*3.1. Statistical Analysis and Aspects of the EHF Index*

Primarily required for the definition of the EHIsig index, the 95th percentile of the Tmean was computed for a standard period of 30 years from 1971 to 2000. It was found that the T95 values varied from 24.8 ◦C to 28.7 ◦C. The lowest percentile values were observed over Kozani, a station in the west continental part of Greece, followed by Alexandroupoli (25.4 ◦C) in the north. Kozani is a typical continental station, of quite high altitude (400 m), and that is the main reason for the low Tmean 95th percentile values found. On the other hand, the second minimum of Alexandroupoli can be explained by the fact that this station is found in the northeastern part of Greece, and even though it is a coastal station, it presents an intense continental influence especially during winter, which explains the low 95th percentiles. The highest values were found in Elliniko in central Greece (28.5 ◦C) and the maximum was found in Samos, an island station over the southeastern Aegean Sea (28.7 ◦C) (Figure 2).

As mentioned in the previous section, the analysis focused on the hot period of the year (JJAS) for the years from 1958 to 2012. The average number of days in the hot period of the year in Greece when the EHF index presented positive values varied from 17.9 (Larissa) to 23.9 (Kozani). This indicates that for approximately 20 days per year the stations under study experienced heat wave conditions (Table 1). In addition, although Kozani was the station with the lowest T95 percentile value, generally characterized by a relatively colder climate in comparison to the rest of the stations, it showed the highest average positive EHF days.



Moreover, the results regarding the year with the maximum number of days with EHF > 0 indicated that, for all of the stations, they occurred at the end of the examined time period, most of them being in 2012. It seems that during that year, Greece was characterized, in summer, by very intense warm conditions that lasted up to 80 days (Kalamata). However, the absolute maximum was observed in Naxos (89 days) two years earlier (2010). These abnormally hot days tended to become more frequent throughout the examined period, since positive trends were found in all the stations ranging from +3.4 days/decade in Rodos to +11.4 days/decade in Naxos (Table 1). The smaller trend in Rodos could be attributed to the geographical position of the station, in the northwest part of the island, which is highly influenced by the Etesian winds during the summer. The maximum in Naxos could also be related to the location of the station, which is more "protected" from the Etesian winds. Regardless of the trend values, it is worth mentioning that after the application of Kendall's tau test at a significance level of 95%, all of them were found statistically significant. This comes as a robust indication that the days of abnormally increased temperatures do significantly increase during the examined period and it is not just a random rise (Table 2). This finding encouraged the application of another statistical analysis, based on the application of the Mann–Kendall t test method [39], in order to identify breakpoints on the EHF time series. The results from this test are presented in Figure 3. It can be clearly seen from the normal curve in all stations that the time series of the positive EHF days present a statistically significant positive trend that exceeds the statistical significance level (95%) during the last years of the examined period. In addition, according to the criteria of this test [39], in all stations under study, an abrupt change (breakpoint) of the specific parameter is observed (a clear "X" shape between the normal and the retrograde curve). The actual year of the breakpoint is not the same in every station but it can be placed from the mid-90s until the first years of the 21st century. More specifically, the earliest breakpoint is in Samos (1993) and the latest one is in the Thessaloniki station (2002) (Figure 3).

Apart from the examination of the number of days with positive EHF values, the study of the spells of positive EHF is also included (Table 2). These heat wave spells were classified in four classes (1st class: 3 to 10 days, 2nd class: 11 to 20 days, 3rd class: 21 to 30 days, and 4th class: >30 days). In addition, the maximum spell duration was computed in order to provide a magnitude of the most extreme heat waves in terms of duration.

As expected, the most frequent spells are the ones belonging to the first class, with an average number of 122.7 spells during the years of study. The most heat waves with 3 to 10 days duration were found for the Kalamata station (146 spells) followed by the Kozani station (140 spells). These "shorter" heat waves were less frequent for two stations in the north of Greece, namely, Alexandroupoli and Kerkira with 106 and 105 spells, respectively. As the duration of the heat waves becomes longer, their frequency decreases. For the second heat wave class, the highest number of spells was found in Alexandroupoli and Elliniko (26 spells) and the lowest one was found in Larissa and Souda (15 spells), whereas for the third class the frequency of the heat waves did not exceed 7 (heat waves with a duration from 21 to 30 days) which was recorded at the Mitilini station over the eastern Aegean Sea. Regarding heat waves with a duration longer than 30 days, none were observed in the Heraklio, Larissa, Skiros, and Souda stations during the 55-year time period. On the other hand, seven (7) such heat waves were found for the Naxos station in the central Aegean Sea and five (5) were found in Kerkira in the Ionian Sea. Finally, calculating the maximum heat wave in each station during the examined period, it should be noted that during the year 2012, Kerkira and Kozani experienced 58 and 57 consecutive days, respectively, of abnormal hot conditions with positive EHF values. The maximum for this parameter (the duration of the maximum spell) is found in general over continental stations, over the western parts of the country, where the Etesian winds lack influence, during the summer months. Conversely, the minimum is observed for island stations where the sea probably plays an important role in cooling (temperature drop), especially during the night. It should also be highlighted that for most of the stations, this extremely long heat wave was detected during the last three years of the study period (2010, 2011, and 2012), especially the last year. This is in agreement with the general finding that 2012 is considered, on a planetary scale, as one of the warmest years all over the world according to the World Meteorological Organization (WMO).


**Table 2.** Number of heat wave spells (consecutive days with EHF values > 0.0) during the study period 1958–2012 for the 14 stations under study.

#### *3.2. Examination of the EHF Index in "Capturing" Two Characteristic Heat Waves in Greece*
