Evaluation of the Blossom Infection Risk of Fire Blight in Apple Trees Based on Short-Range Weather Forecasts in FBcastS

The FBcastS automated system determines optimal control timing for fire blight blossom infection by running the K-Maryblyt model with observed weather data and short-term forecasts for today, tomorrow, and the day after tomorrow, and issuing recommendations based on blossom infection risk (BIR). From 2022 to 2024, BIR during the entire flowering period was analyzed at the Chungju site using observed and forecast data. The first control recommendation dates at Chungju were April 23 in 2022, April 19 in 2023, and April 15 in 2024. Except in 2024, first recommendation dates at 31 major apple-producing and fire blight-affected sites in Korea were the same as or within 2 days of those at Chungju. Confusion matrix analyses were conducted for the entire flowering period across the 31 sites and for the two in-bloom spray dates recommended by the Rural Development Administration (RDA), and agreement with spray dates derived from observed weather was evaluated using recall (sensitivity). Recall values by RDA recommendations using today’s and tomorrow’s forecasts in 2022 and 2024 were 0.74-0.92, whereas recall using the day-after-tomorrow forecast in 2022 and 2024 were 0.18-0.49. These results indicate that control recommendations based on today’s and tomorrow’s forecasts are reliable, while those based on longer lead-time forecasts require caution. Among the 3 years, 2023 showed both the lowest infection risk and the lowest recall. Improved understanding of K-Maryblyt and FBcastS supports sustainable integrated disease management, including monitoring epiphytic infection potential (EIP) on stigmas and determining copper spray timing based on bud break predictions.