Abstract
Growth prediction models are used nationwide to support rice cultivation management and to predict/evaluate rice production. Development prediction models are also used in Hokkaido. The development index (DVI) model is easy to use and has a high prediction accuracy. However, DVI models are limited to 4 rice cultivars with an earliness of “early to medium”. Therefore, we prepared the parameters of a DVI growth prediction model for rice cultivars with different levels of earliness for the Hokkaido region. During the preparation we encountered an issue because the development prediction model was not able to properly predict the difference in earliness between cultivars. This occurred due to a systematic estimation error, dependent upon the survey area of the growth data which was different between the cultivars. As a solution, we supplemented the growth data of the cultivars with the data of a “standard cultivar” to make the DVI model parameters. This was possible because the correlation between cultivars during the developmental stage of rice plants cultivated in Hokkaido is represented by a linear equation with a slope of almost 1. With the DVI model created using supplemented data, it became possible to accurately predict the difference in earliness between cultivars. This method is effective for creating development prediction models for cultivars whose developmental data is limited to survey areas and new cultivars with little data available.
