Predicting Wheat Growth and Nitrogen Use with an Ensemble of Crop Simulation Models.

Crop growth simulation models are widely used to study the response of environmental factors on cropping systems, but until recently the use of multiple models has been quite rare. Twenty-seven wheat models were used in a Wheat Pilot Study as part of the Agricultural Model Intercomparison and Improvement Project (AgMIP; www.agmip.org). Simulations were carried out for four contrasting wheat growing climates, represented by Wageningen (The Netherlands), Balcarce (Argentina), New Dehli (India) and Wongan Hills (Western Australia). The data include grain yield, total above ground biomass and nitrogen, grain protein concentration, leaf area index, and plant available soil water content. We used the multiple simulations to address two questions. First, does the variability between models give a realistic estimate of the uncertainty in predictions? Second, can one improve predictions by using the mean or the median of the model results?