Base temperature (Tb) selection for the field application of the growing-degree-day (gdd) model for predicting alfalfa (Medicago sativa L.) harvest under a three-cutting management system was investigated, because current usage is based upon extrapolation of growth-chamber data and employs a constant Tb despite microclimatic and physiological changes that take place during the growing season. Three methods of analysis and 11 station-years of climatic and alfalfa one-tenth-flower data from Rosemount and St. Paul, Minnesota, were used for successive growth periods to estimate Tb. The best estimate of Tb in the spring growth period, 3.5°C, was significantly less than the 5°C value currently used. For the succeeding summer periods, Tb estimates were 7.5° and 10.0°C, the latter being significantly higher than the current Tb. The gdd requirement, computed using the best estimates of Tb, were 585, 425 and 425 for alfalfa to reach one-tenth flower in each successive period. When a constant Tb was used, the gdd requirement to flower differed between growth periods, with the late summer period having a greater requirement.