Effect of fine scale spatial climate variability on modelled and observed grapevine phenology

Background and aims: Phenological prediction models can be used when assessing long-term vineyard management requirements in response to climate change. These models have been developed from data derived at a regional scale. However, there is uncertainty around the effect of fine scale variability in climate and factors associated with how climate data is collected and the accuracy of predictions using these models. The aim of this study was to test the efficacy of heat unit accumulation modelling to predict phenological events using two climate data sources, in addition to testing hypotheses related to fine scale climate variability engendered by topography.
Methods and results: Four sites comprising two topographically homogenous vineyards and two topographically diverse vineyards in three wine regions in Victoria (Australia) were studied across three growing seasons. A freely available database of interpolated Australian climate data based on government climate station records (Scientific Information for Land Owners, SILO) provided temperature data for grid cells containing the sites (resolution 0.05° latitude by 0.05° longitude, approximately 5 km × 5 km). In-vineyard data loggers collected temperature data for the same time period. Predicted budbreak and maturity from both sets of data were compared to actual observed budbreak and maturity dates. The results indicated that the only significant difference between the two climate data sources was the minimum temperatures in the topographically varied vineyards where night-time thermal layering is likely to occur. It was found that the two data sources gave very similar predicted budbreak and maturity dates in the topographically homogenous vineyards and fairly similar predicted dates for the topographically diverse vineyards. However, the models with data derived from both sources were only able to accurately predict budbreak in one of the spatially homogenous vineyards. All other predicted budbreak and maturity dates were inconsistent with actual recorded phenological phases.
Conclusion: SILO data closely matched the in-vineyard recorded maximum temperatures in all cases and minimum temperatures for the topographically homogeneous vineyards. However, minimum temperatures were not as accurately predicted by SILO for the topographically heterogeneous sites. Therefore, SILO data are a reasonable substitute for in-vineyard collected data for vineyard sites that are unlikely to experience night-time thermal layering, but in-vineyard monitoring of temperature may be superior for phenological modelling purposes where topography is more varied. As the chosen phenological model using both sources of data was not able to accurately predict budbreak or maturity, accurate phenological predictions may require localised parameterisation for heat accumulation and base temperatures for a range of terroir conditions and cultivars.

Significance of the study: Simulations of the effect of climate change on grapevine phenology can aid long-term vineyard management planning. Access to accurate climate data from a free source such as SILO will provide a valuable tool to manage blocks or sections within vineyards more precisely for the vast number of vineyards that do not have a weather station on site. In topographically varied vineyards, care is required to account for discrepancies in minimum temperature data due to the potential for cool air pooling at night. Phenological modelling at individual locations likely requires parameterisation relating to the environmental and cultural terroir specific to the vineyard.