Predicting direct and indirect non-target impacts of biocontrol agents using machine-learning approaches

Abstract

Biological pest control (i.e. ‘biocontrol’) agents can have direct and indirect non-target impacts, and predicting these effects (especially indirect impacts) remains a central challenge in biocontrol risk assessment. The analysis of ecological networks offers a promising approach to understanding the community-wide impacts of biocontrol agents (via direct and indirect interactions). Independently, species traits and phylogenies have been shown to successfully predict species interactions and network structure (alleviating the need to collect quantitative interaction data), but whether these approaches can be combined to predict indirect impacts of natural enemies remains untested. Whether predictions of interactions (i.e. direct effects) can be made equally well for generalists vs. specialists and across different habitat types is also untested for consumer-prey interactions, though previous work on mutualist networks suggests that interactions among generalists may be more difficult to predict. Here, we used two machine learning techniques (random forest and k-nearest neighbour; KNN) to test whether we could accurately predict empirically-observed quantitative host-parasitoid networks using trait, abundance, and phylogenetic information. Then, we tested whether the accuracy of machine-learning-predicted interactions depended on the generality of the interacting partners or on the source (habitat type) of the training data. Finally, we used these predicted networks to generate predictions of indirect effects via shared natural enemies (i.e. apparent competition), and tested these predictions against empirically observed indirect effects between hosts. We found that random-forest models predicted host-parasitoid pairwise interactions (which could be used to predict attack of non-target host species) more successfully than KNN, and this predictive ability depended on the generality of the interacting partners, but not the source (habitat type) of data used to train the models. Further, although our machine-learning informed methods could significantly predict indirect effects, the explanatory power for both direct and indirect interactions was reasonably low. Combining machine-learning and network approaches provides a starting point for reducing risk in biocontrol.

Characterising uncertainty in risk assessments for biological control: using case studies from New Zealand to inform future research

“…Using qualitative food webs, Todd et al (2021) found that showing all known connections between species in a food web diagram could help to reduce uncertainty around which species might be at risk. Kotula et al (2021) tested machine-learning approaches for predicting indirect effects and found that while they were not able to predict indirect effects, they did have potential to rank hosts as having low or high risk of indirect effects. These authors suggested that validation of such predictions post-release would be of value to regulators in future decision-making.…”

Predicting predator-prey interactions in terrestrial endotherms using random forest

“…Of these, random forest is emerging as a popular method due to its ability to learn nonlinear interactions between predictor variables, outperform other inference methods (Desjardins-Proulx et al 2017, Laigle et al 2018, Sydenham et al 2021, and its availability to a wide range of users-including those without expert programming skills-through several R packages (e.g., Liaw and Wiener 2002, Wright and Ziegler 2017, Seligman 2022. Random forests have been used to predict various interaction types, including plant-pollinator (Ornai andKeasar 2020, Sydenham et al 2021), parasite-host (Kotula et al 2020), and predator-prey interactions (Desjardins-Proulx et al 2017, Parravicini et al 2020. They have also been used to identify which species' traits are most important for predicting whether species interact (Dellinger et al 2019, Klomberg et al 2022.…”