LAUSR

Abstract. Complex crop models have been developed to simulate the interactions among biophysical processes and to extend experimental results beyond the local soil and climate conditions. However, in-depth studies on a model’s capability to predict crop growth under different conditions are sparse, and the question of whether a crop model outperforms a simple water production function (WPF) has not been answered. The objective of this study was to compare the predictive ability of a complex crop model with simple WPFs for yield and biomass estimation at three sites (Greeley, Fort Collins, and Akron) in eastern Colorado. Specifically, the CERES-Maize crop model in the Root Zone Water Quality Model (RZWQM2), which has been applied extensively in eastern Colorado for simulating maize growth, was compared to crop WPFs based on irrigation and rainfall amounts during growing seasons. Results showed that the predictive ability of CERES-Maize depended on which datasets were used for model parameterization, and that WPFs in general performed as good as or better than CERES-Maize based on a modified F-test after considering experimental uncertainties. The ability of CERES-Maize and the WPF derived from Greeley (2008-2011) to predict maize yield in Greeley (2012-2013), Fort Collins (2006-2010), and Akron (1984-1986) depended on year and site. WPFs outperformed CERES-Maize for Greeley (2012-2013) and Fort Collins (2006-2010) but performed similarly for Akron (1984-1986). This study also identified the need to improve crop model responses to water stress, especially at different growth stages, for cropping systems models to be adequate for estimating the impacts of irrigation management on yield. Ultimately, the choice between the use of a complex crop model and a simpler WPF depends on the purpose of the user and the required accuracy. Keywords: Biomass, CERES-Maize, DSSAT, Grain yield, Irrigation management, RZWQM, Water production function.