Predators can affect prey populations through both consumptive and nonconsumptive effects, and physical factors, such as light levels and temperature can influence the strength of the interactions. However, few studies have quantified how nonconsumptive effects contribute to changes in population densities or the potential influence of physical factors in mediating predator-prey interactions at field-relevant scales. We used state-space models fit to long-term (1994-2016) seasonal field time series data to quantify effects of the invasive predatory zooplankter, Bythotrephes longimanus, and physical factors (light and temperature) on two prey species, Daphnia mendotae and Bosmina longirostris, in offshore Lake Michigan. We developed a suite of state-space models that represent different hypothesized effects of Bythotrephes (e.g., consumptive and nonconsumptive effects), light levels (e.g., effects on predator attack rates), and temperature (e.g., effects on prey birth rates, predator attack rates, and background mortality). We then fit the models to field time series data by maximum likelihood using an iterated filtering algorithm. Our method allowed us to account for challenges to using ecological field data, such as irregular sampling intervals, process stochasticity, and measurement error.
Results/Conclusions
Our results suggest that Bythotrephes have had large, negative effects on populations of both prey species in Lake Michigan. For example, we estimated that population growth rates of Daphnia and Bosmina were reduced by 46% and 24%, respectively, at mean Bythotrephes density. Based on the Akaike Information Criterion (AIC), the best performing models for both species included nonconsumptive effects of Bythotrephes (ΔAIC > 6, compared to models with no predator effect), consistent with experiments and observational studies that suggest that Bythotrephes induces a vertical migratory response in prey to deeper, colder waters, causing reduced individual growth rates. In contrast, inclusion of Bythotrephes consumptive effects offered no improvement in model performance (ΔAIC < 0), even when attack rates were allowed to vary with changing light levels. Incorporating temperature effects for D. mendotae offered improvement only when modeling temperature dependence for prey background mortality (ΔAIC = 4.1), but the estimated temperature effect could plausibly be attributed to other factors that vary seasonably. Our approach thus helped us understand predation mechanisms in the field, but there are challenges, and we will discuss our insights on how to meet them.