Temperature-based grapevine sugar ripeness modelling for a wide range of Vitis vinifera L. cultivars

Increasing temperatures due to climate change are leading to advances in grapevine phenology and sugar accumulation in grape berries. This study aims to (i) determine if a temperature-based model can predict the time to target sugar concentrations from 170 to 220 g/L for Vitis vinifera L., (ii) use the best model to characterise the time to the specified target sugar concentrations for a wide range of cultivars with statistical evaluation of each cultivar's parameterisation, and (iii) establish cultivar classifications based on these thermal times to the specified target sugar concentrations.

The Day of the Year (DOY) to reach the specified target sugar concentrations (170, 180, 190, 200, 210 and 220 g/L) was determined from time series of sugar concentrations collected from research institutes, extension services and private companies. Models were fitted for the species Vitis vinifera L. The two best-fit models for the DOY to reach the target sugar concentrations were selected using the Akaike Criterion (AIC) (evaluates model complexity and goodness of fit within one criterion) and assessed for model efficiency (EF) and error of prediction (RMSE, root means squared error) followed by a sensitivity analysis and model validation. The models were then parameterised for individual cultivars.

The best model across all target sugar concentrations was the non-linear best Sigmoid model "best SIG"’ model (parameters: start date (t0) = 86, d =−0.1294, e = 14.87). The best linear (Growing Degree Days) model was also selected which represents the model that required the least parameters and therefore the simplest in application for winegrowers. This model was termed the “Grapevine Sugar Ripeness” model (GSR) (parameters: base temperature (Tb) = 0 °C, start date (t0) = 91 or 1 April, Northern Hemisphere). Both models performed better than the Winkler and Huglin growing degree day models.

Sixty-five cultivars were classified for the thermal time to one or more of the six sugar targets using these two models. Fifty percent of all combinations of cultivar and time to target sugar concentrations had EF values greater than 0.5 and RMSE values less than seven days. Confidence intervals were calculated for cultivars where there was sufficient data for the thermal time to target sugar concentrations. The classifications generated from both models provides the opportunity to implement either model to support cultivar choice in response to concerns of climate change and may provide cultivar solutions to issues of harvesting grapes at high sugar concentrations with resultant higher alcohol wines.