Taking account of water temperature effects on phenology improves the estimation of rice heading dates: Evidence from 758 field observations across Japan

Water temperature (T_w) plays a key role in growth and development of plants inhabiting flooded environment, but most phenology models use air temperature (T_a) for phenology prediction. Gaps between T_w and T_a are known to differ regionally, but regional differences in the importance of using T_w for phenology prediction are not known. This study attempts to determine whether the use of T_w improves the prediction of heading time, by using 758 field observations across Japan and estimated T_w from a nationwide weather database (MeteoCrop Database). We have confirmed that the use of T_w improved the accuracy of prediction by 1.0-2.4 days as measured by the root-mean-square error, but the degree of improvement was similar at 23-41% across different latitudes, altitudes, or planting times, likely because the T_w-T_a difference is highly variable even at similar latitudes or altitudes. The models proposed here and the nationwide database of future climate projection will help to reduce the uncertainty in predicting crop calendar for a range of climatic conditions.