High Resolution Future Projections of Drought Characteristics in Greece Based on SPI and SPEI Indices

Abstract

Future changes in drought characteristics in Greece were investigated using dynamically downscaled high-resolution simulations of 5 km. The Weather Research and Forecasting model simulations were driven by EC-EARTH output for historical and future periods, under Representative Concentration Pathways 4.5 and 8.5. For the drought analysis, the standardized precipitation index (SPI) and the standardized precipitation-evapotranspiration index (SPEI) were calculated. This work contributed to achieve an improved characterization of the expected high-resolution changes of drought in Greece. Overall, the results indicate that Greece will face severe drought conditions in the upcoming years, particularly under RCP8.5, up to 8/5 y of severity change signal. The results of 6-month timescale indices suggest that more severe and prolonged drought events are expected with an increase of 4 months/5 y, particularly in areas of central and eastern part of the country in near future, and areas of the western parts in far future. The indices obtained in a 12-month timescale for the period 2075–2099 and under RCP8.5 have shown an increase in the mean duration of drought events along the entire country. Drought conditions will be more severe in lowland areas of agricultural interest (e.g., Thessaly and Crete).

1. Introduction

According to the latest IPCC WGII AR6 report [1], climate change is projected to intensify throughout the Mediterranean region, and its impacts include longer and/or more intensive droughts that will become in the future more prevalent in many areas. Drought causes a cascade of effects that will affect many different environmental systems in a region, through direct and indirect natural processes [2]. It is a recurring, inevitable feature of climate that results in serious economic, environmental, and social impacts [3]. For those reasons, the recognition of drought as a climate hazard is becoming an urgent priority in a warming world [4]. Thus, the scope of this work is to provide a highly detailed spatial and temporal drought assessment in the entire domain of Greece in high spatial resolution.

Previous modelling studies on drought projection were derived from the outputs from the Coupled Model Intercomparison Project (CMIP) of phase three (e.g., CMIP3; [5,6] and five (CMIP5; [7,8,9,10,11,12,13]). Those studies revealed similar results, in which droughts were more frequent and severe in the 21st century over identified drought hotspots like the Mediterranean basin and some adjacent areas. The latest CMIP6 phase has produced updated global climate model outputs, using the shared socioeconomic pathways (SSPs) emission scenarios [14,15]. Currently, projections of these global simulations with RCMs remain unavailable at high spatial resolution. In the global scale study of Li et al. 2021 [16], the results indicate that the magnitude and extent of droughts are projected to increase significantly with increasing SSPs and warming in some regions of the world by the late 21st century. However, such projections suffer from high uncertainties due to coarse spatial resolution. Of late years, downscaling methodologies, such as dynamical downscaling using a regional climate model (RCM), have been proposed to produce the high-resolution climate variables that are much needed. Therefore, studies based on the Coordinated Regional Climate Downscaling Experiment (CORDEX) and the Representative Concentration Pathways (RCPs) investigated drought hazard projections not only globally [17,18] but also in selected countries or regions [18,19,20,21] by means of RCMs.

At the regional scale, several studies [19,20,22,23,24] agree on the increase in droughts over the past decades and on projected increases in the duration and intensity of droughts for most parts of the Mediterranean basin, based on future climate scenarios. The authors of [25] emphasized that droughts are not spatially coherent in the Mediterranean basin, demonstrating different spatial patterns even at the regional scale or between eastern and western regions. Drought conditions are established in the southern and eastern regions of the Mediterranean Basin [26]; however, recent studies also report, in high confidence, that the North Mediterranean areas experience more frequent and intense drought events [1].

Greece as a north-eastern Mediterranean country is characterized as a semi-arid region which is experiencing an increasing number of various extreme events during the last decades that could be attributed to climate change directly or indirectly (e.g., fires, floods, heat waves, dry episodes, etc.). The authors of [27] indicated that long persistent droughts over Greece are related to large scale atmospheric circulation patterns, such as the extension of the subtropical anticyclone of the Atlantic (Azores) up to central Mediterranean, characterized by high positive anomaly of geopotential height of 500 mb over North-Eastern Europe or high positive North Atlantic Oscillation (NAO) index.

Considerable drought incidents have been noted during the last 50 years (e.g., 1989–2003, 2007–2008) in Greece, and a number of studies have contributed to the assessment of drought conditions in Greece. The authors of [28] applied the Palmer Drought Severity Index (PSDI) based on station data, in the central and northern region of Greece. Severity–duration–frequency (SDF) relationships of droughts and wet periods over Greece for hydroclimatic and agroclimatic design and planning were also developed by [29]. The authors of [30] studied the drought phenomenon through spatiotemporal analysis of the dry spells. The authors of [31] presented SPI for characterizing drought, while [32,33,34,35] used SPI to detect and study important drought events in spatiotemporal basis based on station data. A few studies have applied statistical analysis to obtain spatiotemporal parameters of drought episodes in Greece [36,37]. The authors of [38] calculated Reconnaissance Drought Index using different PET methods from two reliable meteorological stations of Greece. The authors of [39] estimated the Aridity Index (AI) for three selected areas in Greece based on local observations data and interpolation methods. Other studies used datasets derived from global or regional models to assess historical and projected change of drought. The authors of [40] examined the changes on spatiotemporal drought characteristics of Thessaly region using SPI through GCM output and under SRES scenarios. The authors of [27] applied statistical downscaling method in the outputs of Global Circulation Model for the assessment of climate change on hydrological, agricultural, and water resources droughts in Thessaly. The authors of [41,42] assessed drought in Platis basin and for the island of Crete based on bias corrected historical and future GCM output data under RCPs scenarios. The authors of [43] studied the spatial and temporal variability of the Aridity Index (AI) in Greece, derived from regional model within the ENSEMBLES European Project and under SRES A1B. The authors of [44] showed the projected effects of climate change on meteorological drought at the Greek region using five RCMs from the ENSEMBLES European Project. The authors of [45] estimated projected changes of drought based on SPI using simulated data from the ENSEMBLES European Project for three agricultural areas widespread in Greece. More recently, the authors of [46] studied the consecutive dry days only for the end-of-the-century period under RCP8.5 based on an ensemble of 11 EU-CORDEX models simulations.

The limitation in drought research in Greece in these studies is attributed to the use of output data from global models or of the commonly used horizontal resolutions ranges between 50 and 25 km (usually RCMs from ENSEMMBLES project), 12.5 km from EUCORDEX or application of interpolation methods with limited stations’ numbers. In addition, a number of studies have contributed to the investigation of drought conditions in Greece to assess the drought dimensions based on meteorological, water supply, and demand information for the estimation and quantification of drought conditions in the country’s hydrographic basin or specific areas, but not in the entire country. The authors of [47] reported that for mountainous regions, even 10 km can be considered a coarse resolution because higher resolution is needed to provide useful information for input into basin hydrology studies. Therefore, taking into account the previously commented considerations, there has been limited effort to explicitly examine the potential impacts of future climate change on droughts across Greece at higher spatial resolution. The need for high resolution drought analysis is highly linked to the geomorphological complexity of the country (predominantly a mountainous country with an extended coastal line and numerous scattered islands). Thereafter, the study of future drought events recommends the use of regional models and/or downscaling techniques which are capable on capturing the different processes associated with drought events more precisely at a high-resolution spatial scale. The authors of [45] highlighted the importance of influence of elevation and broadly the topography in the generation of different climatic conditions in different basins which in turn affect the spatial analysis of droughts. Thus, it is imperative to study in higher detail drought characteristics that are expected to vary spatially, as the complex topographic features influence the local climate of the various parts of the country.

To address the above discussed issues, the presented work is a first attempt (to our knowledge) to carry out a very detailed assessment of future drought characteristics in Greece under a changing climate at the very high spatial resolution of 5 km, using dynamically downscaled model gridded simulated data. This work aims to assess the spatial and temporal change of drought characteristics (severity, duration, and intensity), which are thoroughly investigated using two drought indices, the SPI and the SPEI in different timescales (6 and 12 months). Within this context, vulnerable areas to drought will be highlighted in an effort to contribute to the efficient climate change adaptation and mitigation strategies among different sectors for drought risk management.

2. Materials and Methods

In this section, we describe the methodology applied to investigate the impact of climate change on drought characteristics in Greece. Climate gridded datasets of precipitation, minimum and maximum temperatures were derived from regional climate simulations for the area of Greece, with the Weather Research and Forecasting (WRF-ARW) model [48] appropriately setup, driven by the EC-EARTH global model [49]. Output data are converted to monthly values to compute drought indices for two time periods, in the future and under two emission scenarios (RCP4.5 and RCP8.5) [50,51]. Both indices are calculated over the land grid cells of the nested domain for each grid point and for each time period. At a final step, the analysis is performed to determine the modifications in spatial and temporal drought characteristics in terms of severity, intensity and duration under a changing climate. The above methodology follows the same approach applied in recent studies (e.g., [17,21]).

2.1. Model Setup

EC-Earth model demonstrates very good forecasting skill and can also be used for climate studies, supporting emerging seamless prediction strategies. In addition, the model simulates well the tropospheric fields and the dynamic variables, but not as good the surface temperature and fluxes [49,52,53]. Moreover, EC-Earth climate simulations and projections have been widely used for climate studies in the framework of CMIP5, CORDEX (e.g., [54,55,56,57,58]) and, more recently, in CMIP6 [59]. EC-Earth model is a full physics seamless atmosphere–ocean-sea ice coupled earth system prediction model and developed from the operational Integrated Forecast System (IFS) cycle 31 r of ECMWF. The set-up of the atmospheric model in the EC-Earth version 2.3 corresponds to the use of a horizontal spectral resolution of T159 (triangular truncation at wavenumber 159), roughly 125 km, and a vertical grid with vertical 62 levels of a terrain-following mixed sigma-pressure hybrid coordinates, of which about 15 are within the planetary boundary layer (PBL).

The dynamical downscaling technique includes the following procedure: The non-hydrostatic (WRF-ARW) model was used in its version 3.6.1. The spatial configuration of the model was composed of two nested grids. The outer domain (Europe) domain (d01) of 20 km resolution is centered in the Mediterranean basin at 42.5 N and 16.00 E and the high-resolution inner domain was set up at 5 km (D02—Greece) of horizontal grid spacing (Figure 1). The physics options are extensively described in Politi et al., 2021. The boundary conditions for the climate change assessment derive from EC-Earth global model climate simulations for RCP4.5 and RCP8.5 scenario and encompass time slices representative of the historical (1980–2004), midcentury (2025–2049), and end of century (2075–2099) periods. For the future projections, the equivalent-CO2 concentration was updated every year accordingly to the emission scenario in the WRF simulations.

As with any impact assessment related to climate change, uncertainty requires to be considered and discussed. In this study, nevertheless, the high-resolution climate simulation data used (temperatures and precipitation) are based on extensively attentive validation studies [60,61,62,63,64,65,66] from the application of a regional model (WRF) and/or a GCM (EC-EARTH), which proved capable of capturing the spatial and temporal patterns of precipitation and temperature compared to observations. Consequently, any uncertainties in this study can generally be summarized in the reliability of model simulations concerning the emission scenarios (RCPs) and the selected drought indices.

Further information regarding model configuration can be found in the aforementioned study of [62], which has proven the ability of this setup to yield more realistic topographical-induced precipitation and temperature fields with their extremes.

2.2. Drought Analysis Method

In this study, in order to identify dryness or wetness conditions that can cause drought impacts on various sectors, the Standardized Precipitation Index (SPI) [67] and Standardized Precipitation Evapotranspiration Index (SPEI) [68] are estimated. These two indices are among the most widely used indices for drought identification and monitoring in Europe. As no single drought index alone may precisely describe all the attributes of drought conditions, their combination is a common approach in the scientific literature lately [69,70,71,72,73].

The SPI is calculated by fitting a probability density function to a given frequency distribution of precipitation totals for a station or grid point and for an accumulation period and then the probabilities are transformed into a normalized distribution with a mean equal to zero and a variance of one. SPI is calculated as follows in Equation (1):

$$\mathrm{SPI}=\frac{x_{\iota }-x_j}{\sigma }$$

(1)

where, x_i refers to the current precipitation in the examined period, x_j refers to the mean precipitation of the timeseries, and σ refers to the standard deviation of the timeseries.

The inclusion of temperature on SPEI’s calculation (through potential evapotranspiration (PET)) is suggested by WMO and Global Water Partnership (GWP) [74] since it is more suitable for the study of impact of future climate change. Details of the SPEI calculation can be found in [68,75]. SPEI is estimated using the same methodology for SPI, but also includes the climatic water balance, which is the difference between precipitation and potential evapotranspiration. The distribution functions used for computing those indices were the ‘log-Logistic’ for SPEI, and ‘Gamma’ for SPI. The applied herein distributions are the most widely used in literature and recommended by the indices’ original developers [76]. Here, to calculate the SPEI index, the monthly potential evapotranspiration is estimated based on the Samani equation [77] by estimating solar radiation from monthly minimum and maximum temperature along with the location (latitude) of the grid cell. This method is frequently used in drought studies [17,78,79]. The comparison of SPI and SPEI is made to assess the impact of potential evapotranspiration which is a metric of the atmospheric evaporative demand (AED) to determine the drought in the study areas as well as the uncertainty in the results obtained using the SPI. SPI or SPEI values of 6 and 12 months are proposed as more appropriate for denoting droughts in arid and semi-arid regions, applied in several studies (e.g., [76,80,81,82,83]). Accordingly, the SPI-6, SPEI-6, SPI-12, and SPEI-12 are selected for the drought characterization in Greece.

Both indices in 6- and 12-month timescales are calculated over the nested domain for each grid point and all precipitation data are converted to monthly values. Moreover, the time series of the drought indices are calculated over the land grid cells for each time period. Drought conditions are indicated as SPI decreases below −1.0, while increasingly severe excess rainfall is indicated as SPI increases above 1.0. The applied classification is consistent with EU recommendations for this area which is part of Euro-Mediterranean area [74,84]. In this context, the characterization of a drought event is established when dry or near-normal conditions are followed by drought conditions with values of the index below −1 at least for two consecutive months. In the same way, it is considered that the event ends when the value of index corresponds to near normal/wet conditions (index values greater than 0). In order to examine the drought characteristics, three different parameters are used: (1) severity which is determined as the absolute sum of SPI and SPEI values for a drought event; (2) duration as the length of each drought event (in months); and (3) mean intensity which is calculated as the average SPI and SPEI value during a drought event or even defined by the severity divided by duration.

The analysis of projected changes of drought was evaluated through the Delta-Change approach [85] in terms of duration, intensity, frequency, and severity of drought events by comparing indices, their time scales, scenarios, and periods. Thus, future climate changes of drought characteristics are defined as the differences (Delta change) between the projection run (near or far period) and the control run (reference period). Along with the projected changes of drought characteristics, we studied the trends of drought characteristics, as well as their significance. The linear trend is calculated based on the annual values of intensity, severity and duration of drought events. Figure 2 presents the regions of particular interest based on the results.

To calculate the SPI and SPEI, the R software was used, the “SPEI” package [68,75]. In addition, drought characteristics and trends have been analyzed in the R environment (http://www.r-project.org/index.html (accessed date: 3 August 2022)).

3. Results and Discussion

3.1. SPI-SPEI 6

In this section, we analyze the results of drought characteristics and their trends derived from the SPI and SPEI values of the 6-month timescale. The frequency (number of events), duration, severity, and intensity of the events shown in the following figures are expressed per 5 years. The historical conditions of drought characteristics are also depicted for each case.

Figure 3a depicts the frequency of drought events during the 25-year reference period calculated by SPI and SPEI. Both indices yield a similar pattern of drought events however, SPEI presents droughts of increased frequency and spatial coverage compared to SPI. The number of relatively higher frequency drought events (above 8) are found in northern Greece (parts of eastern and central Macedonia), central mainland, the Peloponnese, Evia, several islands (of the northern and central Aegean Sea and the Ionian Sea), and western Crete. Figure 3b shows the SPI and SPEI projected changes in the number of drought events (per 5 years), based on the different emission scenarios and future periods. The two indices show an overall decrease in the frequency of drought events in both future periods over the country and under both emission scenarios, apart from parts of Crete, Thessaly, western central Greece, and the Peloponnese.

It should be clarified that the plain areas of Crete and Thessaly exhibit an outstanding contribution to the country’s agricultural sector and, therefore, potential projected changes in drought frequency are of immense importance. The same holds for the areas of western and north-eastern Peloponnese.

According to SPI and SPEI, the drought events are projected to be more frequent in the near future period than in the recent past under RCP4.5 in the plain areas of Thessaly, Thrace, western-central continental Greece, central Peloponnese and eastern Crete. Overall, changes in the far future period and under RCP4.5 yield decreased frequency of drought events in central and northern parts of the country, particularly those obtained by SPEI. On the other hand, the far future projections using both indices under RCP8.5 present a reduced signal of drought frequency over Thessaly compared to RCP4.5 results. Notably, the SPEI projected changes in the far future and under RCP8.5 show the strongest signal of reduction in drought events compared to all other cases examined.

It could also be mentioned that a reduced frequency of drought events is projected over the highly populated region of Attica except for the near-future SPEI projections under RCP8.5.

The duration of drought events attains values up to 12 months/5 y for the historical period with both indices, but in the case of SPEI, the larger part of the land area is characterized from at least 6–8 months/5 y duration, as it is illustrated in Figure 4a. Our results are in agreement with the findings of [86] for the region of Thessaly and the historical drought investigation of [32] indicating severe droughts slightly increasing from north to south and from west to east.

Regarding the projected duration, under RCP4.5 and according to both indices, the drought events are projected to be longer in the near future than in the past in Thessaly (~4 months/5 y), in north-eastern Greece (Macedonia), western-central Greece and north-western Peloponnese, eastern Crete and eastern Aegean (islands). Shorter drought events are observed in Epirus, the mountainous parts of the mainland, eastern parts of central Greece, and northeastern Peloponnese. Some of these areas like western Macedonia, and Epirus show in the far future an increased duration of drought events and a reduced one in Evia.

In the near future period and under RCP8.5, the increase in drought length is more intense in the eastern parts of the country (with the Aegean islands included), with both indices and less intense in Crete. However, in the far future, the projected RCP8.5 change of increased duration of drought events is more intense and shifted to the western parts of the country (Epirus/western Greece and western Peloponnese) and Crete, with almost the same spatial patterns for the two indices. On the other hand, a notable reduction in the signal of drought duration is observed over the larger part of the central and eastern mainland with Attica included.

Concerning the severity (see Figure 5) and the duration of drought events (Figure 4), the drought events are projected according to both indices to follow, on average, the same spatial patterns, while the spatial patterns of the drought intensity (Figure 6) are less homogeneous and this remark is also made in the study of [7]. The intensity of drought events using both indices, as illustrated in Figure 6a, shows that SPI yields higher values and spatial coverage compared to SPEI. The intensity is projected to be higher and more extended with SPI than SPEI (see Figure 6b), for all periods and RCPs, and particularly for the northern and eastern parts of the country and Crete, with the additional inclusion of the western mainland during the far future under both RCPs.

The analysis of drought characteristics is also studied in terms of spatial trends for the area of Greece, for the two periods, RCPs and indices to examine their differences along with the representation of their statistical significance. Severity and duration trends, as illustrated in Figure 7 and Figure 8, show similar spatial patterns for both scenarios over Greece. In general, the projected results show both positive and negative trends, with larger areas presenting a strong positive (of which in many areas statistically significant) trend by using both indices mainly under RCP8.5 due to a combination of both warming and drying climate change signal (see [61]).

On the other hand, both indices revealed a negative trend under RPC4.5 mainly during the far future period all over the country with a statistically significant trend in some areas in the northern part of the country, western Peloponnese, and the Ionian islands. However, there are some exceptions, where statistically significant positive trends are noticed with SPEI resulting in more intense drought events, for example, in areas of Thessaly, Thrace, Chalkidiki, and Crete but at the same time these drought events are less severe and/or of shorter duration (Figure 8).

Under RPC4.5 and in the near future, statistically significant positive trends are observed with intense drought events in western mainland and eastern Macedonia, with an also statistically significant positive trend in severity and duration in western mountainous parts and western Crete.

Finally, it is also worthy to report that under RCP8.5 for both periods there is a northeast to southwest gradient towards negative trends in some areas of central-eastern Greece, eastern Peloponnese, the Aegean Islands, Crete and Thrace, without being statistically significant though. In the far future and in both emission scenarios, a decrease in the mean duration and severity is observed locally on the eastern coasts of the mainland, probably related to the extreme rainfall events that only persist there in the far future, as it has been indicated in the study of [87]. This outcome is associated with the Arctic amplification and possible connection to the weakening of mid-latitude storm tracks [88]. A profound positive statistically significant trend in changes in drought characteristics is found mainly locally over areas in Macedonia, Thrace, Thessaly, and Peloponnese using both indices and under both periods. The high spatial resolution of the simulations gives the opportunity to determine the specific areas of a rather small extent prone to drought, e.g., northern parts of the island of Rhodes under RCP4.5 in the near future (Figure 7) and the islands of Lesvos and Chios in eastern Aegean Sea under RCP8.5 in the near future (Figure 8). However, when averaged over the whole area of Greece, we proceed with the investigation of the general response to drought tendency.

In this context, boxplots illustrate the trends of the total land area of drought characteristics during each period and emission scenario (Figure 9). The drought characteristics are calculated by spatial and temporal averaging (over five years). Drought intensity, duration and severity follow similar patterns, showing positive trends using both indices in the near future period under both scenarios and only in the far future under RCP8.5. It is deduced that RCP8.5 drought characteristics with SPI-6 and SPEI-6 present a stronger positive trend than those obtained under RCP4.5. The negative projected trend in the far future period under RCP4.5 is probably related to the negative projected trend of maximum temperature and simultaneously the positive projected trend of precipitation (particularly in the west parts of the country for both variables) during the same period (not shown), leading to wetter and milder conditions.

3.2. SPI-SPEI 12

In this section, we proceed to drought investigation on the scale of 12 months that sometimes is used to define hydrological droughts [89].

As depicted in Figure 10a, historical simulations with SPI and SPEI illustrate similar spatial patterns of drought events. Moreover, the two indices agree on the projected decrease or increase in drought frequency (per 5 year) with a higher number of drought events being observed under RCP4.5 during both time periods (Figure 10b).

Drought duration for the historical period shows that SPEI covers more extended areas of longer duration all over the country than the SPI, mainly in eastern Macedonia, Thrace, Epirus, Central Greece (with Attica and Evia included), and northern Peloponnese (Figure 11a). Increases in projected drought duration under RCP4.5 affect many plain areas all over Greece, with maximum values of duration of ~8 months/5 y occurring in Macedonia and in local areas in Thessaly, northern Evia, central Greece, the Ionian and Aegean islands, Crete and the Peloponnese (Figure 11b). In the near future, only Epirus and northeast Peloponnese are the regions with reduced duration of drought events. While, in the far future, only the eastern Peloponnese, Attica, south Evia, and eastern Rhodes will experience drought events of shorter duration. These results are deduced with both indices. Under RCP8.5, the duration will be longer in some parts of the eastern country and Crete in the near future. Nevertheless, the signal will change in the far future with a longer duration in the southern and western Peloponnese, central-west continental parts and Epirus, central Macedonia and Thrace. Thus, the regions of north-east Peloponnese, northern Evia and Thessaly are projected to experience the strongest decrease in drought duration in the far future. Overall, the differences between SPI and SPEI in duration are almost negligible.

Concerning the increase in drought duration (e.g., Figure 11a) and severity (e.g., Figure 12a) for the historical period for the region of Thessaly, our results are in agreement with the findings of [86]. Furthermore, our findings agree with the historical drought investigation of [32] indicating severe droughts slightly increasing from north to south and from west to east (e.g., Figure 4a and Figure 10a).

In general, the overall area projected to be impacted by more severe drought events in the future is much larger according to SPI than the SPEI as depicted in Figure 12b. However, in some locations, the opposite signal in severity is detected between the two indices (e.g., central Macedonia, Attica) in the near future, particularly under RCP8.5. This finding concerning Attica and SPI is consistent with the results of [90] that indicated a reduction of air mass origin up to 45% originating from the cyclogenesis region of the central Mediterranean and the Adriatic Sea. Over the island of Crete and the northern and eastern parts of the country, the drought severity is projected to increase under both RCPs and future periods, but more prominent increases are found with the SPI index. Moreover, both indices indicate more notable increases in drought severity in the south-western parts of the country in the far future, under RCP8.5.

Τhe climate change signal of reduced drought conditions in several western parts of the country and over some locations in the central and northern mainland (e.g., Thrace) is associated with the increased precipitation as derived in the study of [61], during both periods under RCP4.5 and in the near future under RCP8.5. Additionally, in these areas, the increase in precipitation is strong enough to outweigh the effect of increasing temperature (and, thus, the evapotranspiration), explaining why the drought variables decrease according to SPI. Those areas will be characterized by a hot and wet future, potentially being exposed to even more weather precipitation extremes [17].

Regarding drought intensity under RCP4.5, there is a clear climate change signal for more intense drought events derived from SPI in the areas of central and eastern Macedonia, northern Evia, some locations in central Greece, and the Aegean islands and Crete, in both future periods (Figure 13). High intensity of drought events under RCP8.5 will additionally impact several locations of western Greece and Peloponnese in the far future. However, using the SPEI index, the climate signal of drought intensity becomes overall significantly weaker, locally more limited, which yields future droughts of reduced intensity under both RCPs.

Similar patterns are observed regarding the spatial trends of corresponding drought characteristics derived from the 12-month SPI and SPEI, as presented in Figure 14 and Figure 15 respectively; however, the results obtained with SPI show higher positive trends. Regarding the far future, the two scenarios show different tendencies of projected drought conditions, with larger areas of strong positive trends under RCP8.5. In particular, the projections with both indices reveal longer and more intense and severe drought events under RCP8.5. However, the results indicate, with spatially limited statistical significance, an amplified signal for Crete with less intense and severe droughts of shorter duration, in the near future and under RCP8.5.

Under RCP.4.5, projected changes of drought characteristics are milder than those under RCP8.5 for both periods, but indicate some notable remarks where some areas are prone to longer and more severe and intense droughts. In the near future and under RCP4.5, statistically significant drought characteristics over western Crete are projected to increase resulting in longer and more intense and severe events. The majority of the country tends to be exposed to more severe and longer drought events, except for Macedonia, Thrace, islands of north-eastern Aegean, and some local areas in southern Peloponnese. It should be mentioned that areas in the central mainland with a positive trend of statistical significance in drought duration and severity become more spatially extended with the SPEI index under RCP4.5 in the near future. In addition, under RCP4.5, some areas of western Greece and Thrace are prone to a positive statistically significant trend on projected changes in drought intensity in the far future.

On the other hand, a profound positive and statistically significant trend in changes in drought intensity is found using both indices mainly over the areas of Macedonia and Lesvos in the near future, under RCP8.5. Furthermore, in the far future and under RCP8.5, a positive statistically significant trend in drought intensity is obtained with both indices for some areas of central Macedonia and Thessaly. Moreover, a noteworthy positive trend of statistical significance in severity and duration is revealed with SPI only in large parts of Thessaly. In general, drought conditions are related to the local climate conditions which are characterized by low precipitation with high variability. According to the two indices, RCP8.5 drought characteristics, as illustrated in Figure 16, present a stronger positive 5-years trend than those obtained under RCP4.5, similarly as in the study of 6-month indices.

4. Discussion

The presented results based on these two drought indices showed that Greece will experience an increasingly severe climate with increasing drought severity and duration under moderate (RCP4.5) and extreme (RCP8.5) global warming scenarios in almost all parts of the country.

In general, previous works that indicate an increase in the frequency, duration, and severity of drought events, conducted for local areas of Greece under different periods and/or IPCC scenarios, are in agreement with our findings in the context of future projections; ref. [27] for Karla lake in Thessaly, ref. [91] for the eastern part of the mainland (from Thrace down to the Peloponnese) [43] for eastern Greece and northern Aegean Islands, ref. [42] for Crete, and [44,45] for Ardas and Sperchios river basins in north-eastern and central Greece. Similar reports are included on the critical review of water resources in Greece by [92]. More recently, the authors of [46] found that the number of consecutive dry days in a year will increase by 15.4 days (30%) at the end of the century for central-southern Aegean Sea and continental areas around based on an ensemble of EURO-CORDEX regional climate simulations. The authors of [93] indicated that desertification risk in the future is expected to increase in a study of future land degradation for Thessaly, using the RCA4/MPI-ESM-LR models from EURO-CORDEX. The study of Spinoni [19], based on an ensemble of 11 bias-adjusted simulations from the EURO-CORDEX datasets using a composite index (combination of SPI, SPEI, and RDI), showed that an increasing drought trend is projected to continue and grow stronger until the end of the 21st century over southern Europe for both scenarios investigated (RCP4.5 and 8.5).

The positive trends observed in drought intensity, severity, and duration on 12 months’ timescale analysis, are also consistent with the tendency (higher probability of occurrence) of increased long-range southerly flows (40%) under RCP4.5 and, hence, more heatwaves, that can result in drier conditions in the future, as reported by [89] in a study of a comparative assessment of backward trajectories in the near future and both RCPs. It also observed that the drought tendencies of the two indices revealed for some areas contradicting values. The interpretation of these cases (not only for 6-month indices but 12-month as well) is more complicated since we must consider the climate change signal of precipitation (SPI) and temperature (SPEI) and consequently evapotranspiration or both variables in these areas. In some cases, drought characteristics increase with increasing temperature and/or decreasing precipitation, while in other parts of Greece, they may remain the same or even decrease. Each region or river basin may have its own unique response to climate change. Consequently, the sensitivity to continued climate warming becomes very region-specific.

Furthermore, the results derived by both indices, enhance the importance of using their combination for studying drought projections since by excluding temperature could lead to an incomplete interpretation of the situation. Projections of drought events using SPI show more moderate/robust changes or trends than those from the SPEI or the opposite, according to the area, topography, etc. This is because an index based solely on precipitation cannot explain the full magnitude or spatial extent of drying reflected by the SPEI [11]. In fact, the authors of [94] pointed out that as the temperature increases in the future, the evapotranspiration increases also (because of the greater moisture demand by the atmosphere), which is possible to result in even more profound impact than precipitation deficits in a warmer world. Hence, the use of SPEI is imperative in the investigation of climate change impacts on drought. On the other hand, droughts can also be caused by changes in rainfall characteristics in terms of seasonality, dry spells and precipitation intensity. That causes of drying depend on the season and are probably linked to the dominant seasonal precipitation formation mechanisms in winter (synoptic processes, NAO anomalies) and summer (local phenomena due to convection), as suggested by Brogli [95]. Thus, it would also be of great importance to investigate with regard to observed climate change, in general, the possibility of positive trends particularly occurred in case of wet periods in areas prone to drought.

The impact of the global warming and reduced precipitation on the country will become more evident in the far future, as the extreme maximum temperature will become the most significant hazard, particularly under RCP8.5 [87]. This fact can lead to a remarkable increase in evaporative demand resulting in a shift toward more arid climates. In this context and considering that the projected climate change is likely to result in more frequent and severe weather-related extremes [1,96], it is of the highest importance to investigate over which areas meteorological or agricultural droughts are likely to become more frequent, more intense and/or severe.

The present study indicated that Greece will face relatively severe drought conditions in the upcoming years. Owing to the high spatial resolution used, substantial differences in drought characteristics have been found in future projections between areas, highly varying in temporal and spatial terms under the two emission scenarios. Moreover, our findings are comparable with those in other studies conducted for the Mediterranean region [19,97,98]. Overall, our results point towards a warmer and drier future, particularly under RCP8.5 in agreement with the latest IPCC report on the Mediterranean region by [1]. It is also observed that both SPI and SPEI followed similar patterns in what concerns the spatial distribution of drought severity, intensity, and duration. This fact declares an agreement at local level in spatiotemporal resolution; however, a weaker signal is found in the case of SPEI that in some cases minimizes the effect of drought characteristics, particularly at the 12-month timescale.

The results of this study could be used for estimating the impacts of future drought events, and, consequently, for the development of adequate mitigation and adaptation strategies for water management under climate change in Greece. In addition, the importance of these results lies in the calculation of two drought indices to estimate the projected changes on drought characteristics in high resolution that takes into consideration the complex topography of Greece, and how the results can differentiate based on the parameter that is probably most dominant in future change between temperature (potential evapotranspiration) and precipitation in the future. However, the complex topography of the domain or parts of it may imply the requirement of further impact drought assessment studies by downscaling the climate data to even higher than a 5 km resolution. It was found that the drought conditions will be more severe in the lowland areas (plain areas), such as Thessaly, Crete, etc., where all the agricultural activity takes place. Sordo-Ward [99], who studied past and future SPEI droughts in the La Plata Basin, suggested the need for a potential relocation of certain crops from the exposed vulnerable regions towards cooler and wetter regions. This conclusion is reinforced by the increased statistical significance calculated in those areas. The results also point out that special attention needs to be given to avoid water scarcity problems that will have a great impact on the local population and agricultural activities.

5. Conclusions

This work constitutes a first attempt to analyze the projections of drought characteristics (intensity, duration, and severity) using two drought indices over the whole country with the spatial resolution of 5 km, which, to the best of our knowledge, is the highest used so far (higher-resolution data are available only for very few basins). Our analysis was based on previously and extensively validated EC-EARTH GCM and WRF RCM model simulations, obtained with dynamical downscaling for the area of Greece under RCP4.5 and RCP8.5. The combined use of the SPEI and SPI indices allows an in-depth investigation of the critical role of temperature when dealing with drought conditions.

The results using both indices at a 12-month timescale for the period 2075–2099 and under RCP8.5 have shown a profound increase in the mean duration of drought events along with increased severity for the areas of Crete, central Aegean islands (Cyclades), southern Peloponnese, western continental Greece, Attica, central Macedonia, and Thrace. A positive statistically significant trend in drought intensity is also observed with both indices for some areas of central Macedonia and Thessaly Furthermore, the drought events with the 12-month analysis are projected to be more frequent locally in the central to northern parts of the country, under RCP4.5 than RCP8.5, in both future periods. The projected higher frequency and longer duration mean the stabilization in drier conditions. In this context, the 12-month indices represent more suitably the prolonged duration of drought events.

In addition, the indices calculated at a 6-month timescale yield that a more extended area is affected by drought conditions and more severe and prolonged drought events are expected under both scenarios (particularly, in areas of central and eastern part of the country in the near future, and areas of the western parts in the far future). However, the central and eastern mainland will experience a notable reduction in the signal of drought duration along with decreased frequency of drought events in the far future under RCP8.5. Thessaly, a region of high agricultural interest, will experience more frequent and longer drought events in the near future under RCP4.5. In addition, Crete (and mainly the eastern part) is projected to experience increasingly more prolonged and severe drought events under both scenarios and periods. On the contrary, it could be also mentioned that a reduced frequency of drought events is projected over the highly populated region of Attica, except for the near-future SPEI projections under RCP8.5. Moreover, it is deduced that RCP8.5 drought characteristics with both indices present a stronger positive trend than those obtained under RCP4.5.

In summary, the investigation of drought characteristics focused on projected changes in temperature and precipitation in Greece, which can provide a comprehensive attribution of drought events. It is deduced that the study of drought events is not a straightforward task for areas of complex topography that present climatic variations and the corresponding spatial and temporal characteristics may depend on the choice of the index. Some limitations are related to the fact only one GCM and RCM have been used in this study and no bias correction was applied to improve the climate projections regarding the examined variables, given the lack of consistent gridded observational datasets required for such regions of complex topography and climate variation. Future work in climate change studies could also consider trend analysis based on differentiating between wet and dry seasons, the analysis of moisture transport impacted selected regions during extreme conditions, or studies based on analysis of other factors influenced by drought, such as the characteristics of soils, hydrology, production of different types of crops, etc. Moreover, further research could include the comparison to other drought indices or hydrological parameters and soil parameters in the most affected areas of agricultural or tourism interest (e.g., Thessaly, Crete, and other Islands) as derived from our study, to investigate extensively water resource availability, agricultural production, and fire risk assessment. In general, the future changes in drought characteristics are expected to have a significant impact on the country’s ecosystems, as well as on a number of human activity sectors (e.g., health, agriculture, tourism, forest fire risk, loss of biodiversity, etc.). In any case, the produced high resolution projected changes of the present study can serve as a firm and reliable basis for climate change impact assessments based on drought characteristics for the area of Greece.