Climate change affects crop phenology, and this has raised interest in phenology models. Most previous studies have determined the functions and parameters of phenology models through calibrations based on observed historical bloom dates under field conditions. However, if all the functions and parameters are determined through bud response experiments in controlled environments, it becomes possible to develop a model that is independent of observation in specific regions and has a small risk of overfitting. Hence, we grew potted ‘Fuji’ apple trees under fully temperature-controlled condition to quantify the temperature response of endodormancy release, flower bud development, and their interaction to determine the functions and parameters of models. The optimum temperature to release endodormancy was between –3 °C and 6 °C, and it was not released at temperatures ≥12 °C. The developmental rate of the flower bud increased with increasing chilling treatment time at 6 °C; however, it did not change over 2000 h The relationship between the temperature and developmental rate was exponential for the ecodormancy stage. The models constructed from these experimental results were validated using a dataset of historical bloom dates observed under three different climatic field conditions in Japan. The results demonstrated that, regardless of whether the region was cold or warm, the overlap model can predict bloom dates with accuracy, whereas the sequential model can predict bloom dates with accuracy only in cold regions. The overlap model in which the functions and parameters are fixed through bud response experiments in controlled environments, is expected to be applicable for predictions spanning diverse temperature conditions in global apple-growing regions, including future temperature scenarios, without succumbing to overfitting.