1 Introduction

Energy Sector shapes everyone’s life affecting every single community, business, and economic sector in all parts of the world. It contributes about three quarters of global greenhouse gas emissions (Ritchie et al. 2020; World Research Institute https://www.wri.org/energy, accessed 2023/11/08) but, at the same time, it is significantly impacted directly or indirectly by extreme weather events, due to climate change, which can have very serious consequences in case of power outages with cascading effects for telecommunications, transports, water treatment facilities, health services and food supplies, as all these sectors depend on the energy system reliability. For example, in January 2022, a historic heatwave in Buenos Aires caused a massive power outage, affecting around 700,000 people. In November 2020, a freezing rain event coated power lines in the Far East of the Russian Federation, leaving hundreds of thousands of homes without electricity for several days (WMO 2022). An emblematic case regards the severe storm that hit the Australian state of Victoria in June 2021 and had adverse impacts across the state, causing widespread damage to property, felling trees, downing powerlines, impassable roads, and damaging critical infrastructure. About this event CERRE (2023) reports: “Electric power outages were extensive: at peak almost 300,000 customers had no power and, three days later, almost a quarter of those were still off supply. In October 2021, another storm swept over the state with even larger immediate impact: at peak, over half a million customers lost power. In both cases thousands of customers were still off the electric grid a week after the storms”. The risk may become even greater in an increasingly interconnected world. At the beginning of 2021, Europe risked a blackout on an unprecedented scale: a power failure on the electricity grid in Romania could have had serious repercussions on the old continent, with potentially dramatic effects. The danger was only averted thanks to Italy and France, which switched off the energy-intensive plants, such as steel mills and cement factories, which are defined as interruptible activities in the event of an emergency (ENTSO.E 2021). These are just a few examples showing how weather extremes pose significant risks to the energy sector, affecting the conventional and renewable energy production, the physical security of current and future energy infrastructure, and the adequate supply of electricity in relation to energy demand.

Renewable energy resources are directly linked to meteorological variables such as precipitation, temperature, radiation, and wind, and may be susceptible to future climate change, as extensively addressed by Solaun and Cerdà (2019) who report quantitative estimates of climate change impacts on solar, wind, hydro and other renewable energy production at regional and global scale by 2100. Moreover, all energy infrastructure may suffer during their lifetime a plurality of weather-related threats that might cause cascading effects in terms of risk and vulnerability above all if different hazards are overlapping in spatial and/or time. In particular, dry and hot periods have a significant impact not only on hydroelectric production but also on conventional electricity generation and demand: under conditions of water shortage and increasing temperatures the thermal power plants have a reduced cooling efficiency and, in prolonged dry conditions, it is often necessary to impose a partial working regime or even a shutdown of the power plants to ensure the minimum vital flow in rivers, while power demand for civil and industrial cooling is pushing up. Nuclear plants are also heavily dependent on water to produce electricity.

World Research Institute (Astesiano et al. 2018) states that if we overlaid the existing power plant infrastructure with the areas of current water scarcity, we found that 47% of the world’s thermal power plant capacity — mostly coal, natural gas and nuclear — and 11% of hydroelectric capacity are located in highly water-stressed areas. Moreover, conventional power plants are also often located in low-lying coastal areas and hence are vulnerable to sea-level rise and weather-related flooding (WMO 2022). This overview is basically confirmed by the International Energy Agency which also adds that about a quarter of global power grids are currently exposed to a high risk of destructive cyclonic winds (IEA 2021).

Regarding the future, IEA reports that the frequency of extreme heat events would double by 2050 compared to today and they would be around 120% more intense, affecting the performance not only of power plants, but also of the electric grids. These statements are based on extensive analyses of climate projections developed by both global and regional climate models, carried out by considering different anthropogenic radiative forcings described by the so-called Representative Concentration Pathways (RCPs) (Detlef et al. 2011), each providing a plausible description of how the future may evolve with respect to a number of variables such as demographic, socio-economic and technological changes, energy and land use, greenhouse gas emissions and air pollutants. Among the RCPs, RCP8.5 describes a very high emission pathway leading to radiative forcing RF = 8.5 W/m2 at the end of the Century; RCP4.5 is an intermediate mitigation scenario (RF = 4.5 W/m2), RCP 2.6 represents a mitigation scenario leading to a very low forcing level (RF = 2.6 W/m2).

Climate change effects could be particularly relevant for the Mediterranean Region, identified as one of the most prominent hot-spots in the future climate projections (Giorgi et al. 2008): small variations in the general circulation, such as interactions between a sub-tropical regime and mid-latitude synoptic activity, may shift midlatitude storm tracks or sub-tropical high-pressure cells, resulting in changes in the Mediterranean climate variability at multiple timescales and strong seasonal variability of precipitation in many areas, expected to increase with the magnitude of the anthropogenic radiative forcing. Atmospheric variations are also projected on synoptic scale with implications for extreme weather conditions: analysing global CMIP6 models Trevisiol et al. (2022) highlight some changes in the deployment and activity of Rossby Wave Packets and Atmospheric Blocking Events in the Northern Hemisphere at long term. These results deserve some attention as these atmospheric structures are often associated to serious meteorological events, such as intense precipitations with floods, cold/hot spells, and droughts. Europe will be particularly affected by these variations in winter and spring, above all in the pessimistic Shared-Socio-economic Pathway with intense global warming SSP5-8.5. Actually, in the Mediterranean Region mean temperature is expected to increase at a rate larger than the global mean temperature, particularly in summer, and total precipitation will decrease at a rate of about − 4% per degree of global warming (Lionello and Scarascia, 2018). Indeed, a trend towards an increase of warm temperature extremes and drier conditions over Mediterranean Region, especially in the warm season and over the southern areas, is stated with high confidence in the First Mediterranean Assessment Report (MedECC 2020). Lionello and Scarascia (2020) investigated climate extremes on the basis of an extensive CMIP5 global model ensemble and highlight an exacerbation of the existent contrast between Northern and Southern Mediterranean Region in the hydrological regime, as well as they stress a dramatic increase of temperature extremes over the whole region in a global warming scenario of 4 K. Similarly, Molina et al. (2020) indicate a large increase by the end of the Century in both intensity and length of heat waves from the analysis of Euro-CORDEX regional model ensemble for the Mediterranean basin, highlighting that exceptional heat waves observed early on the century could become normal by the end of this period, with heat waves more intense and last longer under the RCP8.5 than in the RCP4.5 scenario, due to higher increase of temperatures.

The Italian Peninsula, located at the centre of the Mediterranean Basin, could also be seriously affected by climate changes in the coming decades, as described in a number of studies (Spano et al. 2020; MATTM 2022; MIMS 2022), which deal with climate change scenarios concerning the urban environment, geo-hydrological risk, water resources, agriculture, forest fires, and infrastructure, based on a sub-set of Euro-CORDEX models. For meso-scale applications, relevant for national (or supra-national) planning, Mysiak et al. (2018) assess a climate risk in Italy through an index-based methodological approach carried out by computing a wide range of climate extreme indices defined by the Expert Team on Sector-specific Climate Indices (ETCCDI, WMO 2009) inferred from an ensemble of bias-corrected Euro-CORDEX models under the RCP4.5 hypothesis. Notwithstanding the model differences, patterns in climate hazards underline the decline of water availability and the increase of incidence of extreme drought episodes in the Southern Italy. Faggian (2021) confirms the exacerbation of precipitation-related hazards in Italy by computing some ETCCDI indices exploiting an ensemble of Euro-CORDEX models. In fact, despite some uncertainties in the climate future scenarios due to the spread among models, she highlights a likely increase of droughts, with the maximum length of dry spell expected to grow until 10 ÷ 15 days during the summer in Southern Italy in RCP4.5 and RCP8.5. Extreme precipitations are likely to increase, too, with different intensities depending on both region and season. Instead, no signals are found out under RCP2.6 configuration.

Focusing on the climate-related impacts in the hypothesis of partial mitigation (RCP4.5), European critical infrastructures could experience a 10-fold increase in damages by the end of the century. Damages for the European Energy Sector are projected to rise 16-fold by 2100 (Forzieri et al. 2018). Concerning Italian electricity infrastructure, Terna (the Italian Transmission Operator) states that in Italy the greatest risks are related to phenomena of wet-snow, due to their loads on overhead lines, strong wind episodes because of their direct or indirect damages, e.g. to pylons or conductors, as well as to hydrogeological disruptions following exceptionally heavy rainfalls. In addition to an increase in energy demand for cooling and an exacerbation of critical situations for energy production (Faggian and Decimi 2019), the progressive intensification of heat waves will cause an increasing risk of wildfires (Faggian 2018), with serious threats to the preservation of grid infrastructure and the security of energy supply (Terna 2020).

One of the priorities of the EU Adaptation Strategy (European Commission 2013, 2021) was to promote a better-informed decision-making by addressing existing gaps in the knowledge about climate change impacts and adaptation. On 24 February 2021, the European Commission set out the Communication “Forging a climate-resilient Europe - the new EU Strategy on Adaptation to Climate Change”, in which outlines a long-term vision for the EU to become a climate-resilient society, fully adapted to the unavoidable impacts of climate change by 2050. Among the objectives to be pursued to cope with the new normal of extreme events, Italy established the goal of strengthening the security of energy supply in the National Adaptation Strategy to Climate Change (Castellari et al. 2014). The National Regulatory Authority for Energy, Networks and Environment (ARERA 2017) delivered guidelines for the presentation of resilience plans for the increase of resilience of power system, and issued resolutions 31/2018/R/eel and 668/2018/R/EEL in which it established that electric stakeholders (transmission and distribution system operators, TSO and DSO) must have a resilience plan, including the actions to be implemented to contain the risk of disconnection in the face of weather threats.

The IPCC, in the Special Report SREX (2012), has defined the key components that lead to climate-risk events: hazard, exposure and vulnerability. The exposure of assets of interest and their degree of vulnerability are influenced by socio-economic processes, while hazards may be exacerbated by climate change. So, the aim of this study is to investigate the climate change hazards critic for the energy sector, considering their evolution in intensity and frequency in Italy in the coming decades. This was addressed by developing some future projections for individual hazards described through specific ETCCDI climate extreme indices, based on the physical driver of climate-related hazards, used as proxies. As mentioned above, energy infrastructure may suffer during their lifetime a plurality of weather-related threats, therefore it becomes necessary to describe the complex interactions among multiple hazards. In general, the need to cope with multiple climate events is internationally recognised as a fundamental step towards the development of resilient societies (UNISDR 2015). More recently, the IPCC in its Sixth Assessment Report (2021) deals with the enhancement of the risk in extreme compound events, that are defined as the combination of multiple drivers and/or hazards that contribute to societal or environmental risk. Several authors (among others Forzieri et al. 2016; Mysiak et al. 2018; Terzi et al. 2019) have addressed the issue of multi-risk assessment, in which the nature and combination of different interactions (e.g. concurrent space-time events, cascading events, increasing/decreasing probability) have been studied. While referring to the cited studies, risk assessment is beyond the scope of this study. Here a multi-hazard approach was implemented to better steer adaptation efforts with regard to the resilience of the energetic system, considering the different interacting hazards that generally can lead to greater impact on electric system than the sum of the single hazard effects.

The multi-hazard assessment was developed by computing some ad-hoc indices defined to investigate: the regions affected by water scarcity and thermal stress because of their impacts on energy supply; the areas hit by windstorms, that can directly or indirectly damage power lines and stations, e.g. due to falling trees or landslides; the zones expected to be affected by a plurality of hazards (not necessarily simultaneous) and, for this reason, of particular concern to guarantee the security of the energy system. In this framework both spatially and spatio-temporally compounding events are addressed.

The analysis was carried out by considering some climate scenarios at short-, medium- and long-term, focused on the Italian territory, inferred from a sub-set of Euro-CORDEX models under the three emission pathways RCP8.5, RCP4.5, and RCP2.6. The study is presented as follows: datasets and methodology are presented in Sect. 2; discussions of the results are drawn in Sect. 3; a short summary of the results and the main conclusions are reported in Sect. 4.

2 Data and methods

2.1 Datasets and variables

The extreme climate scenarios in Italy presented here for the 21st century were elaborated by analysing the daily maximum and minimum temperatures (Tx and Tn), cumulative precipitation (R), and wind speed (W) provided by 12 Euro-CORDEX models at 0.11° spatial resolution (⁓12 km) (https://www.euro-cordex.net; download data 30th June 2021) (Table 1). From the Euro-CORDEX collection available on the Earth System Grid Federation platform at the date of download, models that provided all the daily aforementioned variables and, at least, under both RCP8.5 and RCP4.5 configurations were selected. The resulting subset can be considered a fairly balanced sample to conduct a robust statistical analysis, because it consists of a combination of 5 different global circulation models with 3 different regional climate models. In the validation phase (paragraph 2.3) the model id.7 has been excluded from the ensemble analysis due to its poor performance in reproducing past precipitation climatology. One member per model was considered.

In addition, the reanalysis dataset MERIDA at 7 km spatial resolution (http://merida.rse-web.it; download data 30th June 2021; Bonanno et al. 2019), spanning the period 1986–2019, was considered as reference dataset.

Table 1 List of Euro-CORDEX simulations considered in this study for RCP (x: used; -: discarded). The name of simulations identifies the research institute, the global circulation model and the regional climate model. For each model simulations, only one member has been used

2.2 Climate extreme indices

Referring to the definitions in Table 2, the SU, CDD, HWDI, and R99PTOT indices were calculated to investigate hot-days, dry conditions, heatwaves, and extreme precipitations scenarios, respectively. Moreover, three new multi-hazard indices for a Resilient and Sustainable Energy system (RSE) were defined by combining the above-mentioned ETCCDI indices and two other single-hazard indices W75P and W99P, considered independent of each other. These latter two characterize windy days ranging from moderate (W75P) to strong or very strong (W99P) and were computed referring to analogous ETCCDI definitions of R75P and R99P for rain. The three RSE indices are: Drought days index (DRO) to investigate the concurrence of high temperatures (SU) and water shortage; Storm days index (STO) to characterize very disruptive compound events with strong winds (W75) and extreme precipitations (RR99); Multi-hazards index (MUHZ) to assess the occurrence (even if not simultaneous) of critical situations selected by considering values of R99PTOT, W99P, CDD and HWDI exceeding some given thresholds. Thus, DRO and STO consider the spatio-temporal concurrence of two threats, whereas MUHZ considers the spatial concurrence of multiple threats.

Table 2 Climate extreme indices used in this study, distinguishing between ETCCDI and RSE indices

2.3 Methodology

The scenarios were computed for Italy on both an annual (YEA) and seasonal scale (winter = DJF, spring = MAM, summer = JJA, autumn = SON), since the Mediterranean Region is characterized by a typical seasonal cycle. Because of the uncertainty regarding future anthropogenic forcing, the climate projections were elaborated by considering model simulations realized under the three different emission pathways RCP8.5, RCP4.5, and RCP2.6 (Table 1).

The 30-year period 1971–2000 was considered as the baseline scenario (REF) and climate change was analysed for three future scenarios at short- (2021–2050), medium- (2041–2070) and long-term (2071–2100) in term of anomalies against REF. We choose to overlap the short- and medium- timeframes by ten years in order to better analyse the 2050s horizon which is of particular interest for stakeholders’ investments in electricity infrastructure. First a bilinear remapping of model data was done at 7 km grid domain referring to MERIDA spatial resolution. Then, an evaluation of the performances of each model in reconstructing the current climatology was carried out by investigating the systematic errors (bias) with respect to observations (MERIDA) in the seasonal values of temperature, precipitation, and wind. This analysis was carried out with the aim of filtering out models with some shortcomings in reconstructing the past climate and, therefore, selecting only the most reliable numerical simulations to generate future multi-model projections. Indeed, the model with id = 7 in Table 1 was excluded from the set of 12 models, because it describes a rainfall regime over Italy significantly different from the typical seasonal values (ISPRA 2022a).

As recommended by Rojas et al. (2011) bias must be removed in impact studies. Therefore, the “Equidistant Quantile Mapping” technique (EQM) (Sachindra et al. 2013) was applied to models’ temperatures to estimate quantile correction weights in the training period 1986–2005 referring to MERIDA data. Tn was adjusted by using quantile correction weights calculated at seasonal scale and the same was done for the Diurnal Temperature Range (DTR). Then, Tx was computed according to the expression: Tx = Tn + DTR, to guarantee coherence between minimum and maximum temperatures. Once the effectiveness of the bias-correction was tested by analysing the adjusted results in the verification period 2006–2019, as well as in the training period 1986–2005, EQM was applied to correct model data up to 2100. About precipitation and wind, we chose to use raw data in order to maintain the temporal correlation among the different meteorological variables, even if precipitation bias may exceed 100% the reanalysis data in cold months (Faggian 2021) and wind bias may be not negligible over terrain with complex orography (not shown).

Climate scenarios were elaborated by computing the climate extreme indices (Table 2) for each model, mainly considering the mean values (ensemble or multi-model mean), but the median and standard deviation were also calculated to investigate the robustness of the results over the whole analysis domain. To study future scenarios the climate anomalies were considered, and the statistical significance of climate changes was tested by applying the rank-sum Wilcoxon test at 95% of confidence level. Mann - Kendall test was also applied for each grid point of the Italian domain to investigate if there is or not significant monotonic trends.

3 Discussion of the results

The results are presented in the Figs. 1, where the maps show the current climatology and the spatial and temporal evolution of the above-mentioned indices in the same order as listed in Table 2. Future scenarios are drawn for each emission pathway RCP2.6, RCP4.5, and RCP8.5 (by column), at short-, medium-, and long-term (by row) on an annual and/or seasonal scale.

Fig. 1
figure 1

SU ensemble mean scenario in REF (1971–2000) at annual and seasonal scale from left to right

In accordance with the reference climatology (ISPRA 2022a), SU values (Fig. 1) describe the occurrence of hot days close to 100% in the summer season mostly in Po Valley, coastal sites, and islands. About 30 ÷ 40 SU days are estimated for Central-Southern Italy in autumn, whereas lower values characterize the whole territory in spring. The DJF scenarios are omitted because SU events are absent in winter, as one would expect. Regarding future projections, a significant overall increase is inferred from the short to the long term. Maximum anomalies are expected in summer over Apennine, Alpine valleys, Prealpine areas with values up to 10 days in RCP2.6 (Fig. 2a), more than 25 days in RCP4.5 (Fig. 2b), and even more than 40 days in RCP8.5 (Fig. 2c). In the other two seasons SU is projected to rise especially over coasts, main Islands, and Po Valley. The highest alpine areas, so far spared by hot days, are likely to be affected by SU occurrences starting in the medium-term (long-term) in RCP8.5 (RCP.4.5). Mann-Kendall test, applied to both the ensemble mean and the individual models, confirms the significance of such trends in all RCPs (not-shown).

Fig. 2
figure 2

SU ensemble mean anomalies at seasonal scale (by column) under RCP2.6 (a), RCP4.5 (b) and RCP8.5 (c) hypotheses in the future scenarios 2021–2050 (first row), 2041–2070 (second row) and 2071–2100 (third row) respect to REF (Fig. 1). The anomalies are depicted with a grey colour if they are not significant according to the Wilcoxon test

The patterns of shortage of precipitations in Italy, reconstructed through the index CDD (Fig. 3), is consistent with the current climate: the longest dry periods are found for summer, especially over low-medium orographic regions of the Southern Italy and the islands with values up to 70 days, whereas CDD values less than 30 days characterize the other seasons. These results agree with the climatology inferred by Lionello and Scarascia (2020) from the reanalyses NCEP, ERA40 and ERA-Interim data. In fact, depending on the dataset, their estimated annual CDD values vary in the range 38 ÷ 42 days in the northern Mediterranean region, and 117 ÷ 139 days in the southern Mediterranean region. Obviously, the areas characterized by the longest dry periods are the same affected by few events during the year (10 ÷ 16 N events) because of the long duration of the dry spell (Fig. S1). On the other hand, a slightly higher number of CDD events (~ 18 N) are found over the western Po Valley (Fig. S1), having no CDD periods longer than 25 ÷ 30 days (Fig. 3). Looking at the anomaly maps, in RCP2.6 a shortening of CDD is inferred in summer, whereas there are no significant and valuable trends in the other seasons (Fig. 4a). Instead, a significant lengthening of CDD is expected in the second half of the century up to 5 days in RCP4.5 (Fig. 4b), 15 days in RCP8.5 (Fig. 4c), especially over southern and Tyrrhenian regions, with an exception in winter over Alpine Region due to an increase of precipitation expected in cold months over this area (Faggian 2021). Albeit a heterogeneous behaviour among models, Mann-Kendall test confirm these results (not shown). It is worth noting that the increase of the climate contrast between North and South is also stressed by Lionello and Scarascia (2020), even if they found out a rate of CDD changes more intense, explainable because generally the Euro-CORDEX models project colder and wetter variations at the end of the 21st century over Europe than their driving CMIP5 models (Boé et al. 2020).

Fig. 3
figure 3

CDD ensemble mean scenario in REF (1971–2000) at annual and seasonal scale from left to right

Fig. 4
figure 4

CDD ensemble mean anomalies at seasonal scale (by column) under RCP2.6 (a), RCP4.5 (b) and RCP8.5 (c) hypotheses in the future scenarios 2021–2050 (first row), 2041–2070 (second row) and 2071–2100 (third row) respect to REF (Fig. 3). The anomalies are depicted with a grey colour if they are not significant according to the Wilcoxon test

R99PTOT estimates the annual percentage amount of precipitation that occurs in wet days with extreme precipitation compared to the 99th percentile (RR99) occurred in the REF period, visualized in the “HISTORICAL” panel of Fig. 5 (top left). Consistent with the current climatology, the most wet areas are the Alpine and Prealpine regions and Liguria with values ranging 40 ÷ 70 mm/d. In the future (RCPs maps in Fig. 5), some R99PTOT positive variations are likely to affect especially the eastern Italian coasts and secondarily Po Valley. Moreover, extreme precipitations are also expected to become more frequent over Italy, reaching values up to 15% at the end of the 21st century, with similar patterns as R99PTOT, as a result of estimating R99P (not shown). Considering that CDD scenarios depict more prolonged dry spells in the future, these last outcomes point out that wet days will decrease but extreme precipitations will be more frequent and more destructive, over the above-mentioned areas.

Fig. 5
figure 5

RR99 [mm/day] ensemble mean scenario in REF (HISTORICAL panel on the top-left) and R99PTOT [%] ensemble mean scenarios under RCP2.6, RCP4.5 and RCP8.5 hypotheses (by column) in the future 2021–2050 (first row), 2041–2070 (second row) and 2071–2100 (third row) (panels on the right)

HWDI identifies exceptionally long and hot periods, also known as heatwaves. In REF scenario (“HISTORICAL” panel in Fig. 6), Italy is characterised by very few episodes (≤ 3). In the future (RCPs panels in Fig. 6), a significant increase of HWDI is projected in each emission pathway, more pronounced in RCP4.5 and RCP8.5. An exacerbation of heatwaves is evident mainly in the Alps and Apennines in RCP8.5, with HWDI reaching 8 ÷ 9 episodes by 2100.

Fig. 6
figure 6

HWDI ensemble mean scenario in REF (HISTORICAL panel on the top-left) and anomalies under RCP2.6, RCP4.5 and RCP8.5 hypotheses (by column) in the future scenarios 2021–2050 (first row), 2041–2070 (second row) and 2071–2100 (third row) (panels on the right). The anomalies are depicted with a grey colour if they are not significant according to the Wilcoxon test

According to its definition (Table 2), DRO selects and counts dry days with very high temperatures. The reference scenarios (Fig. 7) describe reasonably the Mediterranean seasonal cycle, with droughts occurring mainly in summer and affecting especially the southern Po Valley, the coastal areas, and the major islands (up to 80 days), and without events in spring and autumn in the mountainous reliefs, and everywhere in winter (not shown). A significant overall increase of DRO is already expected at short-term in RCP4.5 and RCP8.5, mainly over Apennines, western Po Valley, and the major islands (Fig. 8), consistently with SU projections. This increase is particularly strong in summer with positive anomalies somewhere up to 40 days, but the trend is not to be neglected even in spring and autumn (especially in RCP8.5 configuration) since increases of 25 ÷ 40 DRO days are projected for most of Italy in these two seasons as well. It is worth noting that, although the DRO historical patterns are analogues in spring and autumn (Fig. 7), their evolutions are expected to differentiate along the Century: drought summer conditions are likely to extend to autumn already at medium-term, whereas they will heavily affect spring at long-term. The exacerbation of droughts is cause of concern because they will promote the risk of wildfires. Moreover, during vegetative season (spring and summer), will lead to increased water stress as the energy system competes with the agricultural sector. Since not only the hot days (SU), but also the number of hot days without precipitations are likely to increase, the critical situations for electric infrastructure mentioned in Sect. 1 will pose serious risks to the management of the national energy system in the decades to come. In RCP2.6 hypothesis DRO is expected to increase in spring and autumn, too, whereas it is likely to decrease in summer, mainly at low altitudes (Po Valley and coastal areas) from the second part of the Century, in agreement with summer CDD trends in this RCP (Fig. 4).

Fig. 7
figure 7

DRO ensemble mean scenario in REF (1971–2000) at annual and seasonal scale from left to right

Fig. 8
figure 8

DRO ensemble mean anomalies at seasonal scale (by column) under RCP2.6 (a), RCP4.5 (b) and RCP8.5 (c) hypotheses in the future scenarios 2021–2050 (first row), 2041–2070 (second row) and 2071–2100 (third row) respect to REF (Fig. 7). The anomalies are depicted with a grey colour if they are not significant according to the Wilcoxon test

STO identifies strong windstorms by computing the number of days per year, in which strong winds (W75P) and extreme precipitations (R3rm > RR993rm) are concomitant (Table 2). The precipitation threshold has been evaluated as running mean over 3 days to intercept flooding situations. After analysing the W75-90-95-99P scenarios, similar trends were found out over Italy along the Century (for an example see Fig S2 for W75P scenarios), then the wind threshold has been fixed at the 75th percentile in order to have a robust statistic in combing the strong winds and extreme precipitations and to obtain a numerical reconstruction consistently with historical observations. As shown by Fig. 9, in the reference period STO reaches its maximum values in Central-Northern Italy, with the biggest contributions from spring and autumn meteorological conditions, and identifies Italian areas prone to floods with strong winds, such as the floods occurred in Piedmont in 1994 (Grazzini et al. 2020a), and in Liguria-Tuscany in 2011 (Grazzini et al. 2020a), and the VAIA storm happened over North-Eastern Italy in 2018 (Giovannini et al. 2021). The general increase of these phenomena in the recent past is well represented by the STO scenarios computed comparing MERIDA results between the periods 1986–2005 and 2006–2019 (first and second row of Fig. S3 respectively), where the contributions from winter weather conditions and the summer convective thunderstorms in the Alps are highlighted, too. STO values are projected to increase significantly in the Century, threatening the whole Italy in all RCPs (Fig. 10). Po Valley and upper Adriatic are the regions at greatest risk, being affected by a significant increase of about 3 ÷ 4 days per year, respect with the reference period, regardless of RCP, except for the timing of reaching the maximum increase.

Fig. 9
figure 9

STO ensemble mean scenario in REF (1971–2000) at annual and seasonal scale from left to right

Fig. 10
figure 10

STO ensemble mean anomalies at annual scale under RCP2.6, RCP4.5 and RCP8.5 hypotheses (by column) in the future scenarios 2021–2050 (first row), 2041–2070 (second row) and 2071–2100 (third row) respect to REF (Fig. 9 panel 1971–2000 YEA). The anomalies are depicted with a grey colour if they are not significant according to the Wilcoxon test

To take into account scenarios with a plurality of hazards, the MUHZ index was defined. For each time frame and emission pathway, MUHZ considers the annual occurrence of at least one of four hazards whose intensities are tuned by the following threshold values: R99PTOT ≥ 33%, W99P ≥ 3%, CDD ≥ 30 days, and HWDI ≥ 5 N. The thresholds were defined through a sensitivity analysis, finding a trade-off between a threshold value high enough to select on a yearly scale particular severe conditions but low enough to intercept some climatic signals. For each index, starting with typical values on the Italian Peninsula in REF, threshold values were increased and set when interesting variations over time were found in the spatial pattern of areas characterised by threshold exceedances. Such an example in RCP8.5 hypothesis, the areas expected to be affected distinctly by each hazard with the above-mentioned thresholds are depicted in Fig. S4, showing an expansion over time.

In addition, these thresholds are relevant to Energy system and to the population health. In fact: CDD = 30 days identifies the typical REF values in summer, i.e. the season of fire danger and water shortage (Fig. 3); R99PTOT = 33% characterises large area of Italy with heavy precipitations already in the short term (Fig. 5); HWDI = 5 N corresponds roughly to the values reached by 2021–2050, and it is worth noting that already the current 2–3 heatwaves per year in JJA can stress the electric grid and cause local blackouts to the point of activating the national emergency plan for electrical system safety (Terna 2023), which involves the detachment of the electric loads to prevent nationwide blackouts; W99P = 3% was fixed slightly higher than the REF values because of the poor variations projected in the anemological regime,

This approach is a simpler alternative to that applied by Forzieri et al. (2016) based mainly to the extreme value analysis and peak over thresholds technique, which, however, still requires a definition of thresholds with a certain degree of subjectivity. Moreover, extreme value analysis technique can be very sensitive to the sample (corresponding to the selected timeframe), and uncommon and more advanced applications can account properly the non-stationarity of the climate change in 30-year time frame.

The Italian territory was characterised by MUHZ values ranging from 0 (not hazards) to 4 (all hazards) indicating how many hazards impact on each grid point at annual scale in the reference period and in future projections (HISTORICAL and RCPs panels in Fig. 11 respectively). In the recent past Italy was affected very little by the just-mentioned severe weather conditions (0 or 1 hazard at most), but these threats will become more likely in the coming decades throughout Italy in all three RCPs, as MUHZ already rises noticeably in the short term even in RCP2.6. Po Valley, Adriatic coastal areas, Apulia, and central Tyrrhenian regions are found out as the most exposed areas (reaching the maximum value 4 in RCP4.5 and RCP8.5, and 3 in RCP2.6), but no region is spared in the years to come. These results are important to support the planning and design phase of new energy infrastructures (lines or plants) or the renovation of the old ones, because in the future they may be affected by a number and type of hazards that they are not experiencing now.

Fig. 11
figure 11

MUHZ ensemble mean scenario in REF (HISTORICAL panel on the top-left) and under RCP2.6, RCP4.5 and RCP8.5 hypotheses (by column) in the future 2021–2050 (first row), 2041–2070 (second row) and 2071–2100 (third row) (panels on the right)

4 Conclusions

Increasing in frequency and/or intensity of climate extremes is endangering social, environmental, and economic life. Focusing on the impacts of climate changes on the strategic energy sector, an analysis of the climate hazards for plants, transmission and distribution lines, as well as electricity demand, has been carried out by studying the future scenarios of extreme climate indices related to droughts, storms, floods, heavy precipitations and strong winds. In fact, these events can directly or indirectly damage energetic infrastructure causing electric supply disruptions or hindering the producibility of conventional and renewable sources. In particular, heat waves and droughts can affect the management of the infrastructure, because water scarcity and thermal stress reduce the energy production and the safety of transmission and distribution lines, in concurrence a surplus of cooling demand in both the industrial and civil sectors. Italy, located in the “Mediterranean hot-spot”, is expected to face major hazards from a changing climate over the coming decades and, consequently, the Italian Energy System will face increasingly critical situations, that threaten especially the infrastructure with long lifetimes. Investments are therefore indispensable and urgent to increase the resilience of the energy sector, considering that operational practices have to adapt to the “new normal” of extreme events. For this purpose, the Italian Regulatory Authority (ARERA) call for specific plans and actions by TSOs and DSOs (Terna and RSE, 2022).

Since different interacting hazards can lead to greater impact than the sum of the single hazard effects, in this paper we described an index-based methodological approach in which at first multiple single hazards were considered by assuming them to be independent, then a multi-hazards analysis has been carried out by considering spatial and spatio-temporal concurrence of different threats.

To support electric stakeholders to undertake science-based adaptation actions, climate extreme scenarios were inferred from the outputs of a subset of Euro-CORDEX models under the three emission pathways RCP8.5, RCP4.5, and RCP2.6 describing respectively scenarios with high-emission, partial mitigation, and strong mitigation of greenhouse gases. At first some extreme climate scenarios were elaborated by computing some ETCCDI indices to investigate distinctly single hazards related to thermal stress (SU), water stress (CDD), extreme precipitations (R99PTOT and R99P), strong winds (W75P and W99P), and heatwaves (HWDI) over Italy. Hence, the three RSE indices Droughts (DRO), Storms (STO), Multi-hazards (MUHZ) were defined by combining the above-mentioned single indices exceeding some appropriate thresholds. Although they describe a limited set of climate hazards, this set covers the more common and dangerous threats for the energy system. Despite some limits in the choice of reasonable hazard thresholds, RSE indices take into account the spatial and spatio-temporal overlapping of extreme events that may generate particularly critical conditions for strategic energy infrastructure.

The results show that Italy is likely to face a progressive increase in overall climate hazards, especially heatwaves and droughts, as highlighted by the DRO scenarios, whose frequencies are expected to increase significantly throughout whole Italy not only in the summer but also in autumn and spring. In fact, both seasons are projected to have increasingly typical summer weather conditions, with variations strongly dependent on anthropogenic forcing.

Focusing on CDD and R99PTOT outcomes, wet days are generally expected to decrease but extreme precipitations will be more frequent and more destructive, especially over Po Valley and the eastern Italian coasts. As can be inferred from the STO index analysis, increasing extreme rainfalls in combination with moderate to strong winds enhance the risk of direct and indirect damages just in the same areas historically affected by floods and severe storms (Central-Northern Italy).

From a multi-hazard point of view (MUHZ), hot-spots are identified on coastlines (Tyrrhenian and Adriatic Coast) and floodplains (Po Valley). This is reason of concern because these are highly populated and economical pivotal areas. Climate conditions are expected to exacerbate in the absence of mitigation, but it should be noted that even in a strong mitigation scenario, climate risk prevention actions will be necessary to ensure the reliability of the energy system in the future.

The simple approach adopted in this study can be extended to a variety of other hazards that can seriously affect the Italian region, such as mudslides, landslides, coast floods, fires,… by including other indices dealing with them. With regard to fire danger, a method to select critical conditions was proposed in (Faggian 2018) by considering the exceeding of particular threshold values for the weather variables T, R, and W. The outcomes pointed out scenarios congruent with those obtained by means of the Fire Weather Index (FWI) (Van Wagner 1987) widely used to identify weather conditions favourable to fires.

It represents a starting point that can be integrated by considering the specific exposures and vulnerabilities of the different components of the Italian energy sector to develop a multi-hazard risk assessment. In this regard, if more detailed information from climate model and/or observational dataset were available, the methodology is flexible enough to describe the climate-related hazards at local scale and assess their actual impacts on the different elements of the system, in order to support decision-making processes on adaptation action plans.

Moreover, it can be considered as a base for a new multi-risk assessment to be used from national to local administrative level not only for the Energy System. It can be extrapolated to other critical sectors such as agriculture, health, tourism, etc.