摘要：Ongoing climate change is likely to enhance the deterioration of rice quality that has been
observed in western Japan, especially in Kyushu, since the 1990s. Therefore, it is important
to examine the response of rice quality to environmental variation over a wide
geographical domain. To that end, the aims of this study were (i) to propose a statistical
model to predict rice quality based on temperature, total radiation during the
ripening period, and their multiple effects; and (ii) to evaluate the model validity and
uncertainty in prediction. A Bayesian calibration was adopted to account for
uncertainty in the parameter values associated with non-climatic factors. The validation
results showed that the model performed well in capturing the temporal trend and
interannual variation in observed rice quality in all prefectures of Kyushu. We
then performed the prediction experiment for rice quality in the extremely hot
summer of the year 2010, which was omitted from the model calibration data. The
results showed that the predictive capability of the statistical model is somewhat
dependent on the calibration data, but this dependency does not necessarily mean
that useful predictions for climates not in the calibration data are impossible.
