Modelling within-field variations in deoxynivalenol (DON) content in oats using proximal and remote sensing

Abstract

Within-field variations in the mycotoxin deoxynivalenol (DON) in oat grain were investigated at two farms in south-west Sweden. At Sarestad farm (sampled 2012), where one of two fields studied was ploughed annually and the other was under no-till cultivation, the DON concentration varied between 28 and 1 755 ppb. The level was higher (270–5 000 ppb) at Entorp farm (sampled 2013). Within-field prediction models for DON were constructed using a data mining method (multi-variate adaptive regression splines) with satellite data, an ECa sensor and airborne laser scanning. At Sarestad, the no-till field had higher DON content, with the highest values in silty patches in the otherwise clayey soil. Sensor data related to soil and crop conditions had the potential to describe the DON variability within fields. The covariance between DON content and auxiliary data differed at Entorp farm, where high DON values (>2 000 ppb) were found in clayey parts of the field. This pattern was attributed to poor drainage with recurring waterlogging. Within these clayey parts, the highest DON contents coincided with the highest biomass density. South-west Sweden received much less rainfall in 2013 than in 2012, which may have resulted in different DON patterns in relation to soil types. In 2012, more permeable silty soils apparently promoted growth, biomass production and DON production, whereas in 2013 a poorly drained clayey soil with high water-holding capacity favoured development of high DON concentrations.