Abstract
An alternative approach for predicting the heading date of rice plants was developed by using the 30-yr averaged normal heading date for the input of a developmental index model instead of using the information on the rice transplanting date and the developmental index on that date. This approach is based on the fact that the normal heading date has less spatial heterogeneity compared with the standard input variables and facilitates prediction of the spatial variation of the heading date. The model predicted the inter-regional variation of the heading date in the Tohoku district of Japan within 2 days for its root mean square errors (RMSE), which is similar to that of the traditional developmental index (DVI) model, and well reproduced spatial variation of the heading date over Tohoku district in extremely hot/cold years, although further validation is needed to prove the model accuracy.
