Physiological Characterization and Modeling of Soybean in a Maximum Yield Environment.

Mean USA soybean grain yields have increased linearly from 739 kg ha^-1 in 1924 to 2661 kg ha^-1 in 2012 at a rate of 23 kg ha^-1 yr^-1. However, soybean yields up to 10,790 kg ha^-1 have been previously reported in yield contests. Research was undertaken to characterize soybean grown in a maximum yield environment so that crop growth and yield could be modeled with empirical physiological data. From 2011 to 2013, replicated small-plot research was conducted at the Univ. of Arkansas at Fayetteville with 11 to 15 modern, glyphosate-resistant cultivars with a range of relative maturities (RM) from 4.2 to 5.5. Physiological measurements included radiation use efficiency (RUE, g MJ^-1), N accumulation rate (g N m^-2 d^-1), specific leaf N (g N m^-2), dry matter allocation coefficient (DMAC, d^-1), grain yield and harvest index (HI). Crop models, Sinclair-Soybean (Sinclair et al., 2003) and DSSAT-CROPGRO (v.4.5.1.023) were used to model crop growth using weather data taken on site. In 2012, the observed mean RUE, N accumulation rate, specific leaf N, DMAC, grain yield and HI were 1.30 g MJ^-1, 1.041 g N m^-2 d^-1, 2.88 g N m^-2, 0.0061, 6746 kg ha^-1, and 0.47, respectively. Simulations using the Sinclair-Soybean model with default values underpredicted grain yield and HI both by 24%. Modifying the model with observed values of N accumulation rate, specific leaf N, and RUE overpredicted yield by 17% and underpredicted HI by 27%. CROPGRO was able to accurately predict grain yield (NRMSE < 4%) and grain N concentration (NRMSE < 9%) with input of N and water management and minor cultivar coefficients calibration; however, HI and N accumulation rates were underpredicted by 19 and 33%, respectively. These results indicate that both crop models exhibit some shortcomings but are capable of predicting grain yields as great as were observed.