AbstractThe buds of perennial plants become dormant in autumn and must integrate the information related to chilling and forcing temperatures to resume their growth in spring. In many studies, the initial date for chilling accumulation (DCA) is set arbitrarily using various rules resulting in high variability across studies and sites. To test the relevancy of different rules to set DCA, sequential models (taking into account or not the negative effect of warm temperature) were optimized by minimizing the sums of squares between observed and predicted values for 34 endodormancy release and 77 budbreak dates for the walnut Juglans regia L. cv Franquette across France. Optimization of these different models highlighted that many of the DCA rules, incorporating a photoperiod signal on endodormancy induction, were effective (predicted root mean square standard error less than 10 and 8 days for endodormancy onset and bud break, respectively). Furthermore, the use of functions that compute negative chilling accumulation did not improve the performance of the models. Among the different rules, the projections of the best models were explored under different climates (current climate and Representative Concentration Pathways RCP scenarios). The projections revealed a tipping point at a mean annual temperature between 13 and 15°C, beyond which the advance in ontogenic development during ecodormancy does not compensate for the delay in endodormancy release. Although the physiological mechanisms driving the onset of endodormancy may be profoundly altered by global change, they appear to have minimal impact on the way current models predict dormancy and budbreak dates in walnut.