Modeling Formation of Yield Components Based on Apical Development in Wheat

Prediction for the yield formation is one of the most important contents in the research of crop growth simulation. Wheat yield is composed of ear number per area, grain number per ear and grain weight. Based on the apical primordial development and tillering simulation model, we established an interpretative and predictive model on the formation of yield components to help quantify wheat growth and predict yield. Two suits of experiment data reported in literatures were used to establish the model. One came from the experiment at Northeast Agricultural University from 1994 to 1995 and the other was from the experiment at Nanjing Agricultural University from 1997 to 1999. In the model, grain number per ear was depended on the spikelet number per ear and grain number per spikelet. The spikelet number was simulated by spikelets potentially being transferred from leaf primordia before double ridge stage and being differentiated after double ridge. The fertility conditions during the spikelet development period determined the final spikelet number in this model. Grain number was established by the potential florets per spikelet, the temperature factor and average grain number factor which was determined by the total spikelet number. The grain weight was simulated by the characteristic thousand kernel weight (TKW), temperature sensitivity of different genotypes, mean temperature and soil water condition during grain filling. The product of ear number determined by the established tillering dynamic model, spikelet number per ear, grain number per spikelet and grain weight gave the final grain yield in the present simulation model. Seven genetic parameters of wheat varieties were used in the model, i.e. potential spikelets per ear, potential florets per spikelet, characteristic TKW, temperature sensitivity, physiological vernalization time, photoperiod sensitivity and filling duration factor as shown in Table 1. The model was validated with field experimental data from two cultivars and three sowing dates from 1996 to 1998 in Nanjing. The results showed that the RMSE (root mean square error) values of total leaf primordia, total leaf number, total spikelets, total floret primordia, grain number per ear, TKW and yield were 1.33, 0.87, 1.06, 3.21, 2.79, 1.58(g) and 25.6(kg/ha), respectively (Table 2). Good coincidences between the simulated and observed values of spikelet number per ear, active floret per ear and TKW found in figures of 1 to 4. In conclusion, the present model can provide reliable prediction on the grain number, grain weight and yield formation with different variety types under different environmental conditions.