Measuring and modelling of apple flower stigma temperature as a step towards improved fire blight prediction

In many areas, fire blight (Erwinia amylovora) is a sporadic but potentially devastating disease of apples. Infections occur primarily during bloom when warm weather conducive to bacteria multiplication on the stigma of contaminated flowers is followed by a wetting event, facilitating plant entry. Fire blight prediction models which rely on air temperature for disease forecast can help, but currently produce many false positive and some false negative prognoses. The differences between air and apple flower stigma temperature can explain some of the issues. The present study undertakes an introductory step in resolving this matter by being the first of its kind to document apple stigma temperatures and its departure from air temperature. Thermocouples continuously monitored flower temperature for the blooming seasons of 2018 and 2019 in the orchard of Saint-Bruno-deMontarville, Quebec, Canada, while a thermal imager measured the temperature of randomly selected flowers in 2019. Flower stigma temperature measured with thermocouples followed the diurnal pattern of air temperature, but stigma temperature was higher/lower than air with maxima/minima at peak hours of the day/night. Temperatures measured with the thermal imager revealed a mean positive difference with the air temperature during the day (1.6 +/- 1.3 degrees C). Stigma to air differences for both instruments had a strong positive relation with solar radiation during daytime. Under high humidity, this difference was significantly reduced. From these findings, regression models for estimating stigma temperature were developed for fire blight forecasting. When validating with thermal imaging data, the best model utilizes air temperature, radiation and relative humidity to estimate stigma temperature with better results (RMSE = 1.04 degrees C) than air temperature alone (RMSE = 2.05 degrees C). Although the application of these findings for fire blight prediction models was not tested, there is evidence that models that solely rely on air temperature are at risk of errors.