Machine Learning

Could you predict if a particular wine would be of good or poor quality just by knowing certain climatic conditions during a growing season?

July 4, 2017

Ramon Julià – NomarData

Wine composition is determined by various conditions: grape variety, rootstock, soil type, cultivation techniques, and climatic characteristics. The first three conditions are generally constant due of the control exercised by regulating councils.

Climatic variables are the main influence on year-to-year variability. Although improvement in cultivation techniques allows for obtaining acceptable quality levels each year, climatic or meteorological conditions are largely responsible for inter annual variations in production and the quality of wines obtained in specific years.

Climate variability and wine quality

Recent trends of increased temperatures in South Europe mainly in Spain France,Italy and Portugal  are associated with the significant increase in the frequency and duration of heat waves, and the increase in the frequency of hot days and tropical nights, especially in spring and summer, together with a significant decrease in the frequency of cold waves. Moreover a predominantly negative tendency in precipitation indices was also detected. These trends and associated changes in temperature and precipitation regimes may exert strong influences on agriculture systems.

Meteorological conditions strongly affect viticultural activity, modifying grapevine responses and determining the quality and quantity of production. The analysis of meteorological information can provide viticulturists with operational and forecasting tools for improving the management of vineyards.

The present project was performed for the purpose of analyzing the relationship between meteorological information freely available on the Internet and the average quality of wine (defined by vintage ratings). Temperature and precipitation data were analyzed. The presence of statistical relationships and their effect on quality was investigated by surface temperature and meteorological indices such as North Atlantic Oscillation.

Penedes Region : Analysis of Wine Quality Using Freely Available Meteorological Information

The study will be focus in the Region of Penedes (Catalunya-Spain) interrnational know for the Sparkling wine CAVA.

There are three distinct areas: the Upper Penedès, near the Prelitoral Range, the Penedès Marítimo, by the sea and in the Cordillera Litoral, and the Penedès Central, between the two zones. This is why DO Penedès has a great diversity of microclimates, due to its proximity to the coast and its altitude. The climate is typically Mediterranean, soft and warm. The area of Penedès Marítimo (Baix Penedès and Garraf) is more tempered by the influence and proximity to the sea. The Upper Penedès (Alt Penedès, Alt Camp, Anoia and Baix Llobregat) enjoys more frequent precipitations and greater contrast between maximum and minimum temperatures. The Central Penedès (mainly Alt Penedès) is the compendium of both microclimates.

For the analysis we will take as reference the climatic conditions of the Upper Penedes. Given the climatic diversity in the DO Penedes we will use as proxy to measure the quality of the wine, the North Atlantic Index. The North Atlantic Oscillation (NAO) is a weather phenomenon in the North Atlantic Ocean of fluctuations in the difference of atmospheric pressure at sea level (SLP) between the Icelandic low and the Azores high. Westerly winds blowing across the Atlantic bring moist air into Europe. In years when westerlies are strong, summers are cool, winters are mild and rain is frequent. If westerlies are suppressed, the temperature is more extreme in summer and winter leading to heat waves, deep freezes and reduced rainfall.

Material and methods

Wine quality data series

The Cava Regulatory Council (RCC) ( determines Cava quality annually using four different categories: Excellent (EX), Very Good (VG), Good (G) and Fair (F). Table 1 shows the quality for 1970–2014 vintages.

Meteorological variables

Montly temperatures and precipitation used for this study were taken from the Catalan Metereological Service ( Vilafranca del Penedes was choseen as reference location and Monthly NAO index values were taken from


We performed a binary logistic regression (LR) of Cava quality on these averages of the NAO index between 1970 and 2014. The Cava quality was divided into two classes: excellent and very good quality, both regarded as top quality (designated as 1), and good and fair quality vintages (designated as 0).

All the anlalyses and simulation where done using the amazing BigML (, tool in the cloud.


                                                                                                   Table 1

Results and Discussion

The period of greatest biological activity of grape in during March- August in the Penedès area.

We can check if any particular month temperature or monthy NAO index might be particularly important for the quality of the Cava. Similarly, we calculated diferent NAO monthly windows to assess possible seasonal effects.

We performed a binary logistic regression (LR) of Cava quality on these averages of the NAO between 1970 and 2014. The Cava quality was divided into two classes: Excellent and Very Good quality, both regarded as top quality ( designated as 1), and Good and Fair Quality vintages (designated as 0).




The NAO threshold for clearly favoring a top quality Cava was <−0.25 (Fig 1) ,  whereas the NAO threshold for a year being clearly unfavorable for obtaining a top quality Cava was >0.41 .


                                                                                                                   Fig. 2


Results reveal a strong dependence of wine quality on maximum and minimum temperatures during spring and summer (the growing season) as expected. The threshold of the NAO index is strongly conditioned by the local meteorological conditions, as can be seen in Fig 2.

If we consider a NAO threshold <-0,25 and the maximum average temperatures is 21ºC or higher, the model predicts a top quality year with a 70% probability. If a NAO threshold is >0.41 and maximum average temperatures is around 19.6ºC or lower,  the model predicts a unfavorable quality year with a 70% probability.

If we consider a NAO threshold <-0,269 and the minimum average temperatures is 8.9ºC or lower, the model predicts a unfavorable quality year with a 72.89% probability. If a NAO threshold is >0.41 and minimum average temperatures is around 9.82ºC or higher the model predict an unfavorable quality year with a 71.58% probability.

A few interesting results came out of this study:

1)  A significant negative relationship between NAO for the months of March through August and the probability of having a high/top quality Cava for the given vintage.

A negative NAO is associated with increased rainfall in the Cava region.  This increased rainfall may help keep temperatures at a more mild level and help with a more prolonged maturation process.  A positive NAO, on the hand, is associated with warmer conditions and higher daily maximums in the Cava region (and throughout the entire South Europe)

2) Monthly and seasonal NAO values were not related to Cava quality.

While the NAO values over the entire growing season (March through August) proved significant in predicting the quality of Cava, month-by-month values and season-by-season values did not.

3) This results suggest that this type of analysis may be used in developing a tool that may help anticipating a vintage/high quality year, based on already available seasonal climate outlooks.

This results suggest that this type of analysis may be used in developing a tool that may help anticipating a vintage/high quality year, based on already available seasonal climate outlooks.


Wine quality can be evaluated via chemical analysis using objective techniques such as gas chromatography and the olfactometer. An alternative method for assessing wine quality is sensory analysis based on a subjective evaluation of colour, aroma, and taste to determine annual vintage ratings.

It can be said that, the NAO index (North Atlantic Oscillation)  does a good job in predicting the quality of Cava for a given vintage, though sometimes there are other factors involved, like climatic events, environmental conditions or winemaking techniques, that cannot be predicted by the NAO alone.

Deja una respuesta

Tu dirección de correo electrónico no será publicada. Los campos obligatorios están marcados con *