Repository logo
 

Predicting wine prices based on the weather: Bordeaux vineyards in a changing climate

Published version
Peer-reviewed

Type

Article

Change log

Authors

Roucher, A 
Aristodemou, L 

Abstract

Each grapevine cultivar needs a certain amount of cumulated heat over its growing season for its grapes to ripen properly. In the twentieth century’s Bordeaux vineyard, the average growing season temperature was not always sufficient, thus higher than usual summer temperatures were on average linked with higher grape and wine quality. However, over the last 60+ years, global warming gradually increased the vineyard’s temperatures up to the point where additional growing season heat is not required anymore, and can even become detrimental to wine quality: hence the positive effect of higher-than-usual summer temperatures has progressively vanished. In this context, it is unknown whether any weather variable is still a good predictor of a vintage’s quality. Here we provide a predictive model of wine prices, based only on weather data. We establish that it predicts more accurately a vintage’s long-term quality than a world-class expert rating this same vintage in the year following its production. We first design a corpus of features suited to the grapevine lifecycle to extract from them the most powerful drivers of wine quality. We then build a predictive model that leverages Local Least Squares kernel regression (LLS) to factor in the time-varying nature of climate impact on the grapevine. Hence, it is able to outperform previous models and even provides a better predictive ranking of successive vintages than the grades given by world-famous wine critic Robert Parker. This predictive power demonstrates that weather is still a very efficient predictor of wine quality in Bordeaux. The two main features on which this model is built – following grapevine’s phenological calendar and using a LLS architecture to let the inputoutput relationship vary over time – could help model other agricultural systems amidst climate change and adaptation of production processes.

Description

Keywords

climate change, grapevine, machine learning, local least squares kernel regression, phenology

Journal Title

Frontiers in Environmental Science

Conference Name

Journal ISSN

2296-665X
2296-665X

Volume Title

Publisher

Frontiers Media
Sponsorship
The paper was written by a former student from the MPhil programme based on his dissertation.