To survive winter conditions, perennial plants must simultaneously suspend their growth (through the process of dormancy) and increase their frost tolerance (through the process of frost acclimation). Frost tolerance can therefore be modeled by a phenological approach considering a modulation of the hardening ability depending on the phenological stage (endodormancy induction and release, ecodormancy and growth). We tested the effect of the initial date for chilling accumulation to release endodormancy based on a large dataset of bud dormancy measurements (34 and 77 independent observations of endodormancy release and budburst dates, respectively). The model was then calibrated in three contrasted walnut genotypes Chandler, Franquette and Serr in France and Spain to predict current and future frost risks. However, the 'phenological' approach assumes the plant to be under optimal conditions. To account for the cumulative effects of sub-lethal stresses on phenological processes or on the hardening ability through carbon reserves, a osmohydric model has been developed based on the interaction between solute and water contents. Across different tree species, a model taking all soluble sugars and polyols into account was the most efficient. However, a model taking the individual contribution of each soluble carbohydrate molecule into account, was more efficient at the family or species scales. The osmo-hydric model was improved from a static to a dynamic model, simulating the response of carbon metabolism and water balance to environmental factors. The influence of temperature on the inter-conversion between starch and soluble carbohydrate has been modeled by exposing one-year-old branches of Juglans regia to constant temperature. The water content was simulated as the balance between root uptake and evaporative losses. These two flows were characterized experimentally, by controlling soil temperature and by measuring evaporative losses from excised twigs. These results allowed building a comprehensive model of the involved processes. The two complementary modeling approaches can provide key indicators for defining relevant ideotypes with respect to frost stress.