NBR with Nordic Sugar and other actors currently monitor insect populations in the sugar beet crop and reports these as point based information. This system could be expanded to include digital monitoring and population modelling. This project would fit within NBRs Project Theme Beet Growth, and the Focus Area Digitalisation.
DIGITAL MONITORING
There is hope, but a ready made, off-the-shelf solution doesn’t seem to exist yet.
IN MOTION
NBR is already using and involved in the development of digital sensors . It is part of Nika Jachowicz‘s PhD work, with the sensors being sourced from FaunaPhotonics. These sensors use cloud computing and AI to “identify individual insects based on wing beat frequency, body size, and other features unique to each insect” [Source: faunaphotonics.com 2023-07-27]. The stated delay in processing is around 30 minutes. So, not real-time, but pretty close. This is likely the first place we’d start with any sensor network development.
TRAPS
Paper on the analysis of images taken from traps by Batz et al. [PDF]. They state: “Hitherto, to the best of our knowledge, no study presented an adequate solution for AI based expert level aphid identification from mass catches. However, promising models in the broader field of insect identification, demonstrating remarkable results for a variety of different insect species, could be applicable to this specific context.”. The extension of that statement is that this project would need to contribute to the further exploration of AI based ID of aphids. This could include the investigation of whether non-expert level ID is sufficient for modelling.
MODELLING
With data available from sensors, it should be possible to develop some form of spatial digital shadow. So, a model that is updated on psudeo-live data. Modelling insect populations is notoriously difficult, given the huge variation across the landscape they live in. That said, the Broomes Barn green peach aphid first day of flight model used in the UK is reported to be remarkable accurate. The estimates they make in February/ March for the first day that aphids are likely to migrate either are correct or miss by one or two days.
Returning again to Batz et al. [PDF]: these authors give a long list (SM1) of different models that have been developed for different aphids in different crops. For sugar beet, there are six entries, five of which are from the UK, and all of which are concentrated on Myzus persicae. The sixth model is from the Netherlands and focuses on the spread of virus yellows as opposed to the flight date of the aphid [source: Dusi, A. N., Peters, D., and van der Werf, W. (2000). Measuring and modelling the effects of inoculation date and aphid flights on the secondary spread of Beet mosaic virus in sugar beet. Ann. Appl. Biol. 136, 131–146.]. In among all the other listed models, it seems highly likely that a good framework for modelling aphid/ other insect populations over time and space would be found. Batz et al. also emphasise that weather data is critical in population modelling.
CONCLUSION
The time is right: we could start now using the data we have, and will be able to improved expand as more sensor data becomes available. Starting now with modelling will also help guide near-future monitoring efforts to ensure data is being collected on the right scale.