Integrating abundance and diet data to improve inferences of food web dynamics

Hidden variable models reveal the effects of infection from changes in host survival

The impacts of disease on host vital rates can be demonstrated using longitudinal studies, but these studies can be expensive and logistically challenging. We examined the utility of hidden variable models to infer the individual effects of infectious disease from population-level measurements of survival when longitudinal studies are not possible. Our approach seeks to explain temporal deviations in population-level survival after introducing a disease causative agent when disease prevalence cannot be directly measured by coupling survival and epidemiological models. We tested this approach using an experimental host system (Drosophila melanogaster) with multiple distinct pathogens to validate the ability of the hidden variable model to infer per-capita disease rates. We then applied the approach to a disease outbreak in harbor seals (Phoca vituline) that had data on observed strandings but no epidemiological data. We found that our hidden variable modeling approach could successfully detect the per-capita effects of disease from monitored survival rates in both the experimental and wild populations. Our approach may prove useful for detecting epidemics from public health data in regions where standard surveillance techniques are not available and in the study of epidemics in wildlife populations, where longitudinal studies can be especially difficult to implement.

EcoDiet: A hierarchical Bayesian model to combine stomach, biotracer, and literature data into diet matrix estimation

Although quantifying trophic interactions is a critical path to understanding and forecasting ecosystem functioning, fitting trophic models to field data remains challenging. It requires flexible statistical tools to combine different sources of information from the literature and fieldwork samples. We present EcoDiet, a hierarchical Bayesian modeling framework to simultaneously estimate food-web topology and diet composition of all consumers in the food web, by combining (1) a priori knowledge from the literature on both food-web topology and diet proportions; (2) stomach content analyses, with frequencies of prey occurrence used as the primary source of data to update the prior knowledge on the topological food-web structure; (3) and biotracers data through a mixing model (MM). Inferences are derived in a Bayesian probabilistic rationale that provides a formal way to incorporate prior information and quantifies uncertainty around both the topological structure of the food web and the dietary proportions. EcoDiet was implemented as an open-source R package, providing a user-friendly interface to execute the model, as well as examples and guidelines to familiarize with its use. We used simulated data to demonstrate the benefits of EcoDiet and how the framework can improve inferences on diet matrix by comparison with classical network MM. We applied EcoDiet to the Celtic Sea ecosystem, and showed how combining multiple data types within an integrated approach provides a more robust and holistic picture of the food-web topology and diet matrices than the literature or classical MM approach alone. EcoDiet has the potential to become a reference method for building diet matrices as a preliminary step of ecosystem modeling and to improve our understanding of prey-predator interactions.

A novel approach to quantifying trophic interaction strengths and impact of invasive species in food webs

Measuring ecological and economic impacts of invasive species is necessary for managing invaded food webs. Based on abundance, biomass and diet data of autochthonous and allochthonous fish species, we proposed a novel approach to quantifying trophic interaction strengths in terms of number of individuals and biomass that each species subtract to the others in the food web. This allowed to estimate the economic loss associated to the impact of an invasive species on commercial fish stocks, as well as the resilience of invaded food webs to further perturbations. As case study, we measured the impact of the invasive bass Micropterus salmoides in two lake communities differing in food web complexity and species richness, as well as the biotic resistance of autochthonous and allochthonous fish species against the invader. Resistance to the invader was higher, while its ecological and economic impact was lower, in the more complex and species-rich food web. The percid Perca fluviatilis and the whitefish Coregonus lavaretus were the two species that most limited the invader, representing meaningful targets for conservation biological control strategies. In both food webs, the limiting effect of allochthonous species against M. salmoides was higher than the effect of autochthonous ones. Simulations predicted that the eradication of the invader would increase food web resilience, while that an increase in fish diversity would preserve resilience also at high abundances of M. salmoides. Our results support the conservation of biodiverse food webs as a way to mitigate the impact of bass invasion in lake ecosystems. Notably, the proposed approach could be applied to any habitat and animal species whenever biomass and diet data can be obtained.

Fitting stochastic predator-prey models using both population density and kill rate data

Most mechanistic predator-prey modelling has involved either parameterization from process rate data or inverse modelling. Here, we take a median road: we aim at identifying the potential benefits of combining datasets, when both population growth and predation processes are viewed as stochastic. We fit a discrete-time, stochastic predator-prey model of the Leslie type to simulated time series of densities and kill rate data. Our model has both environmental stochasticity in the growth rates and interaction stochasticity, i.e., a stochastic functional response. We examine what the kill rate data brings to the quality of the estimates, and whether estimation is possible (for various time series lengths) solely with time series of population counts or biomass data. Both Bayesian and frequentist estimation are performed, providing multiple ways to check model identifiability. The Fisher Information Matrix suggests that models with and without kill rate data are all identifiable, although correlations remain between parameters that belong to the same functional form. However, our results show that if the attractor is a fixed point in the absence of stochasticity, identifying parameters in practice requires kill rate data as a complement to the time series of population densities, due to the relatively flat likelihood. Only noisy limit cycle attractors can be identified directly from population count data (as in inverse modelling), although even in this case, adding kill rate data - including in small amounts - can make the estimates much more precise. Overall, we show that under process stochasticity in interaction rates, interaction data might be essential to obtain identifiable dynamical models for multiple species. These results may extend to other biotic interactions than predation, for which similar models combining interaction rates and population counts could be developed.