|Evett, Steven - Steve|
Submitted to: Agricultural Water Management
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/18/2008
Publication Date: 2/18/2009
Citation: Kang, S., Payne, W.A., Evett, S.R., Robinson, C.A. 2009. Simulation of winter wheat evapotranspiration in Texas and Henan using three models of differing complexity. Agricultural Water Management. 96(1):167-178. Interpretive Summary: Winter wheat is an important crop in the U.S. Great Plains and is often irrigated, bringing into question the amount of irrigation needed for wheat as well as the water use efficiency of wheat production under irrigation – that is, how much wheat is produced per unit of water applied. Answers to these questions are important for producers and policy makers in deciding how to allocate scarce water resources. We evaluated three computer models to see how well winter wheat growth and water use were estimated compared to data measured at Bushland, Texas. To add a stronger test of the models, we added comparisons of model outputs to winter wheat data from northern China, also an important winter wheat production area. None of the models estimated winter wheat water consumption well. We showed that the poor estimation of water use was tied to poor estimation of crop growth by the models, and we made recommendations to model developers as to which computer algorithms needed to be strengthened to improve the usefulness of these widely used models.
Technical Abstract: Crop evapotranspiration (ET) is an important component of simulation models with many practical applications related to the efficient management of crop water supply. The algorithms used by models to simulate ET are of various complexity and robustness, and often have to be modified for particular environments. We used three crop models, CROPWAT, MODWht and CERES-Wheat, to simulate ET of winter wheat (Triticum aestivum L.) grown at Bushland, Texas, and Zhengzhou, China, during multiple growing seasons in which ET was measured using lysimeter or soil water balance techniques. None of the models simulated daily ET well at Bushland or Zhengzhou as indicated by high mean absolute differences (MAD > 1.1 mm) and root mean squared errors (RMSE > 2.0 mm). The three models tended to overestimate daily ET when actual ET was small, and to underestimate daily ET when actual ET was large. CROPWAT better predicted total ET than MODWht and CERES-Wheat, with respective MAD and RMSE values of 35.1 mm and 73.3 mm. The fitted values of daily crop coefficients (Kc), calculated from daily ET and reference ET, were very similar to those of Allen et al. (1998) although some Kc were overestimated ( >/- 1.0). Leaf area index (LAI) was poorly simulated by MODWht and CERES-Wheat, especially when using the Priestly-Taylor method to estimate reference ET. Poor overall ET simulation of three models was associated with poorly estimated values of reference ET, Kc and LAI as well as their interactions. The careful calibration of ET parameters in the model algorithms and possible in-season calibration such as LAI would help further improving the performances of crop model simulations.