Comments on “Testing winter wheat simulation models predictions against observed UK grain yields” by Landau et al. (1998)

• Published in: Agricultural and forest meteorology 1999-08, Vol.96 (1-3), p.157-161
• Main Authors: Jamieson, P.D, Porter, J.R, Semenov, M.A, Brooks, R.J, Ewert, F, Ritchie, J.T
• Format: Article
• Language: English
• ISSN: 0168-1923
• DOI: 10.1016/S0168-1923(99)00060-X

In a recent paper in this journal, Landau et al. (1998) tested the ability of three wheat simulation models to predict the variation in yield of winter wheat from a large number of sites in the UK from 1976 to 1993. They found no correlation between modelled and observed yields, and only weak correlation between the predictions of the individual models. They concluded that current models are unreliable for basing advice or policy on, in particular, the impacts of climate change on cereal production. Given the cloud-like scatter in plots comparing observations with predictions (Fig. 3, Landau et al., 1998), these conclusions seem superficially reasonable. However, their reasonableness depends to a large extent on the quality of the data used; whether or not assumptions about the conditions under which crops were grown were valid, and if it was legitimate to expect the models to mimic the observed variation. We contend that the observed yield variations were not caused by direct effects of weather on wheat physiology, but by secondary effects, such as the presence and effects of pests and/or diseases or anaerobic soil conditions, factors that the models have never claimed to address (e.g. Porter, 1993). We further contend that Landau et al. (1998) used versions of two of the models where their calibration for the chosen variety was incorrect, and that this caused the apparent differences in response among the models.