The MS is a mostly deep (1,500 m on average) semi-enclosed regional sea, connected to the west to the Atlantic Ocean by the narrow Gibraltar Strait and to the east to the Black Sea by the Bosporus and Gallipoli (Dardanelles) Straits. Geographically, it is located between the European and the African continents and, from a climatic perspective, at the transition between a midlatitude regime zone, to the north, and a subtropical regime zone, to the south. Meteorologically, the MS region is interested by an intense synoptic-scale activity, with a separate branch of the northern hemisphere storm track passing over the region (Lionello and Sanna, 2005). The MS has a complex morphology, with many islands and peninsulas dividing the basin into sub-basins, some of them connected by straits (Figure 1). Together with the presence of steep mountain ridges close to the coast, this morphological complexity provides to the MS environment with a spatial heterogeneity, which also reflects on the MS region meteorology and climate, making their successful modeling far from being trivial (Lionello, 2012). In addition, and most important for the topic of this study, the orographic complexity of the MS region plays a fundamental role in the genesis of cyclones. In this context, cyclones typically originate (mostly in the western MS) from the interaction of synoptic-scale flows with the mountains, constraining the flow in the lower troposphere and producing local and often strong winds (Lionello, 2012).

In general, the MS wind climate is characterized by the occurrence of winds with regional spatial structure (200–1,000 km) and seasonal variability (Zecchetto and De Biasio, 2007; Lionello, 2012). In Figure 1 we have indicated the most relevant winds for the different sub-basins. Levanter and Vendaval winds blow in the westernmost part of the MS (Western sub-basin), respectively, from the north-east and the south-west: the first during all seasons (but most intensely in winter), the second mainly during autumn and spring. Mistral is a cold dry wind jet, stronger during winter, blowing from the north/north-west over the western and Tyrrhenian sub-basins and occasionally reaching the African coasts. Libeccio and Sirocco are southerly winds blowing in the Tyrrhenian and Ionian-Meridional sub-basins: the first mainly in winter from the south-west, the second mainly in autumn and spring from the south-east. Sirocco is also channeled into the Adriatic sub-basin, and it contributes to the intense storm surges that take place in the northern Adriatic Sea and in the Venice lagoon (Cavaleri et al., 2019). The other dominant wind in the Adriatic climate is Bora, a cold and dry north-easterly wind producing jets over the sea after interacting with the local orography. Northerly winds also dominate the climate in the Levantine-Aegean sub-basin: Etesian winds blow the from north/north-west and are particularly strong during summer.

The climate of wind waves in the MS, mostly driven by these regional winds, is subject to the morphological complexity constraints described above, resulting in limited fetches in large parts of the basin (with some exceptions) and in coastal shallow regions relatively proximal to the generation areas. Some regions present peculiar characteristics, for instance the northern Adriatic Sea or the Aegean Sea, where wave propagation experiences either the evolution over a relatively shallow flat bottom or the shadowing by numerous islands, respectively. Most of the studies on the MS wave climate (including a wind and wave atlas of the MS, Western European Union and Western European Armaments Organisation Research Cell, 2004) based on hindcast or observed data, focused on the extreme significant wave height (e.g., the monthly 99th percentile: Sartini et al., 2015a; Sartini et al., 2015b, 2017; De Leo et al., 2020; Morales-Márquez et al., 2020) with only a few exceptions that focused on the average significant wave height (for instance, Lionello and Sanna, 2005; De Leo et al., 2020). Others were dedicated to the assessment of the wave energy climate at specific locations or over the whole basin, with energy harvesting purposes (Barbariol et al., 2013; Liberti et al., 2013; Arena et al., 2015; Besio et al., 2016). A similar approach to the one presented in this paper, although with a different purpose, can be found in von Schukmann et al. (2021), where Benetazzo et al. have used a shorter portion of our dataset for an assessment of the climate of the maximum individual waves in the MS in the last quarter of a century, focusing on the 2019 anomalies and their physical interpretation. The complexity of the MS climate has also been exploited to test novel approaches for wave climate analysis (Sartini et al., 2015b; Portilla-Yandún et al., 2019). Besides providing the statistics over decadal periods (generally one to three decades), some of the previous studies used statistical and data analysis approaches to obtain accurate return level estimates of the significant wave height (e.g., Sartini et al., 2015a), and to describe the spatio-temporal variability and the long-term change (trends) of the wave climate (e.g., Sartini et al., 2017; De Leo et al., 2020; Morales-Márquez et al., 2020). To interpret inter-annual and inter-decadal variabilities, some studies also tried to relate them to the modes of atmospheric variability, looking for teleconnection patterns and using climate indices (Lionello and Sanna, 2005; Morales-Márquez et al., 2020). Finally, some studies were devoted to providing projections of the future wave climate under climate change scenarios over the whole basin or sub-basins (see for instance, Benetazzo et al., 2012; De Leo et al., 2021).

The climatology of wind-wave sea states in the MS is presented by using the 50th (i.e., median) and the 99th percentiles of the significant wave height H_s, which may represent typical and extreme sea-state conditions, respectively. Similarly, the climatology of maximum individual waves is based on the 50th and 99th percentiles of the maximum crest-to-trough height H¯max (and maximum crest height C¯max in the Supplementary Material).

In order to highlight the effects of intra-annual variability, the 50th and 99th percentiles are estimated at seasonal scale and averaged over the length of the hindcast to provide an empirical estimate of the expected typical and extreme conditions during the different seasons. In this study, we propose and adopt a novel definition of stormy seasons for the MS wind-wave climate based on the observed H_s data in the MS, rather than using the canonical meteorological definition based on the air temperature (i.e., winter: December, January, February, DJF; spring: March, April, May, MAM; summer: June, July, August; JJA; autumn: September, October, November, SON). To this end, satellite altimeter observations of H_s are used (see Section “Assessment Methodology”). Yearly-averaged monthly H_s values are first computed for each satellite altimeter and then the eight satellites are combined to obtain the ensemble average and standard deviation.

To let the interannual climate variability and climate change due to external forcings emerge, we remove the source of intrinsic intra-annual variability due to seasonality by taking separately into account the different seasons, as done for instance by Morales-Márquez et al. (2020). Therefore, in the result presentation (climate variability and climate change), we shall focus on the two most representative seasons only, i.e., winter and summer. The interannual climate variability is shown in terms of climate anomalies, which are defined as deviations of the winter and summer 50th and 99th percentile values of a generic wave height H in a specific i^th year Hipp (pp = 50, 99) with respect to the winter and summer yearly-averaged 50th and 99th percentiles H¯pp (overbar denoting average over the years), i.e., Hipp-H¯pp. Further, to assess the spatial distribution of the inter-annual climate variability and highlight the MS sub-basins and regions that have experienced the largest variations in the study period, we have also normalized anomalies with respect to the local yearly averages, thus showing the relative anomalies, i.e., (Hipp-H¯pp)/H¯pp. In order to motivate the inter-annual climate variability of the MS waves, we look for possible teleconnections between the wave climate variability and the principal modes of atmospheric variability, taking into account four northern hemisphere indices (downloaded as monthly averaged values from³): North Atlantic Oscillation (NAO), Scandinavian (SCAND), East Atlantic (EA) and East Atlantic-Western Russian (EAWR). Monthly mean data are then combined to obtain seasonal averages.

Finally, in order to assess any potential long-term change in the typical and extreme sea state climate over 1981–2019, the linear trend of winter and summer Hs,C¯max and H¯max are estimated over the whole MS and sub-basins through the Sen’s slope of the 50th and 99th percentiles (Theil, 1950; Sen, 1968), testing the statistical significance at the 90% confidence interval with the Mann–Kendall test (Mann, 1945; Kendall, 1948).

The monthly 50th and 99th percentiles of H_s, as observed over the MS by the eight satellite altimeters used for the hindcast assessment (Supplementary Table 1) are shown in Figure 2A. Observations are also normalized with respect to the maximum monthly H_s percentile (Figure 2B), which occurs in January for both the 50th and 99th percentiles. Both typical and extreme sea states clearly show two dominant regimes: a stormy period, from November to March (with values within 20% of the January H_s; Figure 2B), and a relatively calm period, from June to August (with values within few per cent of the minimum H_s, occurring in June for the 50th percentile and in August for the 99th percentile). In between, there are two transitional seasons. Following this variability, we define the four sea-state “seasons” as winter (NDJFM), spring (AM), summer (JJA) and autumn (SO). Compared to the classic meteorological definition, we therefore propose a longer winter season, starting in November and ending in March (both showing significantly larger H_s compared to the other spring and autumn months), and shorter transitional seasons (2-month long), with summer remaining unchanged. This definition applies to the whole basin but still holds if only northern, southern or western sub-basins are considered. If only the eastern part of MS is accounted for, the relative difference between summer and transitional seasons reduces (not shown here), most likely due to the effect of Etesian wind-waves. This is in agreement with findings of Queffeulou and Bentamy (2007) who pointed out differences in western to eastern MS monthly mean H_s from satellite altimetry, especially during summer season. November and December H_s values clearly belong to the winter season that accounts for the early months of the following year; therefore, to have an equal number of seasons, we start counting seasons from winter 1981 (beginning with November 1980) to autumn 2019 (ending in October 2019), thus analyzing only 39 years in the whole period (hereinafter 1981–2019).

The MS wind-wave hindcast performance in the 1991–2019 period against satellite altimeter observations is summarized in Figure 3, which collects information about the relative model-observation bias RBIAS in the MS and in the sub-basins (Figure 3A), the H_s scatter (Figure 3B) and the relation between modeled and observed climate variability and change (Figures 3C,D). Modeled H_s is simulated in the MS with 0.87 best-fit line slope, BIAS = −0.12 m and MAE = 0.26 m. The relative model-observation bias at the 50th percentile of modeled H_s (0.93 m; blue dot in the Figure 3B) is −8%, while at the 99th percentile (3.78 m; red dot in the Figure 3B) it is −14%, denoting a decrease in model performance with H_s, in agreement with the MS wave hindcast study of Mentaschi et al. (2015; verified against in situ wave buoys). The model-observed bias might be partially corrected by calibrating the input/dissipation source term parameterization in WW3. However, the spatial heterogeneity of the MS (with narrow, semi-enclosed sub-basins or/and surrounding orography) makes it difficult to achieve a set of parameters that is optimized to obtain good performance in the whole MS basin and in the whole H_s range (Lionello and Sanna, 2005; Mentaschi et al., 2015; Sartini et al., 2017; Morales-Márquez et al., 2020). This can be seen by comparing the slope of best-fit lines and the relative bias at different percentiles for the MS with those for the sub-basins. Best performance is achieved in the widest sub-basins (Ionian, Levantine), with larger slope and smaller relative bias compared to those obtained in the whole MS. Worst performance are in the Adriatic Sea, where the ERA5 wind is less effective in correctly reproducing wind and then waves due to spatial resolution effects in a narrow basin. These effects are partially visible also in the Tyrrhenian and Western sub-basins (both with islands and narrow parts), although performance in these sub-basins is comparable to that for the MS.

The average (and standard deviation of) correlation coefficients between the monthly 50th and 99th percentile H_s from satellite altimeters and co-located WW3 hindcast are: 0.98 ± 0.004 for the 50th percentile and 0.97 ± 0.009 for the 99th percentile. An example of the monthly percentiles time-series for JASON-1 is shown in Figures 3C,D, with correlation coefficients larger than or equal to 0.98 for both 50th and 99th percentiles. Also, the linear trends of H_s estimated from the same satellite altimeters and co-located model data are in agreement, with consistent (i.e., both positive or both negative) Sen’s slope signs and a general tendency for larger model trends for the 50th percentiles and smaller for the 99th percentiles (compared to JASON-1, the relative trend differences are 15 and −16% for the 50th and 99th percentile, respectively, while the largest relative difference is −43%, between Envisat and model 99th percentile H_s trends).

The yearly-averaged seasonal 50th percentile of H_s is shown in Figure 4. As expected, the largest values occur during the winter (NDJFM) season when wind speeds are typically stronger (Zecchetto and De Biasio, 2007). Spatial patterns of the 50th percentile H_s are rather different in winter and summer (JJA), on the one hand, and very similar in spring (AM) and autumn (SO), on the other hand. The most energetic sea states occur over the western and southern sub-basins in winter (50th percentile up to 1.5 m) and spring (up to about 1 m), and over the eastern and western sub-basins in summer (up to 1.25 m) and autumn (up to 1.0 m). The western basin experiences the largest typical conditions over all seasons except summer, particularly the Sardinian Sea (amid the Gulf of Lyon, Sardinia, the North African coast and the Balearic Islands) where the north-westerly Mistral wind persistently blows from the French inland (Zecchetto and De Biasio, 2007). In summer, the largest wave heights are attained in the Aegean and Levantine basins (particularly southeast of Crete, with a typical jet-like pattern) with values comparable to those of winter and produced by the north-westerly Etesian winds, typical of late spring, summer and autumn seasons but reaching their maximum intensity during summer (Zecchetto and De Biasio, 2007).

The yearly-averaged seasonal 99th percentile of H_s is shown in Figure 5. The spatial distribution of extreme sea states is different compared to that of typical conditions (Figure 4). Indeed, the 99th percentile of H_s in all seasons have similar spatial patterns, and in particular the sea severities during spring and autumn seem comparable (despite some minor differences, for instance in the Alboran Sea and the Levantine basin). The largest values occur in the Gulf of Lyon and Sardinian Sea (up to 5.5 m in winter, up to 4.0 m in summer, up to 4.5 m in spring and autumn). Other regions with large, though smaller, values are located in the southern (Ionian Sea) and eastern (Levantine Sea) sub-basins, while the summer waves southeast of Crete island (the largest in the summer 50th percentile maps; jet-like pattern) are surpassed by the extreme waves generated by the Mistral wind in the Gulf of Lyon. In the Adriatic Sea, the spatial pattern changes compared to the 50th percentile, with large values more homogeneously distributed over the whole basin (including in the northern part) and with the largest ones (in the southern part) off the eastern coast, as a consequence of the south-easterly Sirocco wind storms hitting the Croatia, Albany and Montenegro shorelines.

The first assessment of the wave climate variability is made by analyzing the intra-seasonal extreme-to-typical condition variations. In Figure 6 we show the winter (NDJFM) and summer (JJA) ratios 99_p,Hs/50_p,Hs of the 99th to the 50th percentile of H_s. This ratio provides a measure of the width of the empirical probability distribution of H_s at seasonal scale and it is yearly-averaged to obtain an empirical estimate of the expectation. We observe that sea states are much more variable during summer than winter, with the largest values of 99_p,Hs/50_p,Hs. There is also a larger spatial variability of the ratio during summer with respect to winter. Indeed, during winter extreme conditions are generally up to four times greater than the typical ones, over all the MS basin, with only spatially limited exceptions in the northern Adriatic, southern Tyrrhenian, western Aegean seas and eastern Levantine sub-basin, where they can be up to six times larger (locally, more than six times). During summer, extreme conditions can be more than six times larger than typical ones (up to eight times) over large parts of the central MS, including the Sardinian, Tyrrhenian and Adriatic Seas. The lowest summer variability occurs in the Levantine basin, where the sea states generated by the steady and low-variability Etesian winds are intense but constant over the season.

We now proceed by assessing the interannual variability of the wave climate in the MS region by showing the standard deviation and range of the anomalies of typical and extreme H_s during winter and summer. We first focus on the standard deviation of the H_s anomaly, which is used to point out the regions with the largest climate variability over the years. Then, we focus on the standard deviation and range (maximum-minus-minimum) of the relative anomaly (i.e., with respect to the yearly-averaged percentiles of H_s, see Section “Wave Climate Analysis” for the definition) to show the regions with the largest variability with respect to the local wind-wave climate (and quantify it), allowing to inter-compare regions with different H_s percentiles. The interannual variability of the 50th percentile of H_s is presented in Figure 7. During winter the largest variability occurs in the Western sub-basin (Mistral wind region, up to 0.25 m standard deviation; Figure 7A); in summer, in the Aegean and Levantine sub-basins (Etesian wind region, up to 0.2 m standard deviation; Figure 7D). These are also the regions with the largest 50th percentile of H_s (Figure 4) although the spatial distribution of the standard deviation of the H_s anomaly does not perfectly reflect that of the yearly-averaged 50th percentile of H_s. Looking at the standard deviation of the relative anomalies (Figures 7B,E), it emerges that the regions with the largest variability with respect to the local wave climate are in the Tyrrhenian, Adriatic and Aegean sub-basins during summer, and in the Western, Tyrrhenian, Adriatic, and Levantine sub-basins during winter. The maximum standard deviation of the relative H_s anomaly, up to 20% of the 50th percentile H_s in both summer and winter, occurs in the northern Adriatic Sea and the southern Tyrrhenian Sea, where during 1981–2019 winters the range of variability has reached 100% of the local 50th percentile H_s.

The interannual variability of the 99th percentile of H_s is presented in Figure 7, and show different characteristics with respect to that of the 50th percentile of H_s. Indeed, in summer the maximum standard deviation of the H_s anomaly is in the Western sub-basin (Sardinian Sea, 0.6 m; Figure 7J), where also the maximum 99th percentile occurs, whereas in winter the maximum standard deviation is in the southernmost (Ionian sub-basin, offshore Libya; Figure 7G) and westernmost (Western sub-basin, southeast and southwest of Balearic Islands) parts of the MS, which are not the locations with the maximum 99th percentile of H_s (Sardinian Sea and Gulf of Lyon, Figure 5). Similarly, the largest standard deviation of the relative anomaly (20% of the 99th percentile H_s) is in the Western, Tyrrhenian, Adriatic, Aegean and Ionian sub-basins during summer (as for the 50th percentile H_s; Figure 7K), while in winter the maximum is offshore Libya and southwest of the Balearic Islands (Figure 7H). In these regions, the range of the relative anomaly is up to the 100% of the local 99th percentile H_s.

To provide a synthetic description of the interannual climate variability at the regional scale (i.e., MS basin and sub-basins), in Figure 8 we show the time series of the spatially averaged relative anomaly of the 50th and 99th percentile of H_s over the MS (spatial averages over the sub-basins are shown in Supplementary Figures 1–5). In general, during 1981–2019 there is an alternation of positive and negative anomalies that, more frequently during winter and especially in some sub-basins, tend to aggregate forming clusters of years with (almost) only positive or negative anomalies. An example of this aggregation is the 1989–1994 anomaly of the winter 50th percentile H_s. During this period the largest negative anomaly in the MS have occurred (winter 1990, −18%), followed by four more seasons with negative, often considerable, anomalies. The largest negative anomaly at sub-basin scale have occurred in the Tyrrhenian Sea in the same period (winter 1989, −36% of the 50th percentile of H_s; Supplementary Figure 1), but similar pattern can be observed in the Ionian, Adriatic and Western sub-basins, even if with smaller anomalies. During 1989–1994 winters, the atmosphere in the MS region has been very stable, in particular over the Western, Tyrrhenian, Adriatic and Ionian sub-basins (Supplementary Figures 1–4; as shown in Supplementary Figure 6, displaying the 500 hPa geopotential height, sea level pressure and wind intensity anomalies over western Atlantic, Europe and western Asia from ERA5 atmospheric reanalysis, for some seasons of interest). The cause of this stability can be found in the localization of the pressure fields on a synoptic (European) scale (Supplementary Figure 6). The conditions that induce strong atmospheric stability over the MS basin have been marked positive 500 hPa geopotential anomalies and positive sea level pressure (SLP) anomalies over Europe, and negative wind intensity anomalies over the MS basin. Such conditions have occurred in all the 1989–1994 winters (as shown in Supplementary Figure 6 for 1989 and 1990), and are responsible for the negative H_s anomalies observed in Figure 8 for the MS and in Supplementary Figures 1–5 for sub-basins. The largest positive H_s anomaly in the MS (+21% of the 99th percentile of H_s) have occurred in winter 1981, while the largest positive anomaly at sub-basin scale occurred in the Tyrrhenian Sea in summer 2014 (+32% of the 50th percentile of H_s). Winter 1981 has been characterized by an average circulation over Europe (Supplementary Figure 6) with positive SLP anomaly over Western Europe, positive 500 hPa geopotential anomaly over Russia and a negative SLP anomaly on central and southern Europe and on Scandinavia. These conditions are associated with the meridional advection of arctic air and intense cyclogenesis over the MS, especially over the western part, as shown in Supplementary Figure 6 with maximum wind intensity anomalies of over 1.5 m/s on the Western and Ionian sub-basin (both showing very large positive extreme H_s anomalies in winter 1981, Supplementary Figures 3–4). Summer 2014 (Supplementary Figure 6) has been characterized by geopotential and SLP anomalies in the western MS, with the minimum pressure positioned on central-western Europe. This configuration generates frequent advections of relatively cold air along the eastern edge of the Azores anticyclone, producing positive wind intensity anomalies (with values between 0.5 and 1.5 m/s, Supplementary Figure 6) in the Western, Tyrrhenian and Adriatic sub-basins, all showing positive H_s anomalies in summer 2014. The temporal climate variability of typical H_s in the MS has been the largest in winter typical sea states as shown by the 8.62% standard deviation. At sub-basin scale the largest temporal variability of typical H_s has occurred in the Tyrrhenian, Adriatic and Western sub-basins, both in summer and winter (for the Tyrrhenian Sea, 11.3 and 13.6% standard deviation, respectively; see Supplementary Figure 1), in agreement with results about typical climate variability in Figure 7. Extreme H_s appears to be most variable in the Tyrrhenian and Adriatic seas during summer, and in the Ionian and Levantine sub-basins during winter, (for the Adriatic Sea summer 11.2% and for the Ionian Sea winter 8.9% standard deviation, respectively; see Supplementary Figures 2, 4), in agreement with results for the climate variability shown in Figure 7.

The inter-annual climate variability of the MS waves observed inFigure 8 (and Supplementary Figures 1–5 for sub-basins) can be partially explained by possible teleconnections between the wave signals and the principal modes of atmospheric variability. To this end, we compare the time series of the relative anomaly with the time series of the NAO, SCAND, EA and EAWR indices. Thus, we have (i) averaged the monthly values of 1981–2019 indices over the winter and summer seasons, respectively, (ii) computed cross-correlation coefficients between the four indices and the MS (and sub-basins) H_s relative anomalies and then (iii) plotted in Figure 8 (and Supplementary Figures 1–5 for sub-basins) the time series of the index showing the highest correlation (in search for teleconnection). Generally, with only a few exceptions, relative anomalies in the MS and sub-basins are mostly correlated to SCAND positive phases, particularly, but not only, for the 50th percentiles. In the MS, this is observed for winter and summer typical and extreme sea states, except for winter extreme H_s, which is related to the EA negative phases. The highest correlations are found for the winter 50th percentiles, although the correlation coefficients are generally small, reaching 0.54 at most for the MS (SCAND and winter 50th percentile, Figure 8) and 0.65 maximum value for the sub-basins (SCAND and Tyrrhenian Sea winter 50th percentile, Supplementary Figure 1). The impact of the SCAND index on the Mediterranean atmosphere dynamics and hence on the MS wave climate can be observed, as an example, in the winter 1989 and 1990 atmospheric configuration (Supplementary Figure 6). The persistent low 500 hPa geopotential height and SLP fields over Scandinavia and Eastern Europe (with −1.2 and −0.8 seasonal SCAND index, respectively; Figure 8), induce negative anomalies of wind speed over the whole MS, which, in turn, generate negative anomalies of the 50th percentile H_s (−15 and −18%, Figure 8).

We assess here what might have been the climate change of the winter and summer MS sea states over 1981–2019 showing the maps of the H_s linear trends (using the Sen’s slope, in cm decade^–1, evaluated on 39 yearly data) in Figure 9 and their spatial averages over the basin and sub-basins in Table 1. We show both the trends at the points where there is statistically significance (at 90% confidence interval according to the Mann–Kendall test) and at all the points in a (sub-) basin, including where there is no statistical significance.

The winter 50th percentile of H_s (Figure 9A and Table 1) shows positive trends over large part of the western and northern sub-basins of the MS, and negative or no trends in the eastern and southern sub-basins, resulting in a substantially net zero trend for the MS typical winter sea states (0.3 cm decade^–1; Table 1). The largest increases are in the Sardinian Sea and offshore Algerian coasts (up to 5 cm decade^–1) and largest decreases in the southern MS (down to −3 cm decade^–1). However, only trends offshore the Algerian coasts, in the central Adriatic Sea (Sirocco and middle Bora jets regions) and in small portions of the Aegean and Balearic seas are statistically significant, bringing the spatial average to 1.1 cm decade^–1 (Table 1). The summer map (Figure 9B) is far different, showing generally smaller Sen’s slopes, except in some limited areas in the eastern part of the basin, with statistically significant increases (up to 4 cm decade^–1) in the Aegean Sea and eastern and western Levantine basin and decreases (down to −4 cm decade^–1) in the central Levantine basin where the summer Etesian winds blow. As a whole, the MS presents a very mild decreasing trend during typical summer sea states (−0.8 cm decade^–1; Table 1).

The 99th percentile H_strends, shown in Figures 9C,D, are significantly larger than the 50th percentile trends shown above. During winter the largest statistically significant changes are in the southern Levantine sub-basin with increases up to 12 cm decade^–1 offshore Egypt and Israel (11.0 cm decade^–1 on average in the sub-basin), in the Gulf of Lyon with increases up to 12 cm decade^–1 (Mistral jet region, despite a smaller average trend in the Western sub-basin), in the northern Adriatic Sea (northern Bora jet, increases up to 8 cm decade^–1) and in the southern Tyrrhenian Sea, the only statistically significant area of negative trend for winter extreme sea states (down to −12 cm decade^–1, −10.0 cm decade^–1 on average in the Tyrrhenian Sea). In the MS the winter 99th percentile H_s has increased 6.5 cm decade^–1 (Table 1) as a whole. In summer, the trend distribution and sign are quite different from the winter ones. Indeed, globally the 99th percentile H_s has decreased (−1.6 cm decade^–1; Table 1), and the spatial distribution is quite similar to the distribution of the summer 50th percentile H_s trends. In particular, a statistically significant increase is observed in the Ionian (up to 12 cm decade^–1, 7.9 cm decade^–1 on average in the sub-basin) and Aegean seas only. The Levantine basin shows a decrease down to −10 cm decade^–1 and of −6.6 cm decade^–1 on average, while the Balearic, Tyrrhenian (including the Gulf of Genoa) and Adriatic seas show moderate decreasing trends in the summer extreme sea states. Concluding, on average, in 1981–2019 the MS wind-wave climate has decreased during summer seasons and increased during winter seasons (both 50th and 99th percentiles) with the largest changes in the extreme sea states.

In this section we change the focus of the analysis from the climate of the significant wave height of sea states to the MS climate of the maximum individual wave heights and the other (beside significant wave height) relevant parameter for their estimation, i.e., wave steepness and narrow bandedness.