Tree Bark

Leaf litter was collected in December 2013 from the Monash University, Clayton campus (Victoria, Australia; 37.9119 °S, 145.1317 °E). Invertebrates were extracted from litter using Berlese-Tullgren funnels39 . Neanura muscorum individuals (n = 31) were identified using morphological characteristics and placed into closed 70 ml containers with a plaster of Paris and charcoal (9:1) substratum. This substratum was regularly saturated with reverse osmosis treated water to maintain humidity40 . Containers were kept in temperature-controlled rooms set at 15 °C (mean/sd: 14.64 ± 0.61 °C; verified with iButton Hygrochron® temperature/humidity loggers, Maxim Integrated, San Jose, USA) on a 12:12 hour light:dark photoperiod.
Parental lines (F0) were provided with a combined diet of slime mould (Physarum polycephalum; cultured on 1.5% agarose media at 25 °C on a diet of oats) and algae-covered plane tree (Platanus sp.) bark ad libitum. The latter is a standard diet used in rearing other Collembola species, which enables individuals to select amongst a range of algae, cyanobacteria, and fungi to reach their optimal nutrient target12 13 27 . Collembola cultures were checked daily and eggs were removed to new containers and incubated in the temperature-controlled room until hatching. Four batches of first-generation (F1) eggs were used for experiments (Fig. 1). Subsequent clutches were combined and raised as a F1 breeding population under identical conditions to the parental line. Five batches of eggs laid by the F1 breeding population were then used to create F2 populations for a repetition of the experiments undertaken with the F1 generation (Fig. 1). Neanura muscorum F2 individuals that were not used for experimentation were pooled and kept as laboratory stocks. These stocks continue to be maintained on a combined slime mould and algae-covered plane tree bark diet.

No-choice experiments were conducted in Ohio using X. germanus and in Virginia using X. crassiusculus to compare colonization success on flood-stressed and non-flooded C. florida. Bottle traps described by Ranger et al. [28 ] were used for obtaining field-collected adults, but instead of propylene glycol, a moistened paper towel rolled into a tube was placed in the bottom collection vessel of the trap to maintain ambrosia beetle specimens during 24 hrs under field conditions. Ambrosia beetles were then returned to the laboratory and transferred to parafilm-sealed petri dishes containing moistened filter paper and stored for 24 to 48 hrs at 3.3°C.
Flooding of container grown trees was established using the previously described pot-in-pot system on 19 May 2014 in a greenhouse on the campus of the OARDC and on 12 May 2014 on the campus of the HRAREC. Three days after initiating flooding, an individual adult female X. germanus field-collected in Ohio or X. crassiusculus field-collected in Virginia were placed inside of a chamber made of polytetrafluoroethylene (PTFE) tubing that was cut longitudinally (2.5 cm × 1 cm × 0.9 cm; l × w × h) and sealed at both ends with Molded Thermogreen LB-2 Septa (Supelco, Bellefonte, PA) cut into a semi-circle (Fig 1). Cables ties were used to snugly secure the chambers against the stem in parallel, thereby confining an individual beetle to 2.5 cm² of bark tissue. Smooth bark of the C. florida trees allowed for close contact between the tissue and chambers, thereby effectively confining the beetle specimens. Ten chambers per tree were placed on six flood-stressed and six non-flooded C. florida trees starting from the base and extending linearly about 60 cm up the main stem with about 1.5−2 cm between cages. Generally, one to two unresponsive or injured beetles per tree were removed and replaced during the first day. Chambers confining foundress ambrosia beetles were held in place for 25 days, after which the stems were cut at the base and temporarily stored at 5°C until further analysis. As an indication of tunneling activity, chambers were carefully removed from the stems 1‒2 d later and ejected sawdust within each cage was weighed. Stem sections associated with each chamber as part of the experiment conducted in Ohio were also dissected to determine if the foundress was still alive and assess the presence/absence of eggs, larvae, pupae, and fungal growth within each tunnel/gallery. Flooded trees were drained 15 days into the experiment to avoid tree death and then watered accordingly for the remaining 10 days.

Parmotrema tinctorum (Despr. ex Nyl.) Hale., a radially growing foliose macrolichen, was used as a model lichen species for the following reasons: it is distributed widely in the tropical and subtropical regions of the world [17 (link),18 (link)]; it is loosely attached to the substrate so that the thallus can be peeled off easily and cleanly from the substrate, e.g., bark and rock; mature thallus of P. tinctorum has sufficient biomass for the test. Five individual lichen thalli were collected from the bark of pine trees (Pinus thunbergii Parl.) at five different sites of Jeju Island located at the subtropical zone of the southern part of South Korea. The climate of Jeju Island is affected by Southeast Asian monsoon [19 (link)]. Heavy rainfall derived from the East China Sea and the West Pacific contributes to high levels of humidity in summer. The climate during winter is characterized by cold and dry conditions. The mean annual temperature is around 25.4 and 5.1 °C in summer and winter, respectively [20 (link)]. Five collection sites were the following: site 1 (33°28′31″ N, 126°21′09″ E) is located at a seaside cliff and directly exposed to strong winds from the sea; site 2 (33°30′28″ N, 126°28′01″ E) is a hill near the sea; site 3 (33°32′59″ N, 126°45′26″ E) is located at the top of the rising small defunct volcano which has been used as a graveyard and surrounded by agricultural fields (citrus farms); site 4 (33°14′1″ N, 126°22′59″ E) is located at a small park nearby the sea; and site 5 (33°16′23″ N, 126°42′14″ E) is located along the seaside with densely-developed viridian forest (Figure 1a).

Since 2012, the Toronto Zoo team and partners have collected Blanding’s Turtle eggs from wild populations across Ontario. Each year, approximately 10–150 eggs are collected and incubated ex-situ using standard protocols (available upon request). Blanding’s Turtles have temperature-dependent sex determination, and males are produced when the eggs are incubated at or below 28°C and females are produced at incubation temperatures above 30°C [37 , 38 ]. At the Toronto Zoo, eggs are incubated at 27.5°C and 29.5°C to yield a 1:1.5 male:female sex ratio. Annual hatching success at the Toronto Zoo has ranged from 72% to 100%.
Hatchlings are reared in groups for two years prior to release. A maximum of 15 hatchlings are housed in each Waterland black plastic tub (91 cm x 45 cm x 40 cm), with shallow water (20 cm) and artificial vegetation for one year. After a year, a maximum of 7 hatchlings are housed in each tub. The water temperature is maintained at 25–27˚C in the first year and 23–24˚C in the second year. A 180 L sump filter with a heater located below the tanks is used for water circulation. The filter is cleaned once a week and the water is changed three times a week. One end of the tub is elevated and lined with rocks and pebbles, and a 50W bulb provides a basking area with 28–35˚C. A ramp is placed to facilitate easy access to the basking area by hatchlings. Two fluorescent UVB bulbs with full-spectrum lighting are provided for proper bone growth. The headstarted turtles are fed three times a week at roughly 5% of the average body mass of the cohort. One-year old hatchlings are fed turtle gel and beef heart gel (gel diets are gelatine-based foods for aquatic species formulated by the Toronto Zoo), earthworms, and romaine lettuce. After a year, all turtles are supplemented with fish (smelt) and live crickets. As part of enrichment and to promote better foraging ability in the wild, turtles are given varied food sizes, live worms, and natural tree bark for cover.
A maximum of 60 turtles are reared in human care for two years at the Toronto Zoo and released in June each year. A month before their release, headstarted turtles are relocated to large outdoor tubs (173 cm x 120 cm x 58 cm) with shallow water (25 cm) and artificial vegetation. Each outdoor tub holds a maximum of 25 turtles, and the number of tubs used varies annually. The water temperature in outdoor tanks is maintained at 28°C. Each tub is equipped with a filter and a water pump to create a current. The outdoor holding area is secured using mesh and roof fencing. All headstarted turtles are weighed using an Ohaus CS-series scale to the nearest 0.1 g and measured monthly for two years using Belt-Art calipers to the nearest 0.1 mm. Standard body measurements are recorded including midline carapace length and width, midline plastron length and width, shell height, and body mass. Prior to release, turtles are individually marked with notches on the marginal scutes [39 ] and a subcutaneous PIT (passive integrated transponder) tag is inserted into the left hind leg.

We developed a hierarchical model in a Bayesian context to jointly model both the dynamics of beetle activity intensity over time within our plots, as well as the occurrence–accounting for imperfect detection–of black-backed woodpeckers at those same plots. The model largely follows the structure of a single-species occupancy model [17 (link)], where woodpecker observations of detection or non-detection, y_jkt, for survey interval k at site j (where sites are individual survey points) in year t, are assumed to be imperfectly observed representations of the true occurrence status, z_jt (present or absent), which is constant across all k survey intervals (i.e., closure is assumed within the <17-minute survey period) but can change from year to year. Observed occurrence of black-backed woodpeckers, y_jkt, is thus modeled as
$$y_{jkt}\sim \mathrm{Bernoulli}\left(z_{jt}{p}_{jkt}\right),$$
where p_jkt is the probability of detection for a given survey at a site. Similarly, the true occurrence status of a site in year t, z_jt, is modeled as
$$z_{jt}\sim \mathrm{Bernoulli}\left({\psi }_{jt}\right),$$
where ψ_jt is the probability of occurrence at a site. In this context, occupancy is defined by the space and time in which the survey is conducted and across which closure is assumed [18 ]. Thus, our calculated probabilities represent the probability of at least one black-backed woodpecker occurring within (or alternatively, ’using’ [19 ]) the detection radius of a survey point (approximately 120 m) during a survey period (median = 7 minutes in duration).
The probabilities of woodpecker detection and occurrence are both modeled as logit-linear functions of covariates chosen a priori. Following previous work studying black-backed woodpeckers with this survey methodology [14 (link), 15 (link), 20 (link)], we expected detection, p_jkt, to vary as a function of an intercept and the linear additive combination of a categorical covariate representing the survey type (passive = 0, broadcast = 1), giving
$$\mathrm{logit}\left(p_{jkt}\right)={\alpha }_0+{\alpha }_{type}{\mathrm{type}}_k.$$
The probability of woodpecker occupancy of a survey point was modeled as a function of five covariates: (1) elevation, (2) latitude, (3) snag density, (4) intensity of beetle larvae activity (as indirectly measured by cumulative beetle sign since the fire; modeled as a latent variable, see below), and (5) an interaction between years-since-fire and the intensity of beetle larvae activity (with the hypothesis that cumulative beetle sign becomes less predictive over time). Snag counts were conducted immediately after completing woodpecker surveys and consisted of counting all snags of different size classes (10–30, 30–60, and >60 cm dbh) within 50 m of each survey point. Size-specific snag counts were aggregated in the field into different categories (≤5, 6–15, 16–30, 31–50, 51–100, >100), which were converted to numerical quantities (1, 6, 16, 31, 51, 101, respectively) for analysis [15 (link)]. Counts across all three size classes were summed to calculate a relative index of snag density (snags/ha). The linear additive model for occupancy in the first year of surveys can be described as,
$$\mathrm{logit}\left({\psi }_{j,t=1}\right)={\beta }_{0,j}+{\beta }_{elev}{\mathrm{elev}}_j+{\beta }_{lat}{\mathrm{lat}}_j+{\beta }_{snag}{\mathrm{snag}}_{jt}+{\beta }_{beetle}{\mathrm{intensity}}_{jt}+{\beta }_{ageXbeetle}{\mathrm{age}}_{jt}{\mathrm{intensity}}_{jt},$$
where β represents intercept and slope parameters. To account for pseudoreplication and temporal autocorrelation derived by sampling sites repeatedly in consecutive years, we added a temporal autocorrelation term [21 (link)], ϕ, which was multiplied by the true occurrence status in year t−1, resulting in the following model for additional post-fire years,
$$\mathrm{logit}\left({\psi }_{j,t>1}\right)={\beta }_{0,j}+{\beta }_{elev}{\mathrm{elev}}_j+{\beta }_{lat}{\mathrm{lat}}_j+{\beta }_{snag}{\mathrm{snag}}_{jt}+{\beta }_{beetle}{\mathrm{intensity}}_{jt}+{\beta }_{ageXbeetle}{\mathrm{age}}_{jt}{\mathrm{intensity}}_{jt}+\varphi z_{j,t-1}$$
As 2018 was the last year of surveys used in this dataset, and also the only year with in situ beetle activity surveys, all surveys conducted in 2018 held the temporal index of t = 10. Surveys in previous years (t = 1,…,9) were treated as missing data if no surveys occurred at a site in that survey-year. Finally, the intercept (β_0,j) was modeled as a random effect for each fire (n = 22), drawn from a hierarchical normal informed by a common mean (μ_β0) and precision (τ_β0).
Previous analyses of black-backed woodpecker occurrence have demonstrated the importance of the number of years since fire in models of the species’ occurrence [14 (link), 17 (link)], yet our model of woodpecker occupancy does not include an independent effect of time since fire (Eqs 4–5). By excluding time since fire from our model of woodpecker occupancy, we assume that all temporal changes in occupancy are due to either (1) temporal changes in habitat quality as indicated by intensity of beetle sign (parameters β_beetle and β_ageXbeetle), or (2) random stochasticity (the frequency of which is governed by the parameter ϕ).
A novel feature of our multi-trophic model is that we treat cumulative beetle larvae sign at a survey point each year (intensity_jt) as a latent (i.e., indirectly observed) variable, which for mathematical simplicity we define as continuous. We are then able to model beetle larvae sign as a function of different environmental variables expected to relate to beetle activity and to account for the known dynamic that beetle sign generally accumulates over time even though overall activity may decline. Thus, we hypothesized that the intensity of beetle sign at a site each year (intensity_jt) varies as a function of: (1) the number of years since fire; (2) the proportion of sampled trees per point that were of the genus Pinus; (3) and an interaction between the proportion of pines and years since fire. Based on previous work [13 (link)], we hypothesized that beetle sign increases over time (as sign is generally cumulative and lasting), but that pines would have greater activity in early post-fire years and lower activity in later post-fire years (as pine bark generally decomposes faster than bark of other trees in our study areas). We thus modeled beetle sign intensity as,
$$\mathrm{logit}\left({intensity}_{jt}\right)={\gamma }_0+{\gamma }_{age}{\mathrm{age}}_{jt}+{\gamma }_{pine}{\mathrm{pine}}_j+{\gamma }_{ageXpine}{\mathrm{age}}_{jt}{\mathrm{pine}}_j.$$
We fit this model to observed data collected in 2018, by treating the total sum of beetle sign scores across all surveyed trees per point (max = 6) as a binomially distributed variable, as follows,
$${\mathrm{activity}}_j\sim binomial\left({intensity}_{j,t=10},{\mathrm{numTrees}}_j*8\right),$$
where the maximum activity score is a product of the number of trees sampled per point (numTrees_j) and the maximum potential activity score per tree (i.e., 8).
We fit the model to the data with JAGS [22 ] using the R statistical programming language version 4.0.2 [23 ] and the package ‘R2jags’ [24 ]. We used vague priors (i.e., normal with μ = 0, τ = 0.1). We ran three chains of 50,000 iterations thinned by 50 with a burn-in of 50,000, yielding a posterior sample of 3,000 across all chains. Convergence was checked visually with traceplots and confirmed with a Gelman-Rubin statistic < 1.1 [25 ]. Inference on parameters was made using 95% Bayesian credible intervals (95 CI).

A stem bark sample of T. brownii tree (Figure 1) was acquired from Kitui, Kenya (1.3099°S 37.7558°E; about 152 km from Nairobi) in May 2021, aided by a native herbalist. Botanical verification was conducted at the East African Herbarium situated at the National Museums of Kenya, and a voucher specimen was deposited there (accession number: JWM001). The plant sample was taken to the animal breeding and experimentation facility at Kenyatta University for preparation and bioassays. The sample was rinsed using running tap water, reduced into tiny shreds, and shade-dried till thoroughly dried and powdered [40 (link)].