Mean changes of temperature and precipitation
To date, the regional climate change projections provided by the EU-FP6 ENSEMBLES multi-model ensemble for the SRES A1B scenario are considered as state-of-the-art for European climate impact research. Therefore, we compared this data set to the new regional EURO-CORDEX data set for RCP8.5 and RCP4.5. Figure 1 shows the ensemble mean of the mean annual temperature and total annual precipitation change until the end of the century. Figure 1b, d shows a robust and statistically significant warming, with regional differences, in the range of 1–4.5 °C for RCP4.5 and of 2.5–5.5 °C for RCP8.5. These ranges encompass the warming range projected for the A1B scenario, where temperature increases between 3 and 4.5 °C (Fig. 1f). The projected spatial patterns are very similar in all scenarios with greater annual mean warming in Southern Europe and towards the northeast. Under RCP8.5, large parts of Northern Scandinavia, Eastern Europe and the Alpine ridge might be exposed to a warming of more than 4.5 °C compared to 1971–2000, which could be avoided by RCP4.5.
Associated with the large increase in temperature in RCP8.5 are robust changes in annual precipitation. The ensemble mean projects a statistically significant increase in large parts of Central Europe and Northern Europe of up to about 25 % and a decrease in Southern Europe. A zone with small changes, which are not significant (however, partially robust in the sign of change), indicates where the climate change signals change the sign (Fig. 1c, white areas). The pattern of the changes is very similar for RCP4.5, but less pronounced (Fig. 1a). The spatial pattern for A1B precipitation changes qualitatively agrees with the described changes for RCP4.5 and RCP8.5, and the magnitude of the changes mostly lies in-between the two RCPs. However, differences in the spatial patterns are seen over the British Isles, Benelux and Germany (Fig. 1e).
For mean temperature and precipitation change, a spatial correlation has been done between RCP8.5 and A1B results (Table 1). For all sub-regions, the spatial correlation between SRES A1B and RCP8.5 is very high, with 0.82–0.97 for temperature changes and 0.59–0.92 for precipitation changes depending on the region for mid-century. Towards the end of the century, the correlation is even stronger for both parameters.
Seasonal changes of mean temperature and precipitation are shown in the supplementary material (Fig. s2-s5). The seasonal temperature change signals show more regional heterogeneity than the annual mean. The zone between regions in which precipitation increases in the north and decreases in the south shifts southwards in summer and northwards in winter.
Seasonal mean changes for heavy precipitation
The projected seasonal mean changes in heavy precipitation for the three emission scenarios are relatively similar, but some regional differences are visible (Figs. 2 (RCP8.5), 3 (A1B), 4 (RCP4.5)). Most obvious differences are the increased regional detail in the RCP8.5 and RCP4.5, which is related to the higher horizontal resolution of about 12.5 km for the RCPs compared to 25 km for A1B, for which more homogeneous changes are calculated. The annual cycle of changes in heavy precipitation is similar in all three scenarios, but the amplitude of the change is stronger in RCP8.5 than in A1B in several regions. The results for RCP8.5 include a possible decrease in heavy summer precipitation by about 25 % over some parts of the Iberian Peninsula and Southern France, accompanied by regional increases in parts of Spain and Portugal. For winter, RCP8.5 projects strongest increases in heavy precipitation (up to 35 %) in Central and Eastern Europe, whereas A1B projects changes up to 25 % only in this region. Only for some parts in Scandinavia, A1B shows similar values as RCP8.5.
Important regional differences in heavy precipitation are projected for the RCP4.5 scenario. Compared to RCP8.5, the seasonal patterns of change are similar, but the amount of change is much smaller (up to 15 % in large areas with isolated spots up to 25 %) and—besides isolated regions in Southern Europe (mostly along coastlines)—no decrease in heavy precipitation is indicated.
GCM simulations tend to underestimate the high precipitation intensities (Sun et al. 2006). An improved distribution of high precipitation intensities is an important advantage of regional climate simulations. Figure 5 shows the relative frequencies of daily precipitation intensities of an ensemble of five GCM simulations and the corresponding regional downscaling experiments for the reference period 1971–2000 analysed over a central land region of the EURO-CORDEX domain (45°N–50°N and 2°E–17°E). For this analysis, data from all grid cells were taken into account. The distribution illustrates that the GCMs generally produce more precipitation intensities with up to 12 mm/day. The RCMs, in contrast, show higher intensities. Strong intensities above 30 mm/day do hardly occur in the GCM simulations. Figure 6 shows the temporal changes of the precipitation frequencies between the near future period and the reference period for both scenarios and the GCM and the RCM ensembles. Striking is that both model types reduce the number of weak precipitation intensities below 9 mm/day in both scenarios and increase the relative frequencies in all higher intensity classes. This shift in daily precipitation intensities, however, turns out much more moderate in the RCM than in the GCM simulations of both scenarios, the RCP4.5 and the RCP8.5. The frequency changes of the RCM simulations are especially in the range between 10 and 20 mm/day less than half of the GCM changes. Above 30 mm/day, however, the increase in the RCM ensembles exceeds the climate change signal of the GCMs. Figure 6 also demonstrates that both effects—the reduction in weak intensities and the increase in strong intensities—are more pronounced in the RCP8.5 scenario. The analyses of the processes leading to the different behaviour are beyond the scope of this paper and will be studied in a separate paper.
This analysis proves two effects of an increased resolution, which can be regarded as an added value of regional climate simulations. On the one hand, the RCMs provide higher daily precipitation intensities, which are completely missing in the GCM simulations, and on the other hand, they provide a significantly different climate change of daily precipitation intensities resulting in a smoother shift from weak to moderate and high intensities.
Mean length of dry spells
Projected changes in the 95th percentile of the mean length of dry spells are shown in Fig. 7 for A1B (top), RCP8.5 (middle) and RCP4.5 (bottom) for 2021–2050 and 2071–2100 with respect to 1971–2000. For the early period, the change patterns are very similar in all scenarios (left row), although the number of simulations taken into account for each scenario ensemble is different. Beside some common features in South-West Europe, substantial differences in the projected changes for dry spells lengths are visible until 2100. For RCP4.5 and A1B, a small increase in the length of extended dry spells is projected for Central Europe, which is more pronounced in A1B. A decrease in the length of extended dry spells is calculated in A1B for parts of Scandinavia. This feature is extended towards the Alps in the RCP8.5, in which the number of dry spells increases (not shown). This means that under RCP8.5 more but shorter dry spells are projected in the alpine region. For regions with a large increase in the length of extended dry spells, the number of dry spells is decreasing (not shown).
Mean number of heat waves
Projected changes in the mean number of heat waves during May–September are presented in Fig. 8, for RCP4.5 and RCP8.5, for the two future time periods and for two different definitions of heat waves. From the upper four panels, displaying the p99-heat wave definition, it is obvious that with less warming (see “Mean changes of temperature and precipitation” section) in RCP4.5, the increase in number of heat waves is smaller than in RCP8.5. This is more pronounced towards the end of the century (Fig. 8c, d) than for the earlier time period (Fig. 8a, b). For both scenarios, the increase is strongest in Southern Europe, but towards the end of the century the number of heat waves increases all over Europe. The number of heat waves for Southern Europe is projected to increase by more than 45. The increase is mostly robust and significant. The change in the number of heat waves considerably depends on the definition (thresholds and duration), which is used. Therefore, a second definition was used based on that developed by the World Meteorological Organization (see “Definitions of impact-relevant indices” section). Under this definition, the increase in the mean number of heat waves is much less (Fig. 8e, f). For the WMO-heat wave definition not a single heat wave is detected in the ensemble mean for the reference period as well as for mid of the century, because the criteria are much stricter. Also the duration of the heat wave is two days longer than in the p99-heat wave definition. For RCP8.5, meaning under the strongest projected warming, towards the end of the century, an increase is only projected for some parts of Southern Europe with additional 5 to more than 9 heat waves. The increase is significant and robust south of 55° latitude.
Indices by sub-region
Projected changes of several impact indices, which could be of interest for impact studies in different sectors, are listed in Tables 2 (A1B) and 3 (RCP4.5 and RCP8.5) for the 5 sub-regions. For all sub-regions and indices, the median shifts into the same direction, independent of the scenario. For almost all indices, a substantially larger change in the median is projected in RCP8.5, compared to RCP4.5, however, the likely ranges frequently overlap. Exceptions to this are annual total precipitation in the Atlantic and Continental sub-regions, tropical nights in the Northern sub-region and the cold spell duration everywhere. Here the projected changes in RCP4.5 and RCP8.5 are rather similar. Differences between the RCP scenarios are most pronounced for growing season length and warm spell duration index, with no overlap between the likely ranges over all sub-regions (even the full range seldom overlaps).
The median change in A1B is generally centred within RCP4.5 and RCP8.5. For some cases like annual total precipitation in the Southern sub-region or tropical nights in the Atlantic, Northern and Southern sub-regions, however, the median change in A1B is even stronger than the median change in the RCP8.5 and only for change in annual total precipitation in the Continental sub-region, the median change in A1B is lower than the median change in RCP4.5. The spreads of the projected changes defined as the likely ranges are generally the same between RCP4.5 and RCP8.5 or slightly larger in RCP8.5. Exceptions to this are frost days in the Continental sub-region, tropical nights in the Northern sub-region and total rainfall amount above the 99th percentile of daily rain (wet days only) in the Southern sub-region. Here the likely ranges of projected changes are larger in the RCP4.5 scenario.