Crop forecasting is critical to AWB’s business operations. Current quantitative modelling capability, however, is limited to the prediction of crop yield. This study aimed to investigate the feasibility of developing mathematical models for the regional prediction of grain quality, with focus on protein as the key quality attribute. Historic receival data sets (1988-2001) provided quality data for ten representative shires in the Australian wheat belt. The data sets for each shire were consolidated to derive the distribution of protein received each harvest. Distributions were found to vary significantly between both shires and years. Crop simulations were also performed using historical rainfall records over this time period and results were used to characterise the nature of the season depending on the timing and extent of water limitation experienced by the crop in each year. Cluster analysis identified three distinct types of season for each shire, categorised as Mild (late-onset stress) through to Severe (early-onset stress). Classification of season type generated significant discrimination for protein distribution in 8 of the 10 shires studied, with substantial discrimination likely in the last third of the crop cycle. Discrimination was associated with a shift in the mean percent protein with season type, whilst spread of the distribution was unaffected.

This study has shown that potential exists for development of a predictive system for crop quality (protein). The analysis is being extended to examine spatial associations at a regional and port zone scale, and prediction of screenings.