PS 32-183 - A hierarchical Bayesian approach to quantify the predation of carabid beetles in an agricultural system based on data from molecular trophic gut-content analysis

Abstract: Studies in ecology using molecular gut-content data to infer the diet of animals remains mostly qualitative. Yet, the qualitative nature of molecular data allows inferring the existence of predation events, but does not allow inferring their number. In addition, due to the digestion process by the predator, the detection probability decreases with the time elapsed since the predation event. To fill this gap, we proposed a model in which the consumption and the digestion process are combined to infer the predation rates of Amara aenea, Anchomenus dorsalis, Brachinus sclopeta et Poecilus cupreus on pests, decomposers and spiders in agricultural fields. The predation process was modelled as a homogeneous Poisson process. The digestion process was modelled as a DNA detection probability decreasing logistically with the time elapsed since prey predation. An inhomogeneous Poisson process has emerged from the combination of these two processes. Two datasets, including (1) the detection of prey DNA in a sample of carabids collected in the field and (2) the experimental assessment of the decrease in detection rate with the time elapsed from the predation event, were used to fit the model using a hierarchical Bayesian approach. Our findings suggested that Brachinus sclopeta provides services by preying mostly on pests, especially on slugs, without importantly producing disservices by preying on decomposers and spiders. This makes this species highly beneficial to agriculture. On the contrary, if Amara aenea, Anchomenus dorsalis and Poecilus cupreus can contribute to biocontrol by attacking pests, they could at the same time have a non-negligible impact on other service providers as earthworms and springtails. This model could be applied to estimate ideal composition of species that both minimize the predation on decomposers and maximize the predation on pests. In conclusion, this hierarchical Bayesian approach is a first step towards the quantification of predation rates from molecular gut-content data.