The study was carried out in cocoa-growing villages in the peripheral zone of the Ebo Forest Reserve, within the Ebo Landscape in the Littoral Region of Cameroon. Ebo forest is a proposed National Park that is co-managed by the Ebo Forest Research Project (EFRP) and the Cameroon Ministry of Forestry and Wildlife (MINFOF) [34].

The Ebo landscape is located between the UTM coordinates 446,622N - 503,984N and 621,032E - 672,785E. The Ebo landscape which is a timidly cocoa growing area in the Littoral Region of Cameroon was selected because it has the fundamental characteristics of a typical cocoa frontier relevant to establishing a baseline for reconciling cocoa expansion and biodiversity conservation. It is part of the Congolian Coastal Forest and covers a total surface area of 2067.78 km² of evergreen lowland and submontane forest with EFR constituting 1424 km² [35] and parallel to the line of the Cameroon Highlands. Ebo forest extends 50 km both North to South, and West to East. The forest belongs to a transition zone between the equatorial and tropical zones with a variation in rainfall patterns of prolonged periods of the dry season and short rainy season [34].

The topography of the Ebo Forest is undulating with gentle to moderate slopes of 100 - 500 m, to steep slopes of up to 1400 m asl. Much of the Northern section is relatively mountainous whereas the Southern half is considerably flatter. The mean monthly temperature ranges between 26˚C - 28˚C with monthly peaks from December-April (28.3˚C - 28.1˚C). Average annual precipitation ranges from 2496 mm to 2950 mm with the wettest months being July to October [34]. The peripheral zone of the Ebo Forest Reserve is comprised of many small villages, which until the late 1950s and early 1960s were inhabited mostly by two major ethnic groups; the people of the Banen and Bassa tribes (Figure 1).

Study sites were selected based on a multi-stage sampling technique of cocoa growing villages at varied altitudinal gradients around the coordinates (North, East, South, and West) of the reserve. In this study, peripheral villages with altitudinal and cultural affiliations around each coordinate were grouped and described as clusters [32]. A total of ten villages (Table 1) distributed in four clusters were selected for the study (Figure 1).

Purposive random sampling was conducted to select the farm of a willing farmer for a demonstration of the effects of management practices on agronomic parameters. A cocoa field was earmarked upon a collaboration agreement (CA) with the farm owner. A demonstration plot of at least 200 trees was mapped out from the field as described by [19] [20] [21]. The plot was further divided into two subplots of 100 trees each; representing the farmers’ management practice (FP) treatment and integrated crop pest and disease management practice (ICPM) treatment. The two treatments were replicated in ten sites grouped into four clusters.

The first Agroecosystem analysis (AESA) data was collected upon establishment of the plot. Subsequently, the response variables (Table 2) were monitored weekly for the 100 trees of each subplot.

Activities in the FP treatment were the usual approach of the farmer using the calendar cropping method while good agricultural practice using cultural operations and the Observe, Learn, Decide and Act (OLDA) model adapted from [25] was applied in the ICPM treatment only.

Based on the observations recorded during AESA data collection, the frequency of occurrence of the response variables, and analysis of the farm situation, a decision was made and actions were taken on the ICPM treatment. Following is the OLDA model (Figure 2) as described by [25]. In this model, there are different criteria for which a farm management decision could be taken.

Some of these actions include pruning, farm sanitation, phytosanitary harvest, rational use of homologated pesticides (spot treatment or general), harvesting, breaking, fermentation, and drying. Each action requires different strategies and management prescriptions.

Harvesting of mature ripe cocoa pods was done regularly in the ICPM treatment following the OLDA model while in the FP treatment, it was done during low peak and high peak seasons following the calendar cropping method. After each harvest, the pods of each treatment were assembled separately, broken and the beans removed. The beans were fermented using banana leaves for 6 days before drying. During fermentation, the beans were mixed (inside-out, upside-down) every other day to ensure proper mixture and fermentation of all the beans. After fermentation, the beans were sundried for 5 days. The weight of dry cocoa beans was recorded and summed at the end of the season.

The data obtained from the FP and the ICPM management practices were subjected to ANOVA to assess if it is statistically significant or not. The effects of the two management practices on variables were analyzed using a two-way analysis of variance (ANOVA) and means were compared with Turkey’s test. Statistical differences concerning agronomic parameters were compared between the identified management practices and across clusters. Simple component analyses were used to show the contribution of the parameters across the axis of categorical variables.

Results of the first AESA showed that the farms selected varied in ages, planting distances of cocoa trees, distances of cocoa trees to other trees, and in number of shade trees per plot. All the plots were on a gentle slope and 90% had good drainage. The shade coverage was medium (i.e. ≥25 ˂50%) in 90% of the plots and 10% had heavy shade. Figure 3 showed that 40% of the experimental plots were between 5 - 10 years and there was no plot between 30 - 40 years.

The planting distances of the ten experimental plots ranged from 3 × 3 m, 3.5 × 3.5 m, and 4 × 4 m (Figure 4).

Table 3 shows the results of the Analysis of Variance (ANOVA) on the variation of 26 response variables recorded over 2 categorical variables of clusters and management practices on cocoa agroecosystems. Four (15%) of the response variables were statistically significant across clusters and 57.7% of the response

Values in the Table represent P-values, the level of significance, ANOVA conducted at α = 0.05. P-values less than 0.05 are highlighted and indicate a significant difference in the response variable for the factor concerned. NILPs = Number of immature large pods, NHSPT = Number of healthy small pods on trunk, NLRPU = Number of large ripe pods that can be used, NLRPBPs = Number of large ripe pods attacked by black pods, NUnLPBPR = Number of unusable large pods attacked by black pods and removed, NUnLPRs = Number of unusable large pods attacked by rodents, NLRPIs = Number of large ripe pods attacked by insects, NUnLPIR = Number of unusable large pods attacked by insects and removed, NCMBr = Number of chupons on main branch, NBsCs = Number of basal chupons, PrNF = Presence of new flushes (N = None, L = Low, M = Medium, H = Heavy), PCrMst = Presence of creepers and mistle toes (N = None, L = Low, M = Medium, H = Heavy), NPsDD/incidences = Number of pods damaged by diseas/incidencee (Bp = Blackpod, Ck = Canker, Rr = root rot), NPsDIn = Number of pods damaged by insects/incidence (Cp = Capsid, Sb = Stem borer, Ps = Psyelles), NEO = Natural enemies observed (Ra = Red ant, Pm = Praying mantis).

variables were statistically significant between the management practices.

Variation in agronomic parameters in cocoa agroecosystems across clusters showed statistically significant differences for number of large ripe pods attacked by insects (NLRPIs) (P = 0.036), number of pods damaged by insects due to incidence of capsid, NPsDIn-Cp (P = 0.000) and red ants observed as natural enemies, NEO-Ra (P = 0.000). The expression of the rest of the variables was not statistically significant across clusters (P > 0.05).

With respect to the management practices, there were significant variations in number of healthy small pods on tree trunk (NHSPT) (P = 0.037), number of large ripe pods that can be used, NLRPU (P = 0.002), number of large ripe pods attacked by blackpod disease, NLRPBPs (P = 0.007), number of unusable large pods attacked by blackpods and removed, NUnLPBPR (P = 0.000), number of large ripe pods attacked by insects, NLRPIs (P = 0.025), number of unusable large pods attacked by insects and removed, NUnLPIR (P = 0.000), number of chupons on main branch, NCMBr (P = 0.000), number of basal chupons, NBsCs (P = 0.000), light presence of new flushes PrNF-L (P = 0.005), medium presence of new flushes PrNF-M (P = 0.037), heavy presence of new flushes PrNF-H (P = 0.007), no presence of creepers and mistle toes, PCrMst_N (P = 0.001), low presence of creepers and mistle toes, PCrMst_L (P = 0.000), medium presence of creepers and mistle toes, PCrMst_M (P = 0.001), and number of pods damaged by disease due to incidence of blackpod, NPsDD_Bp (P = 0.015).

Four out of twenty-six variables analyzed were statistically significant between the four clusters. The number of unusable large pods attacked by black pods and removed (NUnLPBPR) ranged from means of 16 ± 6.34 in the North Cluster to 88.8 ± 31.1 in the North East cluster (NEC), and this difference was statistically significant. Similarly, the number of large ripe pods attacked by insects (NLRPIs) ranged from 126 ± 22.1 in the North cluster to 207.6 ± 35.6 in the East Cluster. The East and South West clusters were not statistically significant in the observed number of large ripe pods attacked by insects (NLRPIs) (Figure 5).

Number of basal chupons (NBsCs) was statistically significant across clusters with the means higher in the East cluster (EC) and South West Cluster (SWC) and lower in North cluster (NC) and North East (NEC). The presence of new flushes (PrNF) was dominantly low and medium across clusters with a low frequency of heavy flushes. The presence of creepers and mistletoes (PrCrMst) was mainly none and low. Frequency of medium and heavycreepers and mistletoes was minimal. Incidences of damage caused by blackpod disease and capsid pest were statistically significant. Damage to crop health caused by blackpod was least in NC and predominant in EC, NEC, and SWC while damage caused by capsid was least in NEC which is at higher altitude with a significant level of natural enemies. Natural enemies with a regulating role to prey on pests were recurrent in NC and SWC. NC recorded highest incidences of red ants while SWC recorded highest incidences of praying mantis. Response variables that were statistically significant are represented in Figure 5.

Bars represent mean occurrence (frequency) ± SE. Means separated through GLM ANOVA with Tukey pairwise comparisons at α = 0.05. Bars with the same letter for each response variable are not significantly different. Vertical axis represents Mean Frequency of occurrence of each response variable and horizontal axis represents the clusters.

Figure 6 shows the expression of the response variable between two management practices implemented in the study sites. It showed that 53.8% of the response variables monitored were statistically significant between the two management practices. They include; number of healthy small pods on the tree trunk, number of large ripe pods that can be used, number of large ripe pods attacked by black pods, number of unusable large pods attacked by black pod disease and removed, number of large ripe pods attacked by insects, number of unusable large pods attacked by insects and removed, number of basal chupons, number of chupons on main branches, low, medium and heavy presence of new flushes, the incidence of pods damaged by blackpod disease, as well as none, low and moderate presence of creepers and mistletoes. Variation of statistically significant response variables across two management practices in the PZ of EFR are represented in Figure 6 below.

From the bar charts, the number of healthy small pods on trunk (NHSPT) ranged from 1846.4 ± 87.2 in the FP treatment to 2260 ± 169 in the ICPM treatment. Similarly, number of large ripe pods that can be used (NLRPU) ranged from 507.3 ± 24.8 in the FP treatment to 706 ± 54.1 for the ICPM treatment. The number of chupons on the main branch (NCMBr) was higher in the FP (1063.9 ± 53.5) than in ICPM (600.3 ± 57.5). This goes to reflect the inverse observed in the NHSPT.

The number of large ripe pods attacked by black pods (NLRPBPs) and number

of large ripe pods attacked by insects (NLRPIs) were higher in the FP than in the ICPM. In addition, number of unusable large pods attacked by black pods and removed (NUnLPBPR) and number of unusable large pods attacked by insects and removed (NUnLPIR) were higher in the ICPM treatment than in the FP treatment. There were no records in the FP treatment because it was allowed for the farm owner to perform his usual approach.

In the ICPM treatment, OLDA model was incorporated in the agroecosystem analysis and preventive measures were applied to minimize pest infestation and disease infection. Should they occur and the pods could no longer be used, sanitary harvest was implemented wherein, the diseased/infested pods were harvested, counted, and dumped in a pit to prevent further spread to healthy pods. Number of unusable large pods attacked by rodents was not statistically significant in the two management practices. This shows that management practice has no effect on rodents’ attacks on the crop.

Number of basal chupons (NBsCs) was higher in FP than in ICPM practice. From observations in ICPM, basal chupons were pruned during the maintenance of the cocoa trees. The no-presence of new flushes (PrNF-N) was minimal in the two MP and similar incidences were recorded. When the heavy presence of new flushes (PrNF-H) is observed, it causes the humidity of the microenvironment to increase and it is conducive for black pod disease. Incidences of no presence of creepers and mistletoes were higher in ICPM while low, medium, and heavy were higher in FP.

One hundred trees were monitored per plot from each study site making a total of 1000trees for each treatment (FP and ICPM) from all the ten sites. An average number of mature ripe pods harvested and weight of dry cocoa beans from the two plots were recorded. Table 4 shows that the average total production varied from 385.83 kg in FP plots to 572.8 kg in ICPM plots.

From the first AESA, 40% of the experimental plots were between 5 - 10 years and there was no plot between 30 - 40 years. The farmers with young farms were willing to allow the research to be conducted in their farms so they could gather experience from the two practices and eventually make sound crop management decisions [19]. This is in harmony with findings of [1] [4] who attest that, a significant proportion of farmers in FFS indicated not having any other source of information. Though the research was not FFS per se, the host farmers with limited access to Agricultural extension Services could testify to gaining experience from the study. The variation in planting distances at 3 × 3 m, 3.5 × 3.5 m and 4 × 4 m showed they have limited information on cocoa production.

Agroecosystem analysis (AESA) data collection is one of the key instruments in sustainable tree crop production [1] [20] [25]. The results show that the appropriation of OLDA model in cocoa AESA could improve quality and quantity of production [36]. This is in harmony with the findings of [22] who described agricultural development as a function of agroecosystem properties enhancing productivity, stability, sustainability and equitability. Farmers monitored cocoa trees and fruits through the collection of AESA data. The farmers adopt the owner management system which [37] regarded as pessimistic in their maximin criterion. Understanding the behaviour and important properties of an agroecosystem requires knowledge of only a few key functional relationships [17] [22] [25] [37]. The length of time that a cocoa farm remains productive and financially viable is determined by the application of good maintenance practices, guided by keen observation, analysis of the situation, deciding on what to do and taking action [25], in particular pruning, farm sanitation, phytosanitary harvest, soil enrichment, and pest and disease control through preventive measures or rational use of pesticides [11] [16] [29]. Pruning was conducted for high observations of basal chupons or chupons on main branches, heavy presence of new flushes and presence of creepers and mistletoes [5] [9] [11]. Farm sanitation was carried out to clean/redress observations on poor drainage, disease/insect zoo, overgrown fields leading to high humidity [21] [25]. Phytosanitary harvest was to remove and discard pods damaged by disease or pest and cannot be used. When absolutely necessary, rational use of pesticides was conducted for observations on usable pods affected by pests and diseases. If less than 20% of the plot was affected, spot treatment was implemented on affected area while general treatment was done when over 20% of the plot was affected [25]. Soil enrichment through fertilizer application was an option when there were high observations on no or low presence of new flushes as well as high number of immature large pods [31]. It is therefore important to maintain a high standard of farm management [36] so that the cocoa tree is less susceptible to parasites, disease, and insect attacks, as well as to ensure an appropriate response to specific outbreaks when they do occur [3] [10] [38].

Significant changes were noted between the management practices and across the clusters. The effect of insects on cocoa pods varied with altitudinal gradient and management practice. The number of pods damaged by insects (NPsDIs) was significant across clusters and of the different insect pests observed, only pods damaged by capsid (Sahlbergella singularis and Distanfiella theobroma) were statistically significant across clusters. This shows that capsid is the main pest causing destruction in cocoa agroecosystems in the study site. This is in harmony with the assertions of [36] [38] that the most important pest in cocoa agroecosystem is capsid.

The number of pods damaged by black pod disease (NPsDDs) was significantly different between management practices. This affirms the findings by many other researchers that black pod disease caused by Phytophthora megakarya and other Phytophthora sps, is the main disease affecting yields in cocoa agroecosystems [1] [2] [4]. A healthy crop is a more productive crop [9] [17] and the growth is more vigorous, yields are generally higher and the plant is better able to resist or compensate for pest/disease attacks [28] [31].

In the ICPM treatment, pods were harvested every two weeks if there were few ripe pods, and every week during peak periods. Likewise, it is important to do a separate round of the farm every week to remove sick pods with a cocoa hook that is used only for removing diseased material [16]. It is essential that the pods do not become over-ripe as they are more likely to become infected with diseases, and the beans inside over-ripe pods will germinate. [36] stipulates that evidence to date suggests that Ochratoxin “A” (OTA) producing organisms enter the cocoa supply chain via damaged pods. OTA is a naturally occurring foodborne mycotoxin found in a wide variety of agricultural commodities [16]. To reduce Ochratoxin “A” in the cocoa supply chain, it is recommended that farmers should not wound pods with a machete. Wounded pods of any kind should not be stored for any longer than one day. Every step in the process contributes to the final quantity and quality of the cocoa produce [8] [9] [12].

The cocoa harvested, fermented, and dried from the FP /treatments had some broken pieces while the dried beans from ICPM treatment were not adulterated. This was probably caused by the materials used. Though the plots were in similar locations, there was significant variation in agronomic parameters monitored, and production in the two management practices. The ICPM practice is in line with the model Ordinance of the International Cocoa Standards [16] [17] which provides that cocoa of merchant quality must be: “(a) Fermented, thoroughly dry, free from smoky beans, free from abnormal or foreign odor and free from any evidence of adulteration. (b) Reasonably uniform in size, reasonably free from broken beans, fragments and pieces of shell, and be virtually free from foreign matter”. Thus, this is similar to the results of [24] [40], which showed that on-farm demonstrations are vital for the adoption of sustainable agricultural practices. The results of this study showed that 80% of the farm owners were not acquainted with sustainable management practices, and could not implement adequate measures to ensure the production of quality cocoa beans [7] [8]. The cumulative production from the ICPM treatments was higher (572.8 kg) than that from FP treatment (382.83 kg) though none attained the required standard. The yields in both ICPM and FP practices were generally below the expected standard of 1000 kg giving a yield gap of 427.2 kg and 617.17 kg respectively [9]. This outcome was probably due to dominance of local varieties, availability of shade trees, age of fields with 2 plots of over 40 years, and appropriation of agrochemicals and fertilizers [9] [24].