A simulation model of the wheat crop in response to water and nitrogen supply: II. Model validation

Abstract

The simulation model of soil water, mineral nitrogen supply and crop response of the wheat crop described in Part I was validated against field data from two semi-arid sites in northwest Victoria, Australia, over a period of 4 yr. The experimental data consisted of the four fallow combinations of stubble retention and zero tillage plus a tilled summer fallow prior to pea and wheat crops in rotation. Total soil water was simulated well (R 2 = 0·95; RMSE = 0·02 Mg/m 3) with a generally good characterization of the soil water profile. Some under-estimation of water content occurred at depth during the final crop growth stage as a consequence of over-estimation of crop growth and water use. Total soil mineral nitrogen (SMN) was simulated with less accuracy than soil water (R 2 = 0·61; RMSE = 1·84 g/m 3). Although the simulation of surface (0–10 cm) SMN was generally poorer, the model gave a good account of SMN at depth that ensures good simulation of nitrogen uptake late in the growing season when nitrogen responses are important to crop yield. The simulation of biomass clearly followed the general pattern observed throughout the season (R 2 = 0·77; RMSE = 1655 kg/ha) with final biomass closely matching the observed data in most years. Grain yield was simulated with less accuracy than biomass (R 2 = 0·42; RMSE = 756 kg/ha). This was largely because grain number was poorly simulated (R 2 = 0·07; RMSE = 1425 m −2) and when observed grain number was used in the model the accuracy of yield was increased (R 2 = 0·59; RMSE = 615 kg/ha). Simulated grain size (R 2 = 0·61; RMSE = 5·6 mg) was more accurate than grain nitrogen concentration (R 2 = 0·29; RMSE = 0·31% N). Seasonal crop water use was simulated well (R 2 = 0·83; RMSE = 34 mm). There was no evidence of bias between the various tillage methods tested in the simulations with respect to the soil water and nitrogen and crop response.