Ant-mediated ecosystem services and disservices on marketable yield in cocoa agroforestry systems

Agriculture, Ecosystems & Environment Volume 247, 1 September 2017, Pages 409-417 Agriculture, Ecosystems & Environment Author links open overlay panelD.H.B.BisseleuaabDibierBegoudecHenriTonnangdS.Vidale a World Agroforestry Centre, UN Avenue Gigiri P.O. Box 30677–00100, Nairobi, Kenya b CGIAR Research Program on Integrated Systems for the Humid Tropics (Humidtropics), IITA, Ibadan, Nigeria c Institute of Agricultural Research for Development, Nkolbisson Yaoundé, Cameroon d International Maize and Wheat Improvement Center (CIMMYT) ICRAF House, United Nation, Avenue, Gigiri, P. O. Box 1041 Village Market, 00621, Nairobi, Kenya e Georg-August University Goettingen, Department of Crop Science, Entomological Section, Goettingen, Germany Received 14 December 2016, Revised 28 June 2017, Accepted 3 July 2017, Available online 20 July 2017. crossmark-logo https://doi.org/10.1016/j.agee.2017.07.004 Get rights and content Highlights • Ants provided ecosystem services (e.g. pest control) but also disservices (pathogen spread). • Positive impact of species-rich ant communities on marketable cocoa yield. • Strong link between species rich community performance and income. • Need to consider multiple interactions in agroecosystems than single ecosystem services. Abstract The impact of complex direct and indirect interactions between multiple functional groups on plants is poorly documented. In tropical agroecosystems, ants interact with crop mutualists and antagonists, however, little is known about effects of dominants ant community properties on the consequence of such cascading interactions which can be measured through final ecosystem service, crop yield. Here, we present a replicated ant fauna manipulation experiment in cocoa agroecosystems, where we used ant exclusion treatments to test the economic importance of the presence of ants, and two additional treatments where we experimentally introduced one of two common dominant ant species which allowed comparing their effects with those of the naturally occurring ant fauna. The proximate aim was to assess the impact of ants Crematogaster sp., Camponotus brutus Auguste-Henri Forel and Oecophylla longinoda Latreille on cacao yield, which is known to depend on several, cascading intermediate ecosystem services (e.g. control of specific pests). Ants provided ecosystem services in term of reduced pest damage caused by Salhbergella singularis Hagh (Hemiptera: Miridae) and Characoma stictigrapta Hmps (Lepidoptera: Noctuidae) but also disservices such as increased pathogen disease caused by Phyphthora megakarya Brasier and Griffin dissemination and indirectly enhanced damage of other pest species. Yields were highest in non-manipulated and species-rich ant communities, whereas ant exclusion and communities dominated by a single species decreased yield by more than 30%. Associated ant communities between dominant and non-dominant species resulted in the same yields as in non-manipulated controls. Dominant ant communities maximized the control of a particular pest species, but not cacao yield. We show a positive relationship between ant species rich communities and financial performance and we postulate that complex agroecosystems can offer competitive business opportunities for small-scale farmers, while contributing to biodiversity conservation. However, more interdisciplinary studies are needed to quantify financial and biodiversity performance opportunities to allow up-scaling of these findings. Previous article Next article Keywords Ant exclusion Indirect interactions Marketable yield Multitrophic interactions System research Trade-off analysis Tropical agroecosystems 1. Introduction In tropical agroecosystems, direct and indirect interactions between multiple functional groups such as herbivores (Bisseleua et al., 2013), pathogens (Evans, 2007), predators and pollinators (Klein et al., 2007) generate interactions between ecosystem services and disservices. The role of insect species in delivering individual ecosystem services such as herbivore pest and plant disease control is documented by only a few biodiversity-ecosystem studies (Bisseleua et al., 2009; Lundin et al., 2013; Wielgoss et al., 2014; Rusch et al., 2016). Some of these studies document the importance of synchronized interactions between many animal groups and different element of the ecosystem through trophic and non-trophic interactions with strong impact on intermediate ecosystem services and disservices. In cocoa production, the effects of ants on crop yield involve ecosystem services such as predation (Way and Khoo, 1992) and ecosystem disservices such as the dissemination of spores of plant pathogen (Evans, 2007) are reported. Little is known about effects of dominant ant community properties on the consequence of such cascading interactions, which can be measured through final yield. We expect ant communities with high richness and high evenness to be beneficial by maintaining services (Bihn et al., 2010) thus minimizing potential disservices associated with ant dominance (Hillebrand and Bennett, 2008). The fungal plant pathogen Phytopthora megakarya Brasier and Griffin is the main cause of cocoa (Theobroma cacao L.) pod losses in the West African cocoa belt (Nyasse et al., 2007). Dissemination by wind is often assumed to be the predominant long distance dispersal mechanism for Phytophthora spores (Nyasse et al., 2007) whereas rain splash transport is regarded the primary inoculum responsible for P. megakarya infections near the soil surface (Ndoumbè et al., 2004). Trajectory splash can disperse the spores from 1.5–2.0 m, while wind-blown rain droplets can disperse the spores higher into the tree canopy (Ndoumbè et al., 2004). Insects are not usually viewed as vectors of Phytophthora spores. However, transmission of Phytophthora spores has been reported for a number of insects such as Drosophila (Hunter and Buddenhagen, 1969), the weevil Scyphophorus interstitialis Gyll. and ants (Taylor and Griffin, 1981; McGregor and Moxon,1985; Medeiros et al., 1993). Ants, specifically those dominant in cocoa canopies, may be incidentally associated with the transmission of diseases, directly by transport of contaminated soil or of spores from infected to healthy pods, and indirectly through protection of Homoptera (Yede et al., 2012; Wielgloss et al., 2014). In cocoa agroforests of West Africa, arboreal ants, such as Crematogaster sp., Camponotus brutus and Oecophylla longinoda, are common and mutually co-existing (Padie and Owusu, 2003; Bisseleua et al., 2009 Bisseleua et al., 2009). Crematogaster species are the most abundant and colonize the upper and the lower parts of cocoa trees (Bisseleua, 2007). C. brutus is mainly found on trunks of cocoa trees between 50 and 100 m above ground, while O. longinoda is exclusively found in the canopy of cocoa trees (Bisseleua et al., 2013). Crematogaster species and O. longinoda are also known as important predators of insect pests of cocoa (Way and Kho, 1992; Padie and Owusu., 2003). However, they are also suspected to transmit black pod disease (Evans, 2007). Crematogaster species as the most abundant in cocoa agroforests could be regarded of major importance as vectors of black pod disease due to their higher numbers and densities on cocoa trees. The role of ants in intermediate ecosystem services, such as natural control of a complex of pest and plant disease species has been well documented (Way and Khoo, 1992; Philpott and Armbrecht, 2006; Otto et al., 2008; Zovi et al., 2008). However, their numbers and ecological status (e.g. dominance and invasion) are regulated by anthropogenic disturbance (Gibb and Hochuli, 2003). Dominant ant species will have several nests and will develop mutualistic interactions with hemipterans for honeydew as a sugar source in exchange of protection against potential predators (Bluethgen et al., 2004). Some will reduce species richness and evenness of ant communities by aggressively excluding other species from their territory and from food sources, while others will be more tolerant to conspecifics and do not reduce species richness (Majer, 1976, 1994; Philpott and Armbrecht, 2006). Predicting differences in the dissemination of spores of P. megakarya by ant communities with evenly distributed species abundances and those dominated by single-species has not been investigated in detail yet. Ant communities with low diversity but higher abundance within a tree (e.g. those dominated by an aggressive species) may contribute to a rapid dissemination of spores of P. megakarya as a consequence of a reduction in ant community traits and functional diversity (Bihn et al., 2010). On the contrary more diverse ant communities occupying different niche on a tree and thus reducing the foraging behavior within tree may be less effective in spreading the spores of P. megakarya owing to habitat restriction and buffer areas. In cocoa agroforestry systems, the importance of species richness for ecosystem functioning increased with the number of functions considered (Bisseleua et al., 2013; Asare and Raebild, 2016). For ants, this relationship is less direct because they use different resources, the exploitation of some of which may be an ecosystem service (predation of pests) (Bommarco et al., 2013; Chumacero de Schawe, 2014), others can cause a disservice (vector of plant pathogens such as that of P. megakarya (Nyasse et al., 2007; Nyasse et al., 2007; Ndoumbè et al., 2004) or tending of mealybugs vector of cocoa swollen shoot virus). Though, we expect ant communities with high species richness and high evenness to be beneficial by maintaining services (Wielgoss et al., 2014) and diluting potential disservices associated with ant dominance (Babacauh, 1982), because most ants are predatory to some extent, but only a subset of the species are likely to transmit spores of P. megakarya (Evans, 2007) or to engage in mutualisms with sap-sucking pests of economic importance (e.g. mealybug transmitting cocoa swollen shoot virus disease). Little is known to date about effects of dominant ants and ant community traits on the outcome of such set of interactions which can be measured through final yield (Vandermeer et al., 2010, Wielgoss et al., 2014). In cocoa plantations in the region sampled, Crematogaster sp. and O. longinoda species have recently become invasive and ecologically dominant. The former reduce ant species richness (Bisseleua, 2007) and can displace other dominant species, such as O. longinoda which is an effective predator of cocoa pests (Way and Kho, 1999). Both species form large colonies and can be numerically dominant, but they differ in their ecological traits. O. longinoda workers are more active in the cocoa canopy building their nests by folding cocoa leaves. They also aggressively expel other ants from baits (Padi and Owusu, 2003). Crematogaster sp. workers are active in the whole cocoa tree including the canopy and are relatively less tolerant towards other ant species such as C. brutus. They build their nests using soil materials. C. brutus workers are active on the trunk of cocoa trees, are never found in the canopy and are relatively tolerant towards other ant species such as the Crematogaster sp. or O. longinoda. Their tent material is of plant debris-type. In this study we manipulated the ant fauna in smallholder cocoa agroforests in Cameroon, using ant exclusion treatments, to test the role of ant in the transmission of black pod disease and the related impact on cocoa productivity and marketable yield. We hypothesized that ant exclusion negatively affects cocoa productivity and marketable yield. We expected that dominant species maximize individual intermediate ecosystem disservices and the final integrated ecosystem services (measured as marketable yield), whereas species-rich ant communities with high evenness should result in the highest marketable yield, due to high functional diversity which maintains major ecosystem services while buffering potential disservices of single species. We analyze the effects of ant community structure on the incidence of P. megakarya and how different ant communities affect intermediate ecosystem services and disservices and impact marketable yield. 2. Material and methods 2.1. Study area and study plots All sites were located in the central region of Cameroon located between 4°12′ and 4°30′ latitude north, and 10°6′ and 11°15′ longitude east. The altitude varies between 450 and 715 m above sea level. We selected 20 cocoa plots (30 × 30 m) without insecticide application within the last two years and differing in shade intensity and with proofed absence of the ant species Crematogaster sp., Camponotus brutus and O. longinoda. The plots were located in Boumnyebel (03°53′01″N 10°50′56″E, 402 m alt.), where cocoa is grown under a dense cover of many forest near pristine forests with very old cocoa plantation (> 30 years); Obala (04° 15′82″N 11° 53′62″E, 557 m alt.) where cocoa is grown with a diversity of forests and fruit trees species, with no original remnant forests because of very high human population density; Talba (04°34′421″ N 11°28′33″ E, 462 m alt.) where cocoa is grown in larger farms under less shade; Bakoa (4°°56′42″ N 11°16′47″E, 469 m alt.), where cocoa is grown under very low shade in forest galleries at the forest-savannah transition zone and Kedia (4°°50′.46″N, 11°07′87″E, 459 m alt.) where cocoa is grown under full sun in the savannah. The geographic coordinates of the sites were taken using a GPS (GPSMAP 60CSx). In each plot we established six sub-plots (10 × 10m) with a minimum distance of 10 m to each other and containing ten neighboring cocoa trees. Annual rainfall varied from 1000 mm to 1627 mm. Diurnal temperature (°C), relative humidity (%) and daylight intensity (Lux) were measured during standardized conditions (sunny days, 8–10 am). A combined Electronic hand-held hygro-thermometer (TECPEL CO LTD, Taiwan: Model DTM 321_ DTM 322) was used to measure temperature and relative humidity. Canopy cover was measured at 10 points per site using a hand-held spherical densitometer © (R.E. Lemmon Forest Densiometers, USA). Spatial distribution of cocoa trees was evaluated by measuring the distance between two consecutive trees. We have designed our experimental approaches to reduce confounding effects of environmental variables or land-use practices on ant community structure. 2.2. Ant treatments We evaluated four distinctive ant communities: (i) non-manipulated ant communities with a high evenness and species richness. (ii) ant communities dominated by Crematogaster species with high abundance and evenness; (iii) ant communities dominated by the aggressive species O. longinoda with high abundances, but low species density/evenness; (iv) Exclusion of all ants (with extremely low abundances and low species density/evenness). The ant fauna manipulation treatments were assigned randomly to six subplots: (i) undisturbed naturally occurring ant fauna as controls; (ii) cocoa trees with only Crematogaster sp. (iii) cocoa tree with only O. longinoda specimens; (iv) Crematogaster sp. in association with C. brutus; (v) O. longinoda in association with C. brutus; (vi) the three species together as ecologically dominant ant species on the test trees. Treatments were applied on cocoa trees cultivars (SNK 16 and T79/467) susceptible to P. megakarya infections. 2.3. Data collection During 18 months, from January 2012 to July 2013, and biweekly we monitored ant communities, flowers, cherelles (young pods), pods, incidence of black pod disease and yield on all test trees. We recorded number of flowers, cocoa pods (classified by size) and disease incidence. Ripe pods were harvested, beans were pooled from the pods of one subplot and dry weight of marketable beans was recorded. To account for yield, we separated defective beans and weighed them separately from marketable dry beans. Within each plot we recorded cocoa phenology, growth parameters and wilting. Wilting generally happens at an early development stage of cocoa trees and may be caused by environmental factors or/in combination with pests and diseases incidences. We also recorded temperature and shade cover to correct for confounding effects in different subplots. We surveyed and identified ants to morphospecies one time before and three times after treatment installation in each subplot using standardized tuna and sugar baits on ten cocoa trees and at four ground locations per subplot. Ants and other insects were later removed from sampled trees using nest destruction and insect glue barriers on the stem base. The treatments were installed in January 2012 to July 2013. Treatments were maintained stable in all 10 trees per subplot during that time. 2.4. Collection of ant tents and laboratory inoculation Ant tents of Crematogaster sp. and C. brutus were individually collected in sterilized glass vials to evaluate in the laboratory their potential to passively vector spores of P. megakarya. We use collected tents from Crematogaster sp. and C. brutus to inoculate 10 pods for each ant species. Prior to inoculation, green pods were superficially cleaned with alcohol. One circular well of about 5 mm depth was cut with a 10 mm cork borer in the surface sterilized husk of each pod. Soil and plant debris from individual tents was placed in the wells and moistened with distilled water. The wells were sealed with a humid cotton wad to maintain humidity, and pods were placed individually in sealed sterilized plastic containers. The incubation was conducted at ambient temperature (25° to 27° C) with 10 similarly wounded but untreated pods serving as controls. After 5 days, pods were examined for black spots, diagnostic of black pod disease in the vicinity of the wound. Tissues were removed using a sterilized needle from all pods exhibiting symptoms and plated on a standard V8 growth medium in 90 mm diameter Petri dishes to confirm the presence of P. megakarya. 2.5. Field inoculation experiment For the field inoculation experiment we used the “Boyo strain” of P. megakarya, available in the fungal collections of the Laboratory of Phytopathology of the Agricultural Research Institute of Cameroon. The strain was transferred to a standard V8 growth medium in 90 mm diameter Petri dishes for growth. The white mycelium was recovered after ten days and the virulence of the fungus was confirmed by inoculating a healthy cocoa pod with a piece of PDA cut out from the V8 medium preparation. The inoculated pods were incubated in darkness at ambient temperature and saturated moist environment. The inoculation was considered successful when a brown spot of more or less circular form, spreading from the point of inoculation, was observed. After 5 days, the fungus was isolated from the infected pod by collecting a piece of tissue below the brown spot and transferred to an artificial medium for mass production. Seven day old cultures, from mass production, were used for artificial inoculation of pods in the field. Within each subplot, we inoculated the first two pods at 50 cm from the ground. On each pod one circular well of about 5 mm depth was cut with a 10 mm cork borer in the surface sterilized husks. The wells were covered with a piece of PDA (containing the fungus from the above preparation) and sealed with a humid cotton wad soaked in distilled water. Observations were first based on the success of inoculation and later on biweekly observation on the spreading rate of the disease. Ten pod bearing trees without an artificial inoculation were used as controls per subplots. Observations were terminated at pod harvest. 2.6. Data analysis To correct for potential confounding variables, we fitted linear mixed-effects models with canopy cover, cocoa tree height, diameter at breast height and crown volume as dependent variables, treatment as an explanatory variable and plot as a random effect. We also fitted the linear mixed-effect models for differences in ant community structure using ant abundances at the baits and ant species richness per subplot with plot and survey as random effects. We conducted a Tukey’s contrast test for multiple comparisons of means. To also predict ant treatment effects on the incidence of the main pest Salhbergella singularis Hagh (Hemiptera: Miridae) and Characoma stictigrapta Hmps (Lepidoptera: Noctuidae) we fitted binomial generalized linear mixed model to the amount of infected harvested pods versus the number of healthy harvested pods per subplot and harvesting time. We also fitted a binomial GLMM to the total amount of pods lost to P. megakarya during the experiment time versus the total number of harvested non-infected pods per subplot with plot as a random effect. We aggregated open flower and cherelle (young pods) data from April 2012 to February 2013 and we fitted binomial GLMM to the amount of small pods and the number of open flowers per subplot with plot as random effect. We also fitted binomial GLMMs to the number of aborted pods and the number of healthy pods per subplot to assess differences in the rates of early pod abortion. We aggregated flower and pod data of all observations and the ten trees per subplot and used linear mixed-effects models to test for ant treatment effects on the number of flowers, total number of young cocoa pods, number of pods that survived the early pod abortion and number of harvested pods. We used the first 12 months of data to have estimates for one complete harvesting season for dry yield and total marketable dry cocoa beans. Considering the economic and ecological importance of cocoa, we assessed the relationship between biodiversity performance and financial performance by looking at the effects of ants on marketable yield and revenue from dry cocoa beans. We used data on biodiversity performance in terms of species richness, but also proxy variables known for their correlation with biodiversity, such as shade cover. For financial performance, we used indicators such as productivity (based on average yield of 25 pods per tree or 5 kg of marketable beans per tree) and costs (labour and inputs). Extrapolating the ant community effects on marketable yield observed in our experiment is legitimate since Crematogaster sp. and O. longinoda, when present in a cocoa plantation, can dominate more than 80% of the trees with similar high abundance as in our experimental subplots. We used the mean cocoa world market price (January 2016 was taken as reference to calculate production costs. 1 ha was used as the unit for surface area, 1 year as unit for time, and US dollar as currency) of 2.5 US$ kg−1 and calculated total harvest values per year and hectare using the first 12 months of yield data (1 ha was used as the unit for surface area and 1 year as unit for time) and assuming 1200 trees per hectare (2.0 m planting distance) for each ant community treatment. 3. Results 3.1. Cocoa tree characteristics Cocoa tree height, stem diameter at breast height, crown volume and shade cover did not significantly vary among treatments in each plot (p = 0.67). 3.2. Ant communities On ant exclusion trees, mean ant abundances were reduced to less than 13% compared to the control (t = 2.29; p < 0.05). Ant abundances in subplots with experimentally established single species dominances were lower for O. longinoda than in control subplots but this was not observed for Crematogaster sp. (Fig. 1a, O. longinoda: t = 2.91, p < 0.005; Crematogaster sp.: t = 0.41, p = 0.68). Species richness in O. longinoda treatments was significantly higher than in control subplots (Fig. 1b, t = 3.1; p < 0.005), but significantly lower in the Crematogaster sp. treatments (Fig. 1b, t = 18.1; p < 0.0001). Species richness was 43% lower in the other three treatments (Fig. 1b, O. longinoda + C. brutus: t = 22.6; p < 0.0001, Crematogaster sp. + C. brutus: t = 21.6; p < 0.0001 and ant exclusion: t = 20.4; p < 0.0001). The evenness in Crematogaster sp. treatments was higher than the undisturbed naturally occurring ant communities (Fig. 1c, t = −7.10; p < 0.0001), while presence of O. longinoda or C. brutus reduced evenness (O. longinoda: t = −5.22; p < 0.000; Crematogaster sp. + C. brutus: t = −4.8; p < 0.0001; O. longinoda + C. brutus: t = −5.6; p < 0.0001). Download high-res image (250KB)Download full-size image Fig. 1. (A) Abundance of ant workers at tuna and sugar baits (no significant differences between treatments), (B) ant species richness (F5,23 = 248.5 p < 0.0001) and (C) Pielou’s evenness (F5,23 = 15.8 p < 0.0001) (four repeats; mean of the ten trees per subplot). 3.3. Laboratory studies None of the 20 replicates resulted in positive isolations of P. megakarya from the test pods under ideal incubation conditions indicating a rather marginal contribution of ant tents from Crematogaster sp. and C. brutus to retain viable inoculum. 3.4. Effects on young pod (cherelles) and number of pods No differences were found with regard to the number of pods between the treatments (F5,23 = 0.87; p = 0.52). A similar number of young pod was recorded in undisturbed naturally occurring ant communities compared to O. longinoda treatments (Fig. 2), whereas in the Crematogaster sp. treatments young pod set was significantly higher (F5,23 = 7.80; p < 0.0001). Young pod set positively correlated with ant abundances (F5,23 = 16.74; R2 = 0.42; p < 0.0001). The number of young pods (cherelles) was similar in treatments dominated by O. longinoda. Download high-res image (159KB)Download full-size image Fig. 2. Mean (+/−SE) number of young pod (cherelles), pod aborting, total number of pods produced and total healthy pods harvested in the different ant treatments experiment in cocoa agroforest (Graph from left to right: ant exclusion (black), unmanipulated control (white), dominance of Crematogaster sp. (stripe), association of Crematogaster sp. with C. brutus (grey), association of O. longinoda with C. brutus (white spotted) and dominance of O. longinoda) (coarse). 3.5. Effects on pod wilting (abortion) More than 520 young pods aborted in Crematogaster sp. subplots, 312 in ant exclusion subplots and only 11 in control unmanipulated ant subplots. The number of aborted young pods was highest in Crematogaster sp. subplots, followed by ant exclusion subplots (F5,23 = 6.3; p < 0.001), whereas O. longinoda-dominated ant communities and ant exclusion subplots recorded the lowest abortion rates (Fig. 2). 3.6. Indirect ant community effects on pod losses to insect pests The total pods lost due to the sap-sucking mirid S. singularis was highest in ant exclusion subplots, followed by O. longinoda-C. brutus subplots, whereas Crematogaster sp. dominated subplots resulted in the lowest number of pods damaged by S. singularis (Fig. 3, F5,23 = 7.3; p < 0.001). The number of aborted pods was not correlated with pod damage by S. singularis (F2,23 = 0.17; R = 0.08; P = 0.68). In comparison with control treatments, pod damage by S. singularis was significantly higher (F5,23 = 7.35, p < 0.001) in O. longinoda dominated trees (Fig. 3). Download high-res image (199KB)Download full-size image Fig. 3. Total pods lost due to the pod rot disease P. megakarya and by insect pests S. singularis or C. stictigraphta in ant manipulation experiment in cocoa agroforests. (Refer to previous figure legend). The number of pod damage by the pod borer C. stictigrapta was highest in ant exclusion treatments and in O. longinoda dominated trees (Fig. 3, F = 39.6; n = 5; p < 0.0001). The number of aborted pods was not correlated with pod damage by the pod borer C. stictigrapta (F2,23 = 0.51; R = 0.15; P = 0.48). Crematogaster sp. dominated ant communities and association of Crematogaster sp. with C. brutus reduced pod damage by C. stictigrapta compared to control subplots, while in O. longinoda subplots there was no difference in pod damage by C. stictigrapta. In comparison with control treatments, pod damage by C. stictigrapta was not different in O. longinoda dominated trees (Fig. 3). The number of harvested pods affected by C. stictigrapta was similar in the control and Crematogaster sp. treatments, whereas for S. singularis this number was similar for all treatments. Infestation rates were higher than in control treatments for ant exclusion treatments and trees dominated by O. longinoda. In Crematogaster sp. dominated trees infestation rates were reduced. S. singularis or C. stictigrapta damage rates were not correlated with ant community evenness. 3.7. Indirect ant community effects on pod losses to the pod rot disease P. megakarya In the O. longinoda subplots, a significantly higher number of pods were lost due to P. megakarya compared to all other treatments (Fig. 3). Ant exclusion resulted in significantly less infections on pods compared to the control (t = 9.5; p < 0.0001). The Crematogaster dominated trees reduced pod damage by P. megakarya similar to the control non-manipulated ant communities. The number of aborted pods was not correlated with pod damage by P. megakarya (F2,23 = 0.18; R = 0.09; P = 0.68). 3.8. Harvested pods In ant unmanipulated control and treatments where C. brutus is associated, the number of harvested pods was similar, while in O. longinoda and Crematogaster about 51% and 31% fewer pods were harvested respectively (Fig. 2). Ant exclusion resulted in a significantly fewer number of harvested pods as compared to non-manipulated controls and the association of Crematogaster sp. with C. brutus or the association of O. longinoda with C. brutus (Fig. 2). 3.9. Effects on total dry beans quality and quantity Total dry beans (Kg/ha) was significantly higher in treatments where O. longinoda was associated with C. brutus and where O. longinoda was dominant compared to the other treatments (F5,23 = 21.73; p < 0.00001). Total dry beans (kg/ha) was about 17% lower in Crematogaster sp. treatments and 20% lower in ant exclusion treatments than in the control treatments (Fig. 4). Download high-res image (321KB)Download full-size image Fig. 4. Total dry beans and marketable dry beans in kg per hectare per year and revenue generated in $US in the different ant treatments in cocoa agroforests. (Refer to previous figure legend). 3.10. Effects on marketable yield and revenue Total marketable yield was 38% and thus significantly lower in O. longinoda treatments (F5,23 = 8.7; p < 0.0001), which was equivalent to a loss of 1905 US$ ha−1yr−1 (t = 8.9, p < 0.0001), 34% (1721 US$ ha−1yr−1) lower in ant exclusion treatments and 31% (1548 US$ ha−1yr−1) lower in Crematogaster sp. treatments (F5,23 = 21.8; p < 0.00001) compared with controls (Fig. 4). Yield in treatments where C. brutus was associated Crematogaster sp. and in control plots did not significantly differ. Marketable yield tended to decrease with increasing evenness of the associated ant communities but not significantly so. 3.11. Financial performance The average net revenue ($/ha) of association of O. longinoda with C. brutus was significantly higher than that of treatments with only O. longinoda or ant exclusion (F5,23 = 7.67; p < 0.0001). Additionally, the average net return of diverse ant communities occupying different niches on a tree was 66% significantly higher (t = 14.63; p < 0.00001), showing a higher profit per hectare for farmers having a diverse ant communities on their cocoa trees. 4. Discussion We were able to show that ants, although generally regarded as beneficial in agricultural crops, may contribute to reduce crop yields provided a low diversity and high abundance of dominant species on crop plants is maintained. However, in species-rich and even ant communities ants contribute to crop productivity by driving a complex network of direct and indirect interactions between the crop and its associated pests and pathogens. In such complex interactions, ant exclusion results in significantly lower yields. We demonstrate that ants are important for the maintenance and functioning of ecosystem services and disservices depending on the structure and composition of their community. Ant-mediated disservices are evident when traits of dominant species, such as high abundances and spatial activity patterns (canopy versus trunk foraging behavior) result in uneven and species poor ant communities which are not effective in protecting the cacao crop against some pests. We also unambiguously demonstrate that ant communities differ in their role as vectors of fungal spores. The more ant species inhabit a tree, the lower their spore dissemination capacity is, suggesting that niche restriction through selective increase of ant richness will negatively affect the dissemination of fungal spores. Ant-mediated services in terms of higher marketable yield and revenue generated was recorded in subplots with naturally occurring non-manipulated, species-rich and even ant communities; and provided an additional 1721 US$ ha−1yr−1 to the farmers compared with the ant exclusion treatment, where 34% less marketable yield was harvested. Ant-mediated services in ant communities dominated by single ant species were significantly influenced by the identity of the dominant species. For instance, we recorded yield loss of 38% (-1905 US$ ha−1yr−1) in O. longinoda dominant trees with relatively low evenness, and a yield loss of 31% (-1548 US$ ha−1yr−1) in Crematogaster treatments with relatively high ant community evenness. However, we noted that differences in marketable yield can also be due to the complex effects of ant community structure and dominant species traits on young pod set, pod abortion and interactions among herbivores such as S. singularis and C. stictigrapta. Our experiments demonstrate the importance of ant richness and evenness as determinants of conservation biological control and productivity (Tscharntke et al., 2007; VanMele, 2008). Although the impact of species richness in species-rich, natural ecosystems is poorly understood, we hypothesized that stability and productivity (Tilman et al., 2001) of agricultural systems with low ant species richness may be influenced by two factors; increasing species richness or reducing environmental heterogeneity (Perfecto et al., 2004, Tscharntke et al., 2012). Increasing species richness will have a positive effect on productivity (Tilman et al., 1996; Hector et al., 1999). All the same, species evenness should deserve further attention because evenness is more affected by anthropogenic disturbances than richness, specifically in complex systems with multiple intermediate ecosystem services. This systemic description of multitrophic interactions and combination of species between ants and other organisms was also reported in cocoa agroforest in Indonesia (Wielgoss et al., 2014) and in coffee agroforests (Vendermeer et al., 2010). We have observed that the distribution of niches and niche overlap of one ant species is affected by the presence or absence of another, portraying the importance of each ant species within the system. Thus, the functional redundancy (Gitay et al., 1996) of species in complex systems should not be interpreted as a justification for their elimination. The effect of removal of one species cannot be predicted and in the absence of other species and loss of interactions any dominant species may become a potential vector of fungi spores. Changes in ant community composition may expose complex systems such as cocoa agroforests to vulnerability and invasion. Invasive ants will have specific effects on plants, based on their precise traits, the system and the season (Lach, 2003). It is therefore possible to manipulate the distribution of ant communities by influencing the nature of the habitats so as to balance species density. Our study suggests that in systems with diverse ant communities, such as cocoa agroforests, species-rich and even communities may provide an effective protection to crops (Brittain et al., 2013, Fründ et al., 2013). 4.1. Ant community species traits The ant exclusion treatments, by reducing worker abundances by more than 87%, were comparable to the reduction rates of other ant exclusion experiments (Klimes et al., 2011). The experimental manipulations contributed to an ecological and numerical dominance of Crematogaster sp. and O. longinoda with 53% (Crematogaster sp.) and 19% (O. longinoda) of the workers recruited at baits despite similar total worker abundances of both species. Crematogaster sp and O. longinoda are less tolerant towards other ant species such as C. brutus. Therefore, presence of these species resulted in ant communities with relatively lower species richness and higher evenness than the unmanipulated control ant communities, but Crematogaster sp supported much higher total worker abundance. Nest of Crematogaster sp. and O. longinoda were consistently found in the upper parts of the trees (i.e. above 150 cm) on big branches and on pods for Crematogaster sp. and small branches including leaves for O. longinoda, if nesting alone or together with other ant species such as C. brutus. C. brutus was generally found on the mid-trunk (below 100 cm) active mainly on pods and pod peduncles and usually tending mealybugs or coccids. 4.2. Ant community effects on pod set Young pod set was positively correlated to ant abundance and lowest in control, intermediate in the ant exclusion and O. longinoda treatments and highest in Crematogaster dominated trees. We were not able to provide a clear explanation of the lower number of pod set in the control subplots. Although ants are frequently present near flowers, tending mealybugs, there is no convincing evidence that ants can pollinate cocoa flowers (Glendinning, 1972) or may feed on cocoa flowers. However, ant may disturb pollination by other beneficial. Their active presence may cause pollinators to spend less time on flowers, and thereby reducing pollen transfer, pollen load and number of pollen donors. This may probably reduce pollination success and pod set as observed in non-manipulated control species-rich ant communities. Other studies rather suggested nuisance by ants to enhance pollination success (Wielgoss et al., 2014). 4.3. Indirect ant community effects on pod wilting The ant communities with dominant species enabled higher numbers of early pod abortion than control communities. In the O. longinoda and C. brutus associated communities, we found lower rates of early pod abortion. This can be explained by the preferred worker activity of O. longinoda on branches and leaves than on pods. The reduced number of pods abortion in Crematogaster sp. − C. brutus treatments and controls can be explained by the intensive competitive exclusion activity between the two species around pod peduncles for mealybugs or coccids as well as complex effects of ant-crop-herbivore interactions. Less competition at pod peduncles for mealybugs in Crematogaster sp. treatments could explain the higher number of pod abortion. The percentage of aborted young pods was reported to increase with the amount of mealybug aggregations in cocoa agroforests (Wielgoss et al., 2014). 4.4. Ant community effects on pod losses to the pod rot disease P. megakarya In O. longinoda dominated treatments, about 46% of pod were lost due the pod rot P. megakarya. O. longinoda may tolerate nest building species with plant debris-type such as C. brutus below the canopy (i.e. cocoa stem or trunk) or with soil materials such as Crematogaster sp. Inoculation tests using soils from Crematogaster tents or plant debris from C. brutus failed to initiate any infection when tested in the laboratory on detached cocoa pods. This suggests that ant tents might not be such an important source for pod infections because they are generally made of dry plant particles or arthropod debris. Nevertheless, studies by Wielgoss et al. (2014) in Indonesia, Evans (1973), Taylor and Griffin (1981) in Nigeria; Babacauh (1982) in Côte d’Ivoire incriminated soil and plant debris from ant tents to vector pod rot disease. However, O. longinoda generally is reported as a good biological control agent in cocoa agroforest (vanMele 2008) by tolerating other ant species on the trunk of cocoa trees with frequent contact with contaminated soil material. These species may be regarded as an important vector of P. megakarya spores. Insect vectors of Phytophthora have been reported by several authors (Taylor and Griffin, 1981; Mcgregor and Moxon, 1985; Medeiros et al., 1995) but here we show that the transmission efficiency is highly associated to the distribution of niches and specific behavioral traits of each ant species. 4.5. Ant community effects on total dry beans quality and quantity O. longinoda dominated ant communities were associated with high damage by the major cocoa pest and pod rot diseases (Fig. 3), which significantly reduced the quality and quantity of marketable beans (Fig. 4). O. longinoda may not be efficient in preying upon adult Characoma or disturbing them in their hiding sites in the foliage of cocoa trees. The characteristics of Crematogaster dominated ant communities with higher workers abundance activity in the whole cocoa tree including the canopy, combined with higher evenness enable them to subdue the cocoa pod borer Characoma more effectively than O. longinoda ant communities (Fig. 3). We also noted clear discrimination of S. singularis against cocoa pods infested by C. stictigrapta. Such avoidance may be mediated by vegetation heterogeneity and functional diversity of ant communities. Therefore, efficient ant communities in preventing damages by S. singularis such as those dominated by Crematogaster sp. may indirectly facilitate damage by C. stictigrapta. 4.6. Ant community mediated ecosystem services and disservices We show that high abundances and spatial activity patterns of dominant ant species such as Crematogaster sp. enhanced pest control of S. singularis and C. stictigrapta but did not result in increased yield. The specific traits of the dominant Crematogaster sp. species increased disservices in terms of active dissemination of the cocoa pod rot disease or indirectly facilitated damage by the cocoa pod borer C. stictigrapta, resulting in serious yield losses. On the other hand, O. longinoda dominated ant communities, due to their relative tolerance to other ant species such as C. brutus showed high species richness, high evenness and good services in reducing early pod abortion and pod rot disease by P. megakarya. These services were offset by disservices such as increased damages by the by the cocoa mirids S. singularis and cocoa pod borer C. stictigrapta and also resulted in serious yield losses (compared with the control; Fig. 4). Our study suggests that in systems with diverse ant communities, such as cocoa agroforests, species-rich and −even communities may enhance conservation biological control of pests and plant diseases as a result of positive functional diversity (Tscharntke et al., 2007; VanMele, 2008). However, we were not able to clearly separate evenness effects from effects like dominance and/or aggressivity which alone or in combination contributed to changes in community properties such as low evenness, higher richness or abundances impacting yield. Therefore introducing a single dominant species as a biocontrol agent may enhance disservices and may not result in yield increase compared with communities with more evenly distributed species. Intensive insecticide applications as currently predominantly practiced in the study region may negatively affect evenness and richness of ant communities and may thus be detrimental to species such as O. longinoda with foliar nest-building behavior compared to Crematogaster sp. with special tent-building behavior. 4.7. Ant community and financial performance Although there are only few studies directly linking species rich community performance and financial performance, there are indications that species rich agro-ecosystems potentially combine increases in both types of performance. Clough et al. (2011) did not find a negative relationship between species diversity and income in Indonesian smallholders cocoa agroforestry systems concluding that such system is able to combine high agricultural yield and high biodiversity goals on-farm. Wielgoss et al. (2014) in an ant exclusion experiment on cocoa systems in Indonesia found the highest yield to concur with high level of ant richness, indicating opportunities for increased income with a higher performance on biodiversity; this was also shown by the study of Bisseleua et al. (2009). However, further research is needed that focuses on the trade-offs between economic and biodiversity performance. 5. Conclusion We were able to show that ants, although generally regarded as beneficial in agricultural crops, may contribute to reduce crop yields in agricultural systems with a low diversity and high abundance of a dominant species. We used the final ecosystem services (i.e. crop yield) to evaluate the impact of multitrophic interactions between ant communities of Crematogaster sp., C. brutus and O. longinoda differing in specific traits and structure and showed a negative impact of ant exclusion or the positive impact of species-rich ant communities on yield. We found that species-rich communities provided the best services for agriculture leading to highest marketable yield. The more ant species there are on a tree, the lower the disservices they can provide and the higher the services they can generate in maximizing agricultural yield. Dominant ant communities maximized the control of a particular pest species, but not agricultural yield. Our results suggest that multiple interactions between predators and ecosystems (i.e. the crop and non-crop plant species, plant pathogens and associated organisms) need to be taken into account instead of individual interactions or single ecosystem services. Although case and site-specific conditions need to be considered, species rich agroecosystems offer great potential to reconcile biodiversity conservation and local development. However, further assessments of these relationships are necessary through long-term interdisciplinary studies, to quantify both financial and biodiversity performance in order to gain insight into the opportunities and challenges for upscaling of the outcomes. On a practical note more emphasis should be laid on systemic enquiries (Ison, 2012) about critical final services such as marketable yields. Acknowledgments This study is part of the This work is part of the fellowship project (VW-60420894) of BDHB funded by the Volkswagen Foundation under the shame Funding Initiative Knowledge for Tomorrow − Cooperative Research Projects in Sub-Saharan on Resources, their Dynamics, and Sustainability − Capacity Development in Comparative and Integrated Approaches. We thank the Volkswagen foundation for the financial assistance provided to the first author for field work and for various analyses. We are particularly grateful to the staff of the Institut de Recherche Agricole pour le Development (IRAD) for their support and hospitality. We also thank the Cameroonian cocoa farmers who made their plantations available to us and for their help and support during field work. References Asare and Raebild, 2016 A. Asare, A. Raebild Tree diversity and canopy cover in cocoa systems in Ghana New Forests, 47 (2016), pp. 287-302, 10.1007/s11056-015-9515-3 Babacauh, 1982 K.D. Babacauh Rôle des communautés d’insectes et de l’eau dans la dissémination de Phytophthora palmivora (Butl.) Bult. Emend. Bras. Et Griff. Dans les cacaoyères de la Côte d’Ivoire Café Cacao Thé, 1 (1982), pp. 31-36 View Record in Scopus Bihn et al., 2010 J.H. Bihn, G. Gebauer, R. Brandl Loss of functional diversity of ant assemblages in secondary tropical forests Ecology, 91 (2010), pp. 782-792, 10.1890/08-1276.1) CrossRefView Record in Scopus Bisseleua et al., 2009 D.H.B. Bisseleua, A.D. Missoup, S. Vidal Biodiversity conservation, ecosystem functioning, and economic incentives under cocoa agroforestry intensification Conserv. Biol., 23 (2009), pp. 1176-1184, 10.1111/j.1523-1739.2009.01220. x CrossRefView Record in Scopus Bisseleua et al., 2013 D.H.B. Bisseleua, D. Fotio, Y. Yede, A.D. Missoup, S. Vidal Shade tree diversity, cocoa pest damage, yield compensating inputs and farmers’ net returns in West Africa PLoS One, 8 (3) (2013), p. e56115, 10.1371/journal.pone.0056115 Bisseleua, 2007 D.H.B. Bisseleua Ecological, Social and Economic Determinants in Cocoa Production Systems in Southern Cameroon PhD Thesis Georg-August University of Goettigen (2007) (http://webdoc.sub.gwdg.de/diss/2008/bisseleua_daghela/) Bluethgen et al., 2004 N. Bluethgen, N.E. Stork, K. Fiedler Bottom-up control and co-occurrence in complex communities: honeydew and nectar determine a rainforest ant mosaic Oikos, 106 (2004), pp. 344-358, 10.1111/j.0030-1299.2004.12687. x Bommarco et al., 2013 R. Bommarco, D. Kleijn, S.G. Potts Ecological intensification: harnessing ecosystem services for food security Trends Ecol. Evol. (2013) Brittain et al., 2013 C. Brittain, N. Williams, C. Kremen, A.-M. Klein Synergistic effects of non-Apis bees and honey bees for pollination services Proc. R. Soc. B., 280 (2013), p. 20122767 (1098/rspb.2012.2767) CrossRefView Record in Scopus Chumacero de Schawe, 2014 C. Chumacero de Schawe Pollination ecology and spatial genetic population structure of wild and cultivated species of cacao (theobroma) in Bolivia Doctoral dissertation (2014) http://hdl.handle.net/11858/00-1735-0000-0022-5E65-D Clough et al., 2011 Y. Clough, J. Barkmann, J. Juhrbandt, M. Kessler, T.C. Wanger, A. Anshary, D. Buchori, D. Cicuzza, K. Darras, D.D. Putra, S. Erasmi, R. Pitopang, C. Schmidt, C.H. Schulze, D. Seidel, I. Steffan-Dewenter, K. Stenchly, S. Vidal, M. Weist, A.C. Wielgoss, T. Tscharntke Combining high biodiversity with high yields in tropical agroforests Proc. Natl. Acad. Sci. U. S. A., 108 (2011), pp. 8311-8316, 10.1073/pnas.1016799108 CrossRefView Record in Scopus Evans, 1973 H.C. Evans Invertebrate vectors of Phytophthora palmivora, causing black pod disease of cocoa in Ghana Ann. Appl. Biol., 75 (1973), pp. 331-345 CrossRefView Record in Scopus Evans, 2007 H.C. Evans Cacao diseases: the trilogy revisited Phytopath, 97 (2007), pp. 1640-1643, 10.1094/PHYTO-97-12-1640 CrossRefView Record in Scopus Gibb and Hochuli, 2003 H. Gibb, D.F. Hochuli Colonisation by a dominant ant facilitated by anthropogenic disturbance: effects on ant assemblage composition, biomass and resource use Oikos, 103 (2003), pp. 469-478 CrossRefView Record in Scopus Gitay et al., 1996 H. Gitay, J.B. Wilson, W.G. Lee Species redundancy: a redundant concept? J. Ecol., 4 (1996), pp. 121-124 CrossRefView Record in Scopus Glendinning, 1972 D. Glendinning Natural pollination of cocoa New Phytol., 71 (1972), pp. 719-729 CrossRefView Record in Scopus Hector et al., 1999 A. Hector, B. Schmid, C. Beierkuhnlein, C. Caldeira, M. Diemer, P.G. Dimitrakopoulos, J.A. Finn, H. Freitas, P.S. Giller, J. Good, R. Harris, P. Högberg, K. Huss-Danell, J. Joshi, A. Jumpponen, C. Körner, P.W. Leadley, M. Loreau, A. Minns, C.P. Mulder, G. O'Donovan, S.J. Otway, J.S. Pereira, A. Prinz, D.J. Read, M. Scherer-Lorenzen, E.-D. Schulze, A.D. Siamantziouras, E.M. Spehn, A.C. Terry, A.Y. Troumbis, F.I. Woodward, S. Yachi, J. Lawton Plant diversity and productivity in European grasslands Science, 286 (1999), pp. 1123-1127 CrossRefView Record in Scopus Hillebrand and Bennett, 2008 H. Hillebrand, D. Bennett Consequences of dominance: a review of evenness effects on local and regional ecosystem processes Ecology, 89 (2008), pp. 1510-1520, 10.1890/07-1053.1 CrossRefView Record in Scopus Ison, 2012 R.L. Ison Systems practice: making the systems in farming systems research effective I. Darnhofer, D. Gibbon, D. Dedieu (Eds.), Farming Systems Research into the 21 St Century: The New Dynamic., Springer in press, Dordrecht (2012) Klein et al., 2007 A.-M. Klein, B.E. Vaissière, J.H. Cane, I. Steffan-Dewenter, S.A. Cunningham, C. Kremen, T. Tscharntke Importance of pollinators in changing landscapes for world crops Proc. R. Soc. B, 274 (2007), pp. 303-313, 10.1098/rspb.2006.3721 CrossRefView Record in Scopus Klimes et al., 2011 P. Klimes, M. Janda, S. Ibalim, J. Kua, V. Novotny Experimental suppression of ants foraging on rainforest vegetation in New Guinea: testing methods for a whole-forest manipulation of insect communities Ecol. Entomol., 36 (2011), pp. 94-103, 10.1111/j.1365-2311.2010.01250.x CrossRefView Record in Scopus Lach, 2003 l. Lach Invasive ants: unwanted partners in ant-plant interactions? Ann Missouri Bot. Gard., 90 (2003), pp. 91-108 CrossRefView Record in Scopus Lundin et al., 2013 O. Lundin, H.G. Smith, M. Rundlof, R. Bommarco When ecosystem services interact: crop pollination benefits depend on the level of pest control Proc. R. Soc. B, 280 (2013), p. 20122243, 10.1098/rspb.2012.2243) Majer, 1976 J.D. Majer The influence of ants and ant manipulation on the cocoa farm fauna J. Appl. Ecol., 13 (1976), pp. 157-175 CrossRefView Record in Scopus Majer, 1994 J.D. Majer Arboreal ant community patterns in Brazilian cocoa farms Biotropica, 26 (1994), pp. 73-83 CrossRefView Record in Scopus Mcgregor and Moxon, 1985 A.J. Mcgregor, J.E. Moxon Potential for biological control of tent building species of ants associated with Phytophthora palmivora pod rot of cocoa in Papua New Guinea Ann Appl. Biol., 107 (1985), pp. 271-277 CrossRefView Record in Scopus Medeiros et al., 1993 M.A. Medeiros, H.G. Fowler, J.H.C. Delabie Interaction of the black pod disease (Phytophthora SPP.) and the ant Azteca chartifex spiritii in Bahian cocoa plantations Agrotropica, 5 (3) (1993), pp. 65-68 Nyasse et al., 2007 S. Nyasse, M.I.B. Efombagn, B.I. Kébé, M. Tahi, D. Despréaux, C. Cilas Integrated management of Phytophthora diseases on cocoa (Theobroma cacao L): Impact of plant breeding on pod rot incidence Crop Prot., 26 (2007), pp. 40-45 ArticleDownload PDFView Record in Scopus Otto et al., 2008 S.B. Otto, E.L. Berlow, N.E. Rank, J. Smiley, U. Brose Predator diversity and identity drive interaction strength and trophic cascades in a food web Ecology, 89 (1) (2008), pp. 134-144 CrossRefView Record in Scopus Padi and Owusu, 2003 B. Padi, G.K. Owusu Towards an integrated pest management for sustainable cocoa production in Ghana Paper from Workshop on Shade Grown Cocoa, held in Panama, 3/30-4/2, 1998. Smithsonian institution Washington, D.C (2003) (8p) Perfecto et al., 2004 I. Perfecto, J.H. Vandermeer, G.L. Bautista, G. Ibarra Nuñez, R. Greenberg, P. Bichier, S. Langridge Greater predation in shaded coffee farms: the role of resident neotropical birds Ecology, 85 (2004), pp. 2677-2681 CrossRefView Record in Scopus Philpott and Armbrecht, 2006 S. Philpott, I. Armbrecht Biodiversity in tropical agroforests and the ecological role of ants and ant diversity in predatory function Ecol. Ento., 31 (2006), pp. 369-377 CrossRefView Record in Scopus Rusch et al., 2016 A. Rusch, R. Chaplin-Kramer, M.M. Gardiner, V. Hawro, J. Holland, D. Landis, C. Thies, T. Tscharntke, W.W. Weisser, C. Winqvist, M. Woltz, R. Bommarco Agricultural landscape simplification reduces natural pest control: a quantitative synthesis Agric. Ecosyst. Environ., 221 (2016), pp. 198-204 ArticleDownload PDFView Record in Scopus Taylor and Griffin, 1981 B. Taylor, M.J. Griffin The role and relative importance of different ant species in the dissemination of black pod disease of cocoa P.H. Gregory, A.C. Maddison (Eds.), Epidemiology of Phytophthora on Cocoa in Nigeria., CAB, Wallingford UK (1981), pp. 114-131 View Record in Scopus Tilman et al., 1996 D. Tilman, D. Wedin, J. Knops Productivity and sustainability influenced by biodiversity in grassland ecosystems Nature, 379 (1996), pp. 718-720 CrossRefView Record in Scopus Tilman et al., 2001 D. Tilman, P.B. Reich, J. Knops, D. Wedin, T. Mielke, C. Lehman Diversity and productivity in a long-term grass-land experiment Science, 294 (2001), pp. 843-845 CrossRefView Record in Scopus Tscharntke et al., 2007 T. Tscharntke, R. Bommarco, Y. Clough, O.T. Crist, D. Kleijn, T.A. Rand, J.M. Tylianakis, S. Nouhuys, S. Vidal Conservation biological control and enemy diversity on a landscape scale Biol. Control, 43 (2007), pp. 294-309 ArticleDownload PDFView Record in Scopus Tscharntke et al., 2012 T. Tscharntke, J.M. Tylianakis, T.A. Rand, R.K. Didham, L. Fahrig, P. Batáry, J. Bengtsson, Y. Clough, O.T. Crist, C.F. Dormann, R.W. Ewers, J. Fründ, R.D. Holt, A. Holzschuh, A.M. Klein, D. Kleijn, C. Kremen, D.A. Landis, W. Laurance, D. Lindenmayer, C. Scherber, N. Sodhi, I. Steffan-Dewenter, C. Thies, W.H. van der Putten, C. Westphal Landscape moderation of biodiversity patterns and processes − eight hypotheses Biol. Rev., 87 (2012), pp. 661-685 CrossRefView Record in Scopus VanMele, 2008 P. VanMele A historical review of research on the weaver ant Oecophylla in biological control Agric. Forest Entomol., 10 (2008), pp. 13-22 View Record in Scopus Vandermeer et al., 2010 J. Vandermeer, I. Perfecto, S.M. Philpott Ecological complexity and pest control in organic coffee production: uncovering an autonomous ecosystem service Bioscience, 60 (2010), pp. 527-537, 10.1525/bio.2010.60.7.8 CrossRefView Record in Scopus Way and Khoo, 1992 M.J. Way, K.C. Khoo Role of ants in pest-management Annu. Rev. Entomol., 37 (1992), pp. 479-503 CrossRefView Record in Scopus Wielgoss et al., 2014 A. Wielgoss, T. Tscharntke, A. Rumede, B. Fiala, H. Seidel, S. Shahabuddin, Y. Clough Interaction complexity matters: disentangling services and disservices of ant communities driving yield in tropical agro-ecosystems Proc. R. Soc. B., 281 (2014), p. 20132144, 10.1098/rspb.2013.2144 Yede et al., 2012 B.R. Yede, C. Djieto-Lordon, C. Cilas, L. Dibog, R. Mahob, C.F. Bilong True bug (Heteroptera) impact on cocoa fruit mortality and productivity J. Econ. Entomol., 105 (2012), pp. 1285-1292 View Record in Scopus Zovi et al., 2008 D. Zovi, M. Stastny, A. Battisti, S. Larson Ecological costs on local adaptation of an insect herbivore imposed by host plants and enemies Ecology, 89 (5) (2008), pp. 1388-1398 CrossRefView Record in Scopus © 2017 Elsevier B.V. All rights reserved.