Reverse-engineering phenology models with an ecoinformatics approach: Insights from pea aphid vector biology

Predicting the arrival time of insect vectors for crop pathogens is an important goal in agriculture. However, for many insect vectors, traditional phenology models may not be sufficient to predict the timings or location of outbreaks. Furthermore, while monitoring networks for insect vectors is useful for tracking outbreaks, it is limited to providing information to stakeholders on current conditions rather than forecasts. Pea aphids are an important pest in Pacific Northwestern US pulse crops where they are vectors for a range of devastating pathogens to legumes. In this cool-temperate climate, predicting aphid outbreaks from year to year and within seasons has proven difficult. Consequently, pea aphids have been monitored over this large agricultural region for entire growing seasons across both states.  We analyzed 12 years of these monitoring data for pea aphids in eastern Washington and Idaho using a Generalized Additive Modeling (GAM) approach and long-term atmospheric data. This large-scale, 'ecoinformatics' analysis provided an opportunity to build and validate a single climate-based model that predicts peak arrival time for pea aphids in pulse crops. We believe this approach will be useful in other systems with long-term vector monitoring programs (10+ years) to build better forecasting tools for agricultural entomology extension.