Models of wheat grain quality considering climate, cultivar and nitrogen effects

Year x cultivar means of wheat grain crude protein content (CP), Hagberg falling number (HFN) and grain specific weight, surveyed in England from 1975 to 1995 were investigated with a residual maximum likelihood (REML) analysis. Potential fixed terms included cultivar, nitrogen fertilizer application and roiling three-weekly weather data from 21 meteorological stations selected from within the major wheat growing areas of the UK. Further analyses were conducted to determine how much variance in annual grain quality could be accounted for by successive weeks during the growing season, and also to assess how well the models predicted grain quality at three sites not included in the model derivation. CP was described with a model combining effects for cultivar, nitrogen application, and periods of winter and spring rainfall and summer temperature. This model accounted for 90% of annual variation by harvest time, and 70% by the end of May. Predictions by the model were significantly associated with observations at an independent site where wheat was grown on a similar soil type in all years, but not at a site where soil type varied greatly. HFN was described with a model combining effects for cultivar and periods of summer temperature. This model could account for 85% of annual variation by harvest time, but it was not until August that this figure reached 70%. The model predictions were significantly associated with observations at all three test sites, even when soil type varied. A model for specific weight combined effects for cultivar and periods of autumn rain, winter temperature and summer temperature. This model only accounted for 75% of variation by harvest but predictions from it were significantly associated with observations at two sites. The time trend in changes in cultivar quality during the surveyed period has been slightly detrimental to CP, but beneficial to HFN and specific weight.