RESUMO
Sclerotinia stem rot (SSR) epidemics in soybean, caused by Sclerotinia sclerotiorum, are currently responsible for annual yield reductions in the United States of up to 1 million metric tons. In-season disease management is largely dependent on chemical control but its efficiency and cost-effectiveness depends on both the chemistry used and the risk of apothecia formation, germination, and further dispersal of ascospores during susceptible soybean growth stages. Hence, accurate prediction of the S. sclerotiorum apothecial risk during the soybean flowering period could enable farmers to improve in-season SSR management. From 2014 to 2016, apothecial presence or absence was monitored in three irrigated (n = 1,505 plot-level observations) and six nonirrigated (n = 2,361 plot-level observations) field trials located in Iowa (n = 156), Michigan (n = 1,400), and Wisconsin (n = 2,310), for a total of 3,866 plot-level observations. Hourly air temperature, relative humidity, dew point, wind speed, leaf wetness, and rainfall were also monitored continuously, throughout the season, at each location using high-resolution gridded weather data. Logistic regression models were developed for irrigated and nonirrigated conditions using apothecial presence as a binary response variable. Agronomic variables (row width) and weather-related variables (defined as 30-day moving averages, prior to apothecial presence) were tested for their predictive ability. In irrigated soybean fields, apothecial presence was best explained by row width (r = -0.41, P < 0.0001), 30-day moving averages of daily maximum air temperature (r = 0.27, P < 0.0001), and daily maximum relative humidity (r = 0.16, P < 0.05). In nonirrigated fields, apothecial presence was best explained by using moving averages of daily maximum air temperature (r = -0.30, P < 0.0001) and wind speed (r = -0.27, P < 0.0001). These models correctly predicted (overall accuracy of 67 to 70%) apothecial presence during the soybean flowering period for four independent datasets (n = 1,102 plot-level observations or 30 daily mean observations).
