Dry matter accumulation plays an important role in the wheat yield. Under field condition, wheat cultivars with different tillering abilities have different dry matter accumulation characteristics. So it has great significance to realize the simulation and prediction of wheat dry matter accumulation process. Simulation models can quantitatively describe dry matter accumulation, and the equation based on the normalized method is widely applicable. Accumulated temperature is superior to time parameters as the variable of the prediction model. To investigate a model to simulate wheat dry matter accumulation, three wheat cultivars (Yumai 49-198, Lankaoaizao 8, and Yanzhan 4110) with different tillering abilities were grown at three densities each (750000, 2250000 and 3750000 plants/hm²) in a field experiment. The dry matter accumulation of wheat varieties with higher ear-bearing tiller percentages (YM49-198 and YZ4110) were highest at a density of 3750000 plants/hm², while that of the variety with a lower ear-bearing tiller percentage (LKAZ8) peaked at 2250000 plants/hm². The dry matter accumulation of different varieties differed with density, therefore a suitable density should be chosen for each variety to maximize dry matter accumulation in wheat production. Five simulation models with high correlation coefficients for relative dry matter accumulation were established using normalized accumulated temperature and dry matter accumulation. We tested five models, and the optimal model was found through the limit of all the equations. The best predictive model for dry matter accumulation was the Richard curve equation, i.e., y=1.1435/(1+e^{0.2776-4.6558x})^1/0.1130, r=0.9927, and its characteristic parameters were calculated based on the relative dry matter accumulation model. This Richard equation had relatively small parameters and a straightforward biological interpretation. Normalization overcame changes in model parameters caused by different cultivation techniques and varieties and improved the versatility of the model. The values of parameters b and c changed dramatically among varieties and densities, while parameters a and d varied only slightly. The model was tested with relative dry matter accumulation data from 2010-2011; the correlation coefficient r of simulated dry matter accumulation was above 0.98^{* *}, and the accuracy K was above 0.91^{* *}, showing that this model could accurately predict dry matter accumulation. This model simulated dry matter accumulation of wheat using accumulated temperature in any growth period and predicted well the actual wheat production, making it highly suitable for practical use.
Overall, dry matter accumulation could be divided into early, middle, and late phases based on the two inflexion points in the rate equation. The dry matter accumulation rate was very sensitive to density in the middle phase. The relative accumulated temperature was 0.53 at the maximum dry matter accumulation rate, when the dry matter weight was about one-half of the total weight. These data indicated the importance to improving wheat yields of enhancing field management in the early growth phases, including the cultivation of sound seedlings and the construction of appropriate populations. The average rates of dry matter accumulation were highly significantly correlated with dry matter weight, and they were the most important factor influencing dry matter accumulation according to path analysis. Higher average rates of dry matter accumulation had significant effects on stabilizing and increasing the dry matter weight of wheat.