Abstract: Abstract: Based on the experiments conducted at four eco-sites with six wheat genotypes differing in grain protein content, an ecological model for predicting grain protein content in winter wheat was established. Analyses on correlations between grain protein content and climatic factors showed that five factors, i.e. mean temperature (Tmean), difference of diurnal temperature (DT), total rainfall (R), total sunshine (S) and growing degree days (GDD) from anthesis to maturity, determined the variation of grain protein content. By stepwise regression analysis, grain protein contents of high-protein varieties depended on DT under the environment with significant DT variation, but on interaction of Tmean and total sunshine under the environment with DT variation less than 5%. For middle-protein and low-protein varieties, grain protein content depended on the total sunshine, and interaction of Tmean, R and S. In the model, the effectiveness factors of DT and sunshine depended on the linear relationship between grain protein content and DT, total sunshine. The temperature factor was determined by the quadratic relationship between protein content and Tmean. The factor of rainfall linearly changed with total rainfall from anthesis to maturity if R was less than 50mm, whereas quadratic equation was established when R was more than 50mm. The model was tested by the different data from three experiments: the first experiment was conducted under three sowing dates with three varieties, the second at six eco-sites with eight winter varieties, and the third at two sites with forty varieties. The RMSE values of all varieties and eco-sites tested were less than 10%. The results indicate a good fit between the simulated and observed values of grain protein contents, thus a reliable prediction for the grain protein contents of different winter wheats under different environments can be obtained.

Key words: Winter wheat, Protein content, Climatic factor, Prediction model