*2.2. Statistical Tools and Data Processing*

All model-data processing and figure drawing were executed with R software and the ARC.GIS (MAP) 10.0 environment. Figure 1 depicts the basic steps of the process followed for the assessment of the extreme values of the variables, namely those of temperature

Daily Maximum Precipitation rate

Daily Maximum wind speed value

(maximum and minimum), precipitation rate, snowfall and wind speed. Firstly, the maximum and minimum temperature values were retrieved for the summer and winter seasons, respectively, while data were extracted throughout the year from the gridded datasets to determine the extremes of the precipitation rate, snowfall and wind speed. In this manner, at each grid cell and for each period (see Table 1), the 25 maximum values of each variable were obtained. The process was applied to all three time periods of interest and both RCP scenarios (see Table 1). ature (maximum and minimum), precipitation rate, snowfall and wind speed. Firstly, the maximum and minimum temperature values were retrieved for the summer and winter seasons, respectively, while data were extracted throughout the year from the gridded datasets to determine the extremes of the precipitation rate, snowfall and wind speed. In this manner, at each grid cell and for each period (see Table 1), the 25 maximum values of each variable were obtained. The process was applied to all three time periods of interest and both RCP scenarios (see Table 1).

All model-data processing and figure drawing were executed with R software and the ARC.GIS (MAP) 10.0 environment. Figure 1 depicts the basic steps of the process followed for the assessment of the extreme values of the variables, namely those of temper-

*Appl. Sci.* **2022**, *12*, x FOR PEER REVIEW 4 of 17

*2.2. Statistical Tools and Data Processing*

**Figure 1.** Schematic representation of the main methodological steps for determining a quantified climatic multi-hazard assessment. **Figure 1.** Schematic representation of the main methodological steps for determining a quantified climatic multi-hazard assessment.

Then, the modeled values of the variables of interest to this study were categorized in terms of likelihood, i.e., probability of occurrence, according to Table 2. The likelihood categories, six in total, and the threshold values of Table 2 were retrieved from the EU-CIRCLE project report [51] for the characterization of the hazard (maximum (summer) temperature, minimum (winter) temperature, precipitation rate, snow rate and wind speed). The probability of occurrence of each variable was calculated for all six categories. As the study was aimed at determining the likelihood of extremes, the focus was placed only on the specific categories characterizing the highest threshold values and hence, the probabilities of occurrence were calculated as totals of the three classes "High", "Very Then, the modeled values of the variables of interest to this study were categorized in terms of likelihood, i.e., probability of occurrence, according to Table 2. The likelihood categories, six in total, and the threshold values of Table 2 were retrieved from the EU-CIRCLE project report [51] for the characterization of the hazard (maximum (summer) temperature, minimum (winter) temperature, precipitation rate, snow rate and wind speed). The probability of occurrence of each variable was calculated for all six categories. As the study was aimed at determining the likelihood of extremes, the focus was placed only on the specific categories characterizing the highest threshold values and hence, the probabilities of occurrence were calculated as totals of the three classes "High", "Very High" and "Exceptional".

High" and "Exceptional". **Table 2.** Likelihood categories and threshold values for maximum and minimum temperature, maximum precipitation rate, maximum snow rate and maximum wind speed. In bold, the threshold values to which the probability of occurrence was applied in the current study. **Variables Likelihood Categories Very Low Low Medium High Very High Exceptional** Daily Minimum Temperature [°C] 0< 0–(−2) (−2)–(−5) **(−5)–(−10) (−10)–(−15) <(−15)** In this work, the method for calculating the probability of exceeding a value was based on the Extreme Value Theory (EVT). The determination of the extreme values of the studied variables required the selection and definition of the underlying distribution functions. The estimations were made using the R package "Extremes" [52], fitting a Generalized Extreme Value distribution (GEV) to block maxima data (annual maxima) under the assumption of non-stationarity [53]. The GEV distribution has three parameters: the shape factor ξ, the scale or dispersion parameter σ and the location or mode parameter µ. The GEV-distribution function, *G(y)*, is given by:

$$\text{For } \xi \neq 0, \ G(y) = \exp\left(-\left[1 + \xi \left(\frac{y-\mu}{\sigma}\right)\right]^{-1/\xi}\right) \tag{1}$$

Daily Maximum Snowfall [mm/h] <2.5 2.5–12.7 12.7–25.4 **25.4–76.2 76.2–127 >127** [m/s] <sup>0</sup>–<sup>3</sup> <sup>3</sup>–<sup>12</sup> <sup>12</sup>–<sup>15</sup> **<sup>15</sup>–<sup>20</sup> <sup>20</sup>–<sup>30</sup> >30** For <sup>ξ</sup> <sup>=</sup> 0, *<sup>G</sup>*(*y*) <sup>=</sup> exp <sup>−</sup> exp − *y* − *µ σ* (2)

The GEV has three types depending on shape parameter ξ, as follows:


### 3. When ξ < 0, GEV is known also as Type III Extreme Value Distribution (or Weibull Distribution, upper finite end point). based on the Extreme Value Theory (EVT). The determination of the extreme values of the studied variables required the selection and definition of the underlying distribution functions. The estimations were made using the R package "Extremes" [52], fitting a General-

In this work, the method for calculating the probability of exceeding a value was

**Table 2.** Likelihood categories and threshold values for maximum and minimum temperature, maximum precipitation rate, maximum snow rate and maximum wind speed. In bold, the threshold values to which the probability of occurrence was applied in the current study. ized Extreme Value distribution (GEV) to block maxima data (annual maxima) under the assumption of non-stationarity [53]. The GEV distribution has three parameters: the shape factor ξ, the scale or dispersion parameter σ and the location or mode parameter μ. The


*Appl. Sci.* **2022**, *12*, x FOR PEER REVIEW 5 of 17

The final step of the applied methodology included the calculation of the most frequently appearing hazard at each grid cell, time period and RCP scenario. In this manner, we performed the spatial assessment of the occurrence of multi-hazards in the historical period and in the two studied future periods according to the two RCP projections in order to illustrate the areas susceptible to hazards over long time scales. The final step of the applied methodology included the calculation of the most frequently appearing hazard at each grid cell, time period and RCP scenario. In this manner, we performed the spatial assessment of the occurrence of multi-hazards in the historical period and in the two studied future periods according to the two RCP projections in order to illustrate the areas susceptible to hazards over long time scales.

#### **3. Results and Discussion 3. Results and Discussion**

In this part, we present the results of the previously described applied approach. For reasons of clarity, each subsection provides the results pertinent to each studied hazard. Figure 2a presents the highly resolved topography of the EC-Earth global model downscaled by WRF to the high horizontal resolution of 5 <sup>×</sup> 5 km<sup>2</sup> . Additionally, highlighted in Figure 2b are the regions of the country, where some important findings are more extensively discussed. In this part, we present the results of the previously described applied approach. For reasons of clarity, each subsection provides the results pertinent to each studied hazard. Figure 2a presents the highly resolved topography of the EC-Earth global model downscaled by WRF to the high horizontal resolution of 5 × 5 km<sup>2</sup> . Additionally, highlighted in Figure 2b are the regions of the country, where some important findings are more extensively discussed.

**Figure 2.** (**a**) EC–Earth–WRF topography of the simulated domain of Greece with horizontal resolution (5 × 5 km<sup>2</sup> ). (**b**) Highlighted regions of the country of particular interest for discussion. **Figure 2.** (**a**) EC–Earth–WRF topography of the simulated domain of Greece with horizontal resolution (5 <sup>×</sup> 5 km<sup>2</sup> ). (**b**) Highlighted regions of the country of particular interest for discussion.

### *3.1. Maximum Temperature*

Figure 3a shows the probability of the occurrence of TX exceeding the threshold value (i.e., the probability of exceedance) of 35 ◦C during the summer season of the historical period of 1980–2004. Figure 3b–e depict the differences in the probability of exceedance between the future projections and the historical values (i.e., future–historical) for both RCPs and the studied periods. During the historical period, we may deduce that the majority of areas show low probabilities of exceedance below 2.5% (Figure 3a). On the other hand, the highest occurrence of maximum temperature values with the probability of exceedance above 10% is seen in the plains of the regions of Thessaly, central Macedonia, Peloponnese, the western mainland, the eastern Aegean islands and southern Crete, which are well known as summer hot-spot areas [54]. between the future projections and the historical values (i.e., future–historical) for both RCPs and the studied periods. During the historical period, we may deduce that the majority of areas show low probabilities of exceedance below 2.5% (Figure 3a). On the other hand, the highest occurrence of maximum temperature values with the probability of exceedance above 10% is seen in the plains of the regions of Thessaly, central Macedonia, Peloponnese, the western mainland, the eastern Aegean islands and southern Crete, which are well known as summer hot-spot areas [54].

Figure 3a shows the probability of the occurrence of TX exceeding the threshold value (i.e., the probability of exceedance) of 35 °C during the summer season of the historical period of 1980–2004. Figure 3b–e depict the differences in the probability of exceedance

*Appl. Sci.* **2022**, *12*, x FOR PEER REVIEW 6 of 17

*3.1. Maximum Temperature*

**Figure 3.** (**a**) Spatial distribution of probability of TX exceedance above 35 °C calculated using EC– Earth–WRF downscaled data for the historical summer period 1980–2004. Differences (future–historical) in the probability of TX exceedance for: (**b**) RCP4.5 in near future (2025–2049), (**c**) RCP8.5 in near future (2025–2049), (**d**) RCP4.5 in far future (2075–2099) and (**e**) RCP8.5 in far future (2075– 2099). **Figure 3.** (**a**) Spatial distribution of probability of TX exceedance above 35 ◦C calculated using EC–Earth–WRF downscaled data for the historical summer period 1980–2004. Differences (future– historical) in the probability of TX exceedance for: (**b**) RCP4.5 in near future (2025–2049), (**c**) RCP8.5 in near future (2025–2049), (**d**) RCP4.5 in far future (2075–2099) and (**e**) RCP8.5 in far future (2075–2099).

Overall, in the near future, more areas compared to the historical period were found to be exposed to extreme TX values (Figure 3b,c). In addition, it appears that the probability increase in hot-spot areas is higher in RCP4.5, with these values exceeding 3%. The differences in the referenced areas become much larger, exceeding 10%, and even extend spatially into the far future and more profoundly in RCP8.5 (Figure 3d,e). In fact, the probability of exceedance according to RCP8.5 in the far future increases not only in the plains areas with low topographic heights but also in areas with heights in the range of 500 to 1000 m (Figure 3e).

This could be considered an important change predicted by the worst-case emissions scenario that would cause adverse conditions near the end of the century. Nevertheless, in the near future, insignificant changes are expected over the high mountainous regions according to both RCPs concerning the historical period (Figure 3b,c). In the far future, the insignificant changes are still found over the highest mountains with RCP4.5 (Figure 3d) but the RCP8.5 scenario diminishes these (around zero values), showcasing detectable differences in the probability of exceedances of 3–6% with respect to the historical period, even over the highest summits of the mainland (Figure 3e).

### *3.2. Minimum Temperature*

Figure 4a presents the probability of TN exceeding −5 ◦C towards lower values during the winters of the historical period. Overall, the probability values remain below 5% in most areas of the mainland and the islands. Over the higher topographic heights of the central and northeastern mainland (see Figure 2a), the probability increases to noticeable levels in the range of 20 to 40% that are consistent with the occurrence of very low temperatures and the extreme winter climatology of the country [54]. The differences in the probabilities of exceedance of TN between the two future and historical periods (future–historical) are shown in Figure 4b–e for both RCPs. We may observe that under all scenarios and periods, areas that have historically had very low probability values of extremely cold temperatures preserve these characteristics in the future.

In general, we observe a strong decrease in the probabilities of exceedance of TN over the mountainous areas of the central and northeastern mainland in both future periods with respect to the historical period, which denotes a reduction in the future occurrence of extreme values of TN and thus, fewer winter extremes. In the near-future period, the highest decreases in the central mountainous areas are more intense in RCP4.5 (Figure 4b) than in RCP8.5 (Figure 4c) while in eastern Macedonia, Thrace and Peloponnese there are no noticeable differences between the two scenarios. Stronger decreases in the occurrence of winter extremes are estimated for both scenarios in the far future (Figure 4d,e). The impacted areas according to RCP4.5 remain the same during both periods (Figure 4b,d) but they extend more spatially in the far future with RCP8.5 (Figure 4c,e).

### *3.3. Precipitation Rate*

Figure 5a presents, for the historical period, the probability of precipitation extremes calculated as the probability of precipitation rates exceeding 10 mm/h. The pattern with persisting probabilities over the majority of the domain of less than 0.025% does not yield noticeable spatial variability. Yet, increased probabilities of extreme precipitation with values greater than 0.075% are observable in the very high mountainous areas of the central and eastern mainland and over the summits of Crete and Peloponnese. In addition, some areas known for high-precipitation rates such as the Ionian islands and parts of central and eastern Macedonia and Rhodes reasonably exhibit distinguishable contours of the probability of exceeding 10 mm/h up to 0.075%.

The differences between the future projections and historical simulations indicate a reduction in future extreme precipitation mostly in the eastern parts of the mainland, central Aegean islands and mountainous areas of Crete (Figure 5b–e). The maximum decrease is obtained in the near future (Figure 5b,c). Additionally, with reference to the historical period and both RCPs, we may observe minute and insignificant changes in extreme precipitation rates, more extensively in the regions of Thessaly and central Macedonia and predominantly in the near future (Figure 5b,c). However, interesting patterns of increased probabilities of extreme precipitation can be seen in highly mountainous areas primarily in

the western and northeastern mainland (in the range of 0.051–0.285%) and more vividly for RCP8.5 (Figure 5d,e). The RCP4.5 projections do not show a noticeable change between the two future periods. On the contrary, according to RCP8.5, the far-future period shows on average an increase in the probability of rainfall extremes compared with the near future, both in magnitude and spatial extent. Overall, the patterns of differences showcase increased probabilities of extreme precipitation rates in the studied future periods even at low topographic heights, which may highlight a bothersome climate-change effect for the agricultural economy. *Appl. Sci.* **2022**, *12*, x FOR PEER REVIEW 8 of 17

**Figure 4.** (**a**) Spatial distribution of probability of TN exceedance below −5 °C calculated using EC– Earth–WRF downscaled data for the historical winter period 1980–2004. Differences (future–historical) in the probability of TN exceedance for: (**b**) RCP4.5 in near future (2025–2049), (**c**) RCP8.5 in near future (2025–2049), (**d**) RCP4.5 in far future (2075–2099) and (**e**) RCP8.5 in far future (2075– **Figure 4.** (**a**) Spatial distribution of probability of TN exceedance below −5 ◦C calculated using EC–Earth–WRF downscaled data for the historical winter period 1980–2004. Differences (future– historical) in the probability of TN exceedance for: (**b**) RCP4.5 in near future (2025–2049), (**c**) RCP8.5 in near future (2025–2049), (**d**) RCP4.5 in far future (2075–2099) and (**e**) RCP8.5 in far future (2075–2099).

The differences between the future projections and historical simulations indicate a reduction in future extreme precipitation mostly in the eastern parts of the mainland, central Aegean islands and mountainous areas of Crete (Figure 5b–e). The maximum decrease

period and both RCPs, we may observe minute and insignificant changes in extreme precipitation rates, more extensively in the regions of Thessaly and central Macedonia and predominantly in the near future (Figure 5b,c). However, interesting patterns of increased probabilities of extreme precipitation can be seen in highly mountainous areas primarily in the western and northeastern mainland (in the range of 0.051–0.285%) and more vividly for RCP8.5 (Figure 5d,e). The RCP4.5 projections do not show a noticeable change between the two future periods. On the contrary, according to RCP8.5, the far-future period shows on average an increase in the probability of rainfall extremes compared with the near future, both in magnitude and spatial extent. Overall, the patterns of differences showcase increased probabilities of extreme precipitation rates in the studied future periods even at

2099).

low topographic heights, which may highlight a bothersome climate-change effect for the

**Figure 5.** (**a**) Spatial distribution of probability of RR exceedance above 10 mm/h calculated using EC–Earth–WRF downscaled data for the historical period 1980–2004. Differences (future–historical) in the probability of RR exceedance for: (**b**) RCP4.5 in near future (2025–2049), (**c**) RCP8.5 in near future (2025–2049), (**d**) RCP4.5 in far future (2075–2099) and (**e**) RCP8.5 in far future (2075–2099). **Figure 5.** (**a**) Spatial distribution of probability of RR exceedance above 10 mm/h calculated using EC–Earth–WRF downscaled data for the historical period 1980–2004. Differences (future–historical) in the probability of RR exceedance for: (**b**) RCP4.5 in near future (2025–2049), (**c**) RCP8.5 in near future (2025–2049), (**d**) RCP4.5 in far future (2075–2099) and (**e**) RCP8.5 in far future (2075–2099).

### *3.4. Wind 3.4. Wind*

agricultural economy.

During the historical period, the areas of high wind-speed (exceeding 15 m/s) probability are found to be consistent with other studies in the literature [45,55,56] (Figure 6a) and in accordance with the known synoptic atmospheric systems associated with the prevailing wind patterns. More particularly, the areas of the northeastern Aegean Sea are impacted by north-easterlies and the central Aegean by the Etesians, whereas the topog-During the historical period, the areas of high wind-speed (exceeding 15 m/s) probability are found to be consistent with other studies in the literature [45,55,56] (Figure 6a) and in accordance with the known synoptic atmospheric systems associated with the prevailing wind patterns. More particularly, the areas of the northeastern Aegean Sea are impacted by north-easterlies and the central Aegean by the Etesians, whereas the topography of Crete significantly amplifies the patterns of the extreme winds.

In the near future and under both scenarios, the probability of the occurrence of windstorms is amplified almost all over the domain and mostly over the sea areas and mountain ridges (Figure 6b,c). The Ionian and central Aegean Seas are projected to exhibit

raphy of Crete significantly amplifies the patterns of the extreme winds.

the highest changes in the probabilities in the RCP4.5 scenario.

In the far future, both scenarios show a decrease in the probability of extreme winds over the northeastern Aegean and south Ionian Seas (Figure 6d,e). On the other hand, the

the Aegean Sea projected by both RCPs are in agreement with the findings of [57].

**Figure 6.** (**a**) Spatial distribution of probability of wind speed exceedance over the threshold (15 m/s) calculated using EC–Earth–WRF downscaled data for the historical period 1980–2004. Differences (future–historical) in the probability of wind speed exceedance for: (**b**) RCP4.5 in near future (2025– **Figure 6.** (**a**) Spatial distribution of probability of wind speed exceedance over the threshold (15 m/s) calculated using EC–Earth–WRF downscaled data for the historical period 1980–2004. Differences (future–historical) in the probability of wind speed exceedance for: (**b**) RCP4.5 in near future (2025– 2049), (**c**) RCP8.5 in near future (2025–2049), (**d**) RCP4.5 in far future (2075–2099) and (**e**) RCP8.5 in far future (2075–2099).

In the near future and under both scenarios, the probability of the occurrence of windstorms is amplified almost all over the domain and mostly over the sea areas and mountain ridges (Figure 6b,c). The Ionian and central Aegean Seas are projected to exhibit the highest changes in the probabilities in the RCP4.5 scenario.

In the far future, both scenarios show a decrease in the probability of extreme winds over the northeastern Aegean and south Ionian Seas (Figure 6d,e). On the other hand, the mountainous areas, the central Aegean Sea and Crete present an increase in the probability of extreme winds. The increases in extreme winds associated with the Etesians over the Aegean Sea projected by both RCPs are in agreement with the findings of [57].

### *3.5. Multi-Hazard Probability Maps*

Figure 7 depicts the spatial distribution of the probability of occurrence of the most dominant hazard calculated for the historical period and both future periods and scenarios. The dominant-hazard map for the historic period indicates the dominance of extreme TX over plains and coastal areas and the dominance of extreme TN at topographic heights higher than ~500 m a.m.s.l (Figure 7a). Over the seas, the islands of the central Aegean and the southeastern corners of Evia and Peloponnese, the extreme winds become dominant. However, extreme temperatures preside in the islands of the Ionian and eastern Aegean Seas, while in the northern parts of some of them, the extremely low temperatures or extreme winds dominate. In the case of Crete in particular, we observe the dominance of the three hazards with extreme TX and TN over the plains and high mountainous regions, respectively, and windstorms over the remaining areas of the island. Moreover, the dominance of extreme rainfall is observed in parts of the eastern coasts of the central mainland and Peloponnese.

The projected changes according to RCP4.5 in the near future yield a more extended dominance of extreme TX, particularly in the areas of Peloponnese, the central-eastern mainland and in the north of the country over the plains of central and eastern Macedonia and Thrace (Figure 7b). The same effect is observed in the central Aegean islands and northern parts of Crete. In addition, the extreme rainfall is seen to prevail more extensively in eastern coastal areas of the mainland. On the other hand, according to RCP8.5, the dominance of extreme TX becomes more profound in Thessaly and the plains areas of northern Greece (Figure 7c). Moreover, an interesting pattern of prevailing extreme winds is revealed in parts of the mainland and mostly in the Peloponnese where in the historical period the dominance of extreme temperatures was evident. Furthermore, RCP8.5 presents extremely windier conditions for the Ionian islands and Crete. This result agrees with the findings of Karozis et al. [56] that highlighted the reduction in the persisting anticyclonic activities over Greece and the Balkans in the near future, a change that denotes less frequent stagnant atmospheric conditions. The same study indicated a possible reduction in the passage of cyclones over Greece originating from the cyclogenesis region of the Central Mediterranean and the Adriatic Sea that could explain the reduced extreme-rainfall findings.

In the far future, the projected changes due to RCP4.5 show the dominance of extreme TX in areas of low altitude and more vividly in western Greece, Peloponnese, Thessaly and the northern mainland (Figure 7d). Additionally, in Crete and the central Aegean islands the extreme winds become less dominant and extreme TX predominates. In addition, the extreme rainfall is noted to persist in the far future over coastal parts of the eastern mainland. All the same, the extreme TN dominates in the mountainous areas. The high occurrence of extreme-winter-cold events under high-emission scenarios agrees with other climate-model projections that estimated increases in mid-latitude westerlies and northerly cold-air flow due to the influence of the upper-tropospheric equator-to-pole temperature difference in the storm-track response to climate change [58].

Figure 8 presents the multi-hazard occurrence in Greece for the examined periods. The value indicates the cumulative annual probability that at least one of the studied extreme hazards will occur. It demonstrates that patterns of highly exposed areas in Greece over the historic period will be spatially shifted in the far future as the increase in hotter

In the historic period (Figure 8a), mountainous areas in the northern parts appear to be the most exposed regions in Greece, with TN and RR being the most significant risks (Figure 7a). The plains areas in the mainland and Crete exhibit lower risk levels through a combination of TX and extreme winds. The coastal zones and the Aegean islands appear

climate regimes will be dominating the risk landscape.

to lie at the lower risk level.

**Figure 7.** Dominant-hazard maps using EC–Earth–WRF downscaled data for: (**a**) the historical period 1980–2004, (**b**) RCP4.5 in near future (2025–2049), (**c**) RCP8.5 in near future (2025–2049), (**d**) RCP4.5 in far future (2075–2099) and (**e**) RCP8.5 in far future (2075–2099). **Figure 7.** Dominant-hazard maps using EC–Earth–WRF downscaled data for: (**a**) the historical period 1980–2004, (**b**) RCP4.5 in near future (2025–2049), (**c**) RCP8.5 in near future (2025–2049), (**d**) RCP4.5 in far future (2075–2099) and (**e**) RCP8.5 in far future (2075–2099).

In the near future, the RCP4.5 scenario (Figure 8b) appears to demonstrate the smallest variability in the risk levels when compared to RCP8.5 (Figure 8c). The mountainous areas will be exposed to lower risk levels by a factor between 5 and 14% compared to present times, contrasting a similar increase due to TX in the lowlands and western Greece. These relative changes were determined to be statistically significant at the 95% Furthermore, the RCP8.5 projections of the far future showcase the extended dominance of extreme TX in the islands and mainland except for the highest altitudes (circa 1500 m a.m.s.l.) (Figure 7e). The future intensification of extreme-temperature events is in agreement with other studies due to non-linear interactions, which are presently not quantified, between Arctic teleconnections and other remote and regional feedback processes [59,60]. Furthermore, the extreme winds remain the dominant hazard over the seas while the eastern coasts of the mainland would mostly experience extreme rainfall events that only persist locally in the far future. This outcome is associated with the Arctic amplification and possible connection to the weakening of mid-latitude storm tracks [61].

Figure 8 presents the multi-hazard occurrence in Greece for the examined periods. The value indicates the cumulative annual probability that at least one of the studied extreme hazards will occur. It demonstrates that patterns of highly exposed areas in Greece over the historic period will be spatially shifted in the far future as the increase in hotter climate regimes will be dominating the risk landscape.

exposed to increased risk.

confidence level using the student t-test. The southern parts of the Eastern Aegean islands and Crete will also experience increased level of hazards. For the RCP8.5 scenario, in the near future, a greater number of regions are projected to be exposed to higher levels of risk due to both higher TX values and stronger winds (Figure 8c). These appear to be located in the southern and the western parts of the mainland and the Aegean Sea. For the far future (Figure 8d,e), both scenarios will exhibit similar patterns of risk changes compared to the present period, although RCP8.5 will be associated with higher hazard risk, often exceeding a 5% increase. Plains, agricultural lands and islands will be especially

**Figure 8.** Annual percentage of multi-hazard using EC–Earth–WRF downscaled data for: (**a**) the historical period 1980–2004, (**b**) RCP4.5 in near future (2025–2049), (**c**) RCP8.5 in near future (2025– 2049), (**d**) RCP4.5 in far future (2075–2099) and (**e**) RCP8.5 in far future (2075–2099). **Figure 8.** Annual percentage of multi-hazard using EC–Earth–WRF downscaled data for: (**a**) the historical period 1980–2004, (**b**) RCP4.5 in near future (2025–2049), (**c**) RCP8.5 in near future (2025–2049), (**d**) RCP4.5 in far future (2075–2099) and (**e**) RCP8.5 in far future (2075–2099).

**4. Conclusions** The study presented here aimed to elucidate the highly dynamic changing patterns of climate risk in Greece, a European climate hot spot [62]. The findings highlighted the areas that are exposed to multiple climate hazards in the country, considering the influence of the highly complex topography. In addition, the generated multi-hazard risk maps In the historic period (Figure 8a), mountainous areas in the northern parts appear to be the most exposed regions in Greece, with TN and RR being the most significant risks (Figure 7a). The plains areas in the mainland and Crete exhibit lower risk levels through a combination of TX and extreme winds. The coastal zones and the Aegean islands appear to lie at the lower risk level.

could be used to support disaster-risk-prevention activities such as avoiding future human, natural and material losses and generating economic benefits by reducing climaterelated risks. The introduced method could be easily transferred to other geographical regions provided that the climate simulations are available. In addition, based on the In the near future, the RCP4.5 scenario (Figure 8b) appears to demonstrate the smallest variability in the risk levels when compared to RCP8.5 (Figure 8c). The mountainous areas will be exposed to lower risk levels by a factor between 5 and 14% compared to present times, contrasting a similar increase due to TX in the lowlands and western Greece. These relative changes were determined to be statistically significant at the 95% confidence level using the student t-test. The southern parts of the Eastern Aegean islands and Crete will also experience increased level of hazards. For the RCP8.5 scenario, in the near future, a greater number of regions are projected to be exposed to higher levels of risk due to both higher TX values and stronger winds (Figure 8c). These appear to be located in the southern and the western parts of the mainland and the Aegean Sea. For the far future (Figure 8d,e), both scenarios will exhibit similar patterns of risk changes compared to the present period, although RCP8.5 will be associated with higher hazard risk, often exceeding a 5% increase. Plains, agricultural lands and islands will be especially exposed to increased risk.

### **4. Conclusions**

The study presented here aimed to elucidate the highly dynamic changing patterns of climate risk in Greece, a European climate hot spot [62]. The findings highlighted the areas that are exposed to multiple climate hazards in the country, considering the influence of the highly complex topography. In addition, the generated multi-hazard risk maps could be used to support disaster-risk-prevention activities such as avoiding future human, natural and material losses and generating economic benefits by reducing climate-related risks. The introduced method could be easily transferred to other geographical regions provided that the climate simulations are available. In addition, based on the perceived risk values or values identified in national risk assessments, the likelihood categories of the hazards (Table 2) could be adjusted accordingly. It should be mentioned that a limitation in the approach emanates from the complex topography of a domain or parts of it. In such a case, it may be required to downscale the climate data to even higher than a 5 km resolution over the complex topography areas, where there is the need to study the occurrence of (multi-)hazard(s) in more detail and accuracy.

Overall, the analysis demonstrated that climate change is a highly non-stationary process and the exposed areas, risk level and dominant risk will be significantly changed in the future under both RCP4.5 and RCP8.5 scenarios. More particularly, the impact of global warming on the country will become more evident in the far future (end of the century) when the extreme maximum temperature will dominate all other hazards.

According to the RCP4.5 scenario, a gradual expansion of the extreme maximum temperature can be anticipated from the coastal regions to the higher altitude areas, and this trend will become more persistent towards the end of the century. Under the RCP8.5 scenario, the extreme wind speed was found to be the dominant hazard in the near future, while afterwards, near the end of the century, the extreme maximum temperature becomes the most significant hazard.

**Author Contributions:** Conceptualization, D.V. and A.S.; methodology, D.V. and A.S.; software, I.M., S.K. and N.P.; validation, A.S. and N.P.; formal analysis, D.V. and N.G.; investigation, D.V.; resources, D.V.; writing—original draft preparation, D.V.; writing—review and editing, A.S.; supervision, D.V. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data used in this study can be obtained by request to the authors.

**Acknowledgments:** This work was supported by computational time granted from the Greek Research & Technology Network (GRNET) in the National HPC facility—ARIS—under project ID HRCOG (pr004020 and pr006028).

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**

