[The article below is co-authored by Cornelis van Leeuwen, Laure de Rességuier, Renan Le Roux and Hervé Quénol]
To complete the climatic zoning proposed by Bois et al. (2018) on the whole of the wine-producing Bordeaux area, a more detailed climatic study has been conducted on the appellations Saint-Émilion, Pomerol and their satellites (European project “Life ADVICLIM”; https://www.adviclim.eu). Winegrowers can thus observe climate variability at the plot scale and adapt their grape varieties and vineyard management practices. Within the framework of this project, climate projections were made up to the end of the 21st century for different greenhouse gas emission scenarios and the predicted consequences on the behavior of the vine were evaluated. This work was presented in the form of a “Story Map”, i.e. a set of interactive maps that allow the visualization of climate variability in the study area, climate projections, but also other information such as the soil map (https://www.adviclim.eu/storymap).
To conduct this study, 90 temperature sensors were installed in vineyard plots in 2012 to measure temperature as close to the vegetation as possible (Fig. 1). The sensors were positioned according to the terrain parameters that influence temperature distribution at this scale. They are located in the vegetation to best relate temperatures with vine development. For this reason, the absolute values recorded are not directly comparable with the temperatures measured in a conventional weather station. The temperatures were correlated with topographic data from a digital elevation model (DEM), which allowed to use the topography as a spatialization covariate and to produce maps with a very high accuracy. More details on the methodology used are provided in de Rességuier et al., 2020.
Figure 1 – Delimitation of the study area and location of temperature sensors on a topographic background
LARGE AND SPATIALLY STRUCTURED TEMPERATURE DIFFERENCES
Large differences were recorded between sensors, particularly for minimum temperatures, where a difference of 10°C could be reached between the warmest and coldest sensor in the sector, on anticyclonic days with no wind. Over the growing season (April 1 – September 30), an average difference between sensors of 2.6 °C was observed for minimum temperatures, and 2.1 °C for maximum temperatures (average 2012-2018).
The spatial distribution of minimum temperature over the growing season reveals that plateaus and south-facing slopes are warm areas in contrast to plains and valleys which are cooler areas (Fig. 2A). There is also a west-east gradient, with the areas of Libourne, Pomerol and Lalande-de-Pomerol being warmer despite their low altitudes. The cooler areas for minimum temperatures are more exposed to spring frost.
The spatial distribution of maximum temperature is opposite to that of minimum temperature (Fig. 2B). Altitude strongly impacts this distribution with the warmest temperatures located on the plains and the coolest temperatures on the plateaus. The sectors with the greatest amplitudes between minimum and maximum temperatures are located in the valleys and in the plains, with the exception of the western sector, which is warm for both minimum and maximum temperatures.
Figure 2 – Spatial distribution of average minimum (A) and maximum (B) temperature
over the growing season from April to September (2012-2018)
HUGLIN VEGETATION INDICES VARY GREATLY
The vine needs a certain amount of heat to ensure its development and for the grapes to ripen in good conditions. Bioclimatic indices have been created to characterize the climate of wine growing regions. The Huglin index is a widely used index that takes into account the daily average and maximum temperatures above 10°C from April 1 to September 30 (Huglin and Schneider, 1998). It is a sum of daily temperatures, whose unit is the degree-day (°C.day). It should be noted that the index calculated from the temperatures of the sensors installed in the vegetation is not comparable to the values of the Huglin index calculated from conventional weather stations. For this reason, the index was renamed “Huglin vegetation index” when the temperature data are from sensors installed in vegetation (de Rességuier et al., 2020). The difference between the sensor with the lowest and highest vegetation HI is 295 °C.day (average 2012-2018).
Figure 3 represents the distribution of the Huglin vegetation index over the study area. It can be seen that the southwestern sectors are warm sectors in contrast to the northeastern part of the study area. The south facing slopes are also warm. Spatial modeling tends to reduce the differences, so the amplitude is not the same as that measured at the sensors, but the map allows us to clearly visualize the earliest (because warm) and latest (because cool) sectors. This map can help winegrowers in their choice of plant material (grape varieties and rootstocks).
Figure 3 – Spatial distribution of the Huglin vegetation index (average 2012-2018).
OBSERVED EFFECTS ON VINE DEVELOPMENT
Field observations were made on eighteen Merlot plots located around temperature sensors that contrasted in terms of temperature. The results show that a difference of 295 °C.day on the Huglin vegetation index corresponds to a 25-day difference in technological maturity (Fig. 4). The differences on flowering and veraison are smaller with respectively 9 and 13 days of average duration (2012-2016).
Figure 4 – Observations of phenology (bud break, flowering, veraison, modelled maturity) from 2012 to 2016. Modelled maturity corresponds to the day when the grapes contain 200 g/L of sugar. It should be noted that the harvest usually takes place about two to three weeks later, with higher sugar contents. Nevertheless, the “200 g/L sugar” date enables to compare plots and vintages to each other for the earliness of their technological maturity.
THE RECENT EVOLUTION OF THE CLIMATE IN BORDEAUX
It is interesting to look at the recent evolution of the climate in Bordeaux to better interpret these results. Between the period 1956-1986 and 1987-2017, the mean temperature increased by 1.5 °C and the Huglin index moved from the temperate class to the warm temperate class with an increase of 311 °C.day (Fig. 5). The climatic variability measured over the area is thus as important as the recent evolution of the climate in Bordeaux. This means that today in the study area, the coolest sectors have Huglin indices that correspond to those of the warmest sectors thirty years ago.
Figure 5 – Recent evolution of the average temperature and the Huglin index
calculated from the Mérignac station (Météo-France).
CLIMATE CHANGE SCENARIOS STUDIED
In order to project into the future, temperature modeling is performed according to different scenarios that are more or less optimistic (IPCC, 2013). In the ADVICLIM project, the RCP 4.5 and RCP 8.5 scenarios were used. The first, which projects a temperature increase of 1.1°C to 2.6°C at the end of the 21st century, is a scenario in which there is a stabilization of greenhouse gases (GHGs) emissions by 2040, followed by a decrease. The second is a less virtuous scenario in which GHGs emissions continue to increase until the end of the century, resulting in an increase of 2.6°C to 4.8°C by 2100. Given the difficulty of individual states in meeting their commitments to reduce greenhouse gas emissions, we are now more on the path of the RCP 8.5 scenario.
FUTURE PROJECTIONS AT LOCAL SCALE
Climate change projections with the most accurate spatial resolution (10 km) currently available are those from Euro-Cordex. At this scale and at this study site, it is not possible to visualize local temperature variability since this area is represented by only 4 pixels. Within the framework of the ADVICLIM project, the models developed to spatialize temperatures were used to refine the scale of climate change projections according to the two scenarios selected (RCP 4.5 and RCP 8.5) for three time periods: one historical (1986-2005) and two future (2031-2050 and 2081-2100). These maps make it possible to project into the future at the scale of this territory (Fig. 6). In the medium term (2031-2050), whatever the scenario, the Huglin index should increase by 180 °C.day, which corresponds to the HI of the 2011 vintage in this area. Conversely, there is a strong difference between the two scenarios for the projections at the end of the century (2081-2100). The models predict an increase in HI of 280 °C.day for scenario RCP 4.5 (equivalent HI of the 2003 vintage) and 680 °C.day for scenario RCP 8.5. The HI projected by this scenario has never been recorded in the Bordeaux area. Another important point is that the hottest and coolest sectors remain identical to the historical period, whatever the scenario and the projection period. These maps are useful tools for projecting into the future and anticipating adaptations, notably in terms of plant material (grape varieties and rootstocks) or management practices.
Figure 6 – Downscaling of Huglin Index models from Euro-Cordex data
(Le Roux, R., unpublished data).
ADVANCE IN PHENOLOGY
Since grapevine development is strongly linked to temperature, models have been developed to predict, from temperature sums, the dates of flowering, veraison and the day when a certain sugar concentration is reached (Grapevine Flowering Veraison -GFV- and Grapevine Sugar Ripeness -GSR- models, Parker et al., 2011; 2013; 2020). These models have been calibrated for a large number of varieties. In this study, the GFV model was used to predict the advance of flowering and veraison under the different periods and scenarios studied. The results show little difference between the two scenarios (RCP 4.5 and 8.5) in the near future (2031-2050) (Fig. 7). However, in the long term (2081-2100), there is a strong difference, with important consequences for vine development. Under the scenario RCP 8.5, flowering could be advanced by 13 days and veraison by 20 days over the period 2081-2100.
Figure 7 – Flowering and veraison evolution modeled from Euro-Cordex data using the GFV model (Parker et al., 2011, 2013)
The GSR model (Parker et al., 2020) was used to predict the day when Merlot, Cabernet franc, and Cabernet-Sauvignon will reach 220 g/L sugar. Harvesting typically takes place one week after this sugar level is reached. The dates of technological maturity are very similar for the three grape varieties for all periods in scenario RCP 4.5, and for the period 2031-2050 in scenario RCP 8.5 (Fig. 8). Following the latter scenario over the period 2081-2100, the dates of technological maturity will be much earlier, with technological maturity for Merlot being reached as early as August 12.
Figure 8 – Projection of the evolution of the technological maturity (220 g/L) of the majority grape varieties of the appellations studied according to the different scenarios.
An important lever for adaptation to the consequences of climate change is the choice of grape variety. Among the grape varieties authorized in the appellations of the study area, the later varieties (Cabernet franc and Cabernet-Sauvignon) are more adapted to high temperatures than an early variety such as Merlot. The latter is increasingly likely to produce wines with very high alcohol levels and cooked fruit aromas (van Leeuwen et al., 2022). The integration of a higher proportion of Cabernet franc and Cabernet-Sauvignon in the vineyard is more urgent in the warmer areas of the study site, while Merlot will be able to resist longer in the cooler areas. The RCP 8.5 scenario predicts temperatures that are too high by the end of the century even for Cabernet franc and Cabernet-Sauvignon. In order to produce high-quality wines, it may then be necessary to introduce new grape varieties, even later than Cabernet-Sauvignon. Other adaptations to warmer temperatures are possible, such as growing the vines with taller trunks or using later rootstocks. It would also be interesting to select later-ripening and less sugar-producing clones (van Leeuwen et al., 2019).
