Extended periods without precipitation, observed for example in central Europe including Germany during the seasons from 2018 to 2020, can lead to water deficit and yield and quality losses for grape and wine production. Irrigation infrastructure in these regions to possibly overcome negative effects is largely non-existent. Regional climate models project changes in precipitation amounts and patterns, indicating an increase in frequency of the occurrence of comparable situations in the future. In order to assess possible impacts of climate change on the water budget of grapevines, a water balance model was developed, which accounts for the large heterogeneity of vineyards with respect to their soil water storage capacity, evapotranspiration as a function of slope and aspect, and viticultural management practices. The model was fed with data from soil maps (soil type and plant-available water capacity), a digital elevation model, the European Union (EU) vineyard-register, observed weather data, and future weather data simulated by regional climate models and downscaled by a stochastic weather generator. This allowed conducting a risk assessment of the drought stress occurrence for the wine-producing regions Rheingau and Hessische Bergstraße in Germany on the scale of individual vineyard plots. The simulations showed that the risk for drought stress varies substantially between vineyard sites but might increase for steep-slope regions in the future. Possible adaptation measures depend highly on local conditions and are needed to make targeted use of water resources, while an intense interplay of different wine-industry stakeholders, research, knowledge transfer, and local authorities will be required.
Global mean temperature has increased and each decade since the 1980s has been warmer than any preceding one since 1850 (WMO, 2020). Accordingly, warming during the growing season (April–October, Northern Hemisphere; October–April, Southern Hemisphere) has been observed in all studied wine regions on several continents over the past 50–60 years (Schultz, 2000; Jones et al., 2005a; Webb et al., 2007, 2011; Santos et al., 2012). Changes in temperature have a pronounced effect on the geographical distribution of where grapevines can be grown (Kenny and Harrison, 1992; Jones et al., 2005b; Schultz and Jones, 2010; Santos et al., 2012), since this crop is highly responsive to environmental conditions (Sadras et al., 2012a). Within the existing production areas, where temperature conditions are generally in favour for cultivation, water shortage is probably the most dominant environmental constraint (Williams and Matthews, 1990). Even in moderate temperate climates, grapevines often face some degree of drought stress during the growing season (Morlat et al., 1992; Van Leeuwen and Seguin, 1994; Gaudillère et al., 2002; Gruber and Schultz, 2005; Gruber, 2012). Soil moisture has decreased across Europe since the beginning of the 20th century (Hanel et al., 2018), and in the most recent decade, the severity of drought events has increased in southwestern Germany (Erfurt et al., 2020). This was in part a consequence of observed recent increases in potential evapotranspiration (Bormann, 2011; Hartmann et al., 2013; Schultz, 2017) and the natural variability of precipitation.
Despite some newly emerging wine regions at extreme latitudes to the north
(Jones and Schultz, 2016), Germany's wine-growing regions are still at the
northern fringe of economically important grape cultivation in Europe.
Historically, viticulture is practised only in climatically favourable regions, mostly located on slopes or lowlands along river valleys in the
southwest of Germany. In many of these areas viticulture is the main
socio-economic factor, determining the cultural landscape, with steep-slope
regions additionally forming biodiversity hotspots (Jäger and Porten,
2018; Petermann et al., 2012). Mean annual precipitation is generally low in
these steep-slope regions (500–770 mm; 1971–2000; Ahr, Mittelrhein, Mosel,
Nahe, Rheingau; DWD, 2020), and available water capacity (AWC) of soils is very heterogeneous. Additionally, the percentage of vineyards with low AWC is relatively high (for example, in the Rheingau region AWC
High spatial-resolution predictions are a challenge in climate impact studies and are mainly limited by the size of one grid box of regional climate
models (RCMs). Although climatic conditions within a grid box may change from being suitable for vineyards to areas unsuitable for the cultivation of grapevines, climate change impact studies for European viticulture were often forced to be performed based on the spatial resolution of the underlying gridded climate model data. Santos et al. (2012) analysed observed shifts in bioclimatic indices (mainly temperature related) by means of the E-OBS gridded data set and the connection with large-scale atmospheric forcing. Projections of bioclimatic indices based on RCMs were analysed by Malheiro et al. (2010) and Fraga et al. (2013), with the latter study also including possible changes in interannual variability. In terms of water supply, both studies projected a strong decrease in water availability for the Mediterranean Basin but their projections differed for central Europe ranging from a slight decrease (Fraga et al., 2013) to an increase (Malheiro et al., 2010). More specific regional aspects were analysed by Santos et al. (2013) for the future of wine production in the Douro Valley (Portugal) and Moriondo et al. (2010) for expected changes in the premium wine quality area of Tuscany at a fine spatial resolution (1 km
In addition to weather conditions, the water balance of grapevines also depends on vineyard geometry (row spacing, canopy height, etc.), the training system (canopy shape), soil management practices, and particularly site-specific factors such as AWC, slope, and aspect (Hofmann and Schultz, 2015). These factors describe the interaction of the vineyard site microclimate with water supply and atmospheric demand (Hoppmann et al., 2017; Sturman et al., 2017). AWC, slope, and aspect are particularly heterogeneous in regions of complex terrain resulting in variability in the supply of and demand for water. Increasing water scarcity can put economic pressure on established growing regions because severe drought stress causes losses of grape quality and yield. Adaptation measures such as the implementation of irrigation systems are expensive, and access to water in many places is restricted and difficult. Although irrigation of grapevines has been allowed since 2002 in Germany, water withdrawal rights may also need to be adapted if water is taken from groundwater or surface water bodies. Since precipitation patterns are highly variable in space and time, it is problematic for growers and stakeholders to assess future developments and to make decisions for long-term mitigation and adaptation measures. Against this background, the identification of those vineyard plots or sites within growing regions likely exposed to an increasing risk of drought stress in the future can support the decision-making process.
Map showing the wine-growing regions Rheingau and Hessische Bergstraße and the locations of weather stations (source of the base map (modified): Esri, 2012).
Therefore, the main objective of the study is to quantify the likelihood of risk of future water deficit on the spatial scale of individual vineyard plots within two German grape growing regions: Rheingau and Hessische Bergstraße. The scientific process included (i) statistical downscaling of an ensemble of climate-model-simulated data to the scale of station data, (ii) combining information from land registers, high-resolution soil maps, and digital elevation models in order to characterise vineyard landscapes and their microclimate, (iii) performing vineyard water balance simulations driven by observed and simulated weather data for all vineyard plots.
The risk analysis was conducted for 2 out of the 13 German wine-growing regions, the Rheingau and the Hessische Bergstraße, both located in the federal state of Hesse (Fig. 1). In the Rheingau, grapevines are cultivated on an area of 3191 ha (Destatis, 2018). The Rheingau is physiographically divided into the regions of upper and lower Rheingau (Löhnertz et al., 2004). The upper Rheingau includes an area of approximately 25 km length and 3–6 km width between Wiesbaden and Rüdesheim, bounded by the Rhine River to the south and the ridge of the Taunus mountain range in the north as well as the vineyards near Hochheim on the Main River. Grapevines are cultivated between approximately 80–280 m altitude on a gently rolling hill scape. For most of the region, the soils developed from loess or sandy loess as parent material. They are fertile and have a balanced water budget. Soil erosion, intensified by agriculture over thousands of years, filled dells and, in conjunction with soil formation by a variety of basement rocks (sand, clay, marl, limestone), led to the further differentiation of soils, where the loess layers were thin. The soils of the lower Rheingau to the west of Rüdesheim are very different. The direction of the Rhine changes northward here into the Upper Middle Rhine Valley with its steep slopes. The parent material of the soil formation consists mainly of shallow glacial solifluction layers containing a lot of basement rock (sandstone, quartzite, slate). These soils are nutrient-poor, stony, and shallow and generally have a low AWC (Löhnertz et al., 2004; Böhm et al., 2007). The second wine-growing region of Hesse, the Hessische Bergstraße, has a cultivated area of 462 ha (Destatis, 2018). The vineyards are located on the western slopes of the Odenwald mountain range and at the eastern edge of the Upper Rhine plain. Soils developed from loess are also dominant here. About 60 % of the soils are deep and rich, with an AWC exceeding 200 mm, while about 20 % of the soils have an AWC below 125 mm, particularly at sites where the rooting depth is limited to 60–100 cm (Löhnertz et al., 2004).
The longest-running weather station at Geisenheim (since 1884 in close proximity to the university and serviced by the Deutscher Wetterdienst, DWD,
German Meteorological Service) had an average growing season temperature
(AGST; April–October) for the reference period 1961–1990 of 14.5
Description and parameters for precipitation (
In order to run the water balance model, transient daily data for temperature, global radiation, relative humidity, wind speed, and precipitation are required. Air temperature is used to model the development
of grapevines and cover crops over annual cycles and, together with global
radiation, wind speed, and relative humidity, to calculate reference
evapotranspiration (ET
The first observed series included daily weather data from 1959–1988 (the recording ended in 1989 at some of the stations) of 10 weather stations (6 in the Rheingau and 4 in the Hessische Bergstraße) distributed across the regions (Fig. 1, Table 1) and was provided by the DWD (2018). Precipitation was recorded at all stations. At the station Bensheim (Hessische Bergstraße) temperature and relative humidity were additionally measured. The station Geisenheim (Rheingau) provided data for all five weather variables. More precisely, sunshine hours (SHs) were measured here over the complete period providing a proxy for global radiation (GR). A parallel measurement period of GR and SHs at Geisenheim between 1981 and 1990 was used to establish correlation coefficients between these parameters (Hofmann et al., 2014) based on the Ångström–Prescott equation, and GR was calculated accordingly. In order to be able to use time series with all five weather variables for each station in the subsequent analysis, missing temperature and relative humidity data at the stations in the Hessische Bergstraße were set equal to the measured data from Bensheim, and those at the stations in the Rheingau were set as equal to the data measured at Geisenheim. Wind speed and GR at all stations were set equal to the data measured at Geisenheim. These data were used as model inputs for an assessment of the drought stress occurrence in the past as well as to calibrate the weather generator with respect to the observed baseline climate for all stations (see details below).
The second series included daily data from 2014–2018 and came from newly established weather stations (Fig. 1) by the Hochschule Geisenheim University. These data were used for an assessment of observed drought stress in the recent past.
Input weather series representing the baseline and future climate conditions
were produced by the parametric stochastic weather generator (WG) M & Rfi,
which is an improved follow-up version of the Met & Roll generator (Dubrovský et al., 2000, 2004). Met & Roll was based on the version by Wilks (1992) (adopted for use in future climate conditions) of the classical parametric generator developed by Richardson (1981). M & Rfi is a single-site multi-variate daily weather generator, in which the precipitation time series is modelled by a first-order Markov chain (occurrence of wet and dry days) and Gamma distribution (precipitation amount on wet days). The non-precipitation variables are simulated by a first-order autoregressive model whose parameters depend on the wet/dry status of a given day. The M & Rfi generator has been used in many climate change impact experiments (e.g. Rötter et al., 2011; Hlavinka et al., 2015; Garofalo et al., 2019). This generator also participated in a complex validation experiment of the so-called VALUE project aiming at comparison of various downscaling approaches (Maraun et al., 2019; Gutiérrez et al., 2019; Hertig et al., 2019). Two types of synthetic time series were produced by M & Rfi (Fig. S1 in the Supplement shows a flow diagram). The first time series representing the present (baseline) climate was used to validate the generator by comparing selected weather statistics derived from synthetic vs. observed weather series. The second one representing the future climate was used to assess changes in the drought stress occurrence for future climate change scenarios. In producing the first time series, WG parameters representing the statistical structure of the weather variability between 1959–1988 were derived from the observed station data (baseline climate), and then a 112-year synthetic series (1989–2100) representing the baseline climate (i.e. assuming no climate change) was produced by the WG. For the climate change scenarios (second series), we modified the WG parameters based on climate change scenarios derived from 10 future climate simulations made within the frame of the ENSEMBLES project (van der Linden and Mitchell, 2009; Table 2). Here, RCMs were used, which were driven by various global climate models (GCMs) (Table 2) and run for the A1B emission scenario and approximately 25 km grid resolution. For each station and climate simulation, the data of the four nearest RCM grid boxes enclosing
the weather stations were used to derive changes in WG parameters representing the RCM-based climate change scenario for 2058–2087. In order
to construct transient time series consisting of observed data from 1961–1988 followed by synthetic weather data until 2100 (assuming a smooth increase in climate change signal), the WG parameters representing a given year
Ensemble of climate models (van der Linden and Mitchell, 2009).
Climate models of the ENSEMBLES project were used instead of the successor project EURO-CORDEX (Jacob et al., 2014) for reasons of data availability at the time the study was started. Since Kotlarski et al. (2014) reported comparable biases for both projects and since it can be deduced from Feldmann et al. (2013) that the benefit from the higher spatial resolution of EURO-CORDEX is small in the area of the study region, we concluded that the ENSEMBLES data were suitable.
The chosen RCMs were evaluated in several studies. Model errors and statistics of precipitation and temperature were analysed by Frei et al. (2003), Kjellström et al. (2010), and Suklitsch et al. (2011). Maule et al. (2013) evaluated the RCMs using drought statistics. The models showed reasonable skills in projecting weather characteristics relevant for our study.
We used the vineyard water balance model of Hofmann et al. (2014), which was
developed and validated with the general growing and cultivation conditions
of the study area presented here. This model accounts for different soil
cultivation (bare soil, use of cover crops, or alternating use of both), and
the impact of slope and aspect of the vineyard plots on received global
radiation and ET
Since the individual geometry of each vineyard plot within the two regions
was unknown, the calculations of radiation interception were performed for a
uniform geometry representing a standard vertical shoot positioning system
in Germany (Hoppmann et al., 2017). We used 2 m row spacing, a foliage width
of 0.4 m, a maximum row height of 2.10 m, and minimum height of the lower end of the canopy of 0.8 m above the soil surface as base data. This is typical for the current and mid-term future situation because more than 80 % of the new planted vineyards from 2002–2013 (
For the simulations, it was assumed that in the Rheingau, soil cultivation and cover crops alternate in every second row, and in the Hessische Bergstraße, the soil is completely covered by vegetation (except for a strip of 0.4 m under the vines). This is currently typical for both regions.
The study is based on the high spatial resolution of individual plots (Fig. S3 shows a flow diagram). The underlying data were provided in digital form as spatial polygons from the local authority in charge of the official EU vineyard register (RPDA, 2012). This resolution can be regarded as high, with a total plot number of planted vineyards of 24 858, with a mean area of 0.15 ha per plot. Plots up to 0.5 ha take up 79 % of total planted area with a maximum plot size of 4.2 ha. Each plot was linked with a digital elevation model at 1 m resolution to calculate the mean slope and aspect of the plots. The water balance model needs values for the available water capacity up to 2 m depth (AWC
For each vineyard plot, the climate data of the nearest weather station were
used to calculate the water balance. For almost all vineyards (
As an indicator of drought stress, we calculated the yearly sum of drought
stress days during the vegetation period (1 May–30 September). A day was
classified as a drought stress day if the remaining soil water content was
smaller than 15 % AWC
The R package multitaper (Karim et al., 2014) was used to compare power spectra of observed and synthetic time series (see the following Sect. 3.1). This package also provides a harmonic
Downscaling data of climate models to station data is not trivial, and all of
the possible methods have pros and cons, which have to be considered in order to interpret the data and results (Maraun et al., 2010). In order to assess specific features of the downscaling approach used, we compared 30 consecutive years of the first synthetic time series (representing the
baseline climate, see Sect. 2.2.2) produced by the WG for the station
Geisenheim with observational weather recorded at Geisenheim from 1959–1988. We compared the characteristics of the weather variables precipitation, global radiation, and the derived ET
Comparison of 30 years of observational weather data (1959–1988,
station Geisenheim, Rheingau, Germany) and 30 years of synthetic weather
data produced by the weather generator calibrated with the observational
weather data.
Further comprehensive statistical validation studies were already performed in the framework of the VALUE project (Maraun et al., 2015), where the M & Rfi WG was a member of a large ensemble of statistical downscaling methods. Briefly summarised, the WG showed small biases for most of the climate characteristics studied, but underestimates were reported for precipitation variability (Gutiérrez et al., 2019), interannual variability (temperature and precipitation; Maraun et al., 2019), and long wet or dry spells (Hertig et al., 2019).
To perform an indirect validation of the WG, we compared the annual sum of
drought stress days during the vegetation period (Sect. 2.5), calculated with the water balance model for both the observed and the synthetic time series of three weather stations: Geisenheim, Hochheim (in the west and east of the Rheingau), and Bensheim (Hessische Bergstraße). In order to get a valid drought stress response to the weather data, the water balance model was parameterised for a vineyard with a comparably low AWC
Figure 3 illustrates the trend and interannual variability of the water
availability expressed as climatic water balance and calculated with data of
the weather station Geisenheim. The climatic water balance represents the
difference between the sum of precipitations and the sum of reference
evapotranspiration over a hydrological year (1 November–31 October). The
presentation of Fig. 3 does not directly allow conclusions on the extent of
drought stress of a certain year, which additionally depends on site factors
and the plant response to limit water use. Nevertheless, the climatic water
balance has decreased by about 90 mm if the two 30-year periods from 1959–1988 and 1989–2018 are compared. Additionally, the frequency of years with a climatic water balance lower than
Climatic water balance, expressed as the difference between the sum of precipitations and the sum of reference evapotranspiration for a hydrological year (1 November–31 October) for the station Geisenheim (Rheingau), Germany. The solid line shows 11-year running mean values. The decreasing trend is significant (
Water balance calculations for both growing regions with the data of the observation period from 1959–1988, showed that the 5 driest years were 1959, 1964, 1973, 1974, and 1976. In 1959 and 1976, this was related to hot and dry summers with many sunshine hours and high evaporative demand and in 1964, 1973, and 1974 because of extreme dry winters despite only average summers. On average, drought stress days were calculated in the range of 0–41 d yr
Figure 4 shows the strong interannual variability of the soil water content dynamics for a typical vineyard (AWC
Seasonal patterns of the fraction of available water capacity (AWC
The map in Fig. 5 shows the simulated spatial distribution of the sum of
drought stress days for the entire Rheingau region for the year 2018 based
on data of the weather station network (Fig. 1). The year 2018 had the
highest sum of annual ET
Number of simulated drought stress days per vineyard plot for the wine-growing region Rheingau, Germany, during the 2018 vegetation period (1 May–30 September). Calculations were conducted with a water balance model based on data from weather stations and a digital soil map on the assumption of alternating soil cultivation (one row bare soil, one row cover crop).
Annual precipitation projected by the ensemble of climate simulations for
the station Geisenheim based on the emission scenario RCP8.5 showed a high
variability (Fig. 6a). The change signal (difference in mean values between
the time period 2071–2100 and the period of observed values 1961–1988)
ranged from a decrease of
In comparison, the ensemble results for RCP4.5 showed substantial smaller
change signals for annual precipitation ranging from
Seasonal precipitation simulated with 10 climate models for the station Geisenheim (Rheingau), Germany, for the emission scenario RCP8.5.
Grey lines show the range of annual values of all models, coloured lines
11-year running means for individual model runs. The period from 1961–1988
represents observed data and the dashed baselines illustrate their mean
values.
Seasonal trends of the model ensemble for RCP8.5 are shown in Fig. 8. In part, the results of precipitation change signals (2071–2100 compared to 1961–1988, Table 3) reflected possible future seasonal shifts. The range of change signals of the transition seasons spring (March, April, May – MAM;
Taking into account reference evapotranspiration by calculating the seasonal
climatic water balance, the picture changes towards dryer conditions (Fig. 9, Table 3). In winter, the plus of precipitation is slightly reduced due to
higher ET
The results for RCP4.5 showed smaller change signals for precipitation and
the climatic water balance (Figs. S11–S12 and Table S4). The projected increase in winter precipitation for RCP4.5 was about half as large as the increase for RCP8.5 for most simulations. Summer precipitation is also projected to decrease less in RCP4.5 compared to RCP8.5 and ranged from
Range of change signals of 10 climate simulations with different
models for the station Geisenheim (Rheingau), Germany, for the emission
scenario RCP8.5. For precipitation (
Seasonal climatic water balance simulated with 10 climate models for the station Geisenheim (Rheingau), Germany, for the emission scenario RCP8.5. Grey lines show the range of annual values of all models, coloured lines 11-year running means for individual model runs. The period from 1961–1988 represents observed data and the dashed baselines illustrate their mean values.
As most of the climate simulations for RCP8.5 showed significant annual
precipitation trends in the second half of the century (Fig. 6b) and indicated changes in climatic water balance, we calculated the average number of drought stress days for the time periods 1989–2018 and 2041–2070 for each vineyard plot and climate model. Based on this calculation, two indices were derived. The first one describes the overall grape-growing surface area affected by drought stress, defined as the sum of the area of all individual vineyard plots with on average per time period 10 or more days of drought stress during the vegetation period. The second one is the drought stress change signal, calculated as the difference of the average number of drought stress days per vineyard plot and climate simulation between both time periods. The calculation of the grape-growing surface area affected by drought showed that three models projected a substantial increase in this area for both regions of possibly 10 % to 30 % (Rheingau) and 16 %–20 % (Hessische Bergstraße) for the future period 2041–2070. Among these three models were the two projecting a decrease in annual precipitation and the largest decrease in annual climatic water balance, described in more detail in Sect. 3.3.1. The third model illustrates further future weather patterns that could lead to a strong increase in drought stress. This model projected increasing precipitation in SON, DJF, and MAM but a strong decrease in precipitation in JJA and additionally a strong increase in ET
Potential drought stress area of two wine-growing regions (Rheingau and Hessische Bergstraße) in Germany for two time periods (1989–2018 and 2041–2070), calculated with a water balance model, soil maps, and 10 climate simulations with different models for the emission scenario RCP8.5. A vineyard plot was allocated to the drought-stress-affected area if on average 10 or more days with drought stress during the vegetation period (1 May–30 September) were calculated. Individual model results are shown as points in the boxplots.
Similarly, for RCP4.5, seven models projected no or only small changes in
the range of
The calculation of the drought stress change signals per vineyard plot allowed the creation of maps, to illustrate spatially the impact of the
projected climate trends. For RCP8.5, the maps for the “dry” and for the
“wet” simulation at the extremes and the simulation close to the median of
the ensemble (Fig. 10) are shown in Fig. 11 (Rheingau) and Fig. 12
(Hessische Bergstraße). In the case of the dry simulation (Fig. 11a), the
vineyards where drought stress already occurred in the past (in the lower
Rheingau, and near Johannisberg 50.0
Projected change in the occurrence of drought stress days for the
growing region Rheingau (Germany), for the emission scenario RCP8.5, calculated with a water balance model on the assumption of alternating soil
cultivation (one row bare soil, one row cover crop). The maps show the
difference between the number of the mean drought stress days per vegetation
period (1 May–30 September) for the periods 2041–2070 minus 1989–2018 at the spatial scale of the individual vineyard plots.
Projected change in the occurrence of drought stress days for the
growing region Hessische Bergstraße (Germany) for the emission scenario
RCP8.5, calculated with a water balance model on the assumption of cover
crop use in every row. The maps show the difference between the number of
the mean drought stress days per vegetation period (1 May–30 September) for the periods 2041–2070 minus 1989–2018 at the spatial scale of the individual vineyard plots.
For RCP4.5 and the Rheingau (Fig. S14), the dry and the “medium” simulation projected a much smaller increase and the wet simulation a stronger decrease in drought stress days compared to RCP8.5. For the dry simulation drought stress would also occur on some vineyards near Geisenheim and Hochheim with high AWC but compared to RCP8.5 on an overall smaller area and less pronounced. The almost negligible increase in drought stress for the medium simulation would affect only sites with low AWC. For RCP4.5 and the Hessische Bergstraße, a smaller increase in drought stress is projected for almost the same areas compared to RCP8.5. No changes in drought stress would occur for the medium simulation and drought stress could decrease on a few plots for the wet simulation.
Climate projections and impact analyses are subject to a number of uncertainties. In the understanding of climate change, these uncertainties are in general related to the uncertain future external forcing by greenhouse gas emissions, the impact of external forcing factors on climate, and the degree of natural variability of the climate system (Kjellström et al., 2011). In impact analyses, methodical imperfections of the impact models result in further uncertainties. This study looked at a comparably small region; thus the ability of the RCMs to reproduce spatial weather patterns is one additional source of uncertainty. The water balance model itself or previous versions have been validated with field observations on different vineyard plots of the current study area as well as other regions and in different climates (Lebon et al., 2003; Pellegrino et al., 2006). Yet, on a regional scale, it requires high-quality soil data, which have a strong influence on the result of the calculations as a possible source of error. The soil data go back mainly to soil mappings conducted from 1947–1958 (Böhm et al., 2007). Since then, based on land consolidation projects and individual interventions in parts of the complete landscape, some attributes might have changed in local spots, but in general the soil maps still describe the current situation quite well as demonstrated in a follow-up study (Zimmer, 1999).
To capture the magnitude of uncertainties related to possible future climate
evolution for the selected emission scenarios, we used climate projections for the period 2058–2087 simulated by 10 climate models of the project. These data were used to derive the climate change scenario, which was further scaled by smoothly increasing change in global mean temperature (as projected by the MAGICC model for the selected RCP8.5 emission scenario) and used to modify the weather generator parameters, in order to produce transient time series for several weather stations. The simulations showed a high range of the future precipitation change at the end of the century. This range is comparable with the results of the REKLIES-DE project (
One water budget simulation driven by the climate models predicted that drought stress would be less problematic in the future. This would not be
expected from observations in the recent past, where drought stress occurrence has increased. The decrease in the climatic water balance (Fig. 3) is related to an increase in ET
Despite the reduced interannual variability, the climate projections showed seasonal shifts. The impact of the seasonality of precipitation on grape quality is not fully understood (Sadras et al., 2012b). Dry conditions during the ripening period and harvest are in general positive for fruit quality and health, but severe drought stress can lead to a cessation of sugar accumulation, as observed in specific plots of the study area during the 2018 and 2019 vintages. Seasonal shifts in precipitation could reduce the impact of dry spells on plots with sufficient capacity to store available water by enhanced refilling in winter. Trömel and Schönwiese (2007) reported that the trends for the probability for observed monthly extreme precipitation in Germany varied substantially on a spatial scale and projected near-future changes in extreme precipitation also showed heterogeneous spatial change patterns in summer (Feldmann et al., 2013). The performance of many downscaling and bias correction methods to represent temporal aspects of the climate has only recently become a topic of research (Maraun et al., 2019).
The water balance model currently does not account for the impact of
increasing
For both emission scenarios, the models close to the median showed a small
increase in the number of drought stress days in the range of 5–20 d for
vineyards of the lower Rheingau and small parts of the upper Rheingau, in
general on plots where drought stress had already occurred in the recent past.
However, the affected area for RCP4.5 is only about half as large as the
area for RCP8.5. From these simulations, some sub-regions with an increased
future risk for drought stress could be identified. For already irrigated
plots, the scenario outcomes mean that growers would have to irrigate between one to three times more per season on average. The threshold value used to classify a day as a drought stress day (
The simulations showed a widespread array of possible changes making it difficult to generalise adaptation strategies. Both viticultural regions are located in areas where nitrate leaching to the groundwater is a severe environmental issue (Löhnertz et al., 2004). This threat would certainly be enhanced in the future because of higher mineralisation rates, caused by increasing temperature (both air and soil) and rainfall in winter (Table 3). The use of cover crops or natural vegetation to cover the soil on the complete vineyard surface area during the winter months is the most important measure to counteract this development (Berthold et al., 2016). Similarly, these measures and possibly reduced tillage are also important for the summer months to protect against leaching and erosion. Cover crops also reduce surface runoff and increase infiltration but compete with the grapevines for nutrients and water. On steep slopes with shallow soils, grapevine roots and cover crops share much of the same soil reservoir. Consequently, tillage in spring and cover crops in alternate rows has become a standard praxis, balancing the advantages and disadvantages of cover crop use. Wide row spacing could reduce the water use due to the lower planting density, but this would increase the risk of erosion in the cultivated rows. The possibilities to influence the water balance by canopy management are therefore limited in these situations and need also to be considered in the context of the cost disadvantages of steep-slope viticulture (Strub and Loose, 2021). A further interesting long-term viticultural adaptation strategy is the use of rootstocks with enhanced drought tolerance (Ollat et al., 2016).
On the other hand, following the climate projections, irrigation should be possible against the background of the projected shifts from summer to winter precipitation amounts and increasing annual precipitation. This generally offers the opportunity to withdraw and store water from surface water bodies during periods with high flow rates, as potential conflicts with the use of drinking water, which is usually withdrawn from groundwater bodies, could be avoided. Expanded use of bank filtration could also help to avoid future resource conflicts. The construction of such extensive infrastructure measures requires an interplay of all actors involved.
Due to increased temperature combined with relatively unchanged but still
highly variable precipitation patterns (Fig. 8c), the increased occurrence
of warm and wet conditions during the ripening period (September, October)
has increased the risk for rot (Schultz and Hofmann, 2015). A similar climatic trend regarding the decoupling of the relationships between temperature, drought, and early wine grape harvests was reported by Cook and
Wolkovich (2016) for France and Switzerland. These types of non-stationarities are reflected in more or less new environmental conditions and weather patterns, which are a challenge for cultivation. Apart from the water balance, these challenges in the Rheingau (like in other regions) primarily span the management of vigour, yield, grape maturity, and disease management, against a background of a high terrain complexity and natural climate variability (Neethling et al., 2019). The need to assess and apply adaption measures at a regional level down to individual plots, is also evident from our study. For future impact research studies, it could be beneficial to apply regional convection-permitting climate modelling (grid spacing
Based on an ensemble of climate model simulations, a water balance model, a digital soil map, an elevation model, and a land register, our study provides a risk assessment with respect to the future occurrence of drought stress, applied to individual vineyard plots of the wine-growing regions Rheingau and Hessische Bergstraße. The results ranged from a small decrease (one simulation) to a moderate increase in drought stress (median of the ensemble), predominantly on plots already temporarily affected by water deficit, up to a drought stress occurrence touching 20 %–30 % of the growing regions. As drought stress is already currently observed in steep-slope vineyards with shallow soils, these sub-regions were identified as future risk areas by most of the simulations. The results illustrate the large heterogeneity of the water supply within growing regions and between neighbouring vineyards and the need to improve high-resolution modelling approaches. Mid- and long-term adaptation measures need to respect local conditions and will necessitate individual, precision-farming-like application of cultivation practices. In combination with weather station networks delivering real-time data, the presented framework may also serve as a decision support tool to growers and consultants in the future.
Observed weather data of the DWD can be found at
The supplement related to this article is available online at:
MH and HS developed the concept and research goals. DM, MD, CV, and MH designed the methodology. CV downloaded and processed the future climate change scenarios. MD programmed and calibrated the weather generator. MH ran the water balance simulations, prepared the original draft, and produced all figures; all authors contributed to writing, review, and editing.
The contact author has declared that neither they nor their co-authors have any competing interests.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This research was funded by the Hessian Agency for Nature Conservation, Environment and Geology (HLNUG) as a part of the INKLIM-A project. We thank Klaus Friedrich, Matthias Schmanke (HLNUG), and Christoph Presser (Regierungspräsidium Darmstadt, RPD) for combining the comprehensive databases of vineyard plots, elevation, and soil maps. We thank Heike Hübener (HLNUG) for fruitful discussions about how to perform climate change impact studies on individual vineyards. We also thank three referees for their careful review and helpful comments. We acknowledge support by the Open Access Publishing Fund of Hochschule Geisenheim University.
This research has been supported by the Hessian Agency for Nature Conservation, Environment and Geology (HLNUG) (grant Inklim-A).
This paper was edited by Daniel Kirk-Davidoff and reviewed by three anonymous referees.