Climate Change and Thermal Comfort in Greece

: Global warming is an environmental issue keeping all nations alert. Under this con-sideration, the present work investigates the future thermal sensation of the Greek population. Three periods are selected (2021–2050, 2046–2075, 2071–2100) and two Intergovernmental Panel for Climate Change (IPCC) representative concentration pathway (RCP) 4.5 and 8.5 scenarios. Use of Thom’s discomfort index (TDI) is made, which is calculated from air temperature and relative humidity included in typical meteorological years (TMYs) derived for 1985–2014 and future periods (both IPCC scenarios) for 33 locations in Greece. TDI is discriminated into 6 classes. The analysis shows that there is no signiﬁcant shift from past to future annual mean TDIs in terms of its classiﬁcation. The same is found for the summer TDI values. Nevertheless, a distribution of the various TDI classes is provided within the TMYs. Maps of annual TDI values are prepared for Greece by using the kriging method; higher values are found in the southern part of Greece and lower values in the northern. Best-ﬁt regression equations derived show the intra-annual TDI variation in all periods. Also, scatter plots of annual TDIs in the future epochs in comparison with the historical period show a linear relationship.


Introduction
The global-warming phenomenon is impacting the surface temperature of the Earth. Indeed, the Intergovernmental Panel for Climate Change (IPCC) [1] highlights that each of the recent 3 decades has become warmer than any other period prior to 1980. Also, the National Oceanic and Atmospheric Administration (NOAA) [2] reports that the 10 warmest years on record at global scale have appeared after 1998. Therefore, it is common sense that the increase in temperature may result in more frequent heat stresses (thermal discomfort) upon people [3][4][5][6], thus affecting the quality of life. This issue has, therefore, been the subject of several studies; these studies are focused on either the thermal comfort of people indoors (within a building) or outdoors (ambient conditions). For expressing the thermal comfort (or alternatively the thermal discomfort), several indices have been developed.
Yang et al. [7] have studied the thermal comfort during winter at the city of Tianjin, N. China. Adegoke and Dombo [8] examined the human thermal comfort in Akure, Nigeria. Cohen et al. [9] investigated the thermal perception in the urban web of Tel Aviv, Israel. Epstein and Moran [10] compared indices for thermal comfort and heat stress (extreme thermal discomfort). Golasi et al. [11] compared the Mediterranean outdoor comfort index (MOCI), as well as the predicted mean vote (PMV), and the popular events tracking (PET) models in Rome and found that the most appropriate is MOCI. A similar study was performed at Chiba, Japan, by comparing the models of PET, PMV, and standard effective temperature (SET) [12]. The outdoor thermal comfort was also studied in a subarctic climate of north Sweden [13]. Yousif and Tahir [14] investigated the thermal effect on people in Khartoum, Sudan. An informative review about the thermal effect on people has been given by Djongyang et al. [15].
Specifically, for Greece there have been several studies about the effect of heat on the population. Angouridakis and Makrogiannis [16] found the environmental thermal sensation for summer in Thessaloniki by analysing air-temperature and relative-humidity data in the period 1950-1957. Giles et al. [17] studied the heat stress in Athens and Thessaloniki during the heat waves of 1987 and 1988. Paliatsos and Nastos [18] compared the thermal effect during air-pollution episodes in Athens for the summers of 1993-1995 and found that it was related to the ozone concentration. Tselepidaki et al. [19] examined the summer thermal perception in Athens for cooling purposes. Tsitoura et al. [20] performed outdoors questionnaire about the thermal-comfort effect on people in 4 cities of Crete. Stathopoulou et al. [21] calculated the thermal effect from satellite-derived meteorological data and compared it with that from in situ measurements. Pantavou et al. [22] used various thermal-comfort indices in Athens (see their Table 2). Also, Pantavou [23] estimated the thermal sensation of pedestrians in 4 locations of Athens. Katavoutas and Founda [24] examined the period of 1960-2017 and found an increasing risk of heat stress in Athens.
It is obvious from the above works that no study about the thermal comfort has been conducted worldwide for a whole region or country. Therefore, the present work aims at filling this gap in the case of Greece. A second innovation of the study is the use of typical meteorological years (TMYs); such TMYs have been developed for 33 sites in Greece [25].
The work is divided into Sections. Section 2 deploys the data collection and analysis. Section 3 gives the results of the study. Section 4 provides a discussion and Section 5 deploys the main conclusions of the study.

Materials and Methods
According to [25] "A TMY is a set of meteorological and irradiance parameters usually consisting of hourly values in a year for a given geographical location (https://e3p.jrc.ec. europa.eu/articles/typical-meteorological-year-tmy). Furthermore, a TMY consists of a set of typical meteorological months (TMMs) selected from individual years integrated into a complete year. A TMY reflects all the climatic information of the location for a period as long as the mean life of the system." The aim of that study [25] was the development of TMYs for 5 different applications, i.e., TMY-meteorology-climatology (TMY-MC), TMY-bio-meteorology (TMY-BM), TMY-agro-meteorology-hydrology (TMY-AMH), TMY-PV applications (TMY-PV), and TMY-energy design of buildings (TMY-EDB); all these TMYs were generated for 33 sites in Greece in the frame of the "Development of synergistic and integrated methods and tools for monitoring, management and forecasting of environmental parameters and pressures" (KRIPIS-THESPIA-II) funded project, by using historical data of various meteorological parameters in the period 1985-2014 (30 years). During this project, TMYs-MC were also derived for future periods (2021-2050, 2046-2075, 2071-2100) for all 33 locations, not yet published. The future TMYs-MC considered two opposing IPCC scenarios, i.e., the representative concentration pathway (RCP) 4.5 (intermediate scenario with an "anthropogenic" radiative forcing of 4.5 Wm −2 in the year 2100) and the RCP 8.5 (extreme scenario with an "anthropogenic" radiative forcing of 8.5 Wm −2 in the year 2100). For the TMY-generation statistical procedure one may refer to [25]. Since TMYs-MC have been derived for both the historical and the future time periods, they are used in the present work.
The historical TMYs-MC consist of hourly mean values of dry-bulb temperature (in • C), relative humidity (in %), atmospheric pressure (in hPa), and global horizontal irradiance (in Wm −2 ) at the sites' altitude, while the future TMYs contain daily mean values of these parameters. To display the thermal comforts (or discomforts) across Greece for the historical and future periods and compare them, use of Thom's discomfort index (TDI in • C) has been made [26]. Its definition is as follows.
where T d is the dry-bulb temperature (in • C) and RH the relative humidity (in %). Thom defined a series of limits for TDI; these are given in Table 1. Therefore, hourly values of TDI were calculated from the corresponding values of T d and RH in the historical TMYs-MC and daily values in the future TMYs-MC for the 33 sites in Greece. These locations cover all of Greece and are given in Table 2; a map with the location of the 33 sites is also shown in Figure 1. The meteorological data for the 33 sites in the historical period 1985-2014 have been obtained from the Hellenic National Meteorological Service (HNMS) network, while the corresponding global horizontal irradiance values have been derived through the meteorological radiation model (MRM) code [27,28]. From these 33 data bases the historical TMYs-MC have been generated. As far as the future TMYs-MC are concerned, data for the closest grid point to the HNMS station's location were employed from a transient realisation of the regional climate model RCA4 of the Swedish Meteorological and Hydrological Institute (SMHI) ( [29], and references therein) driven by the Earth system model of the Max Planck Institute for Meteorology (MPI-M) [30]; simulations were carried out in the framework of the European Coordinated Regional Climate Downscaling Experiment (EURO-CORDEX) modelling experiment (http://www.euro-cordex.net). The horizontal resolution of the RCA4 model was 0.11 • (~12 × 12 km). The meteorological and global horizontal radiation data for the future periods 2021-2050, 2046-2075, and 2071-2100 were provided from ECO-CORDEX to the KRIPIS-THESPIA-II project; therefore, corresponding TMYs-MC were derived following the statistical procedure described in [25]. The presentation of these future TMYs-MC will be the subject of a separate publication, which is in preparation.  [31], which is the adoption of the ISO 843 one [32].  Table 2. The regions of south Ionian Sea, Peloponnese, the Cyclades (islands), Crete, and Megisti complex (of islands) are also shown; the Megisti complex is also known as the Kastellorizo complex, a complex of 14 small islands/rocky islets.

Results
This section is divided in sub-Sections according to the time scale encountered (annual, seasonal, monthly). As far as the seasonal analysis is concerned, this is performed for the summer time (June, July, August-JJA), as it is the one which may affect people negatively because of high ambient temperatures [19] or even heatwaves [17,33]. As mentioned in Section 2, two IPCC scenarios have been adopted in this work. The first is the IPCC RCP 4.5 (stabilisation) scenario; according to it a global temperature increase by 2100 is in the order of 1.7-3.2 • C. The second is the IPCC RCP 8.5 (high-end) scenario, which foresees a global temperature increase in the range of 3.2-5.4 • C.
As mentioned in Sections 1 and 2, use of the derived TMYs for the 33 sites in Greece (both historical and future) has been made in this study. Under this situation, analysis of the TDI tendency during the historical and future periods cannot be done because these long periods have been substituted by a single year (a TMY). As a TMY is a representative year climate-wise for the period from which it has come, so are the TDIs derived this way. Another reason for this controversy is the fact that the TMMs that comprise a single TMY come from different years of the period considered (see Table 3 as example where 4 sites have arbitrarily been chosen). On the other hand, comparison between the historical and future TDIs cannot directly be made because the historical TDIs are hourly values, while the future ones are daily values.  2004200019942001199720001999200419881994Irakleio1995 20012004199419971987200319961996199720042003Kastoria1987 19882005198519951996199619981991199419892008Tripoli 199919851993199319951987198719861986199120041994 3.1. Annual Thom's Discomfort Index (TDI) Values Figure 2 shows the annual mean TDI values for all 33 sites in Greece. It is seen, as anticipated, that the RCP 8.5 scenario produces greater spread of the annual values among the various epochs in comparison with the RCP 4.5 scenario. To show this effect in a better way, differences between the future annual mean TDI values and those in the historical period have been calculated. Figure 3 provides the annual mean differences ∆TDIs = TDI future epoch − TDI historical period . Under the RCP 4.5 scenario the increase in the absolute TDI values is of the order of 0.1-1.7 • C approximately, considering all 3 future epochs. It is interesting to see that the sites of Elliniko and Ioannina present negative differences in all 3 periods, while Kasteli and Lamia in the first future period only, and Naxos and Thira in the first 2 future epochs; this means that their population will experience a "milder" thermal sensation in the future. By contrast, the RCP 8.5 scenario demonstrates an apparent increase in the thermal sensation across Greece in the range of almost 0.4-3.5 • C. Some stations (Elliniko, Ioannina, Kasteli, Kerkyra, Lamia, Naxos, Thira), however, present a dissonance as far as the first (2021-2050) future epoch is concerned, as done in the RCP 4.5 scenario also. Anyhow, in both scenarios, the average annual increase in TDI implies even higher TDI values on the daily or even hourly scale, which can be related to persistent and more intense heat waves that may occur in the future. Table 4 shows the difference between the annual mean TDI values for all sites and periods considered.    If one considers the centres of each 30-year period in Table 4, being the years 2000, 2035, 2060, and 2085, respectively, then the corresponding increase in TDI is found to be ≈0.12 • C/decade, ≈0.11 • C/decade, ≈0.10 • C/decade for the 3 periods under the IPCC RCP 4.5 scenario, and ≈0.09 • C/decade, ≈0.24 • C/decade, ≈0.30 • C/decade for the respective periods under the IPCC RCP 8.5 scenario. Such increases are considered insignificant in terms of the thermal sensation of the Greek population because they do not shift the thermal (dis)comfort from class A to a higher one, since all average annual TDIs lie in the TDI-class A (see Table 1).

Summer TDI Variation
In a similar way to what was presented in Figure 3, Figure 4 shows the summer differences between the future TDI values and the historical ones (∆TDI = TDI future epoch − TDI historical period ). From these figures the following observations can be made: (i) many sites seem to have a better summer thermal sensation (negative ∆TDIs) mostly in the RCP 4.5 2046-2075 period and less in the RCP 4.5 2021-2050 one ( Figure 4a); (ii) higher annual absolute ∆TDIs values of almost 0.5 • C are found under both IPCC scenarios than their respective annual TDI ones (cf. Figure 3); this implies an abrupt increase in the future thermal (dis)comfort at these sites.  Table 4 for the summer case, Table 5 is derived. It is understandable that the summer thermal feeling throughout Greece both in the present and the future is in class B as all mean TDI values vary in the range 21-23.99; the exception is the RCP 8.5 2071-2011 period, which shifts to TDI-class C (24-26.99 • C). Calculations of the trends of TDI during summertime as done for its annual values result in the following conclusions: an increase in TDI of ≈0.03 • C/decade, ≈0.06 • C/decade, ≈0.06 • C/decade for the 3 periods under the IPCC RCP 4.5 scenario, and ≈0.10 • C/decade, ≈0.18 • C/decade, ≈0.25 • C/decade for the respective periods under the IPCC RCP 8.5 scenario. Such increases are considered insignificant in terms of the thermal sensation of the Greek population during summertime because they do not shift the thermal (dis)comfort from class B to a higher one, except for the RCP 8.5 2071-2100 period, which is classified in TDI-class C. Nevertheless, many sites seem to experience milder thermal feeling in the future than now as their ∆TDIs are negative in both RCP scenarios as shown in Figure 4a,b. Another interesting observation from Table 5

Intra-Annual TDI Variation
There is always a question about the in-year variability of a meteorological quantity. The quantity in question in this study is TDI, although it is not an immediate meteorological parameter but a derivative from ambient temperature and relative humidity. Figure 5 shows this variation for all periods of the study. A best-fit curve has also been derived for each case and can be used to calculate the average monthly TDI in Greece. The monthly TDI values in Figure 5 are averages over the 33 sites.   Table 6 gives the equations of the best-fit polynomial lines; their extremely high R 2 (approximately equal to 1) imply an excellent fit to the calculated intra-annual TDI values. Table 6. Best-fit 6th-order polynomial regression equations for the monthly mean TDIs (in • C); t is the month of the year in the range 1-12; R 2 is the coefficient of determination.  Figure 6 repeats the information posed in Figure 3, but in this case the bars refer to the monthly mean TDI differences. In both RCP scenarios the differences are generally reduced during the summer season (JJA) in comparison to the rest of the year. This may be attributed to the fact that the future climate (i.e., the climate change) seems to have a lesser effect on the summer months in comparison to the other seasons. Nevertheless, the increase in TDI under RCP 4.5 is between 0.1 and 1.4 • C and under RCP 8.5 between 0.2 and 2.9 • C. This conclusion excludes the small negative values in January and December (for RCP 8.5 2021-2050) and August (for RCP 4.5 2021-2050); the explanation is not obvious, but it can be taken as a no-real change in the future thermal feeling of the Greek population.  Some examples of the intra-annual TDI variation for selected sites are given in Figure 7. The selection has been based on the lowest and highest annual TDI values of the site during the various periods encountered in this study. It is seen that Kastoria has always the lowest thermal feeling and its population is not, therefore, in danger in the future in this respect. By contrast, Kythira in most periods and to a lesser extent Rodos and Thira shift their thermal sensation from TDI-class A to TDI-class B in the warm season of the year (June-October).

Frequency of TDI Occurrence
Up until now, the analysis has not taken into account the various TDI classes; to the contrary all TDI values have been considered from all 33 HNMS sites. This sub-section, however, gives information about the percentage of each TDI-class occurrence in all periods encountered in the study. This is shown in Figure 8. It is remarkable to observe that the TDI-class E (everyone feels thermal discomfort) appears in the historical period and not in any of the future ones, although at a very low percentage. This observation is true as this refers to almost half of the Greek territory (15 sites out of 33, i.e., Agrinio, Alexandroupoli, Anchialos, Kalamata, Kerkyra, Kozani, Lamia, Lesvos, Limnos, Mikra, Serres, Skyros, Souda, Thira, Zakynthos). The interpretation of this finding may be related to higher ambient temperatures in the future, but less humidity (drier weather) over Greece than now. Indeed, Kambezidis et al. [34] have examined the rain data of 31 HNMS stations in the period 1962-2002 and have found that the rain intensity (in mmh −1 ) shows (i) higher values in western Greece and lower ones in the Aegean Sea, and (ii) positive trends over Attica, Crete, southern Aegean and negative ones in the rest of the country and especially over the western mainland and Ionian islands. This is also confirmed by [35,36]. Furthermore, a technical report from the Bank of Greece [37] about the future impacts of climate change on various aspects of the Greek economy reports an increase in T d of 3.4-5.2 • C and a decrease in RH of 1.6-6.8% by 2100.   Table 1

Discussion
One might wonder whether the relationship between a historical meteorological variable and its evolution in the future is linear or not; in other words, whether there exists such a mathematical relationship for a climatic parameter between now and the future or, to the contrary, whether the climate-change system is chaotic. As a paradigm, an attempt has been made to answer this question in this study by considering the derivative climatic parameter of TDI. A first look at Equation (1) shows that TDI has a non-linear relationship with the independent variables T d and RH; so have T d and RH between each other. Furthermore, if one solves Equation (1) for either T d or RH, he/she gets a 2nd-order polynomial of T d or RH as function of the other parameter in the form y = (ax + b)/(cx + d), considering TDI is constant.
T d = (0.7975 RH + TDI)/(0.0055 RH + 0.4500), From the above expressions, one would, therefore, expect the historical and future TDIs to have a non-linear connection. Nevertheless, Figure 10 shows quite the opposite; there is a fine linear relationship between the historical monthly TDI values and those in any of the other 6 future epochs. No smoothing or deseasonalisation has taken place in the TDI time series. It is worth noticing that the ±95% confidence band around the best-fit line is so narrow that it contains the fewest data points. By contrast, the ±95% prediction zone around the best-fit line includes almost all data points. The difference in interpreting this result is the following. The narrow confidence interval contains some TDI pairs in any of the 6 cases (6 future periods) with a probability of 95%. The existence, however, of the majority of the data pairs outside the confidence interval dictates the fact that the linear best-fit line does not truly represent the relationship of future TDI as function of the historical TDI in the best way. Therefore, one could come to the conclusion that the relationship is not so linear as it appears, but is pseudo-linear. On the other hand, all future TDI pairs will be contained in the prediction interval with a probability of 95%. Then, one could be convinced that the pseudo-linear relation of today will evolve into a linear one in the future. This is confirmed by the high R 2 values in all 6 scatter plots, thus denoting an excellent agreement (correlation) between the historical and future TDI values.

Conclusions
This study investigated the future thermal sensation of the Greek population because of the global-warming effect. Three future periods (2021-2050, 2046-2075, 2071-2100) and two IPCC scenarios (RCP 4.5, RCP 8.5) were chosen for analysis. The thermal (dis)comfort was expressed by Thom's discomfort index (TDI) for 33 sites in Greece. Historical TDI values were also estimated in the period 1985-2014.
The present work presented two innovations. It is the first worldwide (in Greece as well) to (i) compute the future thermal feeling over a country, and (ii) compare it with the past. The second innovation concerns the use of historical as well as future TMYs-MC for all 33 sites, instead of using long-time series of ambient temperature and relative humidity, two parameters that define TDI. The main conclusions of the work can be summarised as follows.
No significant shift from the lowest TDI-class A to the TDI-class B has been shown to occur in the future all over Greece as far as the annual thermal sensation is concerned. TDI was, however, found to increase by ≈0.12 • C/decade, ≈0.11 • C/decade, ≈0.10 • C/decade for the 3 periods under the IPCC RCP 4.5 scenario, and by ≈0.09 • C/decade, ≈0.24 • C/decade, ≈0.30 • C/decade under the IPCC RCP 8.5 one, in comparison with the TDI values in the historical period.
Calculations of the TDI trends during summertime showed an increase of ≈0.03 • C/decade, ≈0.06 • C/decade, ≈0.06 • C/decade for the 3 periods under the IPCC RCP 4.5 scenario, and ≈0.10 • C/decade, ≈0.18 • C/decade, ≈0.25 • C/decade for the respective periods under the IPCC RCP 8.5 scenario, in comparison with the TDI values in the historical period. Such increases were considered insignificant because they did not shift the thermal (dis)comfort from TDI-class B to a higher one, except for the RCP 8.5 2071-2100 period, which was classified in TDI-class C.
For both historical and future epochs best-fit 6th-order regression equations were derived for estimating the monthly mean values of TDI. In all cases R 2 was greater than 0.99.
The decomposition of TDI into its classes and their distribution within the TMYs was performed. It was observed that the TDI-class E (everyone feels thermal discomfort) appeared in the historical period only and not in any future epoch.
Maps of annual mean TDI values were derived; they showed high values over the southern part of Greece (Crete, the Cyclades, south Ionian Sea, Megisti complex), and lower ones over northern Greece and central Peloponnese. This pattern did not seem to change over the decades, from present to future.
A pseudo-linear relationship was shown to exist between the monthly TDI values in any future period with regard to the historical ones. This pseudo-linear expression seemed to evolve into a linear one in the future.
Apart from the conclusions of the present work, it also presents another value: the historical ambient temperature records of 1985-2014 from the 33 sites intrinsically include the recent global surface-warming slowdown, also referred to as hiatus, that occurred in the period 1998-2012 [38]. Data Availability Statement: The present study was based on meteorological data of 33 sites in Greece provided by the Hellenic National Meteorological Service upon specific request. These data were used to generate the TMYs at the 33 locations.