To construct a new calibration method that combines usability and accuracy for estimating herbage mass from rising‐plate meter readings, we derived four models differing in the way their parameters are related to sampling date and compared their estimation accuracies using cross‐validation. The parameters of the linear regression for each sampling date showed seasonal variations, which had a steep decrease from early April to early June and a gradual increase thereafter. The pooled models were less accurate for estimating herbage mass than a separate model, which had specific parameters for each sampling date (S model). Among the pooled models, however, those in which the parameters were assumed to be linear functions (PL model) or combined functions (PC model) of the sampling date showed substantively improved estimation accuracy compared with the traditional pooled model, in which the parameters were assumed to be fixed throughout the year (PF model). Moreover, at the beginning of the season, the models derived from previous years' data were suggested to be applicable as a practical method. Thus, it can be concluded that these types of pooled calibration could be used as ‘compromise methods’ that combine both accuracy and usability.