Risk Assessment and Prediction of Stagonospora nodorum blotch in Winter Wheat

Stagonospora nodorum blotch (SNB), caused by the fungus Parastagonospora nodorum, is a major disease of wheat (Triticum aestivum). Pre-planting factors have been associated with the late-season SNB severity, however, the relative influence of these factors on the risk of the disease has not been determined. The overall goal of this work was to evaluate the combined effect of pre-planting factors and in-season weather on the onset and development of SNB in winter wheat.;Residue from a previously infected wheat crop is one of the pre-planting factors that influences SNB severity and can also be an important source of initial inoculum. Therefore, the effects of infected residue on disease severity and yield were quantified. Experiments were conducted at 10 site-years that included a combination of three locations and three years (2012 to 2014) using a moderately susceptible winter wheat cultivar DG Shirley and a highly susceptible cultivar DG 9012. Across site-years, disease severity ranged from 0 to 50% and increased nonlinearly as residue level increased, with a rapid rise to an upper limit with little change in severity above 20 to 30% soil surface coverage. The effect of residue coverage on yield was only significant for DG Shirley in 2012 and for DG 9012 at Salisbury in 2014. SNB also resulted in significant effects on yield and test weight in the most disease-conducive environments, suggesting that the economic threshold for the disease may be higher than previously assumed and warrants review.;The performance of multiple regression (MR) and three machine learning algorithms (artificial neural networks, categorical and regression trees, and random forests) was examined in predicting the pre-planting risk of SNB in wheat. Pre-planting factors tested were cultivar resistance, latitude, longitude, previous crop, seeding rate, seed treatment, tillage type, and wheat residue. Disease severity assessed at the end of the growing season was used as the response variable. The models were developed using 431 disease cases (unique combinations of predictors) collected from 2012 to 2014, and these cases were randomly divided into training, validation, and test datasets. A strong relationship was observed between specific pre-planting predictors and late-season severity of SNB whereby latitude, longitude, wheat residue, and cultivar resistance were the most important predictors. Results show that the random forest (RF) algorithm, which explained 79% of the variability within the data, was the most accurate in predicting the risk of SNB. The RF algorithm could allow early assessment of the risk of SNB, facilitating sound disease management decisions prior to planting of wheat.;Management of SNB during the season mainly relies on fungicide sprays after flag leaf emergence. However, the disease can occur much earlier than flag leaf emergence, and the relationship between time of SNB onset and yield in winter wheat is not well understood. SNB incidence was recorded in 435 disease cases collected from 11 counties in North Carolina during the 2012 to 2014 growing seasons. SNB onset was defined as day of the year (doy) when SNB incidence reached 50%. Starting at the end of the tillering growth stage, temperature, relative humidity, and precipitation were combined to calculate the daily infection value (DIV), a weather index ranging from no (= 0) to optimum (= 1) growth of P. nodorum. Based on the SNB onset-yield relationship, doy 102 was identified as the onset cutoff point. If onset occurred after doy 102 ('late'), 72% of the cases had above-average yield, while if the onset occurred before doy 102 ('early'), only 13% of the cases had above-average yield. Binary logistic regression was used to predict onset classes as a function of i) cumulative DIV until the flag leaf was visible and ii) pre-planting factors. Wheat residue and cumulative DIV until stem elongation were found to be significant predictors of SNB onset. The binary logistic model had a correct classification rate of 0.94. The performance of the model in cross-validation techniques was also high. Once validated using independent data, the model could serve as a within-season decision support tool to help growers with decisions on fungicide application for SNB management.