Evaluation of the most appropriate spatial resolution of input data for assessing the impact of climate change on rice productivity in Japan

Abstract

　Process-based crop growth models are increasingly utilized as an essential tool for assessing the impact of climate change on crop productivity at field, regional, and national scales. The reliability of model predictions depends strongly on the quality of the meteorological data used as inputs. For evaluations over large areas, the spatial resolution of input data affects the calculation results because factors such as elevation differences between the mean for an entire grid cell and the portion of crop land in the grid can introduce a major temperature bias in the input data. In this study, we attempted to identify the most appropriate spatial resolution to support assessment of the impact of climate change on rice productivity in Japan. We used the Hasegawa-Horie rice growth model under the baseline climate conditions (1981 to 2000) and then applied the model to account for temperature increases to 1 and 3 °C higher than the baseline. First, we calculated the rice yield using inputs at 100-m resolution as the “true value”. We then compared the rice yield calculated using inputs at 10-km and 1-km resolutions with the yield calculated using inputs at 100-m resolution. We found that the yield differences were larger with 10-km resolution than with 1-km resolution in areas that had complex terrain, but the differences were small in homogeneous flat areas. Where the terrain is extremely complex, regional mean yields were underestimated by 11.5% compared with the yield under baseline climatic conditions but were overestimated by 5.4% at increased temperatures using 10-km resolution. These differences are likely to be a major cause of uncertainty in predicting the impacts of climate change on yield at a regional scale. Spatial resolution of input data, using 10-km resolution did not affect the assessment results when yield is aggregated at a national scale.