Lightning strikes are a common source of disturbance in tropical forests, and a typical strike generates large quantities of dead wood. Lightning‐damaged trees are a consistent resource for tropical saproxylic (i.e., dead wood‐dependent) organisms, but patterns of consumer colonization and succession following lightning strikes are not known. Here, we documented the occurrence of four common consumer taxa spanning multiple trophic levels—beetles,
Disturbance often promotes diversity by creating habitat heterogeneity and increasing the local abundance of resources (Sousa 1979, Roxburgh et al., 2004, McClain and Barry 2010, Alencar et al., 2022). Lightning is a common source of disturbance, especially in tropical forests where lightning strikes 35-67 million times per year (Gora et al., 2020, Yanoviak et al., 2020).
Lightning strikes in tropical forests typically damage groups of trees without causing explosive damage or fires (Furtado 1935, Anderson 1964, Magnusson et al., 1996). For example, a typical lightning strike in lowland Panamanian forest kills an average of 5.3 trees while damaging 18.3 others, creates 7.36 metric tons of dead wood biomass, and significantly contributes to canopy gap formation (Yanoviak et al., 2020, Gora et al., 2021). Whereas the role of lightning in forest dynamics is becoming more clear, the ecology of consumer organisms in lightning strike sites is essentially unknown in tropical forests.
Unlike other agents of tree death (e.g., windthrow, pathogens;McDowell et al. 2018), lightning strikes typically generate relatively large volumes of dead wood localized in relatively small patches of forest (Yanoviak et al., 2020, Gora et al., 2021). Trees killed by lightning tend to be disproportionately large (Yanoviak et al., 2020), and often are healthy prior to the strike. Thus, dead wood in lightning strike sites is likely to be relatively high quality in terms of resource volume and nutrient content for saproxylic (i.e., dead wood-dependent) organisms, especially primary consumers of dead wood (xylophages).
The patchiness of dead wood has potentially important consequences for saproxylic organisms (Schiegg 2000). Consumer diversity varies predictably with the amount of dead wood at both the tree-and site-level (Lassauce et al., 2011, Bouget et al., 2013, Adams et al., 2023). In particular, the diversity of saproxylic beetles (Coleoptera), which are among the most abundant colonists of dead wood (Kirkendall et al., 2015, Stork et al., 2015, Berkov 2018), generally increases with increasing volume of an individual dead log (Lassauce et al., 2011, Bouget et al., 2013). At the patch scale, a group of damaged trees is likely to be more attractive to saproxylic beetles than a single damaged tree (Nadeau et al., 2015, Haeler et al., 2023). However, the effects of patch size on the diversity of beetles (and other taxa) are unknown specifically in tropical regions (Seibold et al., 2015;Thorn et al., 2018).
The potential positive and negative effects of lightning-caused disturbance vary among consumer taxa. Beetles, termites, and saprotrophic fungi are abundant saproxylic groups in tropical forests that are likely to benefit from patches of damaged trees generated by lightning strikes (Figure 1; Kirkendall et al., 2015, Berkov 2018). By contrast, lightning-caused tree damage should be detrimental for arboreal scavengers due to the loss of physical structure (e.g., tree branches) following a strike. For example, some tropical arboreal ants require living stems and branches to support their carton nests (e.g., Azteca spp.; Lubin et al., 1977). These and other arboreal ants also function as secondary herbivores by harvesting honeydew from sap-sucking insects (trophobionts; Davidson et al., 2003), and rely on tree crown connectivity to access dispersed resources when foraging (Adams et al., 2019). Thus, the loss of foliage, sap flow, and structural complexity within lightning-caused disturbances (Gora et al., 2021;Gora et al., 2023) should result in the localized loss of scavenger species that depend on these resources (Figure 1).
As a tree dies, the shift in primary consumer composition from folivores to xylophages should be accompanied by compositional changes in consumers at higher trophic levels (i.e., via bottom-up effects; Seibold et al., 2016). In temperate pine forests, the initial colonizers of lightning-damaged trees usually are bark beetles (Curculionidae: Scolytinae), and lightning damage plays a particularly important role in their population dynamics (Schmitz 1969, Taylor 1974, Coulson et al., 1983, Coulson et al., 1986). However, the relevance of lightning to tropical beetle diversity (Sharples 1933;Parlato et al., 2020) and higher trophic level consumers remains unknown. As xylophagous and xylomycetophagous consumers initially aggregate on lightningdamaged trees, specialist predators and secondary consumers also should become more abundant (Kriegel et al., 2023). Similar bottom-up effects occur in other dead wood systems (e.g., Seibold et al., 2016), but remain unexplored in tropical forests.
The diversity of organisms occupying dead wood often is influenced by previous colonization via facilitation or inhibition (i.e., priority effects; Weslien et al., 2011, Seibold et al., 2023). The timing of dead wood colonization, and variation in the order of colonization by saproxylic organisms, can potentially alter ecosystem processes such as wood decomposition rates and nutrient cycling (Seibold et al., 2023). In lightning strike sites, the relatively small amount of contact between dead wood and the ground presumably favors rapid beetle colonization over saprotrophic fungal growth, which is often greater when wood resources are in contact with the forest floor (Gora et al., 2019;Yang et al., 2021). In some systems, colonizing saproxylic beetles modify habitat and provide directed dispersal for subsequent colonists like fungi (Atkinson and Equihua-Martinez 1986, Ulyshen 2016, Jacobsen et al., 2017, Seibold et al., 2019) and termites (Little et al., 2012, Clay et al., 2017). Similar priority effects in dead wood systems remain largely unexplored in tropical forests.
The goal of this study was to determine how four common consumer taxa-beetles, Azteca ants, termites, and fungi-vary in abundance over time and with resource availability in lightning strike sites in a lowland tropical forest. We predicted that saproxylic taxa (beetles, termites, and fungi) would be observed more frequently in lightning-damaged trees than in undamaged trees due to the abundance of dead wood resources, and that Azteca ants would show the opposite pattern given their dependence upon living trees and lianas (woody vines) for nest sites, connectivity, and food (Figure 1). We also expected that the frequency of occupancy by all taxa would increase with increasing tree size. We expected that strike sites would attract unique beetle assemblages dominated by saproxylic taxa, and that beetle species richness would increase with increasing tree necromass within a site. Lastly, we predicted that temporal patterns of colonization by saproxylic taxa would reflect priority effects of beetles.
Field work was conducted on Barro Colorado Island (BCI), Panama (9.152°N, 79.846°W) and nearby mainland sites from 2015-2020 (Appendix S1: Figure S1). BCI is a 15 km 2 seasonally moist forest with a wet season spanning May-December. All field work for this project was conducted during the wet season. More detailed information about BCI is provided elsewhere (Croat 1978, Leigh et al., 1996).
We located lightning-damaged trees (Appendix S1: Figure S2) using a combination of cameras, electric field change meters, and ground-based field surveys. The camera-based monitoring system consisted of multiple surveillance cameras mounted on towers extending above the forest canopy on BCI, and at nearby locations that provide a broad view of BCI. We located strike sites in 2014-2018 by isolating lightning flashes recorded by two or more cameras, and triangulating their positions based on camera azimuths and image details (Yanoviak et al., 2017). We also detected lightning strikes in 2018-2019 using waveform data from four electric field change meters in addition to the cameras. Details regarding the specifications and operation of the camera and field change meter systems are provided elsewhere (Yanoviak et al., 2017, Yanoviak et al., 2020, Gora et al., 2021, Gora et al., 2023). We used ground-based observations and drone imagery (DJI Mavic Pro 2, Shenzhen, China; Appendix S1: Figure S2) to survey tree damage at each strike site. Ultimately, 62 strike sites used in this study were located with camera or field change meter systems, and 13 additional strikes were located solely from ground-based surveys (Yanoviak et al., 2017; Appendix S1: Figure S1).
We identified the centrally-struck tree in each strike site and recorded the condition (as percent crown dieback) of every lightning-damaged tree >10 cm diameter at breast height (DBH) within a 45 m radius (Yanoviak et al., 2020). We recorded 1447 damaged trees and 237 killed trees (i.e., those with 100% crown dieback) among the 75 strike sites used in this study (Appendix S1: Figure S3).
We measured the size (DBH) of each damaged tree and noted the presence of four focal taxa: Azteca spp. ants (hereafter, "Azteca"), beetles, fungal fruiting bodies (hereafter, "fungi"), and carton-building termites (hereafter, "termites"). We selected these focal taxa for three reasons. First, they span a broad range of trophic ecologies within and among groups; beetles are functionally diverse, and Azteca alone can act as predators, scavengers, and secondary herbivores. Second, the different focal taxa use trees in different ways. Specifically, many beetles and fungi directly colonize dead wood, whereas Azteca ants and carton-building termites require intact (living) branches and trunks to support their nests. Azteca further depend on living vegetation within their territories for connectivity (Adams et al., 2019), and to support their sap-sucking trophobionts (Davidson et al., 2003). Finally, all four focal taxa are relatively easily surveyed from the forest floor with the naked eye or with binoculars.
The first census (Census 1) of each strike site was conducted 21-105 days post-strike (mean ± SD: 50 ± 34 days; Appendix S1: Table S1). We determined the presence of termites by searching the branches of each tree for conspicuous carton nests or carton trails (Appendix S1: Figure S4). Two genera of termites on BCI (Microcerotermes, Nasutitermes) build nests and trails covered by similar carton material, which is a reliable indicator of colonization. Similarly, Azteca presence was determined by searching for their distinct carton nests (Appendix S1: Figure S4). Beetle presence was determined by searching individual tree trunks for conspicuous beetle activity (e.g., aggregations of frass, sawdust, beetle holes, or adult beetles; Appendix S1: Figure S5). Beetle holes, frass, and sawdust persist for weeks, making their presence a reliable indicator of beetle occupancy (Appendix S1: Figure S5). Lastly, fungal colonization was determined by the presence of fruiting bodies on the trunk and branches of each tree. We did not attempt to determine the presence of fungi in the absence of fruiting bodies (e.g., endophytic fungi). When beetles and fungal fruiting bodies were observed on a given tree, we assumed that both were present in all subsequent censuses; both develop within dead wood, thus confirming their absence is logistically difficult.
We examined each tree in Census 1 to determine the effects of tree condition and tree size on colonization by the focal taxa (Appendix S1: Table S1). Trees in a subset of strike sites (n = 36 sites) were examined again following the same survey protocol in a subsequent census (Census 2) 300-600 days post-strike (mean ± SD: 453 ± 48 days; Appendix S1: Table S1). We used data from Census 2 to quantify successional patterns by measuring the effects of tree death, tree size, and prior organism occupancy on the presence of the four focal taxa.
We compared insect and fungal occupancy in lightning-damaged trees with those in trees having no obvious lightning damage in nearby forest. Specifically, we paired 17 strike locations with 17 control sites in 2018 (n = 12) and 2019 (n = 5; Appendix S1: Figure S3). Control sites were non-randomly selected and individually paired with strike sites based on the size and identity of the central struck tree, proximity to a strike site (mean ± SD: 293 ± 324 m), and the structure of the surrounding forest (i.e., the composition and range of tree sizes; Gora et al., 2021). We compared similar numbers of trees within strike and control sites, and each site in a pair was censused within a few days of each other. We only used data from Census 1 at each site for comparisons because the 2019 strikes lacked a second census (Appendix S1: Table S1).
We also compared beetle communities between a subset of the control and strike sites (Appendix S1: Figure S3). We hung a single Lindgren flight intercept trap in each site to collect beetles at N = 11 site pairs (4 from 2018 and 7 from 2019; Appendix S1: Figure S6 and Table S2). Each trap was hung in the lower crown of the central tree (Appendix S1: Figure S6), and its collection cup was filled with 250 ml of soapy water. Traps were deployed for an average (± SD) of 54 ± 25 days following a strike event, and remained active for 156 ± 25 days post-strike.
Traps within a given strike and control site pair were deployed and removed on the same days, and trap contents were collected every 4 days following deployment. Collected beetles were stored in 95% ethanol and identified using a variety of resources (Appendix S2). All beetles were identified to at least the family level, with 65% identified to genus and 12% identified to species. Hereafter, we refer to these latter groups collectively as morphospecies.
We categorized each collected beetle into one of six feeding guilds (bark/ambrosia, predator, herbivorous, xylophagous, fungivorous, saprophagous) or as "unknown" if its feeding ecology could not be determined (Grimbacher et al., 2007). Guild assignments were based on published data; however, individuals that were not identified below tribe level generally were assigned to the "unknown" group. In all, 390 individuals in 53 morphospecies (12% of collected individuals) were classified as "unknown". The "bark/ambrosia" group included weevils (Curculionidae) in the subfamilies Scolytinae and Platypodinae, and the "predator" group included generalist predators, and those that specialize on bark or ambrosia beetles.
Given that a patch of damaged trees is likely to be more attractive to saproxylic beetles than a single damaged tree (Nadeau et al., 2015, Haeler et al., 2023), we determined whether beetle abundance and taxonomic richness varied predictably with total tree necromass for each of the 11 strike sites where beetles were collected. We calculated aboveground necromass for each damaged tree in strike sites from diameter-based biomass allometry (Chave et al., 2014) and percent crown dieback, as described elsewhere (Gora et al., 2021). Necromass calculations were estimated as the product of crown biomass and crown dieback (Gora et al., 2021), and were based only on tree data from Census 1 because the majority of beetle collections occurred shortly thereafter (Appendix S1: Table S1).
We constructed generalized linear mixed effects models (GLMMs; glmmTMB function in "glmmTMB" package; Brooks et al., 2017) in the R statistical environment (version 4.2.1; R Core Team 2022) with a binomial error distribution for each of the focal taxa (Azteca, beetles, termites, fungi) to determine if tree occupancy in the 75 lightning strike sites was influenced by tree size (DBH) and status (alive or dead). Three different people (EMG, CG, and JCB) collected the bulk of the census data, so we included observer as a random variable in each model. We log-transformed DBH to meet assumptions of normality. Only data from Census 1 were used in each model to determine the influence of tree size and mortality status on the probability of occupancy by each focal taxon (Appendix S1: Table S1). To more clearly visualize the results apart from model output, we binned tree DBH into three size classes (10-30 cm, 30-60 cm, >60 cm; Yanoviak et al., 2020) and plotted mean probabilities of occurrence along with 95% confidence intervals.
Successional patterns in lightning-damaged trees of 36 strike sites were analyzed using GLMMs for each of the four focal taxa (Appendix S1: Table S1). We used tree size (DBH, logtransformed) and tree location (strike or control) in Census 2 as predictors in the models. We ran two separate binomial GLMMs for each focal taxon, one with "arrival" as the response variable and the other with "departure" as the response variable. "Arrival" indicated cases where trees lacking the taxon in Census 1 gained it in Census 2, and "departure" indicated whether a taxon present in Census 1 was subsequently absent in Census 2. To determine whether Azteca and termites tended to co-occur, we included a term for the presence of Azteca in Census 2 in the termite model, and vice versa. Similarly, to evaluate patterns in the arrival order of termites, beetles, and fungal fruiting bodies, we included terms for presence of all taxa during Census 2 in each model (e.g., the model for beetle arrival included terms for fungi presence and termite presence in Census 2). We used bidirectional stepwise selection to reduce the models to their final form based on differences in AIC values > 2. We also plotted rarefaction curves using presence/absence data to visualize how colonization (and abandonment) of trees occurred among each focal taxon over time post-strike.
We compared lightning-damaged trees to control trees in 17 strike and 17 control sites (Appendix S1: Figure S3) using generalized linear mixed effect models (GLMMs) for each focal taxon. We used occupant presence as the response variable and tree location (strike or control) as the predictor variable for each model. We included site pair as a random variable to account for differences across years and forest locations. We also included observer identity as a random variable to account for potential observer differences across censuses. Because some strikes and controls in 2019 were only censused once, we only made comparisons between strike and control sites using Census 1 (Appendix S1: Table S1).
Beetle assemblage composition was compared between 11 control and 11 strike sites (Appendix S1: Figure S3) using PERMANOVA (PRIMER Version 7.0.21). We included site pair as a nested random effect to account for potential site-specific differences. We used logtransformed abundance data for each morphospecies and constructed a species-similarity matrix using the Bray-Curtis dissimilarity method. P-values were obtained using 9999 random permutations. We used PERMDISP to test for differences in compositional beta diversity and to determine if potential differences among sites were driven by dispersion rather than differences in composition alone. We evaluated whether certain beetle morphospecies were associated with control or strike sites using species indicator analysis. We used the multipatt function in R ("indicspecies" package; Cáceres and Legendre 2009) to calculate indicator values for each morphospecies in the treatment groups (i.e., strike or control). Due to the high frequency of taxa with low abundance, we restricted our analyses to morphospecies with ≥ 50 individuals. This design allowed us to statistically control for multiple comparisons (Cáceres et al., 2010) while simultaneously increasing statistical power to detect significant indicator taxa. Significance was determined from 999 random permutations. We used the specaccum function in R ("vegan" package; Oksanen et al., 2022) to construct rarefaction curves to determine how beetle morphospecies richness differed between the 11 lightning strike and 11 control sites (Appendix S1: Figure S3). Due to the high frequency of zero data (i.e., no beetles caught during a 4-day sampling period), we used hurdle models with a gamma distribution to compare beetle abundance and morphospecies richness between strike and control sites (glmmTMB function in "glmmTMB" package; Brooks et al., 2017). Hurdle models simultaneously test the effects of predictor variables on the binary process of having nonzero values occur (binary response) and on the continuous processes underlying the non-zero data. We constructed two hurdle models with morphospecies richness day -1 and abundance day -1 (hereafter referred to as morphospecies richness and abundance) as the response variables, and included tree location (strike site or control site), days post-strike, and log-transformed DBH as fixed effects. We also included site pair as a random effect to account for site-specific variation.
For both the morphospecies richness and abundance models, we obtained output for a zero-inflated model (the binary process) and a conditional model (the continuous process), resulting in a total of four model outputs.
We constructed two hurdle models with a gamma data distribution to determine how beetle abundance and morphospecies richness were influenced by total site necromass in the 11 lightning strike sites (Appendix S1: Figure S3, Table S1). For both the abundance and morphospecies richness models, we included total site necromass (kg) as a fixed effect and site as a random effect. We used the same model parameters for both the zero-inflated and the conditional models within both the overall abundance and morphospecies richness models. All statistical analyses, with the exception of the PERMANOVA and PERMDISP tests (PRIMER Version 7.0.21), were performed in the R statistical environment version 4.2.1 (R Core Team 2022).
Fungi, termites, Azteca, and beetles all colonized larger trees more frequently than smaller trees in strike sites (Figure 2; Appendix S1: Table S3) in Census 1. Fungi, termites, and beetles had higher occurrence probabilities in dead trees (Figure 2; Appendix S1: Table S3), and tree mortality had no significant effect on Azteca occurrence within strike sites (Z = -1.58, P = 0.12; Figure 2).
Generally, Azteca and termites were the first observed occupants of surveyed trees, followed by beetles and fungal fruiting bodies (Figure 3). The probability of fungal arrival was higher in dead trees (Appendix S1: Table S4 and Figure S7), and was positively associated with prior beetle colonization (Z = 2.83, p = 0.005). Azteca arrival and departure were not associated with any tested predictors, including prior termite colonization (Appendix S1: Table S4). The probability of termite arrival was higher for larger trees, but termite departure was not associated with any of the tested predictors (Appendix S1: Table S4 and Figure S7). Dead trees had slightly higher beetle arrival probabilities (Z = 1.56, p = 0.12; Figure S7), but the trend was not significant. Termite presence had no significant effect on the probability of beetle arrival (Z = 1.31, p = 0.18).
Trees in strike sites were 3.8 times more likely to be occupied by beetles and 12.2 times more likely to be occupied by fungi compared to trees in control sites in Census 1 (Figure 4; Appendix S1: Table S5). There was no significant difference in termite and Azteca occupancy between strike and control sites (Z = 1.02, p = 0.31, and Z = -0.93, p = 0.35, respectively; Figure 4). Overall, termites were the most frequently encountered focal taxon; trees in both strike and control sites had > 20% probability of containing termites (Figure 4). Lindgren flight intercept traps in lightning strike sites caught 10 times more beetles than traps hung in control sites (Figure 5; Appendix S1: Figure S8), with bark/ambrosia beetles (Scolytinae, Platypodinae) being disproportionately represented in strike sites initially (Figure 6, Appendix S1: Table S6; Appendix S2: Table S1). Higher-level consumers (e.g., predators) increased in abundance over time, with a slight lag time in accumulation when compared to their prey (Figure 6). Beetles with unknown feeding ecologies were the third most abundant guild (Figure 6). Lightning strike sites also accumulated unique morphospecies at a faster rate than control sites (Figure 7).
The composition of beetles captured in Lindgren traps in lightning strike sites differed from that of control sites at the morphospecies level (Psuedo-F1,22 = 2.26, p = 0.001; Figure 8) and the family level (Psuedo-F1,22 = 3.10, p = 0.003; Appendix S1: Figure S9). These differences were not driven by differences in beta diversity (i.e., dispersion of the data; morphospecies level: F1,22 = 0.43, p = 0.56; family level: F1,22 = 3.86, p = 0.11). Four morphospecies of beetles were statistically associated with lightning strike sites (Appendix S1: Table S7), while none was associated with control sites. Three of the indicator morphospecies were bark/ambrosia beetles and one was a bark/ambrosia beetle specialist predator (Trypanaeus sp.; Appendix S1: Table S7).
Beetle abundance in Lindgren traps was higher in lightning strike sites vs. control sites for both the zero inflated and the conditional model (Figure 5; Appendix S1: Figure S8, Table S8). Days post-strike had a significant effect on beetle abundance (Appendix S1: Table S8). Specifically, beetle abundance was highly variable among strike sites, but generally increased up to 90 days post-strike before decreasing over the remainder of the sampling period (Figure 5; Appendix S1: Figure S8). By contrast, beetle abundance was consistently low in control sites (Figure 5). Beetle morphospecies richness in Lindgren traps was higher in strike sites than control sites (Figure 7), and increased with days post-strike (Appendix S1: Table S9).
Estimates of strike site total necromass in Census 1 ranged from 29 to 6646 kg (Appendix S1: Figures S10 and S11). Neither beetle abundance nor morphospecies richness varied predictably with total site necromass (Appendix S1: Tables S10 and S11).
Lightning is an important agent of disturbance in many tropical forests (Furtado 1935, Anderson 1964, Magnusson et al., 1996, Yanoviak et al., 2020, Gora et al., 2021), but the role of lightning-damaged trees in consumer dynamics is relatively poorly understood in this setting (e.g., Parlato et al., 2020). Here, we show that consumers spanning various taxonomic groups and ecological roles tend to aggregate in lightning-damaged patches of forest, and follow predictable patterns of colonization of lightning-damaged trees. Collectively, the results suggest that lightning-caused disturbance is an important contributor to consumer metacommunity and metapopulation dynamics in lowland tropical forests.
The increased probability of occupancy by saproxylic organisms with increasing size of dead trees further suggests that large dead trees are particularly relevant to the maintenance of consumer diversity. Large trees harbor the bulk of forest biomass (Lutz et al., 2018), and large dead trees are important resources for saproxylic species in some forests (e.g., Similä et al. 2003), but the role of tree size in shaping tropical forest food webs remains poorly understood.
Lightning accounts for >40% of large-tree death in lowland tropical forest in Panama (Yanoviak et al., 2020), thus lightning-caused disturbance presumably plays an important ecological role for any taxa that depend on large dead trees in this setting.
The lack of difference in the occurrence frequency of Azteca and termites between control and strike sites was not expected. Damage to trees and lianas from lightning (Gora et al., 2021) presumably eliminates the availability of, and access to, local honeydew resources for Azteca. However, the lack of consistent patterns of nest abandonment or relocation by Azteca suggests that they are resistant to disturbance at the spatiotemporal scales associated with lightning strikes. This remains to be studied. Azteca and termite nest abandonment is relatively infrequent (Lubin et al., 1977), and both groups presumably withstand localized disturbances via a variety of behavioral mechanisms (e.g., long-range foraging, polydomous colonies; Roisin et al., 2006, Adams et al., 2019), potentially explaining the lack of a conspicuous lightning effect.
The more frequent occurrence of fungal fruiting bodies in larger, lightning-damaged trees observed in this study is consistent with similar studies conducted in temperate forests (Vasiliauskas 2001, Heilmann-Clausen andChristensen 2004). Lightning-struck trees provide diverse, vertically variable substrates for fungal colonizers. Such vertical heterogeneity affects fungal diversity and composition (Yang et al., 2021), and has consequences for broader microbial community structure and wood decomposition rates (Gora et al. 2019). However, studies comparing fungal assemblages and decomposition rates among structurally different disturbance types in tropical forests (e.g., lightning vs. windthrow) are lacking.
The successional trajectories of organisms colonizing dead wood often exhibit strong priority effects (Weslien et al., 2011), and the tendency for beetle occupancy to precede the appearance of fungal fruiting bodies in this study suggests facilitation effects of beetle activity.
Many groups of beetles, specifically ambrosia beetles (65% of beetles collected in this study), vector ascomycetous and basidiomycetous fungi (Hsiau and Harrington 2003). Many saproxylic beetles also fragment inner portions of the wood, exposing previously inaccessible substrates.
This targeted dispersal and wood fragmentation by saproxylic beetles creates a clear pathway for subsequent fungal colonization (Ulyshen 2016). A comparison of colonization patterns among similar disturbances that generate dead wood (e.g., windthrow) was beyond the scope of this study, but likely would clarify the relative importance of lightning in structuring saproxylic consumer communities.
Differences in the composition of beetles collected in strike and control sites coupled with the indicator species results suggest that lightning-caused disturbance in tropical forests plays an important role in maintaining populations of bark/ambrosia beetles, consumers at higher trophic levels (e.g., bark/ambrosia specialist predators), and other saproxylic taxa (e.g., xylophages). The observed variation in beetle diversity over time likely reflects time lags in aggregation for certain groups (i.e., bark/ambrosia beetles; Økland et al., 2005) and dead wood becoming less attractive to abundant, early colonists as it ages (Ulyshen 2016). The relatively synchronous colonization by bark/ambrosia beetles and predators observed here is consistent with the results of similar studies (e.g., Wermelinger et al., 2013, Weed et al., 2016).
The specific physical structure of lightning-damaged trees potentially enhances consumer diversity relative to other types of forest disturbance (e.g., windthrow) that generate similar volumes of dead wood. Specifically, dead wood from windthrow lacks the vertical structure typical of lightning-caused disturbance (Li et al., 2017, Lettenmaier et al., 2022), and many neotropical beetles exhibit vertical specialization in resource use (Berkov 2018). In general, lightning strike sites should be especially attractive to organisms that rely on dead wood resources that are larger (Grove 2002), standing (Graf et al., 2021), or located in the canopy (Roisin et al., 2006, Ulyshen 2011, Berkov 2018). Understanding the relevance of such factors will require experimetal and comparative data from windthrow and other types of disturbance. Contrary to our expectations, dead wood quantity (necromass) at the site level had no effect on beetle morphospecies richness and abundance. The lack of a strong relationship between necromass and beetle diversity at a site is likely due to low sampling intensity. Certain longhorn beetles (Coleoptera: Cerambycidae) of lowland forest in Panama tend to colonize lessdense host wood, including one of the species damaged by lightning in this study (Gustavia superba; Lanuza-Garay and Barrios 2018, Torres et al., 2023), but patterns for most other beetle groups remain unknown. Future studies could be improved by including more precise measurements of dead wood quantities than were attempted here, or by rearing beetles from bait wood (Lachat et al., 2006, Lee et al., 2014). Understanding consumer diversity patterns in relation to local dead wood mass and wood traits will become increasingly important given expected increases in lightning frequency (Harel and Price 2020), and its associated effects on forest dynamics (Gora et al., 2021).
Data gathered for this study did not allow for analysis of differences in consumer dynamics among tree species. Whereas trees show interspecific differences in response to lightning strikes (Richards et al., 2022), the effect of host tree identity on beetle colonization post-strike remains unknown. Many saproxylic beetles exhibit high specificity for dead wood resources (Li et al., 2017, Torres et al., 2023), and it is likely that tree identity also plays an important role in beetle colonization following a lightning strike. Moreover, some beetles found in the lightning strike sites could not be assigned to a feeding guild due to lack of published information, suggesting that taxa with previously unknown ecologies either are saproxylic or are otherwise attracted to lightning-caused disturbance. Additional basic natural history data could significantly improve our understanding of the importance of tree identity in particular, and lightning-generated dead wood in general, for saproxylic beetles and other taxa in tropical forests.
We note that the timing of Census 2-which was used to quantify consumer succcessional patterns-varied among strike sites. As such, we could not quantify fine-scale succesional patterns given the range of censusing dates. The exact timing of consumer colonization lightning-damaged trees also remians unknown. Although the lightning monitoring system provides accurate locations of strikes (<12 m error in 95% of the study site; Yanoviak et al., 2017) within days of the event, the local distribution of lightning damage generally is not apparent until many weeks following a strike. Similarly, the effects of tree occupancy prior to a strike are not known. These and other important knowledge gaps could be filled via experimental manipulation of lightning strike locations, although this approach is not yet technically feasible.
Finally, understanding the dynamics of consumers associated with dead wood is relevant to understanding potential changes in ecosystem-level processes, especially carbon and nutrient cycling (Edburg et al., 2012, Siegert et al., 2018). Such processes are likely sensitive to predicted increases in the frequency and intensity of storm-related disturbances in the tropics, especially lightning and windthrow (e.g., Harel andPrice 2020, Gora et al., 2021;Feng et al., 2023). Linking consumer dynamics to larger scale patterns following disturbance will become increasingly important as tropical forests begin to shift from carbon sinks to sources (Mitchard 2018). Figure 1. Successional patterns in lightning-damaged trees favor saproxylic taxa over time. Conceptual depiction of tree occupancy by beetles, fungi, carton-building termites, and Azteca ants in (presumably healthy) trees before a lightning strike (a), two months post-strike (b), and 15 months post-strike (c). Silhouettes represent the relative abundance of each taxon. Created with BioRender.com by Kane A. Lawhorn. CI) in lightning-damaged trees during Census 1 as a function of tree size (as DBH; diameter at breast height) and Status (alive, dead). Created with BioRender.com and R by Kane A. Lawhorn and Jeannine H. Richards. were assumed to continue to occupy a tree if their occurrence was documented once, whereas termites and Azteca conspicuously colonized and vacated trees over the observation period. Census 1. Asterisks (*) denote significant differences between means within a taxon. Created with BioRender.com and R by Kane A. Lawhorn and Jeannine H. Richards. represent local regression fitting and shading indicates 95% confidence intervals. Bark/ambrosia beetles were excluded from the xylophagous and fungivorous guild calculations. Unknown = individuals with uncertain ecology that were not assigned to a feeding guild. number of sites sampled. Shading indicates 95% confidence intervals.