Performance of seasonal forecasts of Douro and Port wine production
Introduction
Seasonal weather forecasts are becoming increasingly important across a wide range of sectors, such as agriculture, energy, water resources and insurance (Doblas-Reyes et al., 2013; Turco et al., 2017). Prediction of extreme events (such as the 2003 heatwave) on the seasonal time scale still represents a challenge in the extra-tropical regions (Weisheimer et al., 2011). However, new findings show the potential for a better understanding of the spatial and temporal features of these climatic events, along with associated precursors (Scaife et al., 2014; Prodhomme et al., 2016; Wang et al., 2017). The skill of seasonal forecasts is generally limited in Europe, but there are regions and seasons where significant skill appears as a result of processes like the ongoing climate change and/or soil processes, amongst others. Consequently, seasonal forecasts are an added value for the agricultural sector across Europe (e.g. Ceglar et al., 2018; Falloon et al., 2018). Although the assessments of seasonal forecast skills applied to the winemaking sector are still incipient, they are of foremost relevance, owing to the importance of this sector in the economy of many regions worldwide.
More specifically in Portugal, vitiviniculture is a key socioeconomic sector. According to the most recent data from the Portuguese governmental authorities (Instituto da Vinha e do Vinho, IVV; http://www.ivv.gov.pt/), the vineyard area in Portugal is of roughly 187 000 ha (the 10th largest national vineyard area in the world). Portugal is the 11th wine producer and the 9th wine exporter worldwide (OIV, 2018). Although Mediterranean climatic conditions prevail, a large diversity of viticultural “terroirs” (van Leeuwen et al., 2004) and wine typicity (Drappier et al., 2019) can be found. Different mesoclimates, soils, cultural practices and grapevine varieties explain this diversity (Magalhães, 2008). This study targets the Douro & Port wine region (D&P WR henceforth), located in northern Portugal (Fig. 1), which comprises two conterminous Denominations of Origin (Douro & Port). This region comprises a vineyard area of nearly 45 000 ha, corresponding to approximately 22% of the total vineyard area in Portugal. It is responsible for >40% of the Portuguese wine exports, dominated by the world-famous Port Wine (Gouveia et al., 2018). The D&P WR features very complex orography, with a large diversity of terroirs (Fraga et al., 2017b). These different terroirs have strong implications on the chromatic and aromatic descriptors of the wines (Prata-Sena et al., 2018). A large number of autochthonous varieties can be found, but with low productivity (typically <5 000 kg ha−1).
The D&P WR is characterized by meso‑Mediterranean climates, with annual mean temperatures within the range of 12–15 °C, January means from 5 to 9 °C and July means from 21 to 25 °C (Costa et al., 2017). In terms of aridity/dryness, conditions vary from humid in the “Baixo-Corgo” (westernmost sector) and “Cima-Corgo” (central sector) sub-regions to sub-humid (i.e. annual totals of precipitation below evapotranspiration) in the “Douro Superior” (easternmost sector), with a strong east-west precipitation gradient, ranging from ca. 400 to 1200 mm (Costa et al., 2017). Precipitation is ubiquitously scarce in summer (June–August), typically less than 10% of annual precipitation, as it is strongly concentrated in autumn and winter, with large inter-annual variability (Costa et al., 2017).
Owing to the widely recognised sensitivity of the grapevine physiological development to weather and climate conditions, through several direct and indirect processes (Smart, 1985), grape berry quantity and quality reveal important inter-annual variability. Grapevine phenological timings are largely controlled by temperature (Malheiro et al., 2013), also influencing vineyard management and cultural practices. Hence, viticulture is at risk under climate change, as grapevine responses will be necessarily different under future climates (Moriondo et al., 2015; Fraga et al., 2016a; de Cortazar-Atauri et al., 2017). This may threaten wine typicity and wine balance of a given region, or even, in more extreme circumstances, its viticultural suitability (Santos et al., 2020a).
In Portugal, where significant warming and drying trends are projected for the future, including enhancements in the frequency of occurrence of temperature and precipitation extreme events (Costa et al., 2012; Andrade et al., 2014; Santos et al., 2019b), viticulture is particularly vulnerable to climate change (Fraga et al., 2014a, 2016b, 2017a; Santos et al., 2019a). The strong connection between precipitation in Portugal and the large-scale atmospheric circulation is a major factor to take into account in future climate conditions. Shifts in the large-scale atmospheric patterns within the Euro-Atlantic sector will significantly increase the frequencies of occurrence of severe precipitation deficits and droughts (Santos et al., 2009, 2016). This will challenge the country's water resources (Andrade et al., 2011) and limit irrigation as a potential adaptation measure for viticulture (Fraga et al., 2018). More frequent and intense heatwaves are also an important hazard to be taken into account in future climates (Fraga et al., 2020).
In the D&P WR, the strong inter-annual variability in grapevine yields and wine production has been associated to atmospheric forcing, while long-term trends have been linked to changes in cultural practices and management, as well as in national and regional policies, such as programmes devoted to vineyard plantation, restructuring or replacement (Santos et al., 2011, 2013). The pronounced interannual variability may have important impacts on phenology (e.g. budburst, flowering and veraison), yields, wine acidity and berry sugar content, thus challenging the stable production of high-quality wines (Santos et al., 2020a). Therefore, knowing in advance the potential wine production, based on seasonal climate forecasts, is of utmost relevance for the winemaking sector. Suitable adaptation measures can be applied to mitigate both the annual fluctuations in yields, in the short-term, and the climate change impacts, in the long-term. Changes in agricultural practices, such as pruning, application of sunscreens, cover crops, mulching, soil tillage, phytosanitary treatments, irrigation and genetics, are only a few examples of adaptation options (Duchene, 2016; Mosedale et al., 2016; Bernardo et al., 2018; Fraga et al., 2018; Fraga and Santos, 2018). Stock management in wineries is also a key aspect that can be better planned when predictions are available. All these measures may effectively reduce costs and improve the efficiency of the whole wine production chain, thus highlighting the need for reliable seasonal predictions of wine production. Although this study is focused on a specific wine region, similar methodologies can be implemented in other wine regions worldwide.
Along the previous lines, the present study objectives are threefold: 1) to develop a wine production model for the D&P WR that can be operated with seasonal forecasts of meteorological variables, 2) to assess the skill of the seasonal weather forecasts in the D&P WR, and 3) to evaluate the performance of the seasonal predictions of wine production in the D&P WR. The first objective builds on the study by Santos et al. (2013), where a wine production model for the D&P WR was developed. In that preceding study, climate model data were used to generate long-range projections of wine production, based on different anthropogenic forcing scenarios. Nevertheless, here, we develop a novel empirical wine production model aiming to integrate seasonal weather forecasts into medium-range regional wine production outlooks. To our knowledge, objectives 2) and 3) have not been addressed in any previous study. Data and methods are presented in Section 2, the main results in Section 3, followed by a discussion of the main results and the overall conclusions of the study.
Section snippets
Wine production data
We apply a multinomial logistic linear regression (Wilks, 1995) to model the wine production in the D&P WR, northern Portugal (Fig. 1). Since sub-regional or local/single “Quinta” (wine estates) time series tend to suffer from many inconsistencies and heterogeneities, we use the D&P WR wine production time series (in 103 hl) from the IVV (the entity responsible for monitoring and regulating regional wine production) over the period from 1950 to 2017 (Fig. 2a). This is a relatively long time
Analysis of the wine production time series
Concerning the wine production series (Fig. 2a) from 1950 to 2017, no missing data exists and only a slight long-term linear trend is apparent (dashed line), though it is not statistically significant at a 5% significance level, according to the non-parametric Mann-Kendall test (Wilks, 1995). The 11-year moving averages (thick black line) also hint at the absence of a non-linear trend, thus showing that the background low-frequency variability is mostly represented by the linear regression
Discussion and conclusions
The usefulness of seasonal weather forecasts on predicting wine production has been herein assessed. We have categorized wine production into three classes (low, normal and high production years) and applied multinomial logistic regression to obtain a predictive model of wine production in the Portuguese D&P WR over the period 1950–2017. In this empirical model, temperature and precipitation averaged over the periods of February–March, May–June and July–September, along with the anomalies of
Declaration of Competing Interest
None.
Acknowledgements
This study was supported by the European Commission-funded project “Climate change impact mitigation for European viticulture: knowledge transfer for an integrated approach – Clim4Vitis” [grant number 810176]. This work was also supported by National Funds by FCT - Portuguese Foundation for Science and Technology [grant number UIDB/04033/2020]. Chloe Prodhomme was supported by the Spanish Juan de la Cierva [IJCI-2016–30802] and has received funding from the EU H2020 Framework Programme IMPREX
References (53)
. The land management tool: developing a climate service in Southwest UK
Climate Services
(2018)- et al.
Viticultural irrigation demands under climate change scenarios in Portugal
Agr Water Manage
(2018) - et al.
Climate factors driving wine production in the Portuguese Minho region
Agr Forest Meteorol
(2014) - et al.
Vineyard mulching as a climate change adaptation measure: future simulations for Alentejo
Portugal. Agr Syst
(2018) Modelling olive trees and grapevines in a changing climate
Environ Modell Softw
(2015)- et al.
The terroir of Port wine: two hundred and sixty years of history
Food Chem
(2018) - et al.
The response of maize, sorghum, and soybean yield to growing-phase climate revealed with machine learning
Environ Res Lett.
(2020) - et al.
Climate change multi- model projections for temperature extremes in Portugal
Atmos Sci Lett
(2014) - et al.
Large-scale atmospheric dynamics of the wet winter 2009-2010 and its impact on hydrology in Portugal
Clim Res
(2011) - et al.
Evaluation of the ECMWF ocean reanalysis system ORAS4
Q J Roy Meteor Soc
(2013)
Grapevine abiotic stress assessment and search for sustainable adaptation strategies in Mediterranean-like climates. A review
Agron Sustain Dev
Land surface intialisation improves seasonal climate prediction skill for maize yield forecast
Sci Rep-Uk
An Ensemble Version of the E-OBS Temperature and Precipitation Data Sets
Journal of Geophysical Research: Atmospheres
Climate change scenarios for precipitation extremes in Portugal
Theor Appl Climatol
Implications of future bioclimatic shifts on Portuguese forests
Reg Environ Change
Grapevine Phenology of cv. Touriga Franca and Touriga Nacional in the Douro Wine Region: modelling and Climate Change Projections
Agronomy
Grapevine phenology in France: from past observations to future evolutions in the context of climate change
Oeno One
Seasonal climate predictability and forecasting: status and prospects
Wires Clim Change
Relationship between wine composition and temperature: impact on Bordeaux wine typicity in the context of global warming-Review
Crit Rev Food Sci
Sensitivity of decadal predictions to the initial atmospheric and oceanic perturbations
Clim Dynam
How Can Grapevine Genetics Contribute to the Adaptation to Climate Change?
Oeno One
Viticulture in Portugal: a review of recent trends and climate change projections
Oeno One
Modelling climate change impacts on viticultural yield, phenology and stress conditions in Europe
Global Change Biol
Multivariate Clustering of Viticultural Terroirs in the Douro Winemaking Region
Cienc Tec Vitivinic
Very high resolution bioclimatic zoning of Portuguese wine regions: present and future scenarios
Reg Environ Change
What Is the Impact of Heatwaves on European Viticulture? A Modelling Assessment
Applied Sciences
Cited by (16)
The probability of unprecedented high rainfall in wine regions of northern Portugal
2023, Climate ServicesPerformance of seasonal forecasts for the flowering and veraison of two major Portuguese grapevine varieties
2023, Agricultural and Forest MeteorologyGood practice for knowledge co-development through climate related case studies
2023, Climate Risk ManagementUsing NDVI, climate data and machine learning to estimate yield in the Douro wine region
2022, International Journal of Applied Earth Observation and GeoinformationCitation Excerpt :They need to be calibrated and validated, requiring adaptability for new environments (distinct climate, soil, varieties, and management), making operationality and transferability difficult, complex, and costly in terms of time and biophysical data requirements (Sirsat et al., 2019). The current model outperforms other models, such as the simple grape production model (PGP) based on favorable meteorological conditions, developed by Fraga et al. (Fraga and Santos, 2017), and the empirical model proposed by Santos et al. (Santos et al., 2020) where temperature and precipitation averaged over different periods, along with the anomalies of wine production in the previous five years, were used as predictors. Models based on computer vision and image processing (by extraction of variables that can be related to the actual yield: number of berries, bunch/cluster area, leaf area, number of flowers, stems, and branches), trellis tension, laser, radar, and radio frequency data processing also constitute viable approaches for estimating vineyard yield.