Phenological models are considered key tools for the short-term planning of viticultural activities and long-term impact assessment of climate change. In the present study, statistical phenological models were developed for budburst (BUD), flowering (FLO) and veraison (VER) of 16 grapevine varieties (autochthonous and international) from the Portuguese wine-making regions of Douro, Lisbon and Vinhos Verdes. For model calibration, monthly averages of daily minimum (Tmin), maximum (Tmax) and mean (Tmean) temperatures were selected as potential regressors by a stepwise methodology. Significant predictors included Tmin in January–February–March for BUD, Tmax in March–April for FLO, and Tmin, Tmax and Tmean in March–July for VER. Developed models showed a high degree of accuracy after validation, representing 0·71 of total variance for BUD, 0·83 for FLO and 0·78 for VER. Model errors were in most cases < 5 days, outperforming classic growing degree-day models, including models based on optimized temperature thresholds for each variety. Applied to the future scenarios RCP4·5/8·5, projections indicate earlier phenophase onset and shorter interphases for all varieties. These changes may bring significant challenges to the Portuguese wine-making sector, highlighting the need for suitable adaptation/mitigation strategies, to ensure its future sustainability.