APPLICATION OF CROP MODELLING TO PORTUGUESE VITICULTURE: IMPLEMENTATION AND ADDED-VALUES FOR STRATEGIC PLANNING

Grapevine ( Vitis vinifera L.) is a very important crop in Portugal, where the viticultural sector plays a central role in the national economy. The present study provides a review of most relevant research on grapevine modelling, giving particular emphasis to its past and future applicability to Portuguese conditions. A brief overview of the national sector, as well as of the prevailing physical and biological environments and viticultural practices is provided. Further, the terroir concept is discussed in view of the main controlling factors of grapevine development. Several crop models, either statistical or dynamic, that have reliably simulated grapevine/vineyard parameters, such as phenology, yield and quality, are referred. Statistical models are based on statistically significant relationships between a number of predictors and a given grapevine parameter. Dynamic crop models simulate plant growth and development and holistically integrate crop phenotype, soil profiles, weather and climate data and management practices in their simulations. Dynamic crop models are then becoming important decision support systems in viticulture. Additionally, they allow testing the effects of soils, management decisions and weather on crops. However, only a few dynamic models can properly simulate grapevine performance. Several studies also apply crop models under future conditions to assess the detrimental climate change impacts on grapevines. These crop models can be either applied to real-time monitoring and short-range predictions or to develop long-term climate change projections for the Portuguese viticulture. These studies will represent important added-values for the competitiveness and future sustainability of the winemaking sector in Portugal


CHARACTERIZATION OF PORTUGUESE VITICULTURE
The present study provides a succinct review of the Portuguese viticulture and addresses the application of crop models as key decision supporting systems Article available at http://www.ctv-jve-journal.org or http://dx.doi.org/10.1051/ctv/20153001029 towards a more competitive, efficient and sustainable viticulture, under current and future climates.

The sector
All around the world, viticulture and winemaking are very important activities with relevant impacts on local and regional economies. For the specific case of Portugal, which is the 11th wine producer and the 10th wine exporter in the world (OIV, 2013), viticulture and the winemaking socioeconomic sector play a key role in its economic growth. The vineyard area in Portugal is over 229 kha, with an annual wine production of about 6.7 Mhl (OIV, 2013). Portugal exports almost half of its total national wine production. Table wines represent 55% of the total national exports, followed by DO wines with 36%, which include fortified wines, such as Port wine as well as Regional wines with 12%. These exports have a large impact on the local economies, accounting for nearly 2% of national export income (IVV, 2013).
Portugal is divided in 14 viticultural regions (mainland Portugal, Azores and Madeira), with 25 Denominations of Origin -DO ( Figure 1). The most important winemaking regions are Douro/Porto, Minho ("Vinho Verde"), Alentejo and Lisboa. In terms of production, Douro/Porto is the main region, with almost 1.5 Mhl in 2013, approximately corresponding to 25% of all wine produced in Portugal (IVV, 2013). This region has also the largest vineyard area in the country (ca. 45 kha). Alentejo is, however, the highest yielding region, with an average production of 49 hl/ha, followed by Península-de-Setúbal, with 47 hl/ha, and Lisboa, with 38 hl/ha (2013 estimates).

The physical and biological environment
Portuguese winemaking regions commonly present very specific environmental and climatic characteristics. Portuguese climates show prevailing Mediterranean-like characteristics, with warm dry summers and wet autumn-winter periods. The Portuguese warm-dry summers critically limit crop growth, mainly due to summertime low water availability (Fraga et al., 2014c). Furthermore, extreme winter/spring weather events, such as hail and frost, tend to occur in some northern regions and to infringe important damages to this crop. In general, Portugal presents higher temperatures in the south (e.g. Alentejo) and lower temperatures in the north (e.g. Minho and Douro/Porto) (Figure 2a). The lowest temperatures are found in the mountainous areas of the northern half of the country. With respect to precipitation (Figure 2c), Portugal presents the highest amounts in the northwest and the lowest amounts in its southernmost part. Concerning temperatures and precipitation during the growing season (April-October; Figure 2b and d), similar spatial patterns to the yearly average and cumulative totals are found, but with much lower amounts in precipitation, as it mostly falls in the winter half of the year, a typical feature of the Mediterranean climates.

Regiões vitivinícolas em Portugal Continental
The predominant soils in Portugal (FAO, 2006) are cambisols, mostly in the north, and lithosols/luvisols, in the south (Figure 3a). In the northern half of the country there is a large area of cambisols in Minho, Beiras-Atlântico and Terras-da-Beira, while Douro/Porto predominantly presents lithosols, which has been greatly affected by human activity (anthrosols). In the south, the region of Lisboa also presents cambisols, while Peninsula-de-Setubal and Tejo mostly show podzols. The most representative soil types in Alentejo and Algarve are lithosols and luvisols. With respect to topography (Figure 3b), the most mountainous areas are located in inner-northern Portugal, whereas flatlands and plateaus prevail in coastal and southern areas.  Regarding intraspecific biodiversity, a large number of native varieties can be found in Portugal, according to their adaptation to the different soils, climates and topographic conditions, with red varieties typically predominant in the south and white varieties in the northwest (Fraga et al., 2014b). Some of the most well-known Portuguese winegrapes are: Alvarinho, Castelão, Fernão-Pires, Touriga-Nacional and Touriga-Franca. All these varieties present unique agronomic and oenological characteristics that ultimately result in distinctive wines. They give origin to high quality wines, ranging from the iconic fortified Port wine to the fresh and light "Vinho Verde" (IVV, 2013), which are typically characterized by blending varietals. Therefore, other winemaking regions worldwide are already growing some of these varieties (Anderson, 2014), namely Touriga-Nacional. Regarding the main phenological stages (Real et al., 2014), though depending on the grapevine varieties and environmental conditions in a given year, budburst occurs annually from March to April ( Figure  4). It is then followed by a phase of intensive growth until flowering that generally occurs in May-June. Immediately after flowering, the number of potential, and viable fruits, is determined. This stage is followed by the veraison that initiates the grapevine ripening, usually in July-August. Full maturity is then reached, typically in September to October (Magalhães, 2008). At the end of the developing season, the grape clusters are harvested and finally leaves begin to fall, initiating the dormancy stage. Cultural practices have also several particularities in each region. As an example, in Douro/Porto, due to steep slopes, walled terraces are common. Regarding the training systems, the cordon (unilateral and bi-lateral) prevails, though 'pergola' (in Minho) and 'gobelet' (e.g. Alentejo) can also be found. In this way, vine training determines density, orientation and microclimate of the vines.

FACTORS OF INFLUENCE IN VITICULTURE
A highly complex and interactive system, formed by climate, soil and management practices, greatly influences grapevine development (Magalhães, 2008). This system evolved towards a concept of terroir defined as "an area in which collective knowledge of the interactions between the identifiable physical and biological environment and applied vitivinicultural practices develops, providing distinctive characteristics for the products originating from this area" (OIV, 2010). All these terroir elements strongly influence growth and development of the different varieties of Vitis vinifera L. (van Leeuwen et al., 2004;van Vaudour et al., 2010;Yau et al., 2013;Fraga et al., 2014b), as well as wine type, yield and quality. Climate is considered the main element of terroir (Jones and Davis, 2000;Carbonneau, 2003;van Leeuwen et al., 2004). One of the most important climatic thresholds is the classical 10 °C base temperature for budburst (Winkler, 1974). In fact, temperature conditions drive the timing and length of all grapevine phenological stages (Kose, 2014), thus also affecting the inter-annual variability in yield, production and quality (de Orduna, 2010;Fraga et al., 2014d). Berry colour, flavour and aroma, tannin and sugar levels are also affected by temperature at ripening (Jones and Davis, 2000;Malheiro et al., 2010). Precipitation is another important element, which widely controls soil water availability affecting grapevine water use (Ferreira et al., 2012). Overall, high-quality wines are associated to moderate water stress during maturation (Koundouras et al., 1999;van Leeuwen et al., 2004;Storchi et al., 2005). A severe water stress during early stages may considerably delay growth and grapevine development (Hardie and Considine, 1976). On the other hand, excessive vigour, diseases and other problems negatively affecting wine quality may occur when excessive soil moisture is verified over the growing season (During, 1986;Magalhães, 2008). Extreme weather events, such as late spring frosts, hail and heat waves (above 35 ºC), may severely damage grapevine leaves and berries (e.g. sunscald) and are a significant risk to this crop (Trought et al., 1999;Chuine et al., 2004;Molitor et al., 2014).
Soil is another important factor for viticulture and is an important part of the terroir concept (Magalhães, 2008). Soils are composed by organic and inorganic materials and a source of water and nutrients (e.g. nitrogen) that are crucial for grapevine physiology, growth and yield attributes (Winkler, 1974;Morlat and Jacquet, 2003). In fact, grapevine composition can be influenced by soil structure and chemistry and can thereby affect wine quality (Mackenzie and Christy, 2005). Vineyard installation is usually preferred in deep soils with good drainage, either natural or manmade. Root growth and development is limited in compact and shallow soils that can obstruct their access to oxygen, water and nutrients (Jackson and Lombard, 1993). Soil water holding properties are also essential, as they can have an effect on grapevine performance (Field et al., 2009;Yau et al., 2013).
Topographic elements are another important factor. Elevation, slope and aspect are very relevant in viticulture (Jones et al., 2004;Yau et al., 2013). Elevation influences temperatures on vineyards (through the vertical temperature gradient), exerting this way a strong influence in site and varietal selection (Magalhães, 2008). Slope degree affects grapevines through solar exposure, thus having an impact on canopy microclimate, soil erosion, water drainage and viticultural management (Zsofi et al., 2011).
The quality and growth of grapevines can also be influenced by management practices, such as the choice of rootstock and scion, training systems, crop load, pruning type and cultural timings (Winkler, 1974). Knowledge of the varietal specificities allows optimizing viticultural practices and is required for production of high quality wines (Jones and Davis, 2000). Additionally, oenological practices also greatly affect wine quality (Unwin, 1996).

CROP MODELS IN VITICULTURE
Many studies have been undertaken in order to identify statistical relationships between grapevine parameters and the abovementioned terroir factors. These relationships are often rendered in crop models, which have proven to be useful in predicting yields, phenology, berry development and biomass. Crop model simulations integrate current scientific knowledge from many different areas, including crop physiology, climatology/agrometeorology, plant breeding, agronomy, soil physics/chemistry and pathology. Given the growing awareness for the need to implement these types of models, several international projects have been created with this aim, such as the MACSUR (Modelling European Agriculture with Climate Change for Food Security) or the AgMIP (Agricultural Model Intercomparison and Improvement Project) project. Crop models can be either statistical or dynamic in their formulation. Statistical models are computationally cheaper than dynamic models, but may present inconsistencies among variables (Shin et al., 2009). These weaknesses may happen mainly during non-linear crop-environment interactions that cannot be properly resolved by statistical approaches. As dynamic methods holistically integrate the different environmental interactions, they have the potential to outperform statistical techniques. Nevertheless, statistical models may still be a very useful tool when neither dynamic models nor sufficient computing resources are available (Shin et al., 2009). A more detailed discussion of these models is presented in the following sections.
Regarding yield and quality, some studies have been accomplished to identify statistical relationships between grapevine yield and some environmental variables. For the Portuguese Douro/Porto and Minho regions, statistically significant relations between grapevine yield and monthly mean temperatures or precipitation totals were found (Santos et al., 2011;Santos et al., 2013;Fraga et al., 2014d). These studies demonstrated that anomalously high March rainfall (during budburst), as well as anomalously high temperatures and low precipitation in May and June (flowering and berry development) are needed to achieve moderate-to-high production. Cunha et al. (2010) developed a forecast model for evaluating the annual variability in regional wine yield based on remote sensing indices in vineyards in Alentejo, Minho and Douro. For the latter region, Gouveia et al. (2011) developed a vintage model using climatic and remote sensing data. Cunha et al. (2003) also demonstrated that it is possible to obtain early season estimates of wine production using grapevine airborne pollen concentrations as a predictor.
Still focusing on yield or quality attributes, similar results have also been found for other winemaking regions worldwide. Rovira-Más and Sáiz-Rubio (2013) applied crop metrics (e.g. vegetation amount, berry size, grape yield, elevation, soil compaction and pH) to predict grapevine yield. In another study, Webb et al. (2008b) modelled the climatic sensitivity of premium quality grapevines by developing grapevine quality-temperature models for varieties grown in Australia. Through a statistical approach it was possible to identify key phenological periods influencing phenolic concentration at maturity for Pinot noir (Nicholas et al., 2011). The authors demonstrated that warm temperatures from budburst to flowering increase phenolic concentrations, which are beneficial for wine quality.
Regarding phenology, several studies identified statistically significant relationships with climate. The standard growing degree-day (GDD) model measures accumulated temperatures above 10ºC (Winkler, 1974) and is commonly used for evaluating grapevine phenology (Winkler, 1974;Moncur et al., 1989;Oliveira, 1998;Jones and Davis, 2000;Chuine et al., 2003;van Leeuwen et al., 2008;Duchene et al., 2010;Parker et al., 2011). In Portugal, Lopes et al. (2008) estimated the usefulness of these thermal models for several grapevine varieties in the Lisboa wine region. For the Portuguese Douro/Porto region, Real et al. (2014) also demonstrated their reliability on monitoring phenology, despite the need for local calibrations. For the Lisbon region, Malheiro et al. (2013) and Fraga et al. (2014a) applied a linear regression models to show that phenological timings are deeply tied to air temperatures and remote sensing indices. Many other studies assessing relationships between air temperature, remote sensing indices and grapevine phenology have been conducted worldwide (Williams et al., 1985;Chuine et al., 2003;de Cortazar-Atauri et al., 2009;Caffarra and Eccel, 2010;Bock et al., 2011;Cunha and Richter, 2012;Fila et al., 2012;Parker et al., 2013;Rodrigues et al., 2013).
Concerning vineyard water management, few studies have been devoted to Portugal, while some advances have been made in other winemaking regions in Europe (Lebon et al., 2003;Berdeja et al., 2014;Pellegrino et al., 2014;Roux et al., 2014;Schreiner and Lee, 2014). Pellegrino et al. (2006) used a simple soil water balance model, specifically parameterised for grapevine, to characterize soil water deficits on 24 fields within 4 vineyards in Mediterranean southern France. A similar approach was carried out by Gaudin et al. (2014), using a water balance model to classify water stress in French Mediterranean vineyards. In preparation of a water stress alert system,  developed a water stress model to early detect the best management options as a function of soil water content.
Statistical models have also been developed to estimate the risk of pests and diseases, while none was carried out in the Portuguese vineyards to our knowledge. Caffarra et al. (2012) modelled the impact of insect pests on the eastern Italian Alps. Calonnec et al. (2008) used an epidemiological model to simulate the dynamics of the powdery mildew pathogen affecting the viticultural production systems, while Hoppmann and Berkelmann-Loehnertz (2000) used phenological models to establish suitable plant protection regimes. However, on the whole, the number of statistical models applied to pests or diseases is relatively small compared to other parameters.

Dynamic crop models
Dynamical crop models simulate/monitor plant growth and development on a daily basis and at a given location. These models integrate crop phenotype, soil profiles, weather data and management options in their simulations. Sitespecific parameters for climate, soil, plants and crop management, among others, are defined as input in model runs ( Figure 5). Weather data, such as maximum and minimum near-surface temperatures, rainfall and incoming solar radiation are updated on a daily basis, while other parameters are generally kept invariant throughout the model run (e.g. soils parameters). Dynamic crop models are becoming important decision support systems for monitoring crops and for assessing the impact of soils, management decisions, weather and climate change on crops (Paz et al., 2007;Semenov and Doblas-Reyes, 2007;Challinor and Wheeler, 2008). In fact, the simulation of crop parameters under different conditions, scenarios and stresses are key outcomes from crop models. Although many models have been applied to evaluate crop development and growth, only a few can properly simulate grapevine systems (Valdes-Gomez et al., 2009).
Applied to viticulture, dynamic crop modelling can be either focused on the simulation of a particular process or on the entire plant growth (Moriondo et al., 2015). As an example of the simulation of a particular process, Lebon et al. (2003) used a dynamic model to simulate the seasonal dynamics of soil water balance on vineyards, demonstrating that this process can be adequately replicated by the model. In another study, Nendel and Kersebaum (2004) skilfully simulate nitrogen dynamics in vineyard soils using the NVINE model. Ben-Asher et al. (2006) assessed the skill of the SWAP (soil, water, atmosphere and plant) model to estimate the salinity effects in grapevine production. Their results demonstrate that the model can generate realistic responses to salinity when water quality is the only variable used. To help choose adequate training and pruning strategies, Poni et al. (2006) used the STELLA software to build a model to predicting the daily carbon balance and dry matter accumulation on grapevine vertical shoots. The Walis model was evaluated by Celette et al. (2010) to simulate water partitioning on a intercropped vineyard. Webb et al. (2007) used the VineLOGIC model (Godwin et al., 2002) to determine grapevine phenology. A broader approach was developed by Cola et al. (2014), which developed a new dynamic model MoDeM_IVM DSS (Monitoring and Decision Making in Integrated Vineyard Management Decision Supporting Systems) for predicting grapevine seasonal dynamics, sourcesink balance and yield, showing high potential for its inclusion in viticultural decision supporting systems and technical assistance.
An example of a dynamic crop model that can be used to simulate the whole plant growth process is the STICS (Simulateur mulTIdisciplinaire pour les Cultures Standard) (Brisson et al., 2003). STICS was developed by the French National Institute for Agronomic Research (INRA), comprising a large multidisciplinary community of researchers, from microclimate and soils to crop sciences (Brisson, 2004). This is a generic crop model that can be applied to a wide variety of crops, such as wheat (Brisson et al., 2002;Rodriguez et al., 2004), maize (Bruckler et al., 2000;Brisson et al., 2002;Debaeke, 2004), sugarcane (Valade et al., 2014) and banana (Brisson et al., 1998) and for many other purposes such as irrigation strategies (Katerji et al., 2010), carbon balances (de Noblet-Ducoudre et al., 2004), soil drainage (Tournebize et al., 2004) and nitrate contamination (Ledoux et al., 2007;Jego et al., 2008) and climate change impact assessment (Courault and Ruget, 2001;Juin et al., 2004;Gonzalez-Camacho et al., 2008). de Cortazar-Atauri (2006) adapted this model for grapevines assessing the necessary parameterizations. STICS is one of the few freely available and well-documented crop models integrating grapevine, a short overview of its applications to viticulture will be presented henceforth.
STICS runs on a daily time-step and variables related to climate, soil and crop system are used as input (Brisson et al., 2003). It simulates the whole processes of crop growth and development, also including water and nitrogen balances ( Figure 5). As output, it provides variables related to yield and development parameters. The model dynamics is described as follows (de Cortazar-Atauri, 2006;Brisson et al., 2008): 8 -Plant water stress is calculated based on crop demand and water supply, which influences leaf area and consequently light interception and net assimilation rate (Pellegrino et al., 2005), with implications on grape sugar concentrations (Pellegrino et al., 2006); 9 -Light intercepted by vines is evaluated using a simple geometrical approach, by separating diffuse and direct radiation components.

Representação do modelo cultural dinâmico, categorias principais de parametrização e variáveis de saída.
Focusing on grapevines, STICS has been used for simulating grapevine yield and quality attributes (Valdes-Gomez et al., 2009), soil water balance (Valdes-Gomez et al., 2011), climate change impact assessment (de Cortazar-Atauri, 2006), crop practices , soil management (Brisson et al., 1998) and risks of pests/diseases (Leroy et al., 2013). Valdes-Gomez et al. (2009) assessed STICS skill in replicating grapevine phenology, yield, soil water content and biomass. Irrigated and non-irrigated vineyards in Chile and France were selected for these simulations. The model showed a high skill in estimating grapevine phenology, with differences between simulated and observed timings less than six days. STICS also reliably simulated grapevine water stress, biomass production and yield. In a subsequent study, STICS allowed simulating water and nitrogen balances (Valdes-Gomez et al., 2011) for the "Cabernet Sauvignon" variety in Chile. Celette et al. (2008) also assessed STICS ability to represent vineyard water balance in two situations for Chile and France, differing in annual rainfall and in water management practices. Results revealed an accurate estimation of grapevine phenology, as well as of total dry matter production and yield. Soil water content was also well estimated in both situations. Nevertheless, the authors point out that water balance modelling using STICS may not be appropriate in situations with significant surface runoff. These authors also highlighted some modelling uncertainties, namely in the simulation of surface soil layer humidity, which may have a significant effect on nitrogen balance simulation.
de Cortazar-Atauri (2006) assessed the influences of climate change on French grapevines by coupling STICS with the ARPEGE climate model. This study assesses water balance changes and impacts on production and quality. Results allowed simulating current and future water stress conditions, nitrogen stress, biomass and yield, which may help planning mitigation actions against climate change impacts. In another study, potential climate change impacts on production and phenology were also assessed for two grapevine varieties in Sardinia (Italy) (Muresu, 2012), by coupling climate models with STICS to generate future scenarios. Model performance was also assessed using historical phenology and yield data, showing a high skill.
STICS was also used to assess the environmental impacts of crop fertilization. Ruiz-Ramos et al. (2011) used STICS coupled with a geographic information system to estimate the amount of NO3leaching in La Rioja (Spain). Its performance was examined by comparing simulations and measurements of irrigated vineyards. Using STICS simulations, the authors identified environmentally safe agricultural practices for mitigating NO3pollution. With respect to pests and diseases, Leroy et al. (2013) applied a bioeconomic model, coupled with STICS, to test different fungicide treatment strategies so as to diminish pesticide use in vineyards. In a recent study, Coucheney et al. (2015) evaluated STICS performance for several crops, and found poor performances for N prediction on vine, that can be due to the fact that the model is more focused on carbon functioning making simpler assumptions on N simulation.
In Portugal, some studies have been devoted to the application of STICS model to grapevines. Under the SIAMVITIS project (Climate change in Viticulture: Scenarios, Impacts and Adaptation Measures) an initial application of this model was achieved, showing promising results (Coelho et al., 2013;Pinto, 2013). More recently, Fraga et al. (2015), calibrated the STICS model for three of the most important Portuguese varieties (Aragonez, Touriga-Franca and Touriga-Nacional), obtaining a high model skill in the simulation of yield, phenology and water status.

Monitoring and short-to-medium range prediction
Although grapevine production and phenology in Portugal were already skilfully modelled by several statistical approaches, the use of dynamic crop models in Portuguese viticulture is still in very early stages and future research on this topic is essential, particularly taking into account the relevance of this sector to the national economy, as previously described. After these models are properly calibrated and validated for the Portuguese varieties and environmental conditions, they may become useful decision support systems for stakeholders in the national winemaking sector. These models enable the adjustment of all cultural practices for each specific location, such as vineyard intervention dates, plant density, soil and water management, irrigation efficiency and nutrient management. These models may be used to predict annual yields and phenological timings, allowing a more accurate preparation of vineyard activities, such as spray scheduling and harvest planning. Crop models also provide preventive measures on soil conservation options in order to improve tillage, mulching and application of fertilizers. Due to the wide range of different terroirs throughout Portugal, crop models may also allow identifying the most suitable varieties for a given region based not only on their thermal requirements, but also on their tolerance to stress factors, such as high temperatures, drought, pests and diseases. Given the benefits of the application of these types of models to grapevine monitoring and short-to-medium range prediction they are expect to yield efficiency gains at the vineyard level, resulting in higher profit margins for growers.

Long-range prediction under climate change
Climate change projections are expected to have significant impacts on viticulture, mostly owing to changes in the temperature and precipitation patterns (IPCC, 2013). The expected warming and drying trends in southern Europe may bring some additional challenges for grapevine production (Santos et al., 2011). Temperature projections for the main viticultural regions in Europe highlight increases in the growing-season mean temperatures (Duchene and Schneider, 2005;Jones et al., 2005;Neumann and Matzarakis, 2011). This warming leads to longer growing seasons but earlier phenological events (Chuine et al., 2004;Dalla Marta et al., 2010;Bock et al., 2011), which may have harmful impacts on wine quality (Webb et al., 2008a). Some regions are projected to become excessively dry to grapevine production without adequate irrigation (Kenny and Harrison, 1992;Koundouras et al., 1999;Malheiro et al., 2010;Santos et al., 2013). All these factors suggest a general lowering of the suitability of the winemaking regions in southern Europe (Stock et al., 2005;Malheiro et al., 2010;Santos et al., 2012;Fraga et al., 2013). More specifically for Portugal, projections reveal similar changes to other European regions with Mediterranean-like climates (i.e. warming and drying trends), which can lead to changes in the suitability of the national viticultural regions (Fraga et al., 2012). Climate change may indeed lead to perceivable changes in traditional wine styles. Furthermore, changes on the frequency and strength of precipitation and temperature extremes are also projected for Portugal (Costa et al., 2012;Andrade et al., 2014).
Crop models are a key tool to determine the impact of climate change on crop growth and to evaluate the potential of new viticultural regions (Webb et al., 2007;Bois et al., 2008;Kwon et al., 2008;Scaglione et al., 2008;Caffarra and Eccel, 2010;Duchene et al., 2010;Caffarra and Eccel, 2011) and to prevent, or minimize, the impact of climate extremes (Molitor et al., 2014). Crop models can also be used to assess carbon sequestration and emissions of other greenhouse gases, assisting the design of mitigation measures in conformity with the 20-20-20 commitments of the EU Directive 2009/28/EC. Only by using crop models is possible to fully integrate all these aspects (Tomasi et al., 2011) and to develop adequate mitigation and adaptation measures for the viticultural sector under a changing climate. Therefore, in forthcoming research, dynamic crop models will be applied to the Portuguese viticultural regions under climate change scenarios.

CONCLUSIONS
Given the above description, crop models, such as STICS, can then be considered key decision making tools for short-and long-term strategic planning in viticulture. The possibility of obtaining early predictions of production and development will contribute to a more efficient winemaking process. This knowledge is crucial for a timely planning of field activities, such as determining harvest dates, and for achieving premium quality vintages. Furthermore, several studies use crop models to evaluate the impact of climate change on grapevine. Owing to the abovementioned relevance of the viticultural sector to Portugal, it is vital to improve its resilience and future sustainability. The efficiency gains, of the integration of these types of models in the industry, are expected to increase the competitiveness and sustainability of the wine sector in Portugal.