Meteorological and Climate Modelling Services Tailored to Viticulturists

Grape production is likewise inherently interconnected to climate and weather, and, although grapes may grow worldwide, premium wine-grape production occurs in Mediterranean-like climate ranges. Changes in climate and weather patterns are threatening premium wine-grapes, directly affecting the European wine industry. This is because grapevines are extremely sensitive to their surrounding environment, with seasonal variations in yield much higher than other common crops, such as cereals. With a view to making South European wine industry resilient to climate change, VISCA (Vineyards Integrated Smart Climate Application) project has deployed a Climate Service (CS) Decision Support System (DSS) tool that provides to wine producers with well-founded information to be able to apply correctly adaptation strategies on specific grape varieties and locations, and to achieve optimum production results (e.g., yield and quantity). In this paper we show the meteorological, seasonal and climatic models and data sets used to answer the viticulturist needs; from short-term and mid-term forecast to seasonal forecast and climate projections.


Introduction
Agriculture is a highly dependent sector on heat, sunlight and water, and therefore very sensitive to climate change. According to the current climate projections, weather events worldwide are very likely to become more extreme and frequent. In Europe, Southern countries will be frequently affected by heat ity in the surrounding regions [3] [4].
Agriculture happens to be the major land use of the earth, and this is expected to increase as population growth and diet changes may drive food needs up to 60% by 2050 [5] [6]. Therefore, agriculture should ensure food security in a world where 800 million people are currently chronically [7]. Although agriculture has shown ability to adapt to changing conditions, it is very likely that the above projections overpass agriculture's adaptation limits. Even if policies and efforts to reduce emissions prove effective, some climate change is inevitable; therefore, strategies and actions to adapt to climate change impacts are needed [8] [9].
Grape production is not different from the rest of agricultural activity, and they are likewise inherently interconnected to climate and weather, and, although grapes may grow worldwide, premium wine-grape production occurs in Mediterranean-like climate ranges. Changes in climate and weather patterns are threatening premium wine-grapes (e.g., decrease of the grape quality and quantity, undesirable changes in alcohol production, and acid and sugar concentrations), directly affecting the European wine industry [10]. This is because grapevines are extremely sensitive to their surrounding environment, with seasonal variations in yield much higher than other common crops, such as cereals [11].
As wine growers are facing the growing impact of climate change, the demand for enhanced tools designed and able to support them has grown. In particular, there is a high interest for solutions that can provide accurate weather forecast data together with other useful information (e.g., phenological forecast and irrigation optimisation) would be of the greatest value. Until now the proposed strategies aimed at mitigating and adapting to these conditions involved a range of climatic scales and covered multiple aspects of grapevine management (e.g. change of locations and/or grapevine variety selection), but there was no solution integrating all those aspects in a single adapted tool for winegrowers [12] [13] [14].
Responding to these needs, the VISCA (Vineyards Integrated Smart Climate Application, https://www.visca.eu/) an R & I project co-funded under the Horizon 2020 programme has developed a Climate Service (CS) Decision Support System (DSS) that integrates climate and agricultural modelling, and end-users' specifications, in order to design medium-and long-term adaptation strategies to climate change on vineyards. This objective has been achieved through the integration of climatic data, phenological and irrigation models, and end-users' requirements into a system co-designed with relevant South-European wine companies from Spain (Codorniu), Italy (Mastroberardino) and Portugal (Symington), in order to allow them to take well-founded decisions (e.g., harvesting, defoliation, pruning, minimum water needs, etc.); foresee extreme events at short and medium time-scales; ease the decision-making planning months ahead through the use of seasonal predictions and the development of strategies thanks to the climate projections.
In this paper we show the innovative work done within the VISCA DSS, a platform that integrates meteorological and climate modelling services that will improve the work of the end-users in their daily tasks as to support them long term activities. In particular, we show the meteorological models and data sets used to answer the viticulturist needs from short-term to mid-term forecast and from seasonal forecast to climate projections.

Demo Site Areas and Cellars
With a view to making South European wine industry resilient to climate change, VISCA intends to deploy a climate service tool that provides wine producers with well-founded information to be able to correctly apply adaptation strategies on specific grape varieties and locations and, thus, achieve optimum production results (e.g., yield or quality). It has been validated by real demonstration with end-users on three demo sites belonging to wine stakeholders from Spain, Italy and Portugal, which are likewise partners in the consortium: Codorniu, Mastroberardino and Symington. In Table 1 it is shown the end-users demo site characteristics.

Codorniu (Spain)
Codorniu wine cellar (https://www.codorniu.com/es) is located in the Denomination of Origin (DO) region called Costers del Segre, in the southwest slope of Raimat hills in Catalonia (Spain). This location confers special and differentiated characteristics: higher sunlight exposure in the afternoon, which allows the grapes to accumulate more degree days, achieving an earlier ripening without losing acidity. These are well drained soils; the slight inclination of the fields confers a natural draining of the rainwater. Being on the slope of the hill causes

Modelling Approach
This section summaries the developments carried out for the meteorological components of the VISCA DSS and shows the meteorological and climate models used for each time-scale, from short-term to mid-term weather forecast and from seasonal forecast to climate projections. Figure 1 shows and schema about the different timescales covered by the meteorological and climate models used in the VISCA DSS and the related skill. Since short-term and mid-term weather forecast have pretty good skills the first forecast days, after some days, the skill decreases rapidly. On the other hand, seasonal forecast and climate projections have a constant fair skill along the forecast lead time. The sub-seasonal to seasonal timescale was not considered within the project.

Short and Mid-Term Weather Forecast
During the last years, the use of deterministic Numerical Weather Prediction (NWP) models has increased significantly due to higher accuracy of the models, easier access to community models, computational advances, etc. Despite the inherently uncertainty they can have due to the initial conditions of the atmosphere, several calibrations can be applied to minimize the discrepancies against the real weather conditions. Due to numerical weather prediction models have a wide range of options to set up: physical options, dynamical options, horizontal model resolution, number of vertical layers and density, etc., it is crucial configurate the model with the properly parametrizations and model options [15] [16].
On the other hand, weather forecasts are inherently uncertain because the initial state of the atmosphere can never be known perfectly, and the model equations must be expressed through approximations and simplification in the model system. Furthermore, even the smallest uncertainties in the initial conditions of the forecast model tend to rapidly increase over time because of the chaotic nature of the atmosphere. Therefore, rather than integrating a single forecast from Atmospheric and Climate Sciences a supposedly best guess of the initial state (as was done in short-term forecasting), it has been shown that a better approach would be to start the forecast from several slightly different initial conditions and then derive as many, presumably somewhat different, outcomes from these differing initial conditions ( Figure 2).
This approach is called ensemble forecasting and as outcome, produces forecasts that are given as probability distribution. From these distributions it is possible to calculate local probabilities for different weather events using thresholds.
In this way and within the VISCA framework, the short-term weather data covers the hourly deterministic forecast for the following 48 hours and provides information of the mean temperature, wind speed, accumulated precipitation, relative humidity and downward short-wave flux. The mid-term, on the other hand, it's a 10-day probabilistic forecast of the same variables.

1) Short-term forecast simulation domains
To ensure that the model represents as         tion of humidity and temperature were also studied examining the sensitivity of planetary boundary layer schemes [21].
Secondly, we have investigated the potential improvements in using numerical weather prediction model data assimilation system to set-up better initial conditions [22]. Data assimilation is the technique for combining observational data with the model [23]. Within the VISCA framework, the three-dimensional variational data assimilation (3DVar) system has been implemented into the WRF-ARW, and the analysis system used was the Grid point Statistical Interpolation (https://ral.ucar.edu/solutions/products/gridpoint-statistical-interpolation-gsi).

Seasonal Forecast
The production of statistically consistent and reliable predictions is a necessary condition for the elaboration or climate services. We have also improved the seasonal forecasts trough the combination of seasonal forecasts from several forecast systems in a multi-model ensemble [25] [26]. The main advantage of the multi-model ensembles comes from a better consistency and reliability than those ensembles from individual systems [27], which are essential properties for the probabilistic seasonal forecast to be usable for decision-making [25] [28]. Nevertheless, the skill enhancement of the multi-model ensemble compared to the best available forecast system is rather marginal and limited to some specific regions. The estimation of the seasonal forecast quality based on its past performance is a fundamental step to aid end-user decision-making [25], because it allows quantifying the forecast benefit relative to other prediction approaches. Thus, seasonal predictions must be systematically compared to a reference (reanalysis or observations) to assess their overall quality in a multi-faceted process known as forecast quality assessment [29]. Three sources of uncertainty in common scoring metrics of probabilistic predictions should be considered: improper estimates of probabilities from small-sized ensembles, insufficient number of forecast cases, and imperfect reference values due to observation errors. A way to alleviate these problems is to use several scoring measures to offer a comprehensive picture of the forecast quality of the system [30] and to apply statistical inference as often as required. This information is also valuable to decide about the application of the optimal bias correction and downscaling methods [31]. Hence, this quality assessment framework seeks to provide the end-users with the tools to understand which approaches could be better for their interests considering three different probabilistic metrics: fair ranked probability skill score (FRPSS), fair continuous ranked probability skill score (FCRPSS), reliability diagrams; and one deterministic: ensemble-mean correlation.
Within the VISCA framework, the seasonal forecasts provide information for the temperature and precipitation tercile probability for the next six months (monthly forecasts) and the next two seasons (3-month aggregation forecasts). The forecasts have been bias corrected and downscaled following a calibration procedure. Besides, a multi-model approach has been developed (with little gain compared to the use of single forecast). They have been also verified considering four metrics: ensemble correlation, fair ranked probability skill score, fair continuous probability skill score and reliability diagrams. In all the cases, whenever the skill of the forecast is worse than the climatology, the climate information is provided instead of the model forecast.
Seasonal May. We have used monthly mean data for mean temperature, maximum temperature, minimum temperature and precipitation.
b) Seasonal Forecast System 5 (SEAS5) SEAS5 is the fifth generation of ECMWF's seasonal forecasting system [33]. It replaces the former SEAS4 and uses the Integrated Forecast System (IFS). The re-forecast of SEAS5 covers a 36-year period, from 1981 to 2017, with an ensemble of 25 members. Compared to the SEAS4 it includes a number of enhancements in the atmospheric resolution, land-surface initialisation and in the ocean model. In the atmospheric component, there are 9 vertical levels. Regarding the land-surface initialisation, the SEAS5 includes a new offline recalculation at the native atmospheric resolution with a revised precipitation forcing. Although this initialisation is still not perfect (the reanalysis and real-time assimilation are not the same), the tests performed show a good degree of consistency between the initialisation of SEAS5 re-forecast and real-time predictions. Finally, the SEAS5 uses the new version of ocean model Nucleus for European Modelling of the Ocean (NEMO), with and upgraded model version, ocean physics and resolution. c) Météo-France System-6 (MF6) MF6 is the sixth generation of Météo-France long-range prediction system [34]. The hindcast covers the period 1993-2016 and consists of an ensemble of 25 members. The operational ensemble starts from 2017 and has 51 members. The ensemble spread is generated by a stochastic dynamics' technique in addition to using a lagged initialization. The atmospheric component is the Arpege-IFS [35] which is a complex code designed not only for weather forecast or climate simulation, but also for data assimilation, forecast pre-and post-processing. It has been extended, diversified and complexified since 1986 jointly by Météo-France and ECMWF.

d) Met Office Global Seasonal Forecasting System version 5 (GloSea5)
In VISCA we have used the latest coupled configuration of the GloSea5, in particular the Global Coupled 2.0 (GloSea5-GC2) [36]. The re-forecast of Glo-Sea5-GC2 covers a 36-year period, from 1981 to 2016, with an ensemble of 28 members. Overall, the GloSea5-GC2 is shown to be an improvement on the configurations used currently, particularly in terms of modes of variability (e.g. mid-latitude and tropical cyclone intensities, the Madden-Julian Oscillation and El Niño Southern Oscillation).
More information about the work done and the results can be found in the VISCA project website (seasonal forecasts quality assessment report, https://www.visca.eu/index.php/the-project-4/deliverables-and-scientific-articles ecasts-quality-assessment-report).

Climate Projections
The purpose of this section is to show an overview of the work done within the project. The idea was to analyse the change of climate variables which are impacting the most in the viticulture, namely mean precipitation and temperature, but also temperature and precipitation extreme events for the coming decades until 2100.
Within the VISCA framework, first of all, a set of climate indices were defined giving key information about the impact on the wine sector: seasonal temperature climatology, defined as the mean temperature over a period of a 3-monthly season; seasonal daily precipitation climatology, defined as the mean of maximum daily precipitation over a period of a 3-monthly season; and finally, the The E-OBS dataset is derived through a three-stage process: Monthly means of temperature and precipitation are first interpolated to a 0.1-degree latitude by longitude grid using three-dimensional (latitude, longitude, elevation) thin plate splines.
Daily anomalies, defined as the departure from the monthly mean temperature or precipitation, are interpolated to the same 0. The full dataset covers the period from 1950 to 2017 (both included).
b) EURO-CORDEX climate projections EURO-CORDEX (https://www.euro-cordex.net/) is the European branch of the international CORDEX initiative, which is a program sponsored by the World Climate Research Program (WRCP) to organize an internationally coordinated framework to produce improved regional climate change projections for all land regions world-wide. The CORDEX-results are being used worldwide to assess for climate change impact and to feed adaptation studies within the timeline of the Fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate Change (IPCC) and beyond.
EURO-CORDEX consists of a set of simulations provided by the participating groups. Nowadays, mainly dynamical downscaling simulations are available, namely the output of a regional climate model (RCM) coupled to a Global Climate Model (GCM) or a Reanalysis. The outputs of these simulations are available on rotated-pole regular grids at 0.11 degrees of horizontal resolution (~12 km, depending on the latitude). They are free to download from Earth System Grid Federation (ESGF) nodes like https://esgf-data.dkrz.de/projects/esgf-dkrz/.

Conclusions
The extension and the quality of wine production are strongly related to the environmental conditions of the grape growing season. Grapevine growing factors are determined within a narrow climatic range in which the adequate heat accumulation, the availability of enough water and the low risk of extreme temperature episodes are required. As an example, in 2017 a historically low production (8% less in comparison with 2016) was registered due to climate conditions.
While climate patterns can differ radically from one year to another (climate variability), climate change is more concerning, because a significant shift in the long-term climatology would make the wine business unsustainable. Adaptation measures are most probably needed to adapt the current grape varieties to warmer climate conditions and the presence of more frequent and more intense extreme events such as heat waves, heavy rainy or long dry spells.
Future suitability of a certain viticulture region highly depends on the change on the mean patterns on temperature and precipitation in the coming decades, but also on the impact of extreme temperature and precipitation. For that reason, estimation regional changes in temperature and precipitation and the derived impacts on the suitability of vine cultivation are of paramount importance for the business of gives critical information for making strategic decisions and investments in the near future, which will make the industry more resilient and adapted to climate change.
The VISCA DSS is a tool that integrates meteorological and climate information, among others, in order to provide well-funded information to the end-users to design medium-and long-term adaptation strategies to climate change on vineyards; foresee extreme events at short and medium time-scales; ease the decision-making planning months ahead through the use of seasonal predictions and the development of strategies thanks to the climate projections. In this paper we have shown the meteorological, seasonal and climatic models, and the data sets used to answer the viticulturist needs for all time-scales.
As a short-term weather model, we have used the WRF-ARW version 3.9.1.1, providing hourly forecast information for the following 48 hours for those interesting variables for viticulturists: mean temperature, wind speed, accumulated precipitation, relative humidity and downward short-wave flux.
GEFS information has been used to compute probabilistic forecast up to 10-days in advance for the same variables.
SEAS4, SEAS5, MF6 and GloSea5 seasonal prediction systems have been used to compute the seasonal forecast. This provides information for the temperature and precipitation tercile probability for the next six months (monthly forecasts) and the next two seasons (3-month aggregation forecasts).
Within the climate projections, we have used E-OBS data to perform climatic indices related to the wine sector and EURO-CORDEX climate projections to analyse the projected change over the period 2010-2099.