Interpreting random forest analysis of ecological models to move from prediction to explanation.

As modeling tools and approaches become more advanced, ecological models are becoming more complex. Traditional sensitivity analyses can struggle to identify the nonlinearities and interactions emergent from such complexity, especially across broad swaths of parameter space. This limits understanding of the ecological mechanisms underlying model behavior. Machine learning approaches are a potential answer to this issue, given their predictive ability when applied to complex large datasets. While perceptions that machine learning is a "black box" linger, we seek to illuminate its interpretive potential in ecological modeling. To do so, we detail our process of applying random forests to complex model dynamics to produce both high predictive accuracy and elucidate the ecological mechanisms driving our predictions. Specifically, we employ an empirically rooted ontogenetically stage-structured consumer-resource simulation model. Using simulation parameters as feature inputs and simulation output as dependent variables in our random forests, we extended feature analyses into a simple graphical analysis from which we reduced model behavior to three core ecological mechanisms. These ecological mechanisms reveal the complex interactions between internal plant demography and trophic allocation driving community dynamics while preserving the predictive accuracy achieved by our random forests.

A bioenergetic framework for aboveground terrestrial food webs.

Bioenergetic approaches have been greatly influential for understanding community functioning and stability and predicting effects of environmental changes on biodiversity. These approaches use allometric relationships to establish species' trophic interactions and consumption rates and have been successfully applied to aquatic ecosystems. Terrestrial ecosystems, where body mass is less predictive of plant-consumer interactions, present inherent challenges that these models have yet to meet. Here, we discuss the processes governing terrestrial plant-consumer interactions and develop a bioenergetic framework integrating those processes. Our framework integrates bioenergetics specific to terrestrial plants and their consumers within a food web approach while also considering mutualistic interactions. Such a framework is poised to advance our understanding of terrestrial food webs and to predict their responses to environmental changes.

Phenology and flowering overlap drive specialisation in plant-pollinator networks.

Variation in dietary specialisation stems from fundamental interactions between species and their environment. Consequently, understanding the drivers of this variation is key to understanding ecological and evolutionary processes. Dietary specialisation in wild bees has received attention due to their close mutualistic dependence on plants, and because both groups are threatened by biodiversity loss. Many principles governing pollinator specialisation have been identified, but they remain largely unvalidated. Organismal phenology has the potential to structure realised specialisation by determining concurrent resource availability and pollinator foraging activity. We evaluate this principle using mechanistic models of adaptive foraging in pollinators within plant-pollinator networks. While temporal resource overlap has little impact on specialisation in pollinators with extended flight periods, reduced overlap increases specialisation as pollinator flight periods decrease. These results are corroborated empirically using pollen load data taken from bees with shorter and longer flight periods across environments with high and low temporal resource overlap.

Integrating economic dynamics into ecological networks: The case of fishery sustainability.

Understanding anthropogenic impacts on ecosystems requires investigating feedback processes between ecological and economic dynamics. While network ecology has advanced our understanding of large-scale communities, it has not robustly coupled economic drivers of anthropogenic impact to ecological outcomes. Leveraging allometric trophic network models, we study such integrated economic-ecological dynamics in the case of fishery sustainability. We incorporate economic drivers of fishing effort into food-web network models, evaluating the dynamics of thousands of single-species fisheries across hundreds of simulated food webs under fixed-effort and open-access management strategies. Analyzing simulation results reveals that harvesting species with high population biomass can initially support fishery persistence but threatens long-term economic and ecological sustainability by indirectly inducing extinction cascades in non-harvested species. This dynamic is exacerbated in open-access fisheries where profit-driven growth in fishing effort increases perturbation strength. Our results demonstrate how network theory provides necessary ecological context when considering the sustainability of economically dynamic fishing effort.

Does urbanization favour exotic bee species? Implications for the conservation of native bees in cities.

A growing body of research indicates that cities can support diverse bee communities. However, urbanization may disproportionately benefit exotic bees, potentially to the detriment of native species. We examined the influence of urbanization on exotic and native bees using two datasets from Michigan, USA. We found that urbanization positively influenced exotic-but not native-bee abundance and richness, and that this association could not be explained by proximity to international ports of entry, prevalence of exotic flora or urban warming. We found a negative relationship between native and exotic bee abundance at sites with high total bee abundance, suggesting that exotic bees may negatively affect native bee populations. These effects were not driven by the numerically dominant exotic honeybee, but rather by other exotic bees. Our findings complicate the emerging paradigm of cities as key sites for pollinator conservation.

Changes in adult sex ratio in wild bee communities are linked to urbanization.

Wild bees are indispensable pollinators, supporting global agricultural yield and angiosperm biodiversity. They are experiencing widespread declines, resulting from multiple interacting factors. The effects of urbanization, a major driver of ecological change, on bee populations are not well understood. Studies examining the aggregate response of wild bee abundance and diversity to urbanization tend to document minor changes. However, the use of aggregate metrics may mask trends in particular functional groups. We surveyed bee communities along an urban-to-rural gradient in SE Michigan, USA, and document a large change in observed sex ratio (OSR) along this gradient. OSR became more male biased as urbanization increased, mainly driven by a decline in medium and large bodied ground-nesting female bees. Nest site preference and body size mediated the effects of urbanization on OSR. Our results suggest that previously documented negative effects of urbanization on ground-nesting bees may underestimate the full impact of urbanization, and highlight the need for improved understanding of sex-based differences in the provision of pollination services by wild bees.

Functional reduction in pollination through herbivore-induced pollinator limitation and its potential in mutualist communities.

Plant-pollinator interactions are complex because they are affected by both interactors' phenotypes and external variables. Herbivory is one external variable that can have divergent effects on the individual and the population levels depending on specific phenotypic plastic responses of a plant to herbivory. In the wild tomato, Solanum peruvianum, herbivory limits pollinator visits, which reduces individual plant fitness due to herbivore-induced chemical defenses and signaling on pollinators (herbivore-induced pollinator limitation). We showed these herbivory-induced decreases in pollination to individual plants best match a Type II functional-response curve. We then developed a general model that shows these individual fitness reductions from herbivore-induced changes in plant metabolism can indirectly benefit overall populations and community resilience. These results introduce mechanisms of persistence in antagonized mutualistic communities that were previously found prone to extinction in theoretical models. Results also imply that emergent ecological dynamics of individual fitness reductions may be more complex than previously thought.

Big city Bombus: using natural history and land-use history to find significant environmental drivers in bumble-bee declines in urban development.

Native bee populations are critical sources of pollination. Unfortunately, native bees are declining in abundance and diversity. Much of this decline comes from human land-use change. While the effects of large-scale agriculture on native bees are relatively well understood, the effects of urban development are less clear. Understanding urbanity's effect on native bees requires consideration of specific characteristics of both particular bee species and their urban landscape. We surveyed bumble-bee (Bombus spp.) abundance and diversity in gardens across multiple urban centres in southeastern Michigan. There are significant declines in Bombus abundance and diversity associated with urban development when measured on scales in-line with Bombus flight ability. These declines are entirely driven by declines in females; males showed no response to urbanization. We hypothesize that this is owing to differing foraging strategies between the sexes, and it suggests reduced Bombus colony density in more urban areas. While urbanity reduced Bombus prevalence, results in Detroit imply that 'shrinking cities' potentially offer unique urban paradigms that must be considered when studying wild bee ecology. Results show previously unidentified differences in the effects of urbanity on female and male bumble-bee populations and suggest that urban landscapes can be managed to support native bee conservation.

The evolution of resistance to two-toxin pyramid transgenic crops.

Pyramid transgenic crops that express two Bacillus thuringiensis (Bt) toxins hold great potential for reducing insect damage and slowing the evolution of resistance to the toxins. Here, we analyzed a suite of models for pyramid Bt crops to illustrate factors that should be considered when implementing the high dose-refuge strategy for resistance management; this strategy involves the high expression of toxins in Bt plants and use of non-Bt plants as refuges. Although resistance evolution to pyramid Bt varieties should in general be slower, resistance to pyramid Bt varieties is nonetheless driven by the same evolutionary processes as single Bt-toxin varieties. The main advantage of pyramid varieties is the low survival of insects heterozygous for resistance alleles. We show that there are two modes of resistance evolution. When populations of purely susceptible insects persist, leading to density dependence, the speed of resistance evolution changes slowly with the proportion of refuges. However, once the proportion of non-Bt plants crosses the threshold below which a susceptible population cannot persist, the speed of resistance evolution increases rapidly. This suggests that adaptive management be used to guarantee persistence of susceptible populations. We compared the use of seed mixtures in which Bt and non-Bt plants are sown in the same fields to the use of spatial refuges. As found for single Bt varieties, seed mixtures can speed resistance evolution if larvae move among plants. Devising optimal management plans for deploying spatial refuges is difficult because they depend on crop rotation patterns, whether males or females have limited dispersal, and other characteristics. Nonetheless, the effects of spatial refuges on resistance evolution can be understood by considering the three mechanisms determining the rate of resistance evolution: the force of selection (the proportion of insects killed by Bt), assortative mating (deviations of the proportion of heterozygotes from Hardy-Weinberg equilibrium at the total population level), and male mating success (when males carrying resistance alleles find fewer mates). Of these three, assortative mating is often the least important, even though this mechanism is the most frequently cited explanation for the efficacy of the high dose-refuge strategy.