Potato blackleg is a common bacterial disease that causes serious losses in potato (Solanum tuberosum) production worldwide. Despite this, relatively little is known of the landscape epidemiology of this disease. This study provides the first national-scale analysis of spatial and spatiotemporal patterns of blackleg incidence rates and associated risk factors for disease at the landscape scale. This was achieved through a combination of ArcGIS and interpretable machine learning applied to a longitudinal dataset of naturally infected seed potato crops from across Scotland. We found striking differences in long-term disease outcomes across the country and identified that features (variables) related to the health status and management of mother crops (seed stocks), matching features in daughter crops, and the characteristics of surrounding potato crop distributions were the most important predictors of disease, followed by field, bioclimatic, and soil features. Our approach provides a comprehensive overview of potato blackleg at a national scale, new epidemiological insights, and an accurate model that could serve as the basis of a decision support tool for improved blackleg management.
Keywords: ArcGIS; Pectobacterium; bacterial pathogens; decision support; disease mapping; disease risk; machine learning; seed potato.