CR 64:187-199 (2015)  -  DOI: https://doi.org/10.3354/cr01314

ABSTRACT: Recent studies have shown a generalized impact of multi-decadal climatic variability on the growing-quality of world wines during the twentieth century. This information has been used to predict the effects of climate change, mainly rising temperatures, on the future spatial distribution of wine regions and constraints on wine quality. On the other hand, some wine experts have identified the adaptive technological revolution of the wine industry as the main factor explaining rising wine quality; however, this hypothesis remains untested. Here, a time-varying coefficients regression approach is developed that is capable of partitioning the temporal effects of temperature on wine quality across sequential grape growing seasons, accounting for wine-rating uncertainty. Using data from the Bordeaux wine region from 1920 to 2009, it is shown that this model statistically outperforms the previously used constant regression models: a period characterized by strong effects of temperature on wine quality up to the mid-1960s was followed by an abrupt and steady decline to present marginal temperature effects. This pattern can be explained by key adaptive shifts implemented in the Bordeaux wine industry from the 1960s onwards, which likely dampened the relative impact of temperature on the inter-annual variability of wine quality. Additionally, estimated optimum growing-season temperatures are higher and more uncertain than previously modeled. These results suggest that previous approaches likely overestimated the effects of growing-season temperatures on Bordeaux wine quality during recent years. Consequently, the projected effects of future climate change on viticulture might be biased in this region.

KEY WORDS: Bayesian analysis · Climate change · Grapevine · Growing-season temperature · Hamiltonian Monte Carlo · Observation error · Time-varying coefficients · Viticulture · Wine quality rating