Accurate harvest time prediction is a challenge for developing sustainable fruit production, especially for both dessert and plantain banana varieties. The proposed approach to predict the optimum harvest date was based on estimating appropriate parameter values, namely the threshold temperature and the corresponding degree-day sum, for a given cultivar and targeted market. When the threshold temperature was estimated from a sufficiently wide and contrasted dataset, as illustrated for the variety 938, the model achieves high predictive accuracy. The thermal time sum model can accurately predict the flowering-to- fruit-ripening phase of five dessert banana varieties moderately to strongly resistant to black leaf streak and fusarium wilt TR4 diseases and for the first time of one of plantain banana variety, cv. French Clair, with values of RMSE, MAPE, and R² ranging, respectively, from 3.8 to 8, from 3% to 6.8%, and from 0.51 to 0.85 on the validation datasets. This study proposes a simple and reliable temperature-based method to optimally forecast banana harvest date according to the targeted market. The calculated thermal time sum from the variety-dependent threshold temperature accurately predicted the fruit storability and the fruit size. Moreover, an adapted statistical method was proposed in the case of inability of recording data on the flowering-to-fruit-ripening phase, based on the occurrence of fruit splitting obtained under contrasting temperatures in different geographical areas. Threshold temperature values for each variety were discussed, and it was determined that this is a purely statistical parameter for prediction, without direct biological meaning.