Validation of fusarium infection risk calculator with AHDB mycotoxin risk assessment and actual DON results

Abstract

Concerns in the wheat supply chain about the reliability of farmer AHDB Mycotoxin Risk Assessments (MRA) has indicated a need for increased awareness amongst farmers of mycotoxin risk. The Fusarium Infection Risk Calculator (FIRC) was developed to link average daily rainfall for each county with real time anthesis (flowering) progress information provided by agronomists. In order to validate the forecasts provided by the FIRC, grain samples (and associated AHDB MRAs) were collected in 2015, 2016 and 2017 from sites with FIRC risk scores (175 samples in total). These were tested for deoxynivalenol (DON) and compared to the forecast FIRC and AHDB MRA risk scores.  Camgrain supplied additional DON test results in 2015 (n=105), 2016 (n=106) and in 2017 (n=159) with associated AHDB MRA, giving a total of 545 samples. The DON results from these were also compared to both the FIRC and AHDB MRA scores.

In total there were six samples in 2016 and 2017 that exceeded the 1250 µg/kg threshold for food safety.  The FIRC forecast two of these samples as being high risk, whilst the AHDB MRA forecast one as high risk. The high DON levels in the 2017 samples were thought to be due to showers during anthesis disrupting T3 ear wash fungicide applications. These showers were sufficiently frequent to prevent treatment, but volume of rainfall at the country level did not exceed 40mm. It may also be that rainfall variability within the county meant that these crops actually received more rainfall than the county average and were at greater risk than the country level forecast picked up. 

In 2016, there was no obvious reason why the two samples tested high, bar the fact that the farms had large amounts of maize in the rotation (if not in the previous crop) and had maize game strips. All the other samples that were forecast by both tools to be at high risk, had DON levels well below the 1250µg/kg threshold. The low incidence of fusarium in the three test years (97% of samples <500 mg/kg DON) meant that both the FIRC and the AHDB MRA tended to overestimate the risk of DON being present in the harvested grain. The assumptions around rainfall alone used in the FIRC tended to forecast medium risk infection more frequently than the AHDB MRA forecast medium mycotoxin risk, although the difference was not statistically significant. 

In conclusion, the approach of linking timing of anthesis with rainfall is possible, however, the ability of this to forecast infection risk is limited, as there are multiple factors that influence overall infection, including the presence of sufficient inoculum.  In low infection years, the model based on rainfall alone, tends to over-estimate risk, whilst only occasionally identifying as high risk crops that ended up having mycotoxin levels that exceeded the 1250 µg/kg threshold for human consumption. The rainfall at flowering data used in the FIRC can give an indication whether a season is likely to be higher or lower risk of fusarium infection, but the consistency of the FIRC results in relation to final DON test results indicates that it tends to overestimate risk.