Periodical cicadas disrupt trophic dynamics via community-level shifts in Avian Foraging

Once every 13 or 17 years within eastern North American deciduous forests, billions of periodical cicadas concurrently emerge from the soil and briefly satiate a diverse array of naive consumers, offering a rare opportunity to assess the cascading impacts of an ecosystem-wide resource pulse on a complex food web. Here, we quantify the effects of the 2021 Brood X emergence, and report that >80 bird species opportunistically switched their foraging to include cicadas, releasing herbivorous insects from predation, and essentially doubling both caterpillar densities and accumulated herbivory levels on host oak trees. These short-lived but massive emergence events help us to understand how resource pulses can rewire interaction webs and disrupt energy flows in ecosystems, with potentially long-lasting effects. 

Methods
Avian Foraging Observations (2021)

Throughout the Brood X cicada emergence (May 15 - June 24, 2021), teams of 1-3 researchers used binoculars to observe and record the foraging behavior of individual birds at our main study site, the Izaak Walton League Conservation Farm (IWL hereafter), located in Poolesville, Maryland, USA (39.095, -77.432). Avian foraging observations were made in a variety of habitats including managed fields, a pond, deciduous forest understory, deciduous forest edge, and old field. A foraging observation began when an identifiable bird attempted to consume food items or transport food and ended when the bird was no longer visible. The observations took place between the 0700 and 1400 hours, and ranged in duration from 5 seconds to 4 minutes. Prey items were easily classified as cicada or non-cicada, due to the conspicuous size and color of cicadas in relation to other common prey items. Bird trait data (feeding guild, size) were extracted from the Elton trait dataset (51) or from the Cornell Lab of Ornithology All About Birds website (52). For our focal dataset containing both cicada and non-cicada foraging observations (Table S3), we analyzed the effect of bird size and feeding guild on the likelihood that a foraging bout included a cicada. The analysis used binomial GLMM with bird size (g) and feeding guild (insectivores: ​​N=11 spp., n=69 observations; omnivores: N=10 spp., n=​​76; granivores: N=8 spp., n=72) as fixed effects, and species as a random effect using ​​​​lme4 (53), with emmeans (54) for post-hoc comparisons in R (55). Because we recorded detailed foraging data for only a single frugivorous bird species (the cedar waxwing, Bombycilla cedrorum) and for no species of carnivores/scavengers, we did not include those as categories in this analysis. To supplement our own data collection, we enlisted the assistance of the larger recreational birding community with a campaign entitled ‘Brood X Bird Feast’ (56). Participants were provided protocols and datasheets for observations of both cicada and non-cicada foraging events. A second citizen-science project initiated by the coordinators for the 3rd Maryland-DC Breeding Bird Atlas (57) included observations of bird species observed foraging on cicadas, but did not record incidences of birds foraging on non-cicada food sources (Table S2). We used a complete list of the birds of ​​Maryland (58) to construct a pruned phylogeny of the local avian community (Fig. S1), excluding birds that were extinct or vagrants based on their typical habitat range as detailed on the Cornell Lab of Ornithology All About Birds website (52). We mapped lineages that contained cicada-feeding individuals (one or more observed instances of cicada-feeding from our combined foraging datasets) on this pruned phylogeny using the ape package in R (59). 

Avian Predation (2020-2022)

At IWL, from May to July (2020) and May-August (2021-2022), we conducted weekly inspections of 5-cm long plasticine caterpillar models (Fig. 3a) affixed with a drop of cyanoacrylate adhesive (‘superglue’) to branches of 20 understory white oak trees <3m tall. Focal trees were located adjacent to a small forest trail in the interior of a closed canopy second growth oak-hickory forest approximately 2.5 km upslope from the Potomac River (39.101, -77.420). Four model caterpillars were affixed to each tree every week, and were replaced with fresh plasticine caterpillars immediately following inspection one week later (N= 80 caterpillars observed/week). All plasticine caterpillars were inspected in the field using 10x hand lenses to identify predation marks (60). The vast majority of marks were clearly made by birds (i.e., were distinctive beak imprints); a variety of attacks by non-bird predators, including various arthropods and even a snail, were recorded, but we restricted our predation analysis to those models exhibiting definitive bird strikes. During the spring and summer of 2021 (the Brood X emergence year), we concurrently deployed an additional set of caterpillar models on planted white oak saplings growing in the BiodiversiTREE plots (61) of the Smithsonian Environmental Research Center (SERC; 38.834, -76.557); this second study site is located about 100 km east of IWL in Edgewater, MD. Because the white oak saplings at SERC were somewhat smaller than those at IWL, we placed 2 plasticine caterpillars/sapling on each of 30 saplings (N = 60 caterpillars observed each week) to maintain a comparable density/tree. The proportion of caterpillar models/sapling that were attacked by birds was compared among years and censuses (at the IWL site only), between sites (SERC vs. IWL in 2021), and among censuses within sites using repeated measures GLMM (link = "logit") followed by EMM [emmeans package (54)] post-hoc comparisons in R.  

Insect Herbivore Surveys (2021-2022)

At our IWL site, we non-destructively recorded the abundance and identity of each externally-feeding folivore (predominantly caterpillars, sawflies, and beetles; Table S4) present on the foliage or stems of 25 understory white oak saplings with accessible branches (< 3m tall). Most insects were identified to species; a small number were assigned to morphospecies. These censuses were conducted in mid-May (just prior to the appearance of adult cicadas at IWL in 2021) and again in late June of 2021, 2022, and 2023. Covid-related restrictions precluded the collection of herbivore data in the first year of the study (2020); hence we sampled the oak herbivore community for an extra year (2023), enabling us to compare the herbivore community sampled in the emergence year (2021) to the two subsequent non-emergence years (2022 and 2023). The May census was timed to sample the distinct insect herbivore fauna associated with young expanding oak foliage, which by late May gives way to an entirely different summer herbivore fauna that feeds exclusively on mature oak leaves (62). The June census was timed to sample the summer herbivores that were present during the cicada emergence at the study site in 2021. During the June censuses, we also recorded the size of the most abundant caterpillar species (Acronicta increta: Noctuidae) as small (larval instars 1-2) or large (larval instars 3-5). For each census, herbivore abundance was calculated by summing counts across taxa, and sampling effort was quantified by counting the total number of leaves examined per tree. We compared insect herbivore abundances among years with generalized linear models, using ln-transformed total number of leaves as an offset, to test for among-year differences in abundance of 1) total herbivores and 2) the most common free-feeding caterpillar (Acronicta increta: Noctuidae). The distribution of counts was over-dispersed (variance > mean) and violated Poisson assumptions; we therefore applied negative binomial GLMs with the log-link, followed by posthoc comparisons of means across years with Tukey family-wise adjustments of p-values. One-way ANOVAs on herbivore densities (free-feeding herbivore/leaf) yielded qualitatively similar results and thus are not reported here. Additionally, we used Fisher’s exact tests to compare the frequencies of small (instars 1-2) vs. large (instars 3-5) A. increta caterpillars censused during the late June census of each year.  

Herbivore Damage Surveys (2020-2022)

We conducted non-destructive herbivory censuses in early June and mid-October of each year to estimate the damage to understory white oak trees incurred from feeding by the spring herbivore community (pre-cicada) and from the summer/fall herbivore community (during and post-cicada). In each census, we visually estimated (to the nearest 1.0%) the amount of leaf tissue missing from each of 30 leaves randomly selected from throughout the crown of each tree (N = 30 understory white oak saplings; damage estimated for 900 total leaves per census). Prior to each census, our team of observers trained on a set of damaged leaves with known percent damage (measured in the field using the LeafByte app (63). During the fall census, 30 randomly selected leaves were harvested from each tree and analyzed using LeafByte for a more precise measure of total accumulated herbivory. For each year, the damage acrrued between the spring and fall herbivory surveys was estimated by subtracting the median % damage of the early June census from the median % damage of the October census for each tree. The means of these median differences (accrued herbivory) were then compared among years with one-way ANOVA followed by EMM (emmeans) post-hoc comparisons in R (54). Because the median % damage values were skewed by small values and bounded by zero, we also converted percentages to proportions and applied beta regression using the ‘betareg’ package in R (64). Because the results were qualitatively identical, we report the simpler ANOVA results.