Data-informed Machine Learning Modeling for Infestation Level Prediction of the Sugar Beet Weevil

The notable increase in insect populations over the recent years has been closely linked to rising global temperatures and more frequent drought events, both consequences of climate change. This surge in insect activity has had a significant impact on agricultural production [1]. Among the crops most affected is sugar beet in the eastern part of Austria [2], where outbreaks of the sugar beet weevil (Asproparthenis punctiventris) have been become increasingly common. Identifying regions more prone to such infestations could aid crop planning and management practices, mitigate agricultural losses, improving energy efficiency, and increase crop yield. Previous publications have already shown the influence of weather conditions on the reproduction and survival rates of insects and linked these factors to their distinct life cycle stages [3,4]. These investigations employed simple regression models and statistical frameworks to study the correlation of the infestation level with weather parameters as well as degree-day models that aimed at predicting the time of insects’ outbreak.

In our study we extend this approach by incorporating soil composition data, historical crop records alongside the most relevant meteorological parameters. We use these data to train machine learning algorithms, specifically species distribution models together with random forests, aiming at forecasting infestation levels. By integrating data from diverse and heterogeneous sources, we construct a comprehensive database used as the foundation for developing our machine learning trained prediction algorithm. We propose a multi-layered model in which each layer processes data from a different source, spatially represented on a map. Furthermore, we integrate geospatial information of the previous sugar beet crops and derive a population spread function, which is subsequently used to refine the prediction results. Initial findings validate the feasibility of the proposed approach and its potential for geographically predicting infestation levels of the sugar beet weevil.

 

References

[1] Skendžić S, Zovko M, Živković IP, Lešić V, Lemić D. The Impact of Climate Change on Agricultural Insect Pests. Insects 2021; 12(5).

[2] Strotmann K., Pflanzenschutzverbot: 4.000 ha Rüben in Österreich vernichtet. Agrarmarkt Österreich; Jun.2023.

[3] Drmić Z, Čačija M, Virić Gašparić H, Lemić D, Bažok R. Phenology of the sugar beet weevil, Bothynoderes punctiventris Germar (Coleoptera: Curculionidae), in Croatia. Bull Entomol Res. 2019 Aug;109(4):518-527. doi: 10.1017/S000748531800086X. Epub 2018 Nov 27. PMID: 30477591.

[4] Lydia Jarmer. Masterarbeit: Auftreten des Rübenderbrüsslers (Asproparthenis punctiventris) in Ostösterreich unter besonderer Berücksichtigung von Witterungsverhältnissen. Universität für Bodenkultur; 2022.