A Methodology for Determining Suitable Heat-unit Requirements for Harvest of Highbush Blueberry

Abstract

Fifteen years of Michigan harvest data for highbush blueberry (Vacciniun corymbosum L.) were used in conjunction with daily maximum and minimum temperatures to determine appropriate heat-unit models for first-picking dates of 13 cultivars. For each cultivar, an optimal heat-unit model was chosen after evaluating the performance of a standard method with 72 combinations of three variables: a) starting date for the heat-unit accumulations (SDATE), b) low-temperature threshold (TLOW), and c) high-temperature threshold (THIGH). The optimal model sought to include the most important criteria values with respect to model performance and to minimize the average square of the prediction error (days) and the range in that error. Compared with a strict calendar-day method of estimating harvest dates, the heat-unit models reduced the standard deviation of the prediction error from 22% to 69%, depending on cultivar.