Agroclimatic conditions in Europe under climate change

Study area and data

The current study was confined to datasets of daily weather observations provided by members of the COST734 network. The data cover the period from 1971 to 2000 and were taken from weather stations representing the key agricultural regions of the given countries and provide continuous daily data, including maximum and minimum temperatures, global radiation (or sunshine duration), precipitation, mean daily relative air humidity and wind speed. In addition, these stations (when possible) were located outside urbanized areas. Such requirements significantly reduced the number of suitable sites, especially considering that all of the sites with <95% coverage (for each separate element) were excluded from the analysis. The included sites (Fig. 1) were then grouped according to their presence within particular environmental zones (EnZ) as defined by Metzger et al. (2005) and Jongman et al. (2006). The EnZ definitions herein cover 13 zones (Fig. 1) comprising 84 environmental strata (EnS), which are classified by monthly minimum and maximum temperatures, sum of precipitation, percentage of sunshine in months representing the four seasons (January, April, July and October) together with altitude, slope, northing and oceanicity. Overall, the strata accounting for 72% of all European agricultural land were represented by at least one climate station (Table 1). To simplify the figures, only the mean values for each EnZ are presented. To limit possible bias caused by the uneven representations of EnS, results were first averaged for each individual EnS and these means were used for the calculation of EnZ values.

Complete data were collected from 86 carefully screened sites from a total of 137 provided sites (Appendix S1), and the study domain covered the area between 8.5°W–27°E and 37–63.5°N. Nineteen European states are represented in the database, including the major agricultural producers of the EU; however, several important countries and regions (e.g. the eastern Mediterranean) were not covered due to a lack of data from these areas.

Agroclimatic indices

Figure 2 provides an overview of the methodological approach of using indicators for the evaluation of changes in agroclimatic conditions in Europe under climate change. To describe agroclimatic conditions, the 11 indicators described in Tables 2a and 2b were selected from a plethora of available options to represent the potential effects of weather on crop productivity and management. The selection was made from a ‘short-list’ of approximately 120 indices. The final set of indicators was required not only to represent potential productivity and growing conditions but also field workability as well as the occurrence of extreme events relevant to agriculture. This includes impacts as well as adaptation options for the different agricultural sectors. The study further focused on late frost and drought, as they were identified as major problems across most of Europe (Olesen et al., 2011). In addition, each of the selected indices had to be applicable across all of the sites and be calculable from available datasets; furthermore, the portfolio was chosen so as to complement rather than repeat previous studies.

The daily reference (ET_r) and actual (ET_a) evapotranspiration values were calculated using the Penman–Monteith approach, as described in Allen et al. (1998), using modifications validated by Hlavinka et al. (in press). Crop growth on a given day was considered not to be significantly limited by water if the daily ratios of ET_a to ET_r exceeded 0.4–0.5 (FAO, 1979; Fisher et al., 2002; Eliasson et al., 2007). To limit eventual overestimation of water shortage, the lower end of the range (0.4) was applied here. The temperature thresholds used rely on the works of Chmielewski & Köhn (2000), Mitchell & Hulme (2002) and Larcher (2003), and were similar to those used by Fisher et al. (2002).

The Huglin index represents the thermal suitability for wine production and includes a correction factor for latitude as described by Huglin (1978). This index allows for characterizations of the suitability of viticulture in general and particular grapevine cultivars at a given location. A constraint of this index is that it does not consider cold-temperature limitations, which are critical for continental climates, and other limitations such as sunshine duration, soil conditions and water availability. Local climatic variations based on orography may also alter these conditions significantly.

The thresholds for sowing and harvest suitability (Table 2a) were based on published literature and tested using the observed sowing and harvest dates for spring barley, winter wheat and maize at 30 experimental stations at in the Czech Republic over a period of 20 years. The approach used is broadly in agreement with similar studies by Leenhardt & Lemaire (2002) and Maton et al. (2007). The soil-moisture thresholds used to define the suitable days for sowing and harvesting were stricter than those used by Rounsevell (1993) and Cooper et al. (1997), as no soil compaction or soil-structure damage should occur in sustainable agricultural systems. Across all of the investigated sites, the sowing and harvesting windows were held constant despite the varying relevance of some of these windows.

The agroclimatic parameters listed in Tables 2a and 2b were calculated with the use of a software package, agriclim (Trnka et al., 2010a), which is available from the authors. For all of the ET_r and ET_a calculations, spring barley was used as the reference crop surface because it is grown in all the investigated EnZs. When calculating the status of the available soil water, homogenous soil properties were assumed throughout the profile (top and subsoil). The soil water-holding capacity in the top 0.1 m of soil was assumed to be 0.02 m and the capacity in the entire profile (a 1.3 m soil depth) was 0.27 m. Although soil water-holding capacity (as well as other soil parameters) differed across the investigated sites, a uniform soil profile was used to allow station-to-station comparisons. When calculating evapotranspiration, an adjustment for atmospheric CO₂ concentration was made using the method proposed by Kruijt et al. (2008) using the CO₂ concentrations listed in Table 3.

Creating daily weather series under baseline and climate-change conditions

A restriction on the datasets provided meant that it was not possible to directly apply the observations. Instead, the data were used to train a stochastic weather generator (WG) M&Rfi (Dubrovskýet al., 2004), and a 99-year stochastic daily weather series of global radiation sum, maximum and minimum temperatures, precipitation sum, mean relative air humidity and wind speed were prepared to represent the baseline (1971–2000) climate conditions for each site. In the next step, the baseline WG parameters were perturbed according to the climate-change scenarios (Fig. 3) and used as inputs to the agriclim model.

Climate-change scenarios for this study were developed by means of a ‘pattern-scaling’ technique (Santer et al., 1990) from the outputs of the Global Climate Models (GCMs) and were then used to modify the parameters of the WG (as used in, e.g., Trnka et al., 2004). The ‘pattern-scaling’ technique defines a climate-change scenario as the product of the standardized scenario and the change in global mean temperature. The standardized scenarios (Fig. 3) relate the responses of climatic characteristics to a 1 °C rise in global mean temperature (ΔT_G). They were determined by applying a regression method (Dubrovskýet al., 2005) to the 2000–2099 period, which was obtained from the GCM simulations that had been run with the SRES-A2 (Special Report on Emission Scenarios) emission scenario for the IPCC Fourth Assessment Report (Nakicenovic et al., 2000; Solomon et al., 2007). The projected changes in ΔT_G at 2030 and 2050 were calculated via a simple climate model, MAGICC (Harvey et al., 1997; Hulme et al., 2000), assuming the A2 emission scenario and medium or high climate sensitivity (Table 3). As a ‘worst-case scenario,’ we also assessed changes under a mean global temperature increase of 5 °C. As the role of the climate-sensitivity factor on temperature change by 2050 was relatively small (Table 3), the responses of agroclimatic indices were similar and, therefore, we chose to report the results for the higher sensitivity only.

The three GCMs utilized were ECHAM5/MPI-OM (ECHAM), HadCM3 (HadCM) and NCAR-PCM (NCAR). Apart from representing inter-GCM variability quite well (Table 4), these three GCMs (or previous versions thereof) have been used in a number of impact studies and have generally performed well in reproducing baseline climates in various European regions (e.g. Dubrovskýet al., 2005).

Projected changes in agroclimatic parameters by 2030 and 2050

Figure 3, in combination with Table 4, indicates how overall climatic conditions might change and illustrates change patterns among the seasons and GCMs. More pronounced warming and decreased precipitation between April and September were found for the Mediterranean mountains (MDM), Lusitanian (LUS), Pannonian (PAN), Mediterranean north (MDN) and Mediterranean south (MDS) zones than in the Boreal (BOR) and Alpine north (ALN) zones. The overall patterns of change are consistent for all three GCMs in most zones, except for the colder half of the year. HadCM showed higher changes in temperature and ECHAM more pronounced changes in precipitation, while NCAR showed moderate temperature changes for both in summer, with larger temperature increases in Nemoral (NEM) and ALN during the colder half of the year.

Agroclimatic conditions under 5 °C warming

The projected change patterns in Fig. 5c are similar to those depicted in Fig. 5a and b, although here the changes (especially those that negatively affect the production potential) are more pronounced. In addition to the number of effective growing days, the effective global radiation was projected to decrease for all large agricultural zones investigated in this study except ATN (for the case of changes based on NCAR). Huglin-index values were projected to increase across all zones, reaching unprecedented levels in today's primary wine-growing regions, which may therefore become unsuitable for the currently planted grape varieties. Comparatively, the last frost was projected to occur, on average, much earlier in the year; however, there was also a marked increase in the interannual variability of the last frost date in the ATN, ATC, LUS and CON zones, which might maintain or even increase frost risk, e.g., for fruit trees, due to a concurrent shift to earlier flowering. The overall drying of most of the agriculturally important zones would be severe (especially during summer), with some zones facing the parallel challenge of higher water deficits and larger interannual variability. Most notable were the changes in water balances in the cases of the LUS and CON zones. There were significant improvements in the number of suitable days for harvest in June and July as well as for early sowing, except for the MDN and MDS zones. The late spring sowing window exhibited a large increase in interannual variability. The sowing of winter crops might become problematic because the proportion of suitable sowing days during autumn will vary dramatically in most zones. The areas that will benefit from a longer and more sustained autumn sowing window are those in the ALN, NEM, BOR and ATN zones.

ALN

The ALN zone was represented by a single weather station, which is located in the largest agricultural area within the region. While the variation in the ALN agroclimatic conditions is large (Skjelvåg, 1998), this single site adequately represents the northernmost fringe of European agricultural production. The ALN zone may expect the greatest increase in the number of effective growing days; by 2050, the increase may match the present agroclimatic conditions of the ALS (Fig. 6b). Because of the high latitude of the ALN zone, the relative increase in the effective global radiation will be negligible. Overall, the agricultural potential of this zone is likely to improve; however, this is marginal in a European context due to the relatively small acreage of agricultural land in the zone (Table 1).

BOR

The growing conditions of the BOR region include special features that constrain yield formation (Peltonen-Sainio et al., 2009b). The number of effective growing days under the present climate conditions is strikingly low (Table 5, Fig. 6a); the short growing season is further hampered by a relatively high risk of early-summer night frosts and a high proportion of dry days. Therefore, yields are typically far lower in the BOR zone than in other European regions (Peltonen-Sainio et al., 2009a). Presently, only the late-spring sowing window is used, and most sowing occurs even beyond late spring. This is due to saturated soils that need to dry before sowing is possible with heavy machinery (Fig. 6j), low temperatures that slow germination, seedling establishment and early growth and a greater propensity for night frosts, which make early sowing economically risky (Peltonen-Sainio et al., 2011). The overall low numbers of suitable days during the autumn sowing windows in the ALN, BOR and NEM zones are caused by ample precipitation and/or the early start of the winter season. The BOR zone has the lowest number of such days in late autumn (Table 5) and thus the present sowing window ranges from mid-August to mid-September (Peltonen-Sainio et al., 2009b). Compared with the ALN zone, the increase in the number of effective growing days was projected to be much smaller as a consequence of the projected increase in the proportion of dry days in the BOR zone. Early-summer drought already severely limits yields in some years (Peltonen-Sainio et al., 2009b; Rötter et al., 2009). Of all the investigated zones, the MDS, PAN and BOR zones will have the fewest number of effective growing days by 2050 (Fig. 6a and b). It is likely that the agricultural potential of the BOR zone will remain comparatively low, even in the scenario of a 5 °C climate change.

NEM

Despite the fact that the NEM sites represent the upper limit of the NEM region, the accumulated sum of global radiation is quite similar across the entire NEM region (Skjelvåg, 1998). The low yields in this region are usually attributed to exceptional conditions that cause late maturity and/or pest infestations rather than low radiation input. The fraction of dry days varies across the region, which causes some variation in the suitability of both spring and autumn for sowing. The selected range of indices did not include winter temperature, which is known to be an important yield predictor for perennial and autumn-sown crops in the NEM, BOR and ALN areas (e.g. Samnordisk planteforedling, 1992; Blombäck et al., 2009). Climate changes in the NEM region are likely to increase the crop-yield potential through improvements in the effective global radiation, effective number of growing days, date of last frost and proportions of sowing days (Fig. 6). Only the projected increases in the number of dry days during summer and interseasonal variability could potentially counteract the increases in crop-yield potential. Previous climate-change assessments for grass leys in Sweden have projected a considerably increased production in spring due to increased temperature, which enables an increased use of the high-intensity solar radiation in the spring (Torssell et al., 2008). Using the present climate analogy, the NEM zone would achieve growing conditions that are close to those of present-day ALS, with a frequency of drought days and sowing conditions that are similar to those of the present-day ATN zone. These changes would probably support a shift from spring-sown to autumn-sown cereals (Eckersten et al., 2008) and would enable the expansion of the cultivation of forage maize and similar crops. Under the +5 °C scenario, the NEM area would achieve a Huglin index that is comparable to that observed in the present-day MDM area; the water deficit during dry years (based on a 20-year-return probability) would increase substantially.

ATN

The high yield potential of the north-western ATN zone, which is indicated by the relatively large effective global radiation in these areas, is confirmed by yield statistics for winter cereals (e.g. Schaller & Weigel, 2007). In terms of grassland productivity, Smit et al. (2008) claimed that the ATN zone has the highest production potential among all of the evaluated zones, followed by the ATC and LUS zones. This high productivity results from the relatively long summer days in combination with sufficient precipitation during the growing season, a long grain-filling phase due to moderate summer temperatures and recent increases in the thermal growing season (Chmielewski et al., 2008). The high productivity is particularly evident in fruit-growing regions, e.g., near the Elbe estuary (Henniges et al., 2007). Because of phenological shifts due to recent warming, resulting in earlier bud break and flowering, the risk for frost damage has remained unchanged for grapevines and fruits (Rochette et al., 2004; Stock et al., 2005; Chmielewski et al., 2008; Henniges et al., 2007; Eitzinger et al., 2009) and it is likely to remain unchanged under the projected climate change. Increasing winter and summer temperatures may cause yield reductions in winter cereals (Kristensen et al., 2011), but increasing summer drought may not necessarily reduce yields in this zone, where winter cereals develop deep roots and where current rainfall is generally not limiting. The increasing number of dry days in the June–August period (Fig. 6f) may reduce the yields of spring cereals (Wechsung et al., 2008); however, this phenomenon might be partly compensated for by the earlier sowing of spring cereals (Olesen, 2005). Climate-change studies have generally shown an expansion of warm-season crops (e.g. maize, sunflower, soybean and grapevine) in this zone under climate change (Fronzek & Carter, 2007; Olesen et al., 2007). This was confirmed by the projected changes in growing days, the Huglin index and date of last frost (Fig. 5).

ALS

Mountain chains act as climatic borders for the surrounding regions (e.g. delineating northern from southern EnZs in the Alpine mountain range) and contain a variety of climatic conditions due to strong topographical effects. This must be considered for the mountain regions in the ALS zone (e.g. the Alps and the Massif Central), resulting in a high spatial variability of climates. While there were only two stations selected in the ALS zone, they represent two of the six strata, wherein almost three-quarters of the agricultural area of the zone are located. It should be stressed that these stations represent low elevations that are relatively suitable for crop production. The potential productivities of both sites are at the higher end of all of the analyzed sites (Fig. 6) and the frequency of drought is very low, even during the summer months (Fig. 6e and f). The effect of climate change here was neutral to slightly positive, indicating slight increases in the variability and mean sum of effective radiation (Fig. 6a) and in the mean duration of effective growing days. The Huglin index of this region suggests that it might become suitable for grapevine cultivation; however, additional constraints in the ALS region, such as very low winter temperatures, poor soils and inaccessible terrain, will limit the cultivation of grapevines and other crops. There was a marked increase in projected days with water limitation (Fig. 5) during summer and in summer drought variability (Fig. 6f), threatening the productivity of permanent grasslands, which is one of the largest concerns in the eastern and southern parts of the Alps (Eitzinger et al., 2009). Specifically, a mean global temperature increase of 5 °C would lead to a partial deterioration of productivity (Fig. 5). In the more humid ALS regions (north), an increased grassland biomass production potential can be expected. Similar effects have been projected for arable crop production in recent studies (e.g. Eitzinger et al., 2009), with increasing crop-yield potential via the introduction of higher-yielding and later-ripening cultivars (e.g. maize) or new crops (e.g. soybeans and sunflower).

CON

The CON zone is the EnZ with the largest number of strata (12), the largest acreage of agricultural land (Table 1) and a high degree of variability between sites. The comparable potential productivity of the CON zone (expressed as effective global radiation and growing days) agrees well with the grassland productivity estimated by Smit et al. (2008). For the projected climate change, the overall mean for all CON sites (5, 6) suggests no change, or even a decrease, in the effective global radiation sum and number of effective growing days. Whereas sites north of the Alps mostly showed increases in both indicators (see also Trnka et al., 2010a), those in the southern parts of the CON zone demonstrated decreases of both indicators as a consequence of increased water stress. The projected values of the Huglin index suggest that viticulture will require changes in the cultivars grown (e.g. Stock et al., 2005; Eitzinger et al., 2009). The mean proportion of dry days from April to June did not change appreciably on average (Fig. 6e and f); however, there was a pronounced south-to-north gradient, with sharp increases in the proportion of dry days at southerly sites. The increase in the number of dry days from June to August represents a risk for rainfed agriculture across the present CON area, and this has already partly been reflected in the observed trends of drought since the 1940s–1950s (e.g. Dai et al., 2004; van der Schrier et al., 2006) as well as in national and regional studies (e.g. Wechsung et al., 2008; Dubrovskýet al., 2009). Recent studies (e.g. Jacobeit et al., 2009; Trnka et al., 2009) have also pointed to the fact that changing frequencies of temperature and precipitation extremes are associated with changes in the frequency of particular circulation types. The early-spring sowing window should become longer (on average) and more stable (5, 6). These changes agree well with the shorter duration of snow cover, increasing spring temperatures and earlier start of the spring season (e.g. Chmielewski et al., 2005; Brázdil et al., 2009). Harvesting conditions in June (when the harvest of some crops will take place in the future) are not favorable, making the planning of suitable harvest times more challenging.

PAN

The climate of the PAN zone can be viewed as a variation of the continental climate (CON). The PAN zone primarily consists of flat regions and has warmer and drier summers and higher mean wind speeds compared with the neighboring CON region (e.g. Auer, 2004; Auer & Korus, 2005). This leads to typical steppe-like conditions and high reference evapotranspiration rates during summer (Müller, 1993). Agricultural production in the PAN region under the present climate is primarily restricted by a lack of water, particularly during summer (Table 5). The PAN region was projected to have the sharpest declines in effective global radiation as a consequence of large decreases in water availability (5, 6). The projected trend toward a warmer and drier climate is more pronounced here than in other zones (Fig. 3), and the severe consequences of climatic variability in parts of the PAN zone have been highlighted elsewhere (e.g. Seneviratne et al., 2006). Crop production in the PAN is, to a large degree, dominated by arable production (especially that of maize, sunflower, winter wheat and spring durum wheat) and the results of crop-model-based studies in some countries have shown significant shortening of the growing season and a reduction in crop yields from increases in the frequency of summer drought and heat waves (Alexandrov & Hoogenboom, 2000). This shortening of the growing season could cause a significant loss in crop production and revenue in regions where no additional water sources are available (Eitzinger et al., 2003; Alexandrov & Eitzinger, 2005). The PAN zone is also renowned for viticulture and high-quality white wines; however, the Huglin index in this region in 2050 is projected to become comparable to that of the present MDN zone.

ATC

The present agroclimatic conditions in the ATC zone result from its proximity to the sea, which reduces interseasonal variation in comparison to other zones; however, variability among stations at different altitudes and among seasons is still considerable, particularly for those indices that are associated with soil-moisture content. This can be explained by spatiotemporal differences in rainfall, wherein the oscillatory component in the rainfall series plays a key role (e.g. De Jongh et al., 2006; Ntegeka & Willems, 2008). Frequent high-precipitation events during the late-spring sowing window are the primary cause of the lower number of suitable sowing days. The high number of effective growing days (Table 5 and Fig. 6b) and, to some extent, the effective global radiation levels (Fig. 6a) result in high yields of key field crops here compared with other European regions (e.g. Olesen et al., 2011). The Huglin index of this region suggests only a marginal suitability for wine growing (e.g. Robinson, 2006); however, at some ATC sites, the conditions have been improving over the past few decades, as documented by Schultz (2000) and Eitzinger et al. (2009). The effective global radiation is not expected to change significantly, while the number of dry days is likely to increase (Fig. 6). Whereas Stock et al. (2005) demonstrated the tendency of a northward viticultural shift and an ascent to higher elevations, Schultz et al. (2005) have calculated a similar rate of increase in the Huglin index for Geisenheim/Rheingau (ATC-ATN) as found in our study. For the SRES-A1B scenario, projections have shown average shifts of the latest frost to earlier dates by 28 days for the period of 2071–2100 in Germany (Chmielewski et al., 2008). The earlier start of the growing season results in a higher proportion of suitable sowing days in spring, as was also found by Rötter & van Diepen (1994). The tendency toward more drought stress (5, 6) was also reported by Gupta et al. (2006) for the Netherlands and by Holden et al. (2008) for Ireland.

LUS

Despite having the smallest total area of the zones considered in this study, the LUS zone has one of the largest proportions of agricultural land among all of the investigated zones. The large sum of effective global radiation and large number of effective growing days suggest a high potential productivity (Table 5), which is reflected in crop yields (Reidsma, 2007) and agrees with the findings of Smit et al. (2008) and Fisher et al. (2002) for grassland productivity. This region also contains well-known wine-producing regions, which have historically focused on the production of high-quality wines corresponding with favorable Huglin-index values (Fig. 6c). The risk of late frosts is low (Fig. 6d), as are the risks of drought occurrences during the early growing season and in the summer months, and there is a high proportion of days suitable for harvesting and sowing. The agroclimatic conditions of the LUS zone could potentially worsen through decreases in effective global radiation sums and effective growing days. Despite these changes, the levels of the last two indicators will remain comparatively high in the LUS region, accompanied by low drought risk in the early growing season. The change in the frequency of summer drought stress is quite important, as it will reach levels that are presently seen at PAN sites. The proportion of suitable harvest (Fig. 6g and h) and early sowing days (Fig. 6i) will improve, while the conditions during the late-spring and fall sowing windows will not change. Because the LUS zone hosts key wine-producing regions, an increase in the Huglin index (Fig. 6c) to levels near those presently observed in the MDN zone poses questions about the future of current terroirs (Seguin & Garcia de Cortazar, 2005).

MDM

Although only represented by one station in this study, the analysis of these results offers interesting information regarding potential impacts. Overall, the MDM zone is quite similar to the MDN zone (discussed below). Interestingly, the index that measures the change in last frost did not follow the pattern of MDN and MDS, as it retains the relatively large variability observed in the 1971–2000 period. The proportion of dry days for the period from April to June and June to August is expected to increase considerably for the MDM zone, as also predicted by Iglesias et al. (in press-a).

MDN

The results for the MDN zone reported herein are primarily based on sites in the Central Mediterranean and the Iberian Peninsula (Fig. 1). The current agroclimatic conditions at the analyzed sites suggest high potential productivity (Table 5), which is reflected in the very high grain maize and winter wheat yields in the MDN zone (Iglesias & Quiroga, 2007; Reidsma, 2007; Reidsma et al., 2009; Iglesias et al., in press-a) and in the high values of grassland productivity that have been estimated by various approaches (Fisher et al., 2002; Smit et al., 2008). However, grassland yields based on national statistics (Smit et al., 2008) show that the MDN zone has a significantly lower productivity than the PAN and MDM zones, which may reflect frequent summer droughts (Fig. 6f) in combination with a lack of grassland irrigation (in contrast to arable crops). Harvest (Fig. 6g and h) as well as sowing suitability during early spring and fall were projected to reach very high levels. The late-spring sowing window will become unreliable as a result of spring droughts, which will make sowing or any other tilling operations problematic. Climate change is projected to decrease the sum of effective global radiation and increase the proportion of dry days during the early growing season together with an increase in interseasonal variability. An analysis of the 1955–2007 rainfall series confirms the current trend of reduced rainfall during spring and winter (Bartolini et al., 2008). As a consequence, the proportion of drought days during summer (Fig. 6f) will vary less because almost all years will be affected by severe drought. Aside from drought, one of the perceived threats of climate change is the increasing probability of encountering lethal temperatures close to 40 °C. Crop-survival thresholds are still poorly understood and, thereby, there is a serious risk of future heat-wave-induced crop damage (e.g. Battisti & Naylor, 2009). Consequently, a significant increase in water demand for irrigation can be expected for this and the MDS region, not only for summer crops but also for winter crops, where in some regions the additional demand might not be met by the available water resources (Simota, 2009). The projected increase in temperature and decrease in precipitation in the MDN zone will also significantly decrease the soil-water content and water runoff to the Adriatic coast, resulting in negative consequences for the vegetation and agricultural production therein (Vučetić & Vučetić, 2000). The higher proportion of dry days during the period from April to June indicates a likely earlier onset of the wildfire season and an increased fire risk during summer (Vučetićet al., 2006) as a consequence of longer summer dry spells (Vučetić, 1998). The impact of climate change on wine quality will be very high, as shown by Huglin indices (Fig. 6c) of around 3000, which are indices that are typically associated with the production of dessert wines (Grifoni et al., 2006). An increasing temperature will reduce the occurrence of frost, but the real effect will have to be evaluated by considering the earlier onset of phenological phases and also the possible modification of air circulation (e.g. the possible intrusion of cold air from eastern Europe during March and April).

MDS

The potential rainfed productivity (Table 5) of this zone is limited by drought (Fig. 6e and f), not only during summer (Fig. 6f) but also in spring (Fig. 6e) and autumn, although this could be alleviated by irrigation (Reidsma et al., 2009). A low productivity here was also reported by Smit et al. (2008) for grasslands and for winter wheat and maize by Reidsma (2007) and Reidsma et al. (2009). In terms of harvest suitability (Fig. 6g and h), June and July exhibited the most favorable conditions of all of the investigated zones; however, the durations of the sowing windows (in particular those during early spring and autumn) were particularly low and variable (Fig. 6i and k), mostly as a consequence of increasingly dry soil conditions. Climate-change projections indicated decreases in potential productivity due to increases in the proportion of dry days and a decrease in the interannual variability of these parameters; however, this is hardly surprising given the character of the climate changes in these regions (Fig. 3). More specifically, sharp reductions in precipitation during summer and also in winter months (e.g. Zanis et al., 2009) will likely result in increases in the number of consecutive dry days and heat-wave frequency (Beniston et al., 2007) and the consequent decrease of soil-water content (Calanca et al., 2006). Similarly to the MDN zone, the variability in the proportion of drought days during both evaluated windows (Fig. 6e and f) will decrease during summer and spring, which will further increase the risk of forest fires in the MDS region (Lavalle et al., 2009). The likely impact of climate change on wine quality in the MDS zone is thought to be significant and negative (Fig. 6c). Finally, as in the MDN region, more effective irrigation methods, water management and policy in this region will be the main determinants of future crop distribution and productivity (Iglesias et al., 2007, in press-b; Iglesias 2009; Katerji et al., 2010).

Uncertainties in projected impacts

To date, the existing projections of European crop yields under climate change have been based mainly on the outputs of crop-growth models. While this strategy can be used to estimate the impact of climate change on crop yield, the simulation models usually do not capture crop management or environmental factors (e.g. extreme weather events) in their entirety. Moreover, crop-simulation studies are often limited with respect to the number of crops covered or the spatial coverage. The present study, which is based on selected indices, provides general, although limited, conceptions about fundamental agroclimatic conditions that govern crop-yield potentials and conditions for crop management across Europe. All assumptions and thresholds used in the study were based on published literature, and the sensitivity of our conclusions to the assumptions made was scrutinized by a sensitivity analysis (Table 2b), which showed that changing the thresholds used (e.g. ET_a/ET_r ratio) or modifying assumptions made (e.g. applying different reference surfaces for ET calculations) inevitably leads to variations in the absolute values of the indicators. However, the overall impact of the modified thresholds on the study conclusions was limited, i.e., the relative differences between the baseline conditions and those expected by 2030 and 2050 remained qualitatively the same.

Throughout most of the investigated zones, there were signs of deteriorating agroclimatic conditions and a need for adaptive measures to either increase soil-water availability (e.g. by irrigation or crop-management options) or crop drought resistance in the majority of the zones. While the impacts were demonstrated only for a selection of three GCMs, they represent a wide range of future projections quite well (Table 4).

Perspectives on European agriculture under climate change

Earlier European studies have emphasized that agriculture is expected to potentially benefit from climate change (e.g. Rötter & van Diepen, 1994; Olesen & Bindi, 2002); however, the responses of agricultural systems to changes in the frequency and severity of climatic extremes have rarely been considered in earlier assessments. Recent examples of damage in relation to floods, drought, hail and storms have revealed that the impacts of such extreme events are large (Kabat et al., 2005; Gupta et al., 2006). The present study confirms the substantial northward expansion of the thermal suitability of crop production in Europe under climate change found previously, e.g., by Fisher et al. (2002) and Olesen et al. (2007). The areas where conditions for rainfed crop production will be improved are restricted to the Northern regions (ALN, BOR and NEM), and partly in the ATN and the Alpine Mountains (ALS). This is the result of drier summers in much of central and southern Europe that will limit crop growth during summer unless irrigation is applied. This is not fully consistent with the results of Fischer et al. (2005), who predicted negative impacts on crop productivity only for Western Europe. The projected climate change does not seem to severely interfere with the possibilities for sowing and, to a lesser extent, harvesting, thus generally offering possibilities to adapt by changing sowing and harvesting dates in most European regions. The analysis shows that if the climate patterns evolve according to the assumptions and scenarios we used, some of the currently highly productive agricultural regions in Europe may be at risk of reductions in suitability for rainfed crop production. This is particularly the case for Western France and also parts of South-Eastern Europe (Hungary, Bulgaria, Romania, Serbia, etc.), where summers will become considerably hotter and drier, reducing crop yields and increasing yield variability. In these regions, winters will still be too cold to allow crop growth during winter. The Mediterranean zones will suffer from increases in dryness during spring and sharp declines in rainfed crop-production potential, posing the challenge of added irrigation capacity to irrigated Mediterranean areas, which must therefore become more efficient (Playan & Mateos, 2005). As shown by the Huglin-index values, the conditions for traditional crops such as grapevines will become more challenging, as also found by Jones et al. (2005) and Olesen et al. (2011).