Smart Armyworm Surveillance: Project Technical Report

This project was designed to test three different technologies for monitoring the fall armyworm, a highly migratory moth pest of maize in Kenya. This insect is invasive and is estimated to have caused between US$200 and US$600 million dollars’ worth of crop damage in Africa since it was first observed in western Africa in 2016. The three technologies we piloted were entomological radar, digital pheromone traps and an image identification app. The hypothesis we set out to test was that high-altitude moth migrations are linked to pest incidence at ground level and the later impacts of feeding damage. Our objective was to install and launch all three technologies in the space of a year, with the aim of developing an integrated dataset that would provide an overview of near-ground and upper-atmospheric movements of fall armyworm. This data would be shared with multiple different stakeholders in real-time. Over a 12-month period between March 2019 and February 2020, we installed the entomological radar, 20 digital pheromone traps and launched the Nondo app to test this hypothesis. Although there were several challenges, the outcomes after one year include: 1. Preliminary radar data. We were able to detect insects at heights up to 800m and the preliminary data would suggest these are targets of the appropriate mass to be classed as noctuid moths 2. A network of digital pheromone traps. We observed fall armyworm in almost all of the locations where the traps were installed and received daily automated updates, despite the traps being hosted in very rural parts of Kenya. 3. An image detection algorithm for the fall armyworm and other maize pests. This provided a very high level of accuracy for species where we were able to provide more than 100 images (>90% in some cases). This project has profound potential for impact as it demonstrates the successful implementation of an “internet of things” approach to biological monitoring in very rural parts of Kenya. Indeed, one of the highlights of this project is the deployment of the digital trap network beyond the end of the project. This is the first time such an ambitious, multi-layered monitoring network comprising radar, pheromone traps, machine learning and decision support apps has been established in Sub-Saharan Africa. This project is a proof-of-concept that clearly demonstrates the potential of digital monitoring to deliver major impacts for farmers in Sub-Saharan Africa.