Discussion
Over a period of 36 years, our results indicated few changes to the overall bird community with respect to species abundances or point alpha diversity, although we did observe a decline in overall community biomass. We find these results somewhat surprising considering the plot is located in mature floodplain forest bordering a dynamic whitewater river. For example, in spite of large shifts of the Manu River, there were few changes to species associated with primary successional habitat. Our findings, contrary to other studies in similar locations throughout the Neotropics, suggest that community dynamics and composition in this area are mostly determined by local processes, and that neither local nor regional environmental changes have affected the majority of the bird community. Perhaps most remarkable is the stability in the insectivore community, excepting canopy flocks. Given extensive documentation of the sensitivity of insectivores to forest disturbance and fragmentation, it is somewhat surprising that a group that responds precipitously to perturbations(Blake & Loiselle 2015; Stouffer et al. 2020) shows such little natural variation in undisturbed forest.
In spite of the striking similarity in community composition between the two time periods, we did find changes in some species’ abundances. We found a drastic decrease in Cacicus cela, likely stemming from a loss of a single nesting tree, a decrease in canopy flocks (Fig. 3), and a significant decrease in some bamboo specialists consistent with the complete disappearance of bamboo in the plot after a flowering event circa 2001. One guild that showed a decline in abundance (nectivores) had low sample sizes that led to large variance in model estimations of their abundance. Granivores also showed declines; however, as many of these (e.g., parrots) cue in on patchy ephemeral resources, we speculate that this may be driven by stochastic fluctuations in local resources (e.g., parrots often respond to seasonal changes in fruiting activity on scales much larger than the plot), whereas others (i.e.,Odontophorus stellatus ) may be due to inadequate sampling of group sizes. Some ground dwelling granivores show large scale fluctuations that may be susceptible to nest predation by white-lipped peccaries although this species was absent during both the 1982 and 2018 census periods (Terborgh et al. 1990). The fact that we identified other outlier species when evaluating changes in abundance between the two time periods within ecological guilds merits further exploration, and may be targets for future population dynamics studies.
The strong 1:1 relation between the overall abundances then and now (see Fig. 2A) was not mirrored in biomass. Indeed, after correcting for phylogeny, the biomass totals per species between the two time periods showed a less than 1:1 relationship. Bird species in the lower biomass-per-individual quantiles decreased in abundance, whereas bird species in the upper-biomass-per individual quantiles increased in abundance (Fig. 2B). Additionally, we speculate that some of the small differences in group size estimates for larger species can lead to substantially different biomass estimates. For example, there was a particularly large decline in Cacicus cela . Overall, our results could also suggest possible changes to some habitats as drivers for the few observed changes.
A curious contrast appears in the apparent differences between understory and canopy flocks. Previous research has documented the unique interdependence of these permanent associations, which would suggest that the assemblage of species in these flocks would experience similar trajectories of growth or decline (Munn & Terborgh 1979). Although there is not a clear explanation for the decline in canopy flock densities, this decline is rather strong (see Fig. 3). Because more fruits are found in the canopy, shifts in phenology and the dependent arthropod assemblage therein, may have a greater effect on the omnivores, frugivores and insectivores common to canopy flocks. On a regional scale, there is evidence suggesting vegetation in the canopy may be sensitive to changes in precipitation patterns (Hilker et al. 2014) and that southern Amazonia has experienced a lengthened dry season since 2000 (Marengo et al. 2018). Interpolated precipitation patterns, however, do not seem to have changed in Cocha Cashu during the last 30 years (Fig. S8) (Aybar et al. 2017). Furthermore, the extent to which canopy birds are affected by changes in canopy tree fruiting phenology at the local level at Cocha Cashu is unknown. Evaluating changes in forest phenology would help us evaluate this mechanism as a potential driver of change to canopy species.
Studies in other lowland tropical sites showed marked changes to specific groups in terms of abundances. Blake and Loiselle report long term declines for multiple guilds of understory species over ~ 8-year period from a remote site in Ecuador (Blake & Loiselle 2015). In the absence of any anthropogenic impact, changing regional precipitation events are one of the remaining plausible explanations. Consistent with the results of Blake and Loiselle (2015), declines in several species over 20 years have been confirmed from a remote forest plot in Panama (Brawn et al. 2017). This lowland forest plot has also experienced little to no direct human impact but has undergone species declines related to prolonged seasons of low precipitation. More recently, the same research team has used long-term mist-net data for a larger group of species and found consistent declines in capture rates across the majority of bird species (Pollock et al. unpublished data). In Central Brazil, terrestrial insectivores have steadily declined from undisturbed forests, partially driving the overall shift of the community from a historical baseline (Stoufferet al. 2020).
Human impacts notwithstanding, we find it surprising that the Cocha Cashu site shows little change overall compared to other locally unaffected tierra firme forest plots, given that Cocha Cashu is a mature floodplain plot that experiences succession due to changes in the course of the Manu River. Data on long-term regional environmental variation (i.e., precipitation patterns) at the Cocha Cashu site do not show strong consistent changes that would influence community composition (Figure S8) (Aybar et al. 2017). Given the observed bird community declines in other Neotropical sites that are attributed to changes in rainfall patterns, bird communities experiencing little to no change, such as those in Cocha Cashu, may be good candidates for biodiversity refugia in the face of such global change (Guisan et al. 2013; Morelli et al. 2020).
Unfortunately, because our sampling focused on a snapshot of the ecological process in two time periods, we were not able to evaluate the role of random year-to-year variation on community stabililty/resilience per-se. Given the small changes we found in abundance and territory distribution, however, we predict that these communities are composed of species with low population variability and thus collectively, are in fact stable (sensu ecological theory) (MacArthur 1955; Ives & Carpenter 2007). Long-term temporal population variability has previously been demonstrated for specific components of insectivorous bird communities in several locations. Loiselle and Blake (1992), for example, suggested that community dynamics of Neotropical bird communities are likely guild-dependent, with insectivores showing much less population variability over time compared to frugivores. However, our data extend beyond specific groups of insectivorous birds and show some level of constancy across a far larger percentage of the community irrespective of guild or functional traits.
Ultimately, we cannot speak to the mechanistic processes driving the similarity in community structure that we observed. Evaluating subtle changes to forest structure will be critical in understanding whether the bird community is 1) resilient to local or regional environmental change, or 2) whether overall changes to the forest have been so small that they translate into small effects on the bird community. The first two scenarios would underscore different aspects of community stability that maintain community equilibrium (Ives et al. 2003; Ives & Carpenter 2007). Understanding the likelihood of each these scenarios is critical for understanding how resilient tropical bird communities are to natural environmental dynamics and will allow us to decouple the effects of local processes from those of larger regional scale processes, including those linked to climate change.
A major limitation of our study compared to the approaches of the studies mentioned above is our lack of time-series data (Ives et al. 2003; Robinson & Curtis 2020). No effort has been focused on evaluating bird communities at regular time intervals in Cocha Cashu between the two time periods. This lack of continuity and time-series data-gathering reflects funding challenges for complex, large scale censuses at this remote locale. However, the intense behavior observation efforts that we carried out may allow us to design new analytical approaches that efficiently combine a diverse array of sampling techniques that will hopefully allow for longer-term sustainable sampling of neotropical bird communities (Toms et al.2006; Gomez et al. 2018). For example, it is well-known that point count techniques yield largely unreliable density estimates of rarer species in hyper-diverse bird communities (Robinson et al.2018), yet this problem is now largely surmountable if point counts are coupled with behavioral field observations thanks to the recent development of large families of statistical models (Sauer & Link 2002; Yamaura et al. 2012, 2016; Gomez et al. 2018). Combining geo-referencing of the movement and territory occupancy in obligatory understory flocking species with the spot-mapped territories allowed us to calibrate the density estimates obtained from additional point counts that we carried in the same year, and we found a surprising degree of accuracy among the two techniques (unpublished data). We suggest that using supplementary visual group size data to estimate group means for monospecific flocking species, which are normally excluded from point count estimates, would permit a larger percentage of species to be tracked with point counts than previously thought possible. These combined methodologies, together with the increasing availability of inexpensive data loggers, automatic recorders and camera traps, allows for far more financially feasible rapid efficient sustainable sampling (Burton et al. 2015; Shonfield & Bayne 2017). For future fieldwork efforts here and elsewhere, we strongly advocate for a more intimate design and combination of field and statistical sampling techniques in situ by inter-disciplinary teams.
In contrast to most similar studies, which have found profound impacts of climate change on Neotropical bird communities, we report few significant abundance changes in most bird species in this hyper-diverse community. This pattern of little change seems to be a peculiarity of our sampling study; however, to date there is no easy way to directly compare currently available data in order to understand continental-scale biodiversity changes.
The region of Madre de Dios within which Manu National Park is situated is facing ever-mounting pressure from extractive industries and road construction to increase commerce between Brazil and Peru. Given the current focus on determining and predicting refugia in the face of future environmental change, our results suggest that the consequences of development around such sites should be another consideration among stakeholders within the region.