Six phenological series were available for 'Golden Delicious' apple blooming at six sites in Trentino, an alpine fruit-growing region. Several models were tested to predict flowering dates, all involving a "chilling and forcing" approach. In many cases, application of the models to different climatic conditions results in low accuracy of prediction of flowering date. The aim of this work is to develop a model with more general validity, starting from the six available series, and to test it against five other phenological series outside the original area of model development. A modified version of the "Utah" model was the approach that performed best. In fact, an algorithm using "chill units" for rest completion and a thermal sum for growing-degree-hours (GDH), whose efficiency changes over time depending on the fraction of forcing attained, yielded a very good prediction of flowering. Results were good even if hourly temperatures were reconstructed from daily minimum and maximum values. Errors resulting from prediction of flowering data were relatively small, and root mean square errors were in the range of 1-6 days, being <2 days for the longest phenological series. In the most general form of the model, the summation of GDH required for flowering is not a fixed value, but a function of topoclimatic variables for a particular site: slope, aspect and spring mean temperature. This approach allows extension of application of the model to sites with different climatic features outside the test area.