Ecological Archives E093-160-A1

Yann Clough. 2012. A generalized approach to modeling and estimating indirect effects in ecology. Ecology 93:1809–1815. http://dx.doi.org/10.1890/11-1899.1

Appendix A. Illustrations and details of the study system with a supporting figure and text.

Fig. A1. (A) Theobroma cacao trees, (B) Exit hole of Conopomorpha cramerella on cacao pod (C) Adult individual of Helopeltis sulawesi adult and feeding damage on cacao pod (D) Philidris cf. cordata tent structure, workers and pseudococcid trophobionts. Photo credits: Y. Clough.

Study sites

All sites were smallholder cacao plots situated around the Lore Lindu National Park, Central Sulawesi, Indonesia. The elevation in this region ranges from 400 m to about 1000 m asl. Plots 40 × 40 m in size were managed during two years by local assistants to enable homogeneous management that differed only in the treatments assigned in the experiment. Half of the plots were assigned randomly to frequent manual weeding regime (every 2 months), the other half of the plots to infrequent manual weeding (every 6 months). Each plot was separated into two 20 × 40 m subplots and one half fertilised twice a year with urea fertiliser (46% N). Fertiliser was applied twice a year from December 2006 to June 2008 at a rate of 217 g urea (100 g N) per tree with aliquots placed into 10 concentric holes around each tree, which were subsequently covered. This was done to prevent loss with surface water run-off. Minor pruning of chupons was conducted on a regular basis, and heavy pruning twice a year.

The study species

The cacao tree Theobroma cacao L. (Malvaceae) is a cauliflorous tree, originating from the from the neotropics but now grown around the world's tropics. It produces pods which contain the cocoa beans used to produce chocolate, amongst other products. Cacao is the main cash crop grown in the study region. Hybridization between out-crossing genotypes is common in smallholder cacao, and cacao trees in this study, as well as most of the other trees in the area, originate from hybrids between Forastero and Trinitario types.

The ant Philidris cf. cordata (Formicoidea: Dolichoderidae) tends pseudococcids (Homoptera) on pods, but also on the bark, flowers etc. The pseudococcid aggregations, the paths taken by the ants, and the nesting structures of the ants are covered by so-called "tents", built by the workers using plant and soil material. This species is becoming increasingly common in cacao plantations in the study region (A. Rizali and Y. Clough, unpublished data). It is by far the most numerically dominant arboreal ant where it is present (Wielgoss et al. 2010) and where it is absent, overall ant activity is much reduced.

The bug Helopeltis sulawesi Stonedahl (Hemiptera: Miridae) feeds on the pods of cacao as adults as do other species of mirids in other regions (Entwistle 1972). This particular species is not known to transmit viruses or other diseases to the host plant. It feeds externally on young shoots and developing cacao pods (Entwistle 1972 and personal observation). It causes black necrotic spots to develop where the labial stylet has penetrated the plant tissue.

The moth Conopomorpha cramerella Snellen (Lepidoptera: Gracillaridae) oviposits on cacao pods with a preference for the latter 3 months of the fruits' 5–6 months development time (Day 1989). The larvae tunnel into the pod to feed, before tunneling out and pupating, usually on the ground on the underside of dead leaves.

Data collection

Ants were sampled twice using tuna and sugar baits (for details see Wielgoss et al. 2010).

The cacao pods were surveyed for pest incidence every two weeks. Ripe pods were harvested, checked for symptoms of H. sulawesi feeding and opened to check for internal damage due to C. cramerella.

The choice of hypothesized links between covariates and animal and plant variables

Temperature: The plots differed in their altitude above sea-level, and thereby temperature. Using mean daily temperature as a covariate for the invertebrates and the number of pods produced on a tree is necessary because each of the organisms involved can be expected to have a temperature optimum and the incidence and the growth of these organisms at the plot level could be expected to reflect this. Although optimum curves could have been expected, both exploratory data analysis and residual inspection showed that an assumption of linear relationships over the range of temperatures observed was adequate.

Tree age: is an indicator for nesting site availability for the ants, as older trees have more branch stumps and crevices than younger trees. These structural characters were unfortunately not properly quantified, so I use tree age as an indicator. Tree ase was not included as an explanatory variable for number of pods harvested or incidence of herbiovore damage. Cacao production is expected to decrease after 20–25 years. The trees are between 9 and 20 years in age. Thus while I would expect age to have an effect in general, this was not the case for the tree-age range considered in the study. Whole plot harvest data from the plots studied, as well as additional data on 144 plots in the study region, suggest this is correct. For similar reasons, and given personal observations across the study region, I would not have expected tree age to affect herbivores directly.

Fertilization: Nitrogen fertilization can be expected to influence both the number of pods, as well as the incidence of damage due to herbivores, via effects on host-plant choice and host plant quality. Similarly, nitrogen fertilizer could be expected to benefit ants via enhanced production of honeydew by their trophobionts.

Other potential covariates that were omitted: While one can expect the ants to affect the plant, and possibly the herbivores, via the tending of pseudococcids, quantification of pseudococcids was difficult to do non-destructively because they are generally protected and hidden by the tents built by the ants. Our model may therefore be imperfect in that an effect of the ant on the plants and the herbivores via the pseudococcids could not be included. However, the testing of the model using the d-sep approach (Shipley 2009) suggests that there are no significant effects of the ant on the herbivores that were left unaccounted for. I decided not to include the weeding treatment due to the remaining variability in weed height and cover.


Literature Cited

Day, R. K. 1989. Effect of cocoa pod borer, Conopomorpha cramerella, on cocoa yield and quality in Sabah, Malaysia. Crop Protection 8:332–339.

Entwistle, P. F. 1972. Pests of cocoa. Longman, London, UK. 779 p.

Shipley, B. 2009. Confirmatory path analysis in a generalized multilevel context. Ecology 90:363–368.

Wielgoss A. C., et al. 2010. Temperature and a dominant dolichoderine ant species affect ant diversity in Indonesian cacao plantations. Agriculture Ecosystems and Environment 135:253–259.


[Back to E093-160]