One size does not fit all: Customizing MCMC methods for hierarchical models using NIMBLE

Improved efficiency of Markov chain Monte Carlo facilitates all aspects of statistical analysis with Bayesian hierarchical models. Identifying strategies to improve MCMC performance is becoming increasingly crucial as the complexity of models, and the run times to fit them, increases. We evaluate different strategies for improving MCMC efficiency using the open‐source software NIMBLE (R package nimble) using common ecological models of species occurrence and abundance as examples. We ask how MCMC efficiency depends on model formulation, model size, data, and sampling strategy. For multiseason and/or multispecies occupancy models and for N‐mixture models, we compare the efficiency of sampling discrete latent states vs. integrating over them, including more vs. fewer hierarchical model components, and univariate vs. block‐sampling methods. We include the common MCMC tool JAGS in comparisons. For simple models, there is little practical difference between computational approaches. As model complexity increases, there are strong interactions between model formulation and sampling strategy on MCMC efficiency. There is no one‐size‐fits‐all best strategy, but rather problem‐specific best strategies related to model structure and type. In all but the simplest cases, NIMBLE's default or customized performance achieves much higher efficiency than JAGS. In the two most complex examples, NIMBLE was 10–12 times more efficient than JAGS. We find NIMBLE is a valuable tool for many ecologists utilizing Bayesian inference, particularly for complex models where JAGS is prohibitively slow. Our results highlight the need for more guidelines and customizable approaches to fit hierarchical models to ensure practitioners can make the most of occupancy and other hierarchical models. By implementing model‐generic MCMC procedures in open‐source software, including the NIMBLE extensions for integrating over latent states (implemented in the R package nimbleEcology), we have made progress toward this aim.     

Interspecific variation and elevated CO2 influence the relationship between plant chemical resistance and regrowth tolerance

To understand how comprehensive plant defense phenotypes will respond to global change, we investigated the legacy effects of elevated CO₂ on the relationships between chemical resistance (constitutive and induced via mechanical damage) and regrowth tolerance in four milkweed species (Asclepias). We quantified potential resistance and tolerance trade‐offs at the physiological level following simulated mowing, which are relevant to milkweed ecology and conservation. We examined the legacy effects of elevated CO₂ on four hypothesized trade‐offs between the following: (a) plant growth rate and constitutive chemical resistance (foliar cardenolide concentrations), (b) plant growth rate and mechanically induced chemical resistance, (c) constitutive resistance and regrowth tolerance, and (d) regrowth tolerance and mechanically induced resistance. We observed support for one trade‐off between plant regrowth tolerance and mechanically induced resistance traits that was, surprisingly, independent of CO₂ exposure. Across milkweed species, mechanically induced resistance increased by 28% in those plants previously exposed to elevated CO_2. In contrast, constitutive resistance and the diversity of mechanically induced chemical resistance traits declined in response to elevated CO₂ in two out of four milkweed species. Finally, previous exposure to elevated CO₂ uncoupled the positive relationship between plant growth rate and regrowth tolerance following damage. Our data highlight the complex and dynamic nature of plant defense phenotypes under environmental change and question the generality of physiologically based defense trade‐offs.     

Selection for outbreeding in Varroa parasitising resistant honey bee (Apis mellifera) colonies

Parasitism is expected to select for counter‐adaptations in the host: driving a coevolutionary arms race. However, human interference between honey bees (Apis mellifera) and Varroa mites removes the effect of natural selection and restricts the evolution of host counter‐adaptations. With full‐sibling mating common among Varroa, this can rapidly select for virulent, highly inbred, Varroa populations. We investigated how the evolution of host resistance could affect the infesting population of Varroa mites. We screened a Varroa‐resistant honey bee population near Toulouse, France, for a Varroa resistance trait: the inhibition of Varroa's reproduction in drone pupae. We then genotyped Varroa which had co‐infested a cell using microsatellites. Across all resistant honey bee colonies, Varroa's reproductive success was significantly higher in co‐infested cells but the distribution of Varroa between singly and multiply infested cells was not different from random. While there was a trend for increased reproductive success when Varroa of differing haplotypes co‐infested a cell, this was not significant. This suggests local mate competition, through the presence of another Varroa foundress in a pupal cell, may be enough to help Varroa overcome host resistance traits; with a critical mass of infesting Varroa overwhelming host resistance. However, the fitness trade‐offs associated with preferentially co‐infesting cells may be too high for Varroa to evolve a mechanism to identify already‐infested cells. The increased reproductive success of Varroa when co‐infesting resistant pupal cells may act as a release valve on the selective pressure for the evolution of counter resistance traits: helping to maintain a stable host–parasite relationship.     

Allelopathy and its coevolutionary implications between native and non‐native neighbors of invasive Cynara cardunculus L.

Invasive plants apply new selection pressures on neighbor plant species by different means including allelopathy. Recent evidence shows allelopathy functions as remarkably influential mediator for invaders to be successful in their invaded range. However, few studies have determined whether native and non‐native species co‐occurring with invaders have evolved tolerance to allelopathy. In this study, we conducted germination and growth experiments to evaluate whether co‐occurring native Juncus pallidus and non‐native Lolium rigidum species may evolve tolerance to the allelochemicals induced by Cyanara cardunculus in Australian agricultural fields. The test species were germinated and grown in pots filled with collected invaded and uninvaded rhizosphere soil of C. cardunculus with and without activated carbon (AC). Additionally, a separate experiment was done to differentiate the direct effects of AC on the test species. The soil properties showed invaded rhizosphere soils had higher total phenolic and lower pH compared with uninvaded soils. We found significant reduction of germination percentage and seedling growth in terms of above‐ and belowground biomass, and maximum plant height and root length of native in the invaded rhizosphere soil of C. cardunculus, but little effect on non‐native grass species. Even soil manipulated with AC showed no significant differences in the measured parameters of non‐native except aboveground biomass. Taken together, the results indicate allelochemicals induced by C. cardunculus exert more suppressive effects on native than non‐native linking the coevolved tolerance of those.     

Gut microbiomes of sympatric Amazonian wood‐eating catfishes (Loricariidae) reflect host identity and little role in wood digestion

Neotropical wood‐eating catfishes (family Loricariidae) can occur in diverse assemblages with multiple genera and species feeding on the same woody detritus. As such, they present an intriguing system in which to examine the influence of host species identity on the vertebrate gut microbiome as well as to determine the potential role of gut bacteria in wood digestion. We characterized the gut microbiome of two co‐occurring catfish genera and four species: Panaqolus albomaculatus, Panaqolus gnomus, Panaqolus nocturnus, and Panaque bathyphilus, as well as that of submerged wood on which they feed. The gut bacterial community did not significantly vary across three gut regions (proximal, mid, distal) for any catfish species, although interspecific variation in the gut microbiome was significant, with magnitude of interspecific difference generally reflecting host phylogenetic proximity. Further, the gut microbiome of each species was significantly different to that present on the submerged wood. Inferring the genomic potential of the gut microbiome revealed that the majority of wood digesting pathways were at best equivalent to and more often depleted or nonexistent within the catfish gut compared to the submerged wood, suggesting a minimal role for the gut microbiome in wood digestion. Rather, these fishes are more likely reliant on fiber degradation performed by microbes in the environment, with their gut microbiome determined more by host identity and phylogenetic history.     

Large‐scale patterns of seed removal by small mammals differ between areas of low‐ versus high‐wolf occupancy

Because most tree species recruit from seeds, seed predation by small‐mammal granivores may be important for determining plant distribution and regeneration in forests. Despite the importance of seed predation, large‐scale patterns of small‐mammal granivory are often highly variable and thus difficult to predict. We hypothesize distributions of apex predators can create large‐scale variation in the distribution and abundance of mesopredators that consume small mammals, creating predictable areas of high and low granivory. For example, because gray wolf (Canis lupus) territories are characterized by relatively less use by coyotes (C. latrans) and greater use by foxes (Vulpes vulpes, Urocyon cinereoargentus) that consume a greater proportion of small mammals, wolf territories may be areas of reduced small‐mammal granivory. Using large‐scale, multiyear field trials at 22 sites with high‐ and low‐wolf occupancy in northern Wisconsin, we evaluated whether removal of seeds of four tree species was lower in wolf territories. Consistent with the hypothesized consequences of wolf occupancy, seed removal of three species was more than 25% lower in high‐wolf‐occupancy areas across 2 years and small‐mammal abundance was more than 40% lower in high‐wolf areas during one of two study years. These significant results, in conjunction with evidence of seed consumption in situ and the absence of significant habitat differences between high‐ and low‐wolf areas, suggest that top‐down effects of wolves on small‐mammal granivory and seed survival may occur. Understanding how interactions among carnivores create spatial patterns in interactions among lower trophic levels may allow for more accurate predictions of large‐scale patterns in seed survival and forest composition.     

Examining differences in phylogenetic composition enhances understanding of the phylogenetic structure of the shrub community in the northeastern Qinghai‐Tibetan Plateau

Periodic climatic oscillations and species dispersal during the postglacial period are two important causes of plant assemblage and distribution on the Qinghai‐Tibet Plateau (QTP). To improve our understanding of the bio‐geological histories of shrub communities on the QTP, we tested two hypotheses. First, the intensity of climatic oscillations played a filtering role during community structuring. Second, species dispersal during the postglacial period contributed to the recovery of species and phylogenetic diversity and the emergence of phylogenetic overdispersion. To test these hypotheses, we investigated and compared the shrub communities in the alpine and desert habitats of the northeastern QTP. Notably, we observed higher levels of species and phylogenetic diversity in the alpine habitat than in the desert habitat, leading to phylogenetic overdispersion in the alpine shrub communities versus phylogenetic clustering in the desert shrub communities. This phylogenetic overdispersion increased with greater climate anomalies. These results suggest that (a) although climate anomalies strongly affect shrub communities, these phenomena do not act as a filter for shrub community structuring, and (b) species dispersal increases phylogenetic diversity and overdispersion in a community. Moreover, our investigation of the phylogenetic community composition revealed a larger number of plant clades in the alpine shrub communities than in the desert shrub communities, which provided insights into plant clade‐level differences in the phylogenetic structures of alpine and desert shrub communities in the northeastern QTP.