Developing and testing of a winegrape phenology model for Australian conditions

• Published winegrape phenology models predict phenology with Australian data poorly. • More robust models were developed using a range of single temperature functions. • Non-linear or temperature limited linear models were better predictors of phenology. • Models were further improved when within season phenological observations were used. • Phenology models will support industry to manage logistics and climate impacts. Within season vine management and longer-term strategic decisions for vineyards and wineries require an understanding of vine phenology and tools for its prediction. Using a curated dataset of 3,159 phenological observations for ten varieties and across seventeen regions, the largest reported Australian phenological data base, we aimed to develop phenological models for budburst, flowering, veraison and maturity for an Australian context. Published winegrape phenology models were initially tested with the Australian dataset with poor results. These models were then improved by examining different methods to calculate daily forcing units (F*) and alternative start day of accumulation of forcing units and then modelling a phenophase by commencing the accumulation of forcing units from an earlier phenophase. A range of single temperature function models were developed and evaluated. These models improved predictions, particularly when in season phenology observations were available to update the predictions through the season. The models we developed are robust, relatively easy to use, and better suited to an Australian context. Phenological models provide a useful tool to assist vineyard planning and therefore efficiency in the short term and contribute to long term planning in the vineyard and winery, particularly in a variable and changing climate.