SEARCH

SEARCH BY CITATION

Keywords:

  • climate change;
  • frost risk;
  • impact modelling;
  • phenology;
  • viticulture

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Background and Aims

Late frosts are a significant risk to grape production in frost-prone viticultural regions. Increasing air temperature because of climate change is likely to advance grape budburst and last frost events in spring. So far, it is unclear whether one trend will be more pronounced than the other, and hence, whether the risk of late frost damage will increase or decrease. The aim of this work was to investigate the future frost risk in the Luxembourgish winegrowing region by assessing the effect of simulated future climate conditions on the timing of budburst and last frost date.

Methods and Results

Late frost risk was assessed by combining: (i) a phenological model for budburst of the grapevine (DORMPHOT); and (ii) ensemble-based projections of future air temperature. Analyses indicated that increasing spring temperature will advance the timing of budburst and the date of the last frost. This advancement, however, will be more pronounced for last frost events than for budburst.

Conclusions

Modelled projections showed that the frequency of spring frost damage in the Luxembourgish winegrowing region will decrease, without completely excluding them for the near (2021–2050) or the far future (2069–2098).

Significance of the Study

The application of a combination of a phenological model for grape budburst and ensemble-based projections of future air temperature enables the assessment of the future late frost risk in a frost-prone viticulture region.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Late or spring frost damage is a significant risk to grape production in frost-prone viticulture regions all over the world, such as parts of North America (Johnson and Howell 1981, Poling 2008), Australia (Cashman 2000), New Zealand (Trought et al. 1999), France (Brun and Cellier 1992), Italy (Orlandini et al. 2003), Germany (Hill et al. 2011) or Poland (Lisek 2008). In the Luxembourgish winegrowing region, official records (1948‒2011) reported spring frost damage in the vineyards in 11 out of 64 seasons (Anonymous 2011). In other winegrowing regions, the frequency of severe frost damage is even higher. In Champagne, for example, such events occurred every fourth year from 1875 to 1975 (Brun and Cellier 1992) and in Michigan 11 times from 1957 to 1977 (Johnson and Howell 1981). Spring frost can result in severe crop losses, threatening the financial existence of winegrowers. Molitor and Junk (2011) showed that the average yield in the Luxembourgish winegrowing region was reduced by 39% in years with spring frost damage.

The temperature conditions necessary for freeze damage of plant tissue are dependent on the phenological growth stage reached (Trought et al. 1999). Young grape tissues with a high water content are killed as soon as ice crystals are formed (Poling 2008). Such crystal formation on young leaves and shoots is possible at an air temperature around or slightly below 0°C, depending on the type of the freeze event (radiation freeze or advection freeze) (Trought et al. 1999). Damaged tissues rapidly lose their turgor, darken completely and become water soaked (Poling 2008). Even though freeze damage can happen on closed grape buds, budburst can be seen as the onset of the most susceptible period.

Consequently, for an estimation of the risk of late frost under future climate conditions two pieces of information need to be known: (i) the budburst period; and (ii) the air temperature conditions during and after budburst.

Different approaches for the prediction of the day of budburst are presented in the literature (Amerine and Winkler 1944, Garcia de Cortazar-Atauri et al. 2009, Nendel 2010, Urhausen et al. 2011). An overview of phenological models for grapevine budburst simulation is given in Table 1. Models of Amerine and Winkler (1944), Nendel (2010), and Urhausen et al. (2011) are only of local or regional validity. This is probably due to the models not adequately describing the underlying physiology of dormancy and release from dormancy. Factors not considered by the models may change between sites, so that site-specific estimation of model parameters is required. For example, despite the fact that the early literature described the grapevine as relatively insensitive to photoperiod (Alleweldt 1964), recent literature shows that this factor indeed influences the physiology and phenology of Vitis species (Schnabel and Wample 1987, Fennell and Hoover 1991, Perez et al. 2009, 2011) as with the majority of plant species (Kobayashi and Weigel 2007). To overcome these shortcomings, Caffarra et al. (2011) developed the phenological model DORMPHOT that considers photoperiod and makes an attempt at describing the physiological processes taking place during dormancy. The inclusion of dormancy induction, of the interaction between photoperiod and temperature and the use of an experimentally established relationship for quantifying the action of a warm temperature on growth, ensures that the DORMPHOT model is more process based than phenological models based on degree-days or bioclimatic indices. While the DORMPHOT model was originally calibrated and validated on birch, it can be adapted to different plant species provided an adequately large and varied data set is available (Caffarra et al. 2011).

Table 1. A comparison of different published phenological models to forecast budburst
Model referenceMethod employedModel typeDormancy effects consideredStarting point for temperature sum accumulation (northern hemisphere)
Urhausen et al. (2011)Relationship between temperature drivers and phenology, obtained by multiple regression methodsStatisticalNo
Amerine and Winkler (1944)Heat summationProcess-basedNo1 April
Nendel (2010)Heat summationProcess-basedNo1 March
Garcia de Cortazar-Atauri et al. (2009) (BRIN Model)Chill and heat summationProcess-basedYes1 January
Riou and Pouget (1992)Cumulative thermal daily actionsProcess-basedYes1 January
Caffarra et al. (2011) (DORMPHOT)Relationship between rate of development, temperature and photoperiod during the subphases of dormancyProcess-basedYes1 September

Climate change projections indicate that the date of the last spring frost as well as the date of budburst will advance in response to a higher temperature (Caffarra and Eccel 2011). So far, it is unclear if one of these trends will be more pronounced than the other and, therefore, whether the risk of frost damage will increase or decrease. Molitor and Junk (2011) showed that in Luxembourg, the average time span between the last frost and budburst had tended to increase in the last decades, decreasing spring frost risk. In contrast, under the climatic conditions of Tuscany (Orlandini et al. 2009), earlier budburst is expected to increase the late frost risk in the future. Similarly, Poling (2008) suggests that earlier budburst could result in an elevated risk of damaging frost events in the eastern and mid-west regions of the USA. Further complicating the picture is the prediction by White et al. (2006) that frost risks will be reduced overall. If the estimations in the literature are that contrasting, a precise simulation of the late frost risk under future climate conditions in Luxembourg (as an example for a cool climate viticultural region in central Europe) is justified.

The Vitis vinifera cultivar Müller-Thurgau (synonym Rivaner) is the most common grape cultivar in Luxembourg. Because of its frost vulnerability and its economic importance for Luxembourg's viticulture, this study focused on Müller-Thurgau. Based on lengthy phenological observation, all important cultivars grown in Luxembourg (Müller-Thurgau, Riesling, Pinot Blanc, Pinot Gris, Pinot Noir, Elbling and Auxerrois) show almost identical precocity (Urhausen et al. 2011). Hence, with regard to budburst period, Müller-Thurgau can be considered representative of the majority of V. vinifera cultivars grown in Luxembourg.

This study aimed to: (i) develop and validate a model to predict the date of grape budburst; and (ii) estimate the risk of late frost damage in the vineyards of the Luxembourgish winegrowing region under future climate conditions.

Material and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Phenological and meteorological observations

To fit and validate a model taking into account photoperiod, an input data set containing data from different latitudes is essential. In this case, long-term budburst observation data sets for Müller-Thurgau from Remich (Luxembourg), Geisenheim (Germany), Neustadt (Germany), Veitshöchheim (Germany), Klosterneuburg (Austria), Grand Junction (USA) and Crosano (Italy) were used. The coordinates of the locations and the years of observation are indicated in Table 2. The total size of the combined data set was 102.

Table 2. Input locations and years used in the budburst prediction model for Vitis vinifera cultivar Müller-Thurgau as well as the spread of the seasonal average daily mean temperature in autumn (September, October, November), winter (December, January, February) and spring (March, April, May) in the years of observation
LocationCoordinatesYearsSpread average daily mean temperature (°C)
AutumnWinterSpring
Remich, Luxembourg49.54° N, 6.35° E40 (1972‒2011)8.3–13.5−0.5–6.27.7–12.9
Geisenheim, Germany49.99° N, 7.95° E11 (2000‒2010)9.9–13.30.8–5.99.6–12.6
Neustadt, Germany49.37° N, 8.18° E14 (1996‒2009)9.4–13.40.5–6.19.1–13.1
Veitshöchheim, Germany49.83° N, 9.87° E18 (1992, 1993; 1995‒2010)8.0–12.4−1.3–5.08.1–11.6
Grand Junction, USA39.05° N, 108.47° W5 (1995, 1997, 1998, 2001, 2003)11.2–11.70.9–2.911.0–13.3
Crosano, Italy45.81° N, 10.98° E7 (2002‒2008)9.8–12.1−0.2–4.09.9–13.1
Klosterneuburg, Austria48.29° N, 16.32° E7 (2005‒2011)8.9–12.6−1.2–5.09.8–12.3
Total 1028.0–13.5−1.3–6.27.7–13.3

Budburst, defined as BBCH (Biologische Bundesanstalt, Bundessortenamt und Chemische Industrie) stage 09 (green shoot tips clearly visible) according to Lorenz et al. (1995), was determined by local observers. Meteorological data (daily mean, minimum and maximum air temperature at a height of 2 m) were recorded nearby the vineyards of observation.

Budburst dates for Remich were calculated by averaging observed dates of approximately ten vineyards distributed in a radius of less than 20 km around Remich.

Development of the budburst model for Müller-Thurgau

To calculate budburst date, the DORMPHOT model according to Caffarra et al. (2011) was calibrated for V. vinifera cv. Müller-Thurgau. The model was developed for the simulation of budburst for photoperiod-sensitive species (Caffarra et al. 2011) and considers: (i) the dormancy induction process occurring in late summer‒autumn as the plant ceases growth and develops dormant buds in response to short days and a cool temperature; (ii) the action of chilling temperature for dormancy release; and (iii) the promoting effect that a long photoperiod has on bud development during dormancy release and bud development (Caffarra et al. 2011). In Table 3, the model structure is described in detail along with its equations.

Table 3. Description of the processes simulated by the DORMPHOT model and model equations used to predict the data of budburst of Vitis vinifera cultivar Müller-ThurgauThumbnail image of

To obtain a robust and biologically realistic model, we increased the amount of information during the parameterisation of the model, following the approach used by Caffarra and Eccel (2010). Experimental information was used to restrict or fix possible parameter values to biologically realistic values. As grapevines enter dormancy at mid-latitudes in the northern hemisphere in August‒September (Lavee and May 1997) and according to a recent study, the critical photoperiod for dormancy induction appears to be included between 11:40 and 14:40 h (Kühn et al. 2009), we arbitrarily fixed DLcrit (indicating the photoperiod earlier that the rate of dormancy induction is low or zero) to 13:00 h. The parameter dF could be measured directly from experimental results and was set to −0.26 (Caffarra and Eccel 2010). The range of values that could be adopted by parameters Dcrit (critical state of dormancy induction), Ccrit (critical state of chilling) and c (upper threshold of chilling temperature triggering a chilling rate of 0.5 or more) was restricted using the information available in the literature. Dcrit was bound between 0 and 40, as the model simulates dormancy induction starting on the 1 September, and dormancy in the grapevine is known to be attained in late summer–early autumn (Lavee and May 1997, Caffarra and Eccel 2010). Ccrit was bound between 10 and 50, as grapevine dormancy is reported to be released after a few weeks in chilling conditions (Dookozlian 1999, Botelho et al. 2007). Parameter cC, expressing the upper limit of the optimal chilling temperature range, was bound between 10 and 20, according to the experimental studies by Pouget (1963) and Dookozlian (1999).

The combined phenological data set described previously, including data from different countries and latitudes, was split into two parts of similar size, each containing data selected randomly from each site (except for the station of Klosterneuburg, which was used as a completely external test-data set). Half of these data sets (n = 53) was used to fit the DORMPHOT model parameters using the Metropolis algorithm (Metropolis et al. 1953) following Chuine et al. (1998).

The model fit was tested for significance compared with the null model (mean budburst date) with an F-test. Subsequently, model efficiency (ME; Janssen and Heuberger 1995) and the mean absolute error (MAE) were assessed. In order to evaluate systematic prediction errors by the model, the mean bias errors (MBEs) for each of the locations in the validation data set were calculated according to Mayer and Butler (1993).

The model was then validated using the remaining part of the data set (n = 49). Model performance was evaluated using ME and MAE. The MAE provides a mean measure of the model error, which does not allow for compensation between negative and positive errors.

Regional climate change projections and bias correction

Climate change projections based on different emission scenarios are possible representations of the future climate. The effect of climate change, especially for Luxembourg, was recently examined by Goergen et al. (2013) and Matzarakis et al. (2013).

Because the spatial resolution of Global Climate Models (GCM) is too coarse to force regional impact studies, a statistical or dynamic downscaling was necessary to derive regional-scale information. In our study, we used the dynamic downscaled results of the ENSEMBLES project (van der Linden and Mitchell 2009). To overcome the uncertainties of a single climate change projection, a multimodel approach based on six regional climate change projections (spatial resolution of 25 × 25 km per grid cell) based on the A1B emission scenario was adopted for this study (Table 4).

Table 4. Regional climate change projection data sets used in the study
Model abbreviationGCMRCMInstitution (running the RCM)
  1. Presented are the model abbreviation, the driving global climate model (GCM), the regional climate model (RCM) and the institution, which ran the RCMs. SRES A1B emission scenario; spatial resolution: 25 × 25 km.

M1European Centre Hamburg ModelDanish Meteorological InstituteDanish Meteorological Institute
ECHAM5-r3DMI-HIRHAM5DMI
M2Hadley Centre Coupled Model HadCM3Q0ClimateVersion of Lokal-Modell CLMEidgenössische Technische Hochschule ETHZ
M3Hadley Centre Coupled ModelHadley Centre Regional ModelMet Office Hadley Centre
HadCM3Q0HadRM3Q0METO-HC
M4Hadley Centre Coupled ModelHadley Centre Regional ModelMet Office Hadley Centre
HadCM3Q3HadRM3Q3METO-HC
M5European Centre Hamburg Model ECHAM5-r3Regional Climate Model RegCMInternational Centre for Theoretical Physics ICTP
M6European Centre Hamburg Model ECHAM5-r3Regional Model REMOMax Planck Institute for Meteorology
MPI-M

The A1B emission scenario describes anthropogenic emissions of a future world with rapid economic growth until the middle of this century and a balanced use of fossil and non-fossil energy resources. It is widely used in impact assessments because of data availability (Junk et al. 2012). Due to the multimodel approach in the EU (European Union) ENSEMBLES project and the related computational costs, projections for only one emission scenario could be realised within this project. The Regional Climate Model (RCM) results of the Research Theme 2B of the ENSEMBLES project can be retrieved from the ENSEMBLES data repository at the Danish Meteorological Institute.

The selected models cover the overall bandwidth of the available RCM ensemble in terms of air temperature change signals and account for the most widely used European RCMs (Van Pelt et al. 2012). Similar to Ylhäisi et al. (2010), we applied no weighting scheme to the different models. Daily time series of spatial means were extracted for each RCM from all grid cells of the respective model output grid that overlap with the area of Luxembourg (n = 9) and averaged using a weighting factor per grid element determined by the respective overlap area (Junk et al. 2012, Matzarakis et al. 2013). The following three time spans were used to analyse the projected changes in air temperature: the reference period defined from 1961 to 1990, the near future from 2021 to 2050 and the far future from 2069 to 2098. Because absolute values of daily minimum air temperature were required for the frost risk assessment, a bias correction of the model output was necessary. As reference data, we used long-term measurements (1961‒2011) of a weather station located at the Institute Vini-Viticole in Remich. A quantile mapping approach for the bias correction, as described by Themeßl et al. (2011), of the daily mean and minimum air temperature was applied. With the help of this approach, errors in the shape of the distribution of the daily minimum air temperature can be corrected, which leads to corrections in the variability as well. The advantage of the quantile mapping approach is the better adjustment of the distribution of extreme values to the reference observational data compared with that of a linear approach. The individual correction factors were calculated for the reference period (1961–1990) and then applied for the whole period from 1950 until 2098.

Frost risk assessment

Based on the bias-corrected data, the day of year (DOY) of budburst as well as the DOY of the last frost event (air temperature below 0°C) were calculated for every year and for each of the six regional climate change projections. Air temperature of 0°C represents an approximation of the temperature leading to freeze damage after budburst. Hence, a daily minimum air temperature <0°C occurring after the date of budburst was defined as a frost event. Frost years were defined as a season in which the temperature fell at least once below 0°C in the period of 60 days after budburst.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Budburst model

The DORMPHOT model adjusted to the cultivar Müller-Thurgau showed a high goodness of fit to the calibration data sets (ME, 81%; MAE, 2.7 days). It was significantly better than the average date of budburst for describing the data (F-test, P <0.001). The parameter estimates are given in Table 5.

Table 5. Parameter estimates of the DORMPHOT model for predicting the date of budburst of Vitis vinifera cultivar Müller-Thurgau
ParameterValueRange restrictions
  1. †Parameters fixed before parameterisation. ‡Range of parameter values restricted before calibration. In this case, the upper and lower values of the range are provided in the right column.

aD0.23NA
bD9.89NA
aC0.80NA
cC19.8710‒20
dF−0.26NA
gT0.18NA
hDL0.003NA
Dlcrit13.00NA
Dcrit30.740‒40
Fcrit20.12NA
Ccrit30.0010‒50

When applied to the validation data set, the calibrated model showed good predicting performance, with an overall ME of 75% and a MAE of 3.0 days. The scatter plots of predicted versus observed budburst dates show that the model did not markedly over – or underpredict budburst in the sites analysed (Figure 1). The same was the case for the MBEs caused by the location (Table 6).

figure

Figure 1. Plots of predicted versus observed day of year (DOY) of budburst of Vitis vinifera cultivar Müller-Thurgau for the seven sites analysed: (a) Remich; (b) Veitshöchheim; (c) Klosterneuburg; (d) Geisenheim; (e) Grand Junction; (f) Neustadt; and (g) Crosano. Calibration data set (○); validation data set (●); Dashed line: y = x.

Download figure to PowerPoint

Table 6. Performance of the model to predict the date of budburst of Vitis vinifera cultivar Müller-Thurgau in each of the locations used for validation
LocationMBEMAE
  1. MAE, mean absolute error; MBE, mean bias error. The size of each validation data set is given between parentheses.

Remich (n = 18)−0.53.2
Geisenheim (n = 5)−3.83.8
Neustadt (n = 7)2.32.3
Klosterneuburg (n = 7)3.34.0
Veitshöchheim (n = 6)1.01.6
Grand Junction (n = 3)−1.62.1
Crosano (n = 3)1.63.4

Frost risk

The overall trend of the date of budburst and of the last frost event is presented in Figure 2. The last frost event as well as the date of budburst are both likely to advance in the future. From 1950 up to the middle of this century, the spread of both variables overlaps in most years. From then onwards, an increasing temporal gap between the last frost event and budburst is evident.

figure

Figure 2. Median of annual date of budburst of the Vitis vinifera cultivar Müller-Thurgau (image) and date of the last frost event (image), both in day of year (DOY) for Remich/Luxembourg for the period 1961 to 2098. Spreads (min/max) of the date of budburst (image) and of last frost (image) are indicated as shaded areas and defined by the minimum and maximum values of the frost event as well as the budburst date of the six ensemble members.

Download figure to PowerPoint

In Figure 3, box plots for the reference period and the near and far future (for 30-year time slices) show that both the date of budburst and the date of the last frost event will advance. The median day of budburst will be 11 days earlier in the far future compared with that of the reference period (DOY 102 vs 91), whereas the median day of the last frost event will advance by 28 days (DOY 98 vs 70). The difference of the median values of the DOY of the last frost event and of the DOY of budburst is expected to increase by 17 days.

figure

Figure 3. Boxplots of 30-year time slices of the projected budburst date (day of year, DOY) for the Vitis vinifera cultivar Müller-Thurgau (image) as well as the date of the last frost event (image) for Luxembourg. The reference period is defined (a) from 1961 to 1990; (b) near future from 2021 to 2050; and (c) the far future from 2069 to 2098. The box plots are indicating the medians, and the 25 and 75% percentiles, whiskers are limited to 1.5 times of the interquartile range; outliers are marked with black stars.

Download figure to PowerPoint

Figure 4 shows the absolute cumulated number of frost events during the 60 days after budburst for all six ensemble members. In the reference period, the number of frost events after budburst averaged 24.7, while there were 11.2 seasons with frost events after budburst (frost years) during the time slice of 30 years. Up to eight frost events after the day of budburst occurred per season.

figure

Figure 4. Predicted absolute frequency of frost events (daily minimum temperature <0°C) in a period of 60 days after budburst (Vitis vinifera cultivar Müller-Thurgau) in Luxembourg in 30-year time slices [(a) reference period: 1961‒1990; (b) near future: 2021‒2050; and (c) far future: 2059‒2098]. Frost years were defined as seasons in which the temperature fell below 0°C at least once in the period of 60 days after budburst.

Download figure to PowerPoint

In the future, the absolute frequency of frost events after budburst, the number of seasons with frost events as well as the maximum number of frost events after budburst will decrease. There is likely to be 14.8 frost events after budburst and 6.7 seasons with frost events (frost years) in 30 years as well as a maximum of seven frost events per season in the near future time span (2021‒2050). Whereas, in the far future (2069–2098), only 5.2 frost events after budburst and 3.2 frost years in 30 years as well as a maximum of four frost events can be expected per season.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Budburst model

Models to predict the date of budburst presented in the literature are often of only local (Urhausen et al. 2011) or regional (Nendel 2010) validity because they do not take into account the processes that occur during rest and growth resumption of the plant.

In this study, we adapted a more process-based model, the DORMPHOT model, to the cultivar Müller-Thurgau. The calibration and validation proved that the developed model fits well to the broad range of data sets used from different locations in Europe and the USA with different climatic conditions. For example, seasonal average daily mean temperature in spring varied in the data sets between 7.7 and 13.3°C (Table 2).

The model showed good predictive performance when applied to independent data sets not used for the model set-up, including locations presently warmer than Remich/Luxembourg. Such robustness makes the model suitable for simulating the budburst date in those cases in which phenological observations are not available. The fact that interaction between photoperiod and temperature on the rate of dormancy induction and release is considered might offer more precise predictions as warming is likely to shift phenological phases in conjunction with different photoperiods.

In addition, the presented model might be used to calculate the timing of the start of leaf area development, which is a necessary input in leaf development models. These are in turn used as submodels within more complex decision support systems. For example, the leaf area development model by Schultz (1992) is used as a backbone model in grape disease models, such as in the ‘VitiMeteo’ system (Bleyer et al. 2008, Molitor et al. 2013). Also, the calculated budburst date could be used as the starting date for further phenological model simulations of the grapevine.

Future frost damage risk

Global and regional climate models are promising tools to simulate the potential future climate. Results, however, are associated with different uncertainties and systematic errors ranging from uncertainties in the underlying emission scenarios, the initial model conditions or of the parameterisation of physical processes within the models (Knutti et al. 2010, Suklitsch et al. 2011). To minimise these uncertainties, an ensemble-based approach to simulate future climate conditions was applied instead of a single model result.

Present investigations based on an ensemble of six climate change projections (A1B emission scenario) showed that the date of budburst as well as the date of the last frost event will advance in future. This trend will be stronger for the date of the last frost event than for that of budburst, i.e. the time between the last frost event and budburst will increase. The contrasting magnitude of these two trends might be explained by the assumptions and structures behind the budburst model used in the study; while frost risk simply depends on a temperature threshold, budburst date is simulated using nonlinear functions of air temperature and the photoperiod. In addition, the photoperiod acts as a buffering factor, as its pattern does not vary between years, and it has the same weight on the effect of temperature throughout the projection period.

The variability of the last frost events is significantly higher than that for budburst dates. This means that under extreme weather conditions, neither in the near (2021–2050) nor in the far future (2069–2098), frost events after the date of budburst can be excluded for Luxembourg. However, the frequency of these events is expected to decrease.

Climatic variability and long-term trends have a strong influence on the suitability of viticulture (Santos et al. 2012). If the frequency of frost damage events exceeds a threshold, the cultivation of grapes will no longer be economically viable. This study shows that the absolute frequency of frost events after budburst, the frequency of seasons with frost events (frost years) and the maximum number of frost events per season will decrease in the future. Hence, viticulture in Luxembourg in the current areas will not be further threatened by increasing spring frost events. An extension of the winegrowing region northwards and/or to a higher elevation (Santos et al. 2012) might even be possible. Additionally, the suitability of the traditional Luxembourgish viticultural areas for cultivation of grape cultivars with an earlier budburst, such as Chardonnay, might increase in the future.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Because of increasing spring temperature, the date of budburst and the date of the last frost events are expected to advance in the future in the Luxembourgish winegrowing region. According to data analyses based on an ensemble of six climate change projections (A1B scenario), this trend will be stronger for the date of the last frost event than for the date of budburst. This will reduce the frequency of frost events in the Luxembourgish vineyards. Frost events after budburst, however, are still likely to occur in the near (2021–2050) and far future (2069–2098).

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

We thank the Institut Viti-Vinicole in Remich (Luxembourg) for financial support. Parts of this study were carried out in the framework of the CLIMPACT (FNR C09/SR/16) and REMOD projects. The authors gratefully acknowledge Serge Fischer and Robert Mannes (Institut Viti-Vinicole, Remich, Luxembourg), Ottmar Baus (Geisenheim University, Germany), Bernd Ziegler (DLR Rheinpfalz, Neustadt an der Weinstrasse, Germany), Ulrike Maaß and Heinrich Hofmann (LWG Veitshöchsheim, Germany), Barbara Schildberger (HBLA Klosterneuburg, Austria), and Dr Horst Caspari (Colorado State University, Grand Junction, CO, USA) for providing the historical meteorological and phenological data sets as well as Vanessa Peardon for language support.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  • Alleweldt, G. (1964) Die Wirkung des Störlichtes auf die photoperiodische Reaktion der Reben. Vitis 4, 357364.
  • Amerine, M.A. and Winkler, A.J. (1944) Composition and quality of musts and wines of California grapes. Hilgardia 15, 493675.
  • Anonymous (2011) Die Weinjahre 1966 bis 2011 und ihre Ernteergebnisse (Institute Viti-Vinicole: Remich, Luxembourg).
  • Bleyer, G., Kassemeyer, H.-H., Krause, R., Viret, O. and Siegfried, W. (2008) ‘VitiMeteo Plasmopara’ – Prognosemodell zur Bekämpfung von Plasmopara viticola (Rebenperonospora) im Weinbau. Gesunde Pflanze 60, 91100.
  • Botelho, R.V., Pavanello, A.P., Pires, E.J.P., Terra, M.M. and Muller, M.M.L. (2007) Effects of chilling and garlic extract on bud dormancy release in Cabernet sauvignon grapevine cuttings. American Journal of Viticulture and Enology 58, 402404.
  • Brun, O. and Cellier, P. (1992) Les gelées de printemps dans les vignobles septentrionaux. Progrès Agricole et Viticole 109, 151157.
  • Caffarra, A. and Eccel, E. (2010) Increasing the robustness of phenological models for Vitis vinifera cv. Chardonnay. International Journal of Biometeorology 54, 255267.
  • Caffarra, A. and Eccel, E. (2011) Projecting the impacts of climate change on the phenology of grapevine in a mountain area. Australian Journal of Grape and Wine Research 17, 5261.
  • Caffarra, A., Donnelly, A. and Chuine, I. (2011) Modelling the timing of Betula pubescens budburst. II. Integrating complex effects of photoperiod into process-based models. Climate Research 46, 159170.
  • Cashman, G. (2000) Advection frost: a case study of the freeze of 27–28 October 1998 and its effects in the vineyards of the Canberra district. Australian Grapegrower and Winemaker 438, 7484.
  • Chuine, I. (2000) A unified model for budburst of trees. Journal of Theoretical Biology 207, 337347.
  • Chuine, I., Cour, P. and Rousseau, D.D. (1998) Fitting models predicting dates of flowering of temperate-zone trees using simulated annealing. Plant, Cell and Environment 21, 455466.
  • Dookozlian, N.K. (1999) Chilling temperature and duration interact on the budbreak of ‘Perlette’ cuttings. HortScience 34, 10541056.
  • Fennell, A. and Hoover, E. (1991) Photoperiod influences growth, bud dormancy and cold acclimation in Vitis labruscana and Vitis riparia. Journal of the American Society of Horticultural Science 116, 270273.
  • Garcia de Cortazar-Atauri, I., Brisson, N. and Gaudillere, J.P. (2009) Performance of several models for predicting budburst date of grapevine (Vitis vinifera L). International Journal of Biometeorology 53, 317326.
  • Goergen, K., Beersma, J., Hoffmann, L. and Junk, J. (2013) ENSEMBLES-based assessment of regional climate effects in Luxembourg and their impact on vegetation. Climatic Change 119, 761773.
  • Hill, G.K., Götz, G., Thomas, J., Neumann, L., Schwappach, P. and Fuchs, B. (2011) Frostschäden im Weinbau – Zwischen 2 und 100%. Das Deutsche Weinmagazin 10, 1015.
  • Janssen, P.H.M. and Heuberger, P.S.C. (1995) Calibration of process-oriented models. Ecological Modelling 83, 5566.
  • Johnson, D.E. and Howell, G.S. (1981) Factors influencing critical temperatures for spring freeze damage to developing primary shoots on concord grapevines. American Journal of Enology and Viticulture 32, 144149.
  • Junk, J., Eickermann, M., Görgen, K., Beyer, M. and Hoffmann, L. (2012) Ensemble-based analysis of regional climate change effects on the cabbage stem weevil (Ceutorhynchus pallidactylus (Mrsh.)) in winter oilseed rape (Brassica napus L). The Journal of Agricultural Science 150, 191202.
  • Knutti, R., Furrer, R., Tebaldi, C., Cermak, J. and Meehl, G.A. (2010) Challenges in combining projections from multiple climate models. Journal of Climate 23, 27392758.
  • Kobayashi, Y. and Weigel, D. (2007) Move on up, it's time for change – mobile signals controlling photoperiod-dependent flowering. Genes and Development 21, 23712384.
  • Kühn, N., Ormeno-Nunez, J., Jaque-Zamora, G. and Perez, F.J. (2009) Photoperiod modifies the diurnal expression profile of VvPHYA and VvPHYB transcripts in field-grown grapevine leaves. Journal of Plant Physiology 166, 11721180.
  • Lavee, S. and May, P. (1997) Dormancy of grapevine buds – facts and speculation. Australian Journal of Grape and Wine Research 3, 3146.
  • Lisek, J. (2008) Climatic factors affecting development and yielding of grapevine in central Poland. Journal of Fruit and Ornamental Plant Research 16, 285293.
  • Lorenz, D.H., Eichhorn, K.W., Bleiholder, H., Klose, R., Meier, U. and Weber, E. (1995) Phenological growth stages of the grapevine, Vitis vinifera L. ssp. vinifera. Codes and descriptions according to the extended BBCH scale. Australian Journal of Grape and Wine Research 1, 100103.
  • Matzarakis, A., Rammelberg, J. and Junk, J. (2013) Assessment of thermal bioclimate and tourism climate potential for central Europe – the example of Luxembourg. Theoretical and Applied Climatology 114, 193202.
  • Mayer, D.G. and Butler, D.G. (1993) Statistical validation. Ecological Modelling 68, 2122.
  • Metropolis, N., Rosenbluth, A.V., Rosenbluth, M.N., Teller, A.H. and Teller, E. (1953) Equation of state calculations by fast computing machines. Journal of Chemical Physics 21, 10871092.
  • Molitor, D. and Junk, J. (2011) Spätfrostschäden 2011 – Boten des Klimawandels? Das Deutsche Weinmagazin 22, 3033.
  • Molitor, D., Baron, N., Krause, R., Augenstein, B., Mugnai, L., Rinaldi, P.A., Skaventzou, M., Sofia, J., Hill, G.K., Dubuis, P.-H., Jermini, M., Berkelmann-Loehnertz, B., Beyer, M., Evers, D. and Bleyer, G. (2013) VitiMeteo Black rot': development and validation of a novel grape black rot model. Journal of Plant Pathology 95, S1.80.
  • Nendel, C. (2010) Grapevine bud break prediction for cool winter climates. International Journal of Biometeorology 54, 231241.
  • Orlandini, S., Mancini, M., Maracchi, G. and Moriondo, M. (2003) Le gelate primaverili: metodo di valutazione del rischio per il vigneto. Vignevini 30, 100104.
  • Orlandini, S., Di Stefano, V., Lucchesini, P., Puglisi, A. and Bartololini, G. (2009) Current trends of agroclimatic indices applied to grapevine in Tuscany (Central Italy). Quarterly Journal of the Hungarian Meteorological Service 113, 6978.
  • Perez, F.J., Kühn, N., Ormeno-Nunez, J. and Rubio, S. (2009) The expression of VvPHYA and VvPHYB transcripts is differently regulated by photoperiod in leaves and buds of grapevines. Plant Signaling and Behavior 4, 614616.
  • Perez, F.J., Kühn, N. and Vergara, R. (2011) Expression analysis of phytochromes A, B and floral integrator genes during the entry and exit of grapevine-buds from endodormancy. Journal of Plant Physiology 168, 16591666.
  • Poling, E.B. (2008) Spring cold injury to winegrapes and protection strategies and methods. HortScience 43, 16521662.
  • Pouget, R. (1963) Recherches physiologiques sur le repos végétatif de la vigne (Vitis vinifera L.): la dormance des bourgeons et le mécanisme de sa disparition. Thèse pour l'obtention du grade de docteur de l'Université de Bordeaux.
  • Riou, C. and Pouget, R. (1992) Nouvelles propositions pour évaluer la vitesse dedébourrement des bourgeons de la vigne et modélisation de la date de débourrement. Journal International des Sciences de la Vigne et du Vin 26, 6374.
  • Santos, J.A., Malheiro, A.C., Pinto, J.G. and Jones, G.V. (2012) Macroclimate and viticultural zoning in Europe: observed trends and atmospheric forcing. Climate Research 51, 89103.
  • Schnabel, B.J. and Wample, R.L. (1987) Dormancy and cold hardiness in Vitis vinifera L. cv White Riesling as influenced by photoperiod and temperature. American Journal of Enology and Viticulture 38, 265272.
  • Schultz, H.R. (1992) An empirical-model for the simulation of leaf appearance and leaf-area development of primary shoots of several grapevine (Vitis vinifera L.) canopy-systems. Scientia Horticulturae 52, 179200.
  • Suklitsch, M., Gobiet, A., Truhetz, H., Awan, N., Göttel, H. and Jacob, D. (2011) Error characteristics of high resolution regional climate models over the Alpine area. Climate Dynamics 37, 377390.
  • Themeßl, M.J., Gobiet, A. and Heinrich, G. (2011) Empirical-statistical downscaling and error correction of regional climate models and its impact on the climate change signal. Climatic Change 112, 449468.
  • Trought, M.C.T., Howell, G.S. and Cherry, N. (1999) Practical considerations for reducing frost damage in vineyards. Report to New Zealand Winegrowers (New Zealand Winegrowers: Auckland, New Zealand).
  • Urhausen, S., Brienen, S., Kapala, A. and Simmer, C. (2011) Climatic conditions and their impact on viticulture in the Upper Moselle region. Climatic Change 109, 349373.
  • van der Linden, P. and Mitchell, J.F.B. (2009) ENSEMBLES: Climate change and it's impacts: summary of research and results from the ENSEMBLES Project (Met Office Hadley Centre: Exeter, England).
  • Van Pelt, S.C., Beersma, J.J., Buishand, T.A., van den Hurk, B.J.J.M. and Kabat, P. (2012) Future changes in extreme precipitation in the Rhine basin based on global and regional climate model simulations. Hydrology and Earth System Sciences Discussions 9, 65336568.
  • White, M.A., Diffenbaugh, N.S., Jones, G.V., Pal, J.S. and Giorgi, F. (2006) Extreme heat reduces and shifts United States premium wine production in the 21st century. Proceedings of the National Academy of Sciences of the United States of America 103, 1121711222.
  • Ylhäisi, J.S., Tietäväinen, H., Peltonen-Saino, P., Venäläinen, A., Eklund, J., Räisänen, J. and Jylhä, K. (2010) Growing season precipitation in Finland under recent and projected climate. Natural Hazards and Earth System Sciences 10, 15631574.