Validating Sclerotinia sclerotiorum Apothecial Models to Predict Sclerotinia Stem Rot in Soybean (Glycine max) Fields

Abstract

In soybean, Sclerotinia sclerotiorum apothecia are the sources of primary inoculum (ascospores) critical for Sclerotinia stem rot (SSR) development. We recently developed logistic regression models to predict the presence of apothecia in irrigated and nonirrigated soybean fields. In 2017, small-plot trials were established to validate two weather-based models (one for irrigated fields and one for nonirrigated fields) to predict SSR development. Additionally, apothecial scouting and disease monitoring were conducted in 60 commercial fields in three states between 2016 and 2017 to evaluate model accuracy across the growing region. Site-specific air temperature, relative humidity, and wind speed data were obtained through the Integrated Pest Information Platform for Extension and Education (iPiPE) and Dark Sky weather networks. Across all locations, iPiPE-driven model predictions during the soybean flowering period (R1 to R4 growth stages) explained end-of-season disease observations with an accuracy of 81.8% using a probability action threshold of 35%. Dark Sky data, incorporating bias corrections for weather variables, explained end-of-season disease observations with 87.9% accuracy (in 2017 commercial locations in Wisconsin) using a 40% probability threshold. Overall, these validations indicate that these two weather-based apothecial models, using either weather data source, provide disease risk predictions that both reduce unnecessary chemical application and accurately advise applications at critical times.

An overview of the Sclerotinia sclerotiorum pathosystem in soybean: impact, fungal biology, and current management strategies

“…Willbur et al (2018b) found that data from darksky.net were nearly as accurate as weather from on-site weather stations, however bias was detected in the three weather variables of interest. When bias corrections were included, darksky.net weather data were considered a suitable source to drive SSR prediction models (Willbur et al 2018b). Plant phenology information and canopy and row-spacing parameters (Fall et al, 2018) have subsequently been combined with these prediction models into a smartphone application that can be used predict the risk of apothecial presence during the soybean bloom period.…”

“…Models were used in a set of subsequent field validation experiments to test accuracy of prediction of end-ofseason disease levels. In those validation efforts in Wisconsin, Iowa, and Michigan models predicted SSR over 80% of the time (Willbur et al 2018b). Furthermore, sources of weather data were tested, including data from an open-source weather provider, darksky.net.…”

“…Models generated using these publicly available data allow the model to be accessible and functional in virtually any growing location (Magarey et al 2001). Willbur et al (2018b) found that data from darksky.net were nearly as accurate as weather from on-site weather stations, however bias was detected in the three weather variables of interest. When bias corrections were included, darksky.net weather data were considered a suitable source to drive SSR prediction models (Willbur et al 2018b).…”

Identification of factors associated with white mould in snap bean using tree‐based methods

“…Predictive performance was maintained by remaining with a tree‐based methodology (random forests), which again alludes to an earlier point that the presence of white mould in snap bean is characterized by several interactions not easily captured by linear regression models (Bom & Boland, ). In this case the random forest model had what would be considered a very good estimated test performance (cv AUC of 0.819) for a Sclerotinia ‐based prediction model (Harikrishnan & del Río, ; Willbur et al , , b).…”

Modeling of Yield Losses and Risk Analysis of Fungicide Profitability for Managing Fusarium Head Blight in Brazilian Spring Wheat

“…Under the current scenario, one fungicide spray (tebuconazole) during flowering is more likely a profitable decision than applying two sprays, for which there is greater uncertainty. The damage coefficients can be useful in decision support tools like apps embed with real time predictions of disease severity and economic scenarios (Willbur et al 2018…”