Submitted to: Agriculture, Ecosystems and Environment
Publication Type: Peer reviewed journal
Publication Acceptance Date: 11/1/2008
Publication Date: 5/1/2009
Citation: White, J.W. 2009. COMMENTS ON A REPORT OF REGRESSION-BASED EVIDENCE FOR IMPACT OF RECENT CLIMATE CHANGE ON WINTER WHEAT FIELDS. Agriculture, Ecosystems and Environment, 129:547-548. Interpretive Summary: Climate change research often appears controversial because it requires predicting the future with uncertain information. Rather than predicting the future, one option is to examine recent historic trends. This paper re-analyses data from a study in China to determine whether that paper’s conclusions on trends in wheat yields were valid. Regrettably, the conclusion that a warming trend had increased yield at two locations was found to be incorrect. The trend was only confirmed for one location, and a reported trend with annual rainfall also was not confirmed. These negative results emphasize the care needed in conducting credible research on climate change. The suggestion is further made that the results might be better analyzed using computerized crop simulation models.
Technical Abstract: In their 2008 paper entitled “Impact of recent climatic change ...”, Xiao et al. (Agric. Ecosyst. Environ. 127:37-42) presented evidence for decreasing grain yield of winter wheat (Triticum aestivum L.) from 1981 to 2005 for two locations in China. They further reported non-linear effects of rainfall and mean temperature on grain yield. Inspection of the graphs provided in the paper suggested that the reported responses were overstated. Re-analysis of the data in five graphs indicated that yield had increased only at the lower elevation site, that there was no effect of rainfall, and that mean temperature only affected yield at the higher site. Furthermore, although this effect was modeled with an exponential function (r2 = 0.32), a linear relation described the response well enough (r2 = 0.27) that there was no justification for emphasizing a “non-linear” response of yield to temperature. Overall, the re-analysis highlighted the difficulties with using regression to detect effects of climatic change on agriculture.