm) with uniformly high temperatures (mean monthly minimum and maximum of 27 and 31 °C) and high relative humidity. The ME forest zone is characterized by a semi-equatorial climate that has high rainfall (1500–1750 mm) and daily temperatures that range from 22 to 34 °C. Temperatures are high throughout the year, though March is generally the hottest month. The study area is characterized by two rainy seasons, the major occurring between April and July and peaking in May/June, and the minor occurring between September and October with a short dry period in August. Humidity is high, ranging from 70 to 90% for the monthly means.

Soils in these areas are generally developed from rocks of the Birimian system (middle Pre-Cambrian) (Adu, 1992). These consist mainly of argillaceous sediments metamorphosed into phyllites. The well-drained soils belong to the Forest Ochrosol (MSSE) and Forest Ochrosol-Oxysol Intergrade (ME) Great Soil Group of the Ghanaian soil classification system (Brammer, 1962) and are generally accommodated as Acrisols in the FAO-UNESCO Revised Legend (FAO, 1988) and as Ultisols in the US Soil Taxonomy (OSSD, 1998). These soils under natural conditions contain adequate nutrients that are tied-up within the organic layer in their top soils. According to the Soil Research Institute of Ghana, soils in the MSSE generally contain higher levels of available P, whilst soils in the ME have high N, available Ca²⁺ and organic matter content.

Experimental design and treatments

Data were collected from 11 and 13 farms in the Ashanti and Western regions, respectively. Age of the cocoa trees ranged between 8 and 28 years, often with several ages being present on individual farms. This range is considered to be the economically favourable age of cocoa trees (Obiri et al., 2007). Farms in each region were selected such that they were at least 2 km apart in each community. These farms represent traditional cocoa systems in which cocoa seeds were sown on previously cleared forestlands with extremely variable CC of shade trees, spacing and age. On each farm, four circular plots of radius 10 m (341 m² area) were delineated, two with a tree in the middle forming a canopy above the cocoa trees and two without any shade trees in the plots. One shaded and one no-shade plot were fertilized (see later), leaving one shaded and one no-shade plot as controls. Hence, the experiment can be described as a full factorial design with two factors, shade/no-shade and fertilizer/no-fertilizer, replicated on the 24 farms (blocks). The shade trees comprised 22 different species with varying ecological classifications and guilds (Table S1).

Seventy nine percent of the trees were deciduous, shedding their leaves mostly in the dry period from November to April. Seventeen percent of the trees were non-pioneer light demanders and 74% were pioneers. The trees had varying CC for the various plots in the four locations (Table S2).

Plots were delineated on January 2012 and the circular plots included from 14 to 78 trees with an average of approximately 39 cocoa trees (equivalent to 1242 trees ha⁻¹). The boundaries of the plots were marked with nylon ropes to delineate them from other portions of the farms. Mature cocoa pods from the segregated plots were harvested every two weeks and pooled together to constitute one sample. The number of viable pods in each sample were counted, after which the beans were fermented in small heaps covered by banana leaves for 5–6 days as recommended by CRIG. The fermented beans were then dried and weighed. This was done in two crop seasons (2012/2013 and 2013/2014), with harvested pods during two ‘light crop’ periods (February to August) and two ‘main crop’ periods (September to January). Yield data recorded on plots were extrapolated to kg ha⁻¹ for easy interpretation.

Soil nutrients

Soil sampling was conducted before the start of the experiment. At each farm, two composite soil samples consisting of equal portions of soil from the two no-shade plots at depths of 0–15 cm and 15–30 cm were taken. Similar samples were taken for the shaded plots. The soil samples were analysed at the Soil Science Laboratory of the CRIG. Soil pH was determined in a 1:2.5 soil–water suspension using a glass electrode and pH metre (van Reeuwijk, 2002). Organic carbon (C) was measured by the wet combustion method of Walkley and Black (1934). Total N was determined using the Kjeldahl method (Bremner, 1965). Available P was determined colourimetrically on an UV Visible Spectrophotometer by the Truog method (Truog, 1930). Exchangeable bases were determined by leaching the soil with 1 M ammonium acetate solution (Hanway and Heidel, 1952), and analysing the leachate by atomic absorption spectrometry (AAS) for potassium (K⁺), magnesium (Mg²⁺) and calcium (Ca²⁺). Micronutrients including iron (Fe²⁺), copper (Cu²⁺) and zinc (Zn²⁺) were determined by leaching the soil with Mehlich–3 extractant (Mehlich, 1984), and the leachate was analysed on AAS for the concentrations of those elements. After the soil analysis, Asaase wura fertilizer (NPK 0-22-18 +9 CaO, +7 S, +6 MgO) was broadcasted at a rate of 11.8 kg (equivalent to 375 kg ha⁻¹) on each of the fertilized plots in June 2012 and July 2013. This fertilizer formulation was used due to its recommendation by CRIG and was also preferred by farmers in previous surveys (Aneani and Ofori-Frimpong, 2013).

Determination of air temperatures and crown area

In order to determine the air temperature in the cocoa tree canopy, one farm was selected in each of the four districts except Wassa Amenfi West. Air temperature was measured at the highest point of the cocoa tree canopy in the middle of each plot. A calibrated Tinytag 2 Plus TGP-4017 temperature measuring device manufactured by Gemini Data Loggers Ltd (UK) was mounted above the cocoa canopy in the no-shade stands and above the cocoa canopy but under the shade trees in the shade tree stands. In total, four Tinytags were mounted on each farm in order to compare the air temperatures in the shaded and no-shade plots. The Tinytags were shielded in a 12 × 20 cm nursery pot, which was wrapped with aluminium foil. A nylon thread was used to tie it to a branch at the top of the cocoa canopy. The devices were set to record air temperature in intervals of 5 minutes from January to December 2013.

Finally, to estimate the CC provided by the shade trees on the plots, the crown area of each upper canopy non-cocoa tree within the plot was estimated by measuring the diameter of the crown (CD) in four different directions and calculating the average (Blozan, 2006). The diameter measurements were taken from one tip of the crown to the other. The crown spread was used to calculate the canopy area (CA):

(1) $$\begin{equation} {\rm{CA}}\; = \pi^{*}{\left( {\frac{{{\rm{CD}}}}{2}} \right)^2}. \end{equation}$$

The CC for the upper canopy trees was expressed as a percentage of the size of the plot and used as a proxy for shade cover per plot:

(2) $$\begin{equation} {\rm{CC}}\; = \;\left( {\frac{{{\rm{CA}}}}{{{\rm{Plotsize}}}}} \right) \times 100. \end{equation}$$

Data analysis

Cocoa yields from the light and main crops were analysed separately in an ANCOVA model with random effects. An optimal transformation of the response variable within the Box-Cox family (Box and Cox, 1964) was chosen in the maximal model, which consisted of the main effects of location (four levels: Atwima Nwabiagya, Amansie West, Wassa Amenfi West and Sefwi Wiawso), fertilizer (two levels: yes/no), shade (two levels: yes/no) and crown cover (continuous, 0 when shade = no), and the interactions between location and the three other main effects. The random effects were of year (two levels), farm (24 levels), subplot (96 levels), the interaction between location and year, the interaction between farm and year and the residual term, which may be interpreted as the interaction between subplot and year. In formula, the maximal model is

$$\begin{eqnarray*} {\rm{yiel}}{{\rm{d}}^\lambda } &=& \alpha ({\rm{location}},{\rm{shade}}) + \beta ({\rm{location}},{\rm{fertilizer}}) + \gamma ({\rm{location}})*({\rm{crown}}\nonumber\\ && - {\mu _{{\rm{location}},{\rm{shade}}}}) + {\rm{A}}({\rm{year}}) + {\rm{B}}({\rm{location}},{\rm{year}}) + {\rm{C}}({\rm{farm}}) + {\rm{D}}({\rm{farm}},{\rm{year}})\nonumber\\ && + {\rm{E}}({\rm{subplot}}) + {\rm{F}}({\rm{subplot}},{\rm{year}}), \end{eqnarray*}$$

in which the design parameter μ _{location,shade} quantifies the mean CC in the four locations and shade conditions, i.e. these equal zero when shade = no and are given as estimates in the second part of Table 1 for shade = yes. The inclusion of these design parameters implies that the shade variable describes the difference between the no-shade subplots and the average of shaded subplots. The cocoa yield has substantial variation, and the different sources of variation were quantified using the random effects A to F above. Moreover, the inclusion of these random effects is necessary in order to achieve valid statistical inference on the systematic effects. Model selection was done keeping all the random effects, and selecting the best subset of the systematic effects using the Second-order Akaike Information Criterion (Burnham and Anderson, 2002). Significance tests of the systematic effects in the selected model was done using asymptotic chi-square tests on the likelihood ratio test statistic based on the maximum likelihood fits, and parameter estimates and post hoc analysis was done on the restricted maximum likelihood fits (Martinussen et al., 2012).

Table 1. Estimated variance components on random effects and significance tests and parameter estimates of selected systematic effects.

With respect to air temperature, the minimum, maximum and mean temperatures were calculated for each day and sensor. For each of these three values, the average, minimum and maximum temperatures were calculated, resulting in nine variables for each sensor (Table S3). Analysis of variance tests were performed to determine significant differences of soil chemical properties and air temperature between the locations and treatments (Table S3). These variables were then subjected to two-way analysis of variance according to a model with the effects of location (three or four levels: Atwima Nwabiagya, Amansie West, Sefwi Wiawso and Wassa Amenfi West in the case of soil nutrient analyses) and shade (two levels: yes/no): Y = α(location) + β(treatment), where the tests showed significant differences between locations and pairwise comparisons were made using Tukey's Studentized range test (P < 0.05). In all cases, model assumptions were validated by plots of residuals against predicted values, and by normal quantile plots of residuals.

RESULTS

Cocoa yields in relation to canopy cover and fertilization

Cocoa yields were extremely variable, ranging from an average of 1204 ± 489 kg ha⁻¹ year⁻¹ in Wassa Amenfi West to 386 ± 210 kg ha⁻¹ year⁻¹ in Amansie West. The average yields across all farms were 219 ± 214 kg ha⁻¹ year⁻¹ for the light crop and 674 ± 474 kg ha⁻¹ year⁻¹ for the main crop (mean ± s.d.).

The model for the main crop selected by the Akaike Information Criterion is

$$\begin{eqnarray*} {\rm{yiel}}{{\rm{d}}^{0.4}} &=& \alpha ({\rm{location}},{\rm{shade}}) + \beta ({\rm{fertilizer}}) + {\gamma ^*}({\rm{crown}} - {\mu _{{\rm{location}},{\rm{shade}}}})\nonumber\\ && + {\rm{A}}({\rm{year}}) + {\rm{B}}({\rm{location}},{\rm{year}}) + {\rm{C}}({\rm{farm}}) + {\rm{D}}({\rm{farm}},{\rm{year}}) \nonumber\\ &&+ {\rm{F}}({\rm{subplot,year}}). \end{eqnarray*}$$

Here, the random effect of subplot is removed since the corresponding variance component was estimated to 0. The R-square value of the selected model is 0.32, which shows that the main crop is difficult to predict by the systematic factors. The unexplained variation is decomposed in the variance components as shown in Table 1.

Thirty three percent of the unexplained variation may be attributed to differences between farms not accounted for by the location. The remaining part may be attributed to annual variation, where the largest part comes from annual variation within farms (40%). The annual variation within subplots (13%) also includes the residual variation, and is presumably also due to error in recording yields in the main crop. Significance tests and parameter estimates for the selected systematic effects are shown in Table 1. There is a highly significant positive effect of fertilizer, a borderline significant positive effect of the size of the CC in the shaded subplots and a significant interaction between location and shade. The latter means that the contrast between the no-shade and the typical shaded subplots is different in the four investigated locations.

As shown in Figure 1, there are significant negative effects of shade in Atwima Nwabiagya (P = 0.0382) and Amansie West (P = 0.0040), but no effect of shade in Wassa Amenfi West (P = 0.7551) and Sefwi Wiawso (P = 0.4832). It is important to recall that the model includes two different terms describing the shade: The factor shade (yes/no), which illustrate a negative effect in the two sites mentioned above, and the covariate crown cover, which shows increasing yields with increasing levels of crown cover within the shaded plots. The covariate means that for large crown covers, yields are increased above the level of the no-shade plots, except in the case of the low-yielding location of Amansie West, which consisted of five shaded plots instead of six due to disease on one shade tree that resulted in sudden loss of all leaves on the tree.

Figure 1. Canopy cover and yield of cocoa (triangles = with fertilizer, crosses = without fertilizer) in (a) Amansie West; (b) Atwima Nwabiagya; (c) Sefwi Wiawso and (d) Wassa Amenfi West. The lines show the model predictions (full lines = with fertilizer, dashed line = without fertilizer). These model lines are drawn inside the range of the canopy cover in the four districts and two fertilizer regimes, but for visualization extended around cover = 0% for the subplots without shade.

Fertilizers resulted in an increase of 14.5% in the main crop yield (P < 0.0001) when compared to unfertilized plots (953 ± 65 vs. 832 ± 65 kg ha⁻¹) (mean ± s.e.).

The model for the light crop selected by the Akaike Information Criterion is

$$\begin{eqnarray*} {\rm{yiel}}{{\rm{d}}^{0.2}} &=& \beta ({\rm{location}},{\rm{fertilizer}}) + {\rm{A}}({\rm{year}}) + {\rm{B}}({\rm{location}},{\rm{year}}) + {\rm{C}}({\rm{farm}})\nonumber\\ && + {\rm{D}}({\rm{farm}},{\rm{year}}) + {\rm{E}}({\rm{subplot}}) + {\rm{F}}({\rm{subplot,year}}) \end{eqnarray*}$$

In particular, there is no systematic relation between the light crop yield and shade. The R-square value of the selected model is 0.16, showing that only a small fraction of the light crop is predicted by the systematic factors. The unexplained variation was decomposed in the variance components as shown in Table 1. As for the main crop, we see that there was some variation between farms not accounted for by the location (8% of the total unexplained variation), but besides this most of the unexplained variation may be ascribed to annual variation. In contrast to the main crop, most of the unexplained variation of the light crop is jointly over all farms, namely 40% annually over all locations and 12% within the four locations. The interaction between location and fertilization was significant (P = 0.0365). A fifth power back transformation shows only a significant effect (P = 0.0006) of fertilizer in Amansie West.

Across all treatments, Wassa Amenfi West had the highest mean yield for the main crop at 1012 ± 62 kg ha⁻¹, followed by Antwima Nwabiagya (582 ± 54 kg ha⁻¹), Sefwi Wiawso (483 ± 73 kg ha⁻¹) and Amansie West (313 ± 43 kg ha⁻¹), respectively. For the light crop, Antwima Nwabiagya had the highest mean yield (326 ± 18 kg ha⁻¹), followed by Sefwi Wiawso (188 ± 30 kg ha⁻¹), Wassa Amenfi West (183 ± 17 kg ha⁻¹) and Amansie West (73 ± 19 kg ha⁻¹) (mean ± s.d.), respectively.

Effect of canopy cover and soil nutrients

Analyses of variance showed no significant differences in the soil nutrient content between shaded and no-shade plots. However, there were differences between locations (Table 2). Soils in Wassa Amenfi West had significantly lower pH levels compared to the other three districts, with Sefwi Wiawso recording the highest pH. There were no significant differences in total soil N between locations, but Wassa Amenfi West had the lowest levels of total C. Available P was very low in Amansie West, with the highest values recorded in Sefwi Wiawso. K⁺ was high in both Atwima Nwabiagya and Sefwi Wiawso, whilst Mg²⁺ and Ca²⁺ levels were low in Wassa Amenfi West with Amansie West and Sefwei Wiawso recording the highest levels in Mg²⁺ and Ca²⁺, respectively. Sefwi Wiawso had the highest levels of Zn²⁺, whilst Cu²⁺ levels were low in Wassa Amenfi West. Fe²⁺ was high in Wassa Amenfi West and low in Amansie West.

Table 2. Variability of soil chemical properties (Mean ± SE) within 0–30 cm on cocoa farms in Amansie West, Atwima Nwabiagya, Wassa Amenfi West and Sefwi Wiawso districts in Ghana.

Values with different letters (a–d) in each row represent statistical differences (P < 0.05) according to Tukeys HSD test (n = 112).

^†Soil nutrient thresholds for Cocoa cultivation in Ghana [Adopted from Ahenkorah (1981)].

Effect of canopy cover on temperature

Air temperatures were recorded from January 30 to December 18, 2013. There was a wide temperature variation recorded in the cocoa canopy ranging from 14.7 to 45.4 °C across shaded and no-shade plots (Figure 2). Mean values for the sites were significantly different, but all between 25 and 26 °C. Maxima daily air temperatures were recorded in the main dry season (November–March), whilst the minima were recorded at the end of the main rainy season (August) and at one occasion in September. There were also significant differences between location for the lowest daily mean air temperature, the lowest maximum, the highest minimum and the absolute minimum, but differences were small and within 1–2 °C (Table S3). The analyses showed significant effects of shade on the lowest daily mean (P = 0.023), the highest minimum air temperature (P = 0.02) and the absolute minimum (P = 0.0298). However, differences were always less than 1 °C and mostly less than 0.5 °C. Maximum air temperatures were not significantly affected by shade.

Figure 2. Average daily maximum temperatures within shaded and unshaded plots in (a) Amansie West; (b) Atwima Nwabiagya and (c) Sefwi Wiawso: n = 2.

DISCUSSION

The effect of canopy cover on cocoa yields, soil nutrients, air temperature at the plot level

Our results illustrate the complexity of analysing cocoa yields in agroforestry systems. Not only did results vary between sites and seasons; the statistical analysis also revealed an apparent paradox in the relationship between shade and yield. Looking at the main crop, an increasing crown cover resulted in higher yields in the shaded plots. However, for the two locations in the Ashanti region, non-shaded plots on average had higher yields than the shaded plots, indicating a negative effect of the shade trees on yields. We can only speculate on the reasons for this, but it seems logical to assume that it is related to competition between cocoa and shade trees. Below ground competition for water and/or nutrients could reduce yields although other studies have indicated that roots from cocoa trees and shade trees occupy different soil layers and thus may access different resources (Isaac et al., 2014; Moser et al., 2010). Reduced irradiation due to shade from the canopy trees may have a negative effect through reduced photosynthesis, but also a positive effect through reduction of photoinhibition. The effect of reduced radiation is thus probably dependent on the position in the canopy, leaves in the deep shade being more likely to experience light limitations, whilst leaves at the top of the canopy may experience less photoinhibition. It is interesting to note that the reduced yields of shaded plots were only found in the Ashanti region, which has lower precipitation than the Western region, and was especially pronounced in Amansie West, which also had a very low soil P-status. In the two locations in the Western Region, results were easier to interpret as inclusion of shade cover seemed to have positive net effects on yields. It is possible that this is due to the improvement of nutrient uptake by cocoa trees under shade trees as documented by Isaac et al. (2007), which ensure the long-term sustainability of shaded cocoa systems alluded to by Obiri et al. (2007). The shade tree species composition was fairly similar between regions and does not seem to be able to explain differences between regions (Table S1).

Shade appeared to have no effect on yields in the light crop. The light cropping season coincides with a dry period, in which many of the shade trees shed their leaves. Of the shade tree species located within the sampled plots the majority are deciduous pioneers that typically shed leaves in the dry seasons as a strategy to avoid moisture loss by evapotranspiration during the driest months of the year. Hence, cocoa trees in shaded and no-shade plots may experience the same light climate during the light crop, whilst during the main cropping season, most trees would be foliated. The data suggest that when all trees have leaves in the wetter months there is a positive impact on cocoa yields, as represented on Figure 1. In any case, there is a need to more fully elucidate the responses of cocoa to shade and competition. In our study, we have ignored that different shade trees may be more or less compatible with the cocoa trees. Taking the analysis to species level, thus considering guilds and ecological characteristics of the species must be one of the next steps in improving cocoa agroforestry. Likewise, it would seem that simple experiments with suspended shade nets above the cocoa canopy could help in separating the effects of below-ground competition from the effects of shade. Finally, extrapolating the studies to the farm scale may help to elucidate whether there is an overall effect of tree density or CC on yields.

Variability between plots

In order to explore the role of fertilizer application on cocoa yields, it is important to understand the baseline soil conditions and how they relate to the recommended soil nutrient minimum thresholds (Ahenkorah, 1981). We found no effect of shade cover on soil nutrient levels, which is partly consistent with studies by Isaac et al. (2007) who found no effect of shade on nutrients like N in Ghana. Discussions with farmers on the study plots indicated that cocoa farm fertilization was done in an inconsistent manner over the previous years, depending on access and cost. This falls in line with Appiah et al. (1997), who found that cocoa in Ghana is mainly produced by small-scale farmers using few fertilizers, and with Ogunlade et al. (2009), who reported that most Nigerian Cocoa farmers do not use fertilizers. Hence, it is no surprise that the baseline soil information (Table 2) shows mixed and low availability of nutrients.

Mean pH, N, K⁺ and micronutrients were within recommended thresholds for cocoa cultivation at all locations, but P, C and Ca²⁺ contents were lower than the critical thresholds. Authors have consistently reported low levels of P in West African cocoa systems (Aikpokpodion 2010; Hartemink, 2005; Ogunlade and Aikpokpodion, 2006), partly due to the relatively low use of inorganic fertilizers. P was the most limiting nutrient, with levels below the recommended threshold at all locations, but especially in Amansie West. The low yields in Amansie West may partially be a result of the exceptional P limitation in the soil, which could not be sufficiently ameliorated by the recommended fertilizer application. This highlights the importance of target fertilization with regards to locations as opposed to the current regime of blanket recommendation. In general, fertilizer application increased yields, with the best result in Wassa Amenfi West, Atwima Nwabiagya and Sefwi Wiawso and an overall increase of 121 kg ha⁻¹.

Cocoa is growing around the world in temperatures ranging from 18 to 21 °C mean minimum and 30 to 33 °C mean maximum (Wood and Lass, 2008). The overall maximum air temperatures recorded in our study were above 40 °C, with the highest temperatures of 43 and 42 °C found in Nwabiagya and Amansie West, respectively. High air temperatures are cause for concern given that high transpiration rates, as a result of high temperatures, produces water stress, which can reduce productivity in cocoa (Anim-Kwapong and Frimpong, 2008; Daymond and Hadley 2004). However, differences in air temperature between shaded and no-shade plots were small; suggesting that the CC provided by shade trees play only a minor role in moderating air temperature in the cocoa canopy. Based on these preliminary results, cocoa agroforestry may not be the only solution to the higher temperatures projected for the cocoa growing belt of Ghana by Anim-Kwapong and Frimpong (2008). However, it should be emphasized that the dataset is limited to three farms, and that increasing the sample size and expanding to other regions may produce clearer results. It should also be acknowledged that shade trees differ widely in the CC that they provide (Asare and Ræbild, 2016) meaning that some species may have larger effects on air temperatures than others.

CONCLUSION

These on-farm experiments showed that increased CC of shade tree had a positive effect on cocoa yield in some locations and a negative effect in other ones. Although this does not provide sufficient clarity on the effect of shade trees across different locations, the findings indicate that a reassessment of the long standing paradigm that shade trees limit cocoa productivity is necessary. More work is required to explain the relationship between CC of shade trees and cocoa yields, especially at the farm level.

Even though the air temperature results confirm some buffering effects of CC on cocoa farm plots, they also show that CC alone is inadequate in ameliorating the microclimate for cocoa. The extreme air temperatures reaching up to 43 °C in the cocoa canopy widely exceed the recommended temperatures for cocoa.

The baseline soil nutrient results showed no effect of CC on soil nutrients, but P, C, Mg²⁺ and Ca²⁺ were below recommended levels. Even though fertilizer was applied at the recommended rates, it did not produce the dramatic increments documented by other authors. There is a need for fertilizer recommendations to take into consideration targeted nutrient deficiencies, in order to ensure efficiency in application across locations.