Ecological Archives E092-023-A2

L. B. Marczak, C-K. Ho, K. Więski, H. Vu, R. F. Denno, and S.C. Pennings. 2011. Latitudinal variation in top-down and bottom-up control of a salt marsh food web. Ecology 92:276–281.

Appendix B. Expanded methods and additional analyses.

Plant collection and culture

Small Iva plants were collected in the fall of 2007 from five high and five low latitude sites (Table B1) and potted individually in 30 cm diameter pots in a 1:1 mixture of sand and potting soil. Plants were acclimated for 4 months in an open-air (roof only) greenhouse at the University of Georgia Marine Institute on Sapelo Island, Georgia, USA. In March of 2008, we added 9.5 g of fertilizer (Ultra Vigoro Plant Food, 12-5-7) to each plant. Plants averaged 57.65 cm ± 19.33 (SD) tall at the start of the experiment in May 2008. Mesocosms were distributed haphazardly across the greenhouse at the start of the experiment (May 28, 2008). We collected more plants than were needed for the mesocosm experiments; extra plants were used for measurements of initial leaf toughness and nitrogen content so that these destructive measurements would not affect the condition of experimental plants.

TABLE B1. Sources of Iva plants used in mesocosms: site names and locations.

Site Name State Decimal latitude Latitude category
Nelson Island MA 42.44 high
Great Neck MA 42.42 high
100 Acres RI 41.46 high
Rumstick Cove RI 41.43 high
Cotrell Marsh CT 41.20 high
Baruch SC 33.22 low
Ace Basin SC 32.33 low
Eulonia GA 31.54 low
Airport/Dean Creek GA 31.23 low
Amelia FL 30.40 low

Animal collection and culture

We established a culture of Ophraella beetles using ~50 adults collected from Sapelo Island on May 9, 2008 and maintained in the laboratory on a diet of fresh Iva leaves. Larval Ophraella used in the experiment were first instar larvae hatched from this culture in the laboratory between May 22 and May 24. Adult Ophraella used in the experiment were wild-collected from the field between May 20 and May 22 and maintained in individual Petri dishes on fresh Iva leaves for a minimum of 3 days. Groups of adult Ophraella added to mesocosms contained at least one known female (determined by observing egg laying in petri dishes).

Cultures of Uroleucon aphids were initiated from field-collected individuals and maintained in cages on additional Iva plants until needed. On May 20, 2008, 10 individuals were added to each Iva plant intended for use in the mesocosm experiment. On May 25, we thinned aphids by hand until 30 individuals remained on each plant.

Ladybugs (Hippodamia convergens) were purchased from Arbico Organics on May 21 and were fed an ad libitum diet of Uroleucon aphids until the beginning of the experiment. It was difficult to observe ladybug mortality without heavily disturbing the mesocosms, and so we did not attempt to replace ladybugs that died during the course of the experiment.

Armases were field collected, weighed, sexed and placed in mesocosms on the same day that they were collected. Only female Armases were used to standardize for any possible effects of sexual dimorphism (Buck et al. 2003). Crabs that died during the experiment were replaced. The relative growth rate (RGR) of Armases at the end of the experiment was determined as the natural log + 1 of the initial mass minus the natural log +1 of the final mass of green leaves, divided by the duration (days) an individual Armases was in the experiment.

Data standardization and presentation.

In the field, plants from low latitudes are consistently larger than those from high latitudes. Despite efforts to collect plants that were roughly equal in size, this difference was also evident in experimental plants; plants from low latitudes were larger by volume (F1,78 = 41.9, P ≤ 0.0001; calculated as the volume of an ellipse) and had more green leaves at the start of the experiment (F1,78 = 97.8, P = < 0.0001) than plants from high latitudes. Accordingly, we standardized most response variables, with the exception of leaf toughness and N content, by dividing by the initial number of green leaves in each mesocosm. The relative growth rate (RGR) of plants was determined as the natural log + 1 of the initial number of green leaves minus the natural log +1 of the final number of green leaves, divided by the duration (days) of the experiment (Arnott et al. 1974, Groeneveld 1998). Herbivores on Iva rarely consume entire leaves; rather, once a leaf is heavily damaged we have observed that the herbivores used in this experiment abandon that leaf in favour of less damaged leaves (Pennings, personal observation). Since whole Iva leaves are not consumed during herbivory, we were able to use counts of damaged leaves in addition to the percent area of damage on individual leaves to estimate herbivore pressure on Iva plants in our mesocosm experiment. We calculated the relative growth rate (RGR, see Arnott et al. 1974, Groeneveld 1998) of green leaf number as an index of plant growth in order to account for initial variation in plant size. We followed Ho and Pennings (2008) in using the number of green leaves as an index of plant size because this measure directly relates to photosynthetic potential, whereas other potential measures (such as height) in this shrub are strongly affected by plant shape and branching pattern.

Means for all treatment combinations are provided in Appendix C, Table C4. Where appropriate we employed Tukey’s HSD to determine post hoc differences between levels within main factors. We employed Satterthwaite’s approximation when data were unbalanced (e.g. plants with no green leaves could not be used for determination of damage or %N and were excluded from analyses) which results in fractional denominator degrees of freedom (Quinn and Keough 2002).

Predation trials

To inform our understanding of the outcome of the mesocosm experiment, and to help us extrapolate the results from the experiment to the more diverse field community, we conducted predation trials in the laboratory to assess which species from the Iva food web were most likely to feed on each other. Trials were conducted in June and July of 2008. Animals were collected from the field or, in the case of aphids and Ophraella larvae, from stock cultures maintained in the greenhouse or laboratory, and replicates of the different treatments were run as individuals of different species became available. Upon collecting, animals were acclimated for 24 hours to the laboratory conditions. Trials were run until all herbivores were consumed or to a maximum of 24 hours in 850 ml glass jars at a constant room temperature of 25°C and photoperiod 14:8 (day:night). Each jar was stocked with a 15 cm long Iva twig (~9 leaves) which served as a substrate for the animals. Individual consumers were used only once. In each case, we offered consumers more potential prey individuals than they were able to eat (based on pilot trials), but otherwise assumed that encounter rates in the small containers were so high that predation rates were not limited by the number of prey offered (i.e., encounter rates), but only by attack and consumption rates. Because we were interested in generalizing beyond the simplified food web used in the mesocom experiment, we included predation trials using hunting spiders as predators (a single common species, Marpissa sp., and a group of rarer species), and included the beetle Paria and the grasshopper Hesperotettix floridensis as potential prey. The number of replicates varied per trial series due to the availability of animals. Results were standardized to 8 hours of each single feeding trial.


In one-choice predation trials, a single Armases crab ate roughly ten times more aphids than did a single Hippodamia ladybug; spiders ate almost no aphids (F2,14 = 124.3, P < 0.0001, Fig. B1a). Ophraella ate no aphids, confirming that it is completely herbivorous (Fig. B1a). Armases also ate several times more Ophraella larvae than did Hippodamia, with spiders eating essentially no beetle larvae (F9,49 = 17.1, P < 0.0001. Fig. B1b). Armases was the only predator that successfully captured and consumed adult Ophraella (Fig. B1b), although in lower numbers than they consumed Ophraella larvae. Armases were also able to consume moderate numbers of adult Paria beetles (F2,14 = 8.0, P = 0.003, Fig. B1c) and small numbers of Hesperotettix grasshopper nymphs (F3,42 = 15.6, P < 0.0001, Fig. B1d); ladybugs were unable to consume adult Paria (Fig. B1c). Finally, Armases were moderately successful intraguild predators on Hippodamia (Fig. B1c), but were able to consume only a few spiders (Fig. B1d).

FIG. B1. Predation trials showing the number of prey consumed by focal marsh predators in one-choice predation experiments. Numbers below axes indicate sample size for each predator-prey combination. Predators for each combination are indicated at the bottom of the plot. Data are grouped by prey type: (A) prey = aphids (B) prey are small (S), medium (M), large (L) and adult (A) Ophraella. (C) prey = other Coleoptera (D) other common prey from salt marsh habitats.

Relative Interaction Intensity

To better compare treatment effect sizes, we calculated the relative interaction intensity (RII) (Armas et al. 2004) of all pairs of predator and prey (ladybugs and Armases alone and together against aphids, larval Ophraella and adult Ophraella) for data collected after 2 weeks and at the end of the experiment. RII was calculated as the number of prey in the presence of the predator minus the number of prey in the absence of the predator divided by the total number of prey (with and without the predator). We compared the RII of these predator combinations on each herbivore prey in separate ANOVAs with predator combination and latitude as fixed factors.


Relative interaction intensity (RII) between single predators and their combination did not differ by latitude of plant origin for any herbivore prey (aphids, larval Ophraella, adult Ophraella) or time period (after two weeks and at the end of the experiment). However, the RII for each predator type (LB alone, crab alone or LB and crab together) did differ with each herbivore prey. After two weeks in experimental mesocosms crabs showed the weakest interaction intensity against aphids relative to ladybugs or the combination of ladybugs and crabs (predator, F2,5 = 7.02 P = 0.004; Fig. B2c; by the end of the experiment, this difference in RII was no longer significant, largely due to an increase in variability rather than a change in the mean RII (predator F2,5 = 1.98, P = 0.16; Fig. B2d). The RII of all predators against larval Ophraella were equivalent (predator, F2,5 = 1.47 P = 0.25; Fig. B2a) while ladybugs had the weakest RII with adult Ophraella compared to crabs alone or both predators together (predator combination, F2,5 = 6.22 P = 0.007; Fig. B2b).

FIG. B2. Relative interaction intensity (RII) between predators and (A) final Ophraella larvae density, (B) final Ophraella adult density, (C) aphid density after two weeks in experimental mesocosms, and (D) aphid density at the conclusion of the mesocosm experiment.

Structural Equation Modeling

A potential weakness of the ANOVA approach is that it does not account for the fact that some variables (e.g., herbivore numbers, plant traits) are both responding to and simultaneously driving other variables. Thus, we also analyzed results of the mesocosm experiment using structural equation modeling (SEM) which allows a variable to be both influenced by other variables and cause variation in a dependent variable (Grace 2006). Building an SEM model consists of several consecutive steps. It starts with a priori identification of the causal relationships between the interplaying variables, followed by estimation of the path parameters performed by screening the matrix of covariances over the hypothetical model. Finally, model fit is determined by comparing the predicted matrix of covariances with that from the original data. Parameters of the model were estimated using AMOS 7.0 with the maximum likelihood method, and the model fit was tested by the likelihood χ2 value.


SEM analysis indicated that Armases had direct negative effects on both Ophraella beetles and aphids (Figure B3). In addition, aphids were negatively affected by Hippodamia ladybugs and Ophraella . Note that this implies an indirect positive effect of Armases on aphids, because Armaes reduces competition from Ophraella. An observed positive correlation between aphids and Ophraella likely represents shared positive responses to unmeasured aspects of plant quality. SEM indicated that Ophraella had multiple effects on Iva, increasing herbivore damage, decreasing nitrogen content and increasing chlorophyll content. These three plant traits were also affected by latitude, with all values larger for high-latitude plants. Measures of plant size or growth were not included in the final SEM model.

FIG. B3. SEM model of the Iva food web in the mesocosm experiment. The model is consistent with the data (P = 0.43, χ2/df = 1.02). Path coefficients describe standardized values showing relative effects of variables upon each other. Arrow width is proportional to the strength of the path coefficient; one headed arrows represent causal relationships; two-headed arrows represent correlations.

Potential cage effects

We found that Armases suppressed beetle densities in our experimental mesocosms, but herbivorous beetles were not affected by Armases exclusion in the field (Ho and Pennings 2008). The most likely explanation for this difference is that beetle mobility was constrained by cages in our mesocosm experiments, but not in the field experiments of Ho and Pennings (2008), where the presence of Armases was manipulated without cages. This difference in beetle mobility could have affected the experiments in different ways. In the mesocosm experiments, beetle escape behavior (dropping from branches, flying away) might have been rendered less effective by the cages. Conversely, in the field experiments, movement of beetles among experimental and adjacent Iva plants might have swamped crab effects. We conclude, based on the predation trials and mesocosm experiment, that both ladybugs and Armases can exert top-down control of beetles in the field, but that the net effects of this at the landscape scale likely depend on the proportion of Iva plants occupied by ladybugs and Armases, and on beetle movement patterns.


Armas, C., R. Ordiales, and F. I. Pugnaire. 2004. Measuring plant interactions: a new comparative index. Ecology 85:2682–2686.

Arnott, R. A., N. R. Brockington, and C. R. W. Spedding. 1974. A non-destructive method for the measurement of growth in grasses in nutrient solutions. Journal of Experimental Biology 25:1124–1136.

Buck, T. L., G. A. Breed, S. C. Pennings, M. E. Chase, M. Zimmer, and T. H. Carefoot. 2003. Diet choice in an omnivorous salt-marsh crab: different food types, body size, and habitat complexity. Journal of Experimental Marine Biology and Ecology 292:103–116.

Grace, J. B. 2006. Structural equation modelling and natural systems. Cambridge University Press, Cambridge.

Groeneveld, H. W. 1998. Measuring the RGR of individual grass plants. Annals of Botany 82:803–808.

Ho, C. K., and S. C. Pennings. 2008. Consequences of omnivory for trophic interactions on a salt marsh shrub. Ecology 89:1714–1722.

Quinn, G. P., and M. J. Keough. 2002. Experimental design and data analysis for biologists. Cambridge University Press, Cambridge.

[Back to E092-023]