Thermal bioclimates in Europe
Monthly-averaged time-specific UTCI maps derived from 1979 to 2016 ERA-Interim data show that the bioclimate in Europe varies in time and in space (Fig. 1). Heat stress follows a diurnal pattern with UTCI values at 06:00 or 18:00 generally lower than UTCI values at 12:00 or 15:00. Heat stress also shows a latitude gradient, with UTCI values generally increasing towards the south. As a result, two main thermal climates can be identified in Europe. One thermal climate is associated to heat stress conditions and it is predominant in the southern part of Europe, from the Iberian Peninsula to the Urals across the Mediterranean region, the Balkans and the Caucasus, where moderate and strong heat stress are achieved at central day-time hours. One thermal climate is associated to heat-neutral conditions and affects the more northern parts of Europe where on average stress due to heat load is not experienced and early/later day-time hours are characterized by cold stress instead. The definition of the two thermal climates reflects the general relationship between heat load and insolation. The UTCI has been shown to have some linear dependency on the 2 m air temperature and, through the mean radiant temperature, on the solar elevation angle and the surface solar/thermal radiation (Pappenberger et al. 2015). However, the UTCI also shows sensitivity to other atmospheric variables, namely, wind and humidity. Herein lies the added value of calculating UTCI as an indicator of heat stress.
The prevalence of heat stress conditions in Southern Europe is reflected by the frequent occurrence of elevated heat stress levels in that area over the summer season. The thermal bioclimatic plots of European capitals show that the frequency of occurrence of UTCI classes differs from city to city and depends on the climate (Fig. 2). Considering central daytimes as a reference, capital cities such as Athens, Lisbon, Madrid, Roma, Tirana, Belgrade, Bucharest and Podgorica, experience either moderate or strong heat stress every day. This is in agreement with the Köppen-Geiger climate classification (Kottek et al. 2006) of the areas where those cities are located, that is, temperate climates with hot summers (Cfa, Csa). Capitals with temperate or continental climates characterized by warm summer seasons (Cfb, Csb, Dfb) are instead dominated by the frequent occurrence of conditions of no thermal stress or moderate heat stress. Reykjavik, characterized by a cold-summer climate (Cfc), is exposed to conditions of moderate cold stress up to no thermal stress.
A further insight into the dependence of heat stress on daytimes is provided by a period-by-period analysis. For each of the three 10-year periods included in the study, namely 1980–1989, 1990–1999, 2000–2009 and the 7-year period 2010–2016, Europe-averaged UTCI values fall in the no thermal stress range and follow a diurnal trend which peaks at around 12:00–15:00. The intra-period variability is due to the spatial variations of UTCI across Europe (Fig. 3, left panel). As for the period-by-period variability recent summer seasons have been characterized by higher UTCI values than past ones. With respect to the UTCI reference value from the 1979–2016 climatology, the UTCI at 12UTC was about 0.5 °C colder in the 1980–1999 period, while it was 0.5 and 1 °C warmer in the 2000–2009 period and the 2010–2016 period, respectively. The differences between the periods are statistically significant. Similar deviations are observed at the other time points (data not shown). A year-by-year analysis confirms this result and attributes it to the occurrence of years, such as 2010, where UTCI median yearly values deviate positively from climatology (Fig. 3, right panel). These findings extend to the pan-European scale the results of previous studies conducted on the UTCI at the local scale, specifically the increase of UTCI values in the last three 30-year period and the occurrence of unfavourable thermal conditions in the summer months for Hungary and Central Europe (Nemeth 2011; Błażejczyk and Błażejczyk 2014).
In order to assess the impact of heat stress on human health across Europe, the distribution of all-cause death counts at different UTCI maximum values was investigated in 17 countries (Fig. 4).
The UTCI-mortality scatterplot shows that data generally group into clusters. A cluster falls into one or more heat stress categories according to the country. In Denmark, Finland, Ireland, Norway, Sweden and United Kingdom, for instance, deaths are generally associated with conditions of no thermal stress. In Greece, Italy, Portugal and Spain, deaths occur mostly in conditions of moderate and strong heat stress. In Austria, Belgium, France, Germany, Netherlands and Switzerland, deaths are related to conditions of no thermal stress and moderate heat stress. For each of these groups of countries, the heat stress categories spanned by a cluster reflect the thermal bioclimate of the single country, which is the heat-neutral bioclimate for the first group, the heat-stressful bioclimate for the second group and an “intermediate” thermal bioclimate for the third group.
The influence of one heat stress category on the number of deaths is revealed by the trend of the LOWESS function used to fit the clusters. The UTCI-mortality trend depends in general on the stress category a cluster falls into. For clusters in the no thermal stress range, the association between the UTCI and mortality is either flat, i.e. the number of deaths do not increase or decrease with UTCI values (Norway, Iceland), or slightly V-shaped, i.e. the number of deaths drops as the UTCI increases up to 15–20 °C after which the number of deaths starts to rise as the UTCI increases (Denmark, Finland, Ireland, Sweden, United Kingdom). As clusters move towards UTCI domains characterized by conditions of moderate and strong heat stress, the association between UTCI and mortality assumes a more pronounced U shape or a J shape, with slopes getting steeper at about 26 and 32 °C, respectively (Austria, Belgium, France, Germany, Greece, Italy, Portugal, Spain and Switzerland). The importance of this result is twofold. First, it highlights the relevance of UTCI stress categories as ranges defined according to physiologically significant values. Second, it confirms the progressive, larger effect of heat load in the increase of mortality as conditions become more thermally stressful. This effect might be particularly notable in countries dominated by heat-stressful thermal bioclimates and strong UTCI-mortality relationships, whereas short-term extreme heat events might have greater impacts in countries characterized by more heat-neutral thermal bioclimates and weaker UTCI-mortality relationships. Further studies are needed in this regard. An analysis based on monthly mortality data may fail to capture the effect of heat-related illnesses that arise from relatively brief (day scale) but intense heatwaves. Using daily mortality data, for instance, could provide additional insights into the UTCI-mortality relationship here illustrated.
The present analysis demonstrates that the UTCI-mortality relationship is strictly connected with the thermal bioclimate to which a population is exposed and adapted. It also represents the ability of a population to cope with extreme events, such as the 2003 European heatwave.
Case study: the heatwave of summer 2003
From June through July until mid-August, Europe experienced consecutive episodes of intense anticyclonic blockings. Areas of anomalous high pressure, firmly anchored over most of Western Europe, conveyed a very hot dry air mass up from south of the Mediterranean whilst preventing the progression of rain-bearing low pressure systems from the Atlantic Ocean. Air temperatures increased up to 12.5 °C more than seasonal average and extreme maximum temperatures were repeatedly recorded in most of the southern and central European countries (Trigo et al. 2005; García-Herrera et al. 2010).
The exceptionality of the 2003 heatwave is reflected in heat stress levels that affected the European bioclimatic conditions in that period. In August 2003, western and central areas experienced UTCI values up to 10 °C higher-than-seasonal average and heat stress up to 2 categories higher-than-seasonal average (Fig. 5). In France, for instance, areas usually characterized by no or moderate heat stress underwent moderate to strong heat stress for most of the daytime. In the city of Paris, the number of summer days with no thermal stress decreased by 62% as the number of days with moderate, strong and very strong stress increased by 36, 16 and 10%, respectively. The number of days with higher-than-average heat stress increased in all European capitals (Fig. 6).
The heat stress caused an unprecedented increase in mortality and morbidity, making the 2003 heatwave the deadliest natural disaster in Europe in the last 50 years (Robine et al. 2008). In August 2003 higher-than-average UTCI values were mostly associated with higher-than-average mortality, especially in France, Portugal and Spain (Fig. 7) where the heatwave exacerbated heat stress above climatological and adaptation levels. Ireland also experienced higher UTCI values, but differently from other European countries this was associated with lower death counts. This is in agreement with previous epidemiological results which state no excess deaths were reported in Ireland during the 2003 heatwave (Pascal et al. 2013).
The impact of the 2003 heatwave on human health was particularly evident in cities such as Paris (Vandentorren et al. 2004). In the first half of August 2003, Paris experienced exceptional maximum temperatures and high minimum temperatures with the latter providing low night-time relief (Fig. 8, left panel). UTCI maximum values were above 32 °C, indicating a continuous condition of strong and very strong heat stress levels. Humidex values, computed from dewpoint temperature and relative humidity (Masterton and Richardson 1979), fell in the great discomfort category, with some peaks reaching the dangerous level. With regards to mortality, the number of daily deaths started to increase on 6th–8th August, rose to a maximum between 9th and 13th August, and returned to pre-heatwave values around 15th–16th August. The mortality peak (+ 142%, Vandentorren et al. 2004) was reached once the heatwave had become established. The fairly immediate effect of elevated temperatures on mortality is demonstrated by the fact that high correlation values (r > 0.65) between 2 m air temperature and daily death counts are achieved for a time lag equal to 1 to 3 days (Fig. 8, right panel). Within the same lag period the UTCI also shows a strong correlation (r > 0.55) with mortality, especially at 06:00 and 12:00. This result confirms that the heat stress experienced very early in the day and at central daytimes played a role in the excess number of deaths recorded in Paris during the 2003 heatwave event. It also provides insights in the correlation between UTCI and mortality with respect to 2 m temperature and humidex. At 06:00, 12:00 and 15:00 the UTCI is generally more closely associated with daily death counts than humidex. However, correlation values between UTCI and mortality are generally lower than the correlation values between 2 m air temperature and mortality. This might be due to the approximated nature of the operational procedure used to calculate UTCI or the reliability of ERA-Interim input data such as wind at finer resolution scales, i.e. at urban level. As the UTCI is by definition valid in all climates, seasons and spatiotemporal scales, and has thermo-physiologically significance in the whole range of heat exchange conditions (Błażejczyk et al. 2013), future improvements in the quality of the operational procedure as well as of UTCI’s meteorological inputs will help to shed new light in the correlation between the UTCI and mortality.