ArcGIS and machine learning are used to provide a comprehensive overview of potato viruses in Scotland, a deeper understanding of landscape epidemiology, and a model that could serve as the basis of a decision support tool.

Abstract

Virus diseases represent important economic threats to seed potato production worldwide, yet relatively little is known of their epidemiology at the landscape-scale. In this study, data was compiled from the Scottish national seed potato classification scheme on the incidence of 10 different potato viruses for the years 2009–2022. A co-occurrence analysis identified that 12 virus species pairs occurred together more often than expected by chance, and potato blackleg was positively associated with eight potato viruses. ArcGIS was used to investigate spatial and spatiotemporal variation in incidence rates of the three most prevalent viruses (potato virus Y, potato leaf roll virus and potato virus A), and this revealed prominent geographic differences in long-term disease outcomes. Focusing on potato virus Y as the most commonly occurring single infection, interpretable machine-learning techniques were used to investigate the influence of key crop, management and environmental factors on patterns of incidence in space and time. The results showed that health characteristics of seed stocks were among the most important predictors of incidence, along with blackleg infection, several management features, cultivar resistance, distance to the nearest seed and ware crop, temperature variables and several soil features. This approach provides a comprehensive overview of potato viruses in Scotland, a deeper understanding of epidemiological risk factors at the landscape-scale and a forecast model that could serve as the basis of a decision support tool for improved management of potato virus Y.