An evaluation of the chill overlap model to predict flowering time in apple tree

- The sequential and chill overlap phenology models were compared.
- The sequential model did not perform better than modelling to average day-of-year.
- The sequential model predicted flowering earlier that observed at warm winter sites and later than observed at cool winter sites.
- The chill overlap model better captured the variability in flowering timing.

Reliable models of fruit tree flowering phenology that are stable across different climate conditions and for a range of cultivars are required to better anticipate and adapt to anthropogenically induced climate change. Two phenology models, the commonly applied sequential model and the newly defined chill overlap model, were investigated to evaluate their ability to represent 'Cripps Pink' apple flowering phenology data from a wide range of Australian climates. This is the first study to directly compare the chill overlap and sequential phenology models and the first flowering phenology study of this spatial size for apple in the Southern Hemisphere. The sequential model did not adequately explain the variability in the dataset (RMSE 14.7 days) and was a worse approach than modelling to mean full bloom day-of-year (RMSE 9.6 days). The chill overlap model fitted with a chilling requirement of 34 chill portions was better able to represent the observations with RMSE of 6.9 and 5.8 days for parameterising and validating datasets, respectively. Based on the results of this study, further exploration and application of the chill overlap model is warranted, particularly in relation to climate impact assessments and for the development of adaptation strategies.