Model-based forecasting of twig canker incidence of bacterial spot of peach in Fukushima Prefecture

Abstract

Bacterial spot caused by Xanthomonas arboricola pv. pruni (Xap) is the most important disease that affects peach production. A disease-forecasting model was developed to help growers decide when to apply bactericides and remove diseased, last-year twigs. To predict the incidence of “spring cankers”, peach twigs damaged by Xap, we used 12 years (2009–2020) of data from Fukushima Prefecture to develop a forecasting system using a hierarchical Bayesian model (HBM). The model included the number of fields with a bacterial spot incidence (BSI) on leaves ≥ 10% in late September of the previous season and the number of days with rain (≥ 10 mm/day) and maximum wind speed (≥ 5 m/s) during the previous October as predictors. Using a best-fit cutoff value based on a receiver operating characteristic (ROC) curve, the model achieved a 0.836 accuracy, 0.804 sensitivity, 0.847 specificity, 0.847 precision, and 0.712 F-measure. The model was validated using a fourfold cross-validation (CV) procedure and achieved an average accuracy of 0.847. Thus, the model explained 65.7% of the variability compared to observed frequencies with predicted probabilities of twig canker incidence (TCI) ≥ 2% from April to May 2009 to 2020 in Fukushima Prefecture. These results suggest that this disease-forecasting model using HBM based on 12 years of historical data can be used to predict the risk of twig cankers of bacterial spot of peach.