Prediction of deoxynivalenol and zearalenone in winter wheat grain in a maize-free crop rotation based on cultivar susceptibility and meteorological factors

Grains of the three differentially Fusarium-susceptible winter wheat cultivars “Ritmo” (highly susceptible), “Inspiration” (moderately to highly susceptible) and “Dekan” (lowly to moderately susceptible) from up to eight trial locations in Schleswig-Holstein (Northern Germany) and maize-free crop rotations were analysed for their mycotoxin concentration from 2008 to 2017 (“Inspiration” and “Dekan” since 2012). The deoxynivalenol (DON) and zearalenone (ZEA) concentrations of wheat grain samples differed significantly between individual years and within each year between the trial locations due to weather conditions during flowering. Significant relationships were found between the two weather variables cumulative precipitation (added up by all daily cumulative precipitation) and average temperature (averaged for all daily means of temperature) during the period of wheat flowering and DON and ZEA concentrations in wheat grain at harvest. These relationships were determined for “Ritmo” from 2008 to 2014 (\(R^{2}_{{{\text{adj}} .}}\) = 0.81 for DON; \(R^{2}_{{{\text{adj}} .}}\) = 0.75 for ZEA) and for both “Inspiration” (\(R^{2}_{{{\text{adj}} .}}\) = 0.84 for DON; \(R^{2}_{{{\text{adj}} .}}\) = 0.82 for ZEA) and “Dekan” (\(R^{2}_{{{\text{adj}} .}}\) = 0.78 for DON; \(R^{2}_{{{\text{adj}} .}}\) = 0.77 for ZEA) from 2012 to 2016. Based on this, multiple regression models were developed for the three cultivars for the prediction of DON and ZEA contamination in wheat grain: model 1 = highly susceptible; model 2 = moderately to highly susceptible; model 3 = lowly to moderately susceptible. The models included the covariates cumulative precipitation and average temperature during wheat flowering and the interaction term of precipitation and temperature. The predictive power of the three models was evaluated with data not utilized in the development of the models, i.e. weather conditions during wheat flowering and DON and ZEA concentrations in wheat grain at harvest at the same trial locations in the years 2015 to 2017 for model 1 and in 2017 for model 2 and model 3. The models showed a high predictive power by regressing observed versus predicted values (model 1: R2 = 0.89 for DON, R2 = 0.91 for ZEA; model 2: R2 = 0.91 for DON, R2 = 0.84 for ZEA; model 3: R2 = 0.86 for DON, R2 = 0.89 for ZEA). Model 1 predicted correctly whether the concentrations of DON and ZEA were either lower or higher than the European maximum levels of 1250 µg DON/kg and 100 µg ZEA/kg in 95.2% of the cases. Models 2 and 3 performed 85.7% and 100% correct predictions for DON in 2017, respectively. Model 2 predicted correctly whether the ZEA concentration was either lower or higher than the maximum level of 100 µg ZEA/kg in 100% of the cases in 2017, whereas model 3 performed 85.7% correct predictions. The models are therefore useful for the prediction of DON and ZEA concentrations in wheat grain from maize-free crop rotations based on cumulative precipitation and average temperature during wheat flowering and for differentially susceptible wheat cultivars.