The mean EKE in the Mediterranean Sea exhibits a heterogeneous spatial distribution, with most of the basin characterized by relatively low values of EKE (Fig. ). The overall spatiotemporal mean is 66.10 cm² s⁻². In the western Mediterranean, the Alboran Sea stands out with intense eddy activity around the two semi-permanent anticyclonic gyres , reaching a spatiotemporal mean EKE of 235.58 cm² s⁻². This vigorous eddy activity, up to 400 cm² s⁻² , plays a key role in modulating heat and salt transport through the Strait of Gibraltar . In the eastern Mediterranean, the most energetic feature is the Ierapetra eddy, a long-lived anticyclonic structure that forms southeast of Crete , with a spatiotemporal mean EKE of 319.63 cm² s⁻², up to 500 cm² s⁻² . For comparison, EKE values over the Gulf Stream and the Kuroshio Extension can exceed 3000 cm² s⁻² .

The EKE averaged over the Mediterranean Sea (Fig. a) from all-sat-glo shows a significant increase during the altimetric era, from a mean of 44.91 cm² s⁻² in 1993 to 68.04 cm² s⁻² in 2022. The computed trend is 0.75±0.06 cm² s⁻² yr⁻¹ (Fig. e) and is statistically significant. This result is consistent with all-sat-euro which exhibits a similar trend of 0.87±0.05 cm² s⁻² yr⁻¹ (Fig. e). Although comparable, the all-sat-euro trend is slightly stronger as it has a higher spatial resolution (1/8° compared to 1/4° for all-sat-glo), which allows for better representation of mesoscale eddies and their associated energy . However, the two-sat-glo product shows a markedly weaker and not statistically significant trend of 0.06±0.04 cm² s⁻² yr⁻¹ over the Mediterranean Sea (Fig. a and e). The larger trend detected in the all-sat products can be explained by the fact that they estimate nearly twice as much EKE as the two-sat-glo product in recent years. Despite these notable differences in trends magnitude, all three L4 products show strong agreement on seasonal cycles and interannual variability (r=0.98 between all-sat-glo and all-sat-euro and r=0.79 between all-sat-glo and two-sat-glo after removing their linear trends).

In the Alboran Sea region (Fig. b and  e), all three products exhibit positive and statistically significant EKE trends, in contrast to the Mediterranean basin as a whole where the two-sat-glo trend is not significant. The trends are also higher in magnitude than those observed for the whole Mediterranean, with a particularly pronounced gap between all-sat-glo (3.04±0.23 cm² s⁻² yr⁻¹) and two-sat-glo (0.43±0.18 cm² s⁻² yr⁻¹). This indicates that the impact of the increasing number of satellites is especially evident in energetic regions such as the Alboran, where all-sat products capture much stronger EKE levels than the two-sat product. A focus on the eastern Alboran gyre (Fig. c) reveals similar trends to those observed across the broader Alboran Sea for all-sat products (3.64±0.32 cm² s⁻² yr⁻¹ for all-sat-euro and 3.44±0.31 cm² s⁻² yr⁻¹ for all-sat-glo), while revealing a net increase for the two-sat-glo product: 1.16±0.28 cm² s⁻² yr⁻¹. This suggests that the alignment of altimeter tracks relative to energetic features (Fig. ) plays a significant role in capturing variability accurately.

Unlike the other regions, the Ierapetra region displays strong statistically significant negative trends (Fig. d and e, the boundary of the Ierapetra area was defined as an EKE-isoline of the 30 years mean of all-sat-euro product in order to have the same region for all the products, value of the isoline at 200 cm² s⁻²). In addition, the EKE time series reveal intense peaks compared to the rest of the basin. The Ierapetra eddy itself is not a permanent feature , but rather a seasonal anticyclonic structure with peak intensity in late summer (Fig. d). We notice a marked decrease in peak intensity, with EKE maxima reaching ∼1500 cm² s⁻² before 2007 but only ∼500 cm² s⁻² after 2012. In this region, all products capture similar trends, likely because the spatial coverage of the two-sat product is already good near a crossover point of the reference altimeter tracks, which enhances the accuracy of the gridded fields and reduces interpolation errors . Additionally, the lower short-term variability in this region (compared to the Alboran Sea) means that the temporal resolution of the two-sat product is sufficient to capture the SLA signal accurately, thereby narrowing the difference with the all-sat products.

For robustness, similar analysis were performed in more dynamically complex regions, such as the Ionian Sea and the eastern Algerian basin. The corresponding EKE time series and trends are provided in the Supplement (Fig. S1), where larger short-term variability makes the interpretation of long-term trends more challenging.

Part of the differences observed between EKE trends computed with all-sat and two-sat altimetric products over the Mediterranean Sea could be explained by the increasing number of merged altimeters in the all-sat products, which may amplify EKE signals over time. To further explore this hypothesis, we examined the temporal evolution of EKE differences between products in relation to the number of satellites incorporated in all-sat. The evolution of the difference in the area-weighted mean EKE over the Mediterranean Sea between all-sat-glo and two-sat-glo exhibits a marked increase over time (Fig. a). This increase is strongly correlated with the temporal evolution of the number of altimetry missions merged into the all-sat-glo product, with a Pearson correlation coefficient of 0.85. As more altimeters are included into the all-sat product, spatial and temporal sampling improve, which enhance the detection of mesoscale signals . This improved observational capability introduces an artificial positive trend of EKE. This strong correlation suggests that a significant portion of the observed trend in all-sat products seems to arise from the increasing number of satellites, rather than from changes in oceanic variability. Note that once more than five altimeters are operational, the rate of increase in the EKE difference slows down. To isolate this artificial trend, we performed a second degree polynomial regression between the signal difference and the number of satellites. By subtracting this fitted signal from the original all-sat-glo time series, we obtained a version of the product without the artificial trend, shown in green in Fig. b. This new signal, denoted as all-sat-glo-corrected, represents the all-sat-glo EKE with the artificial satellite-driven trend removed. The resulting trend for all-sat-glo-corrected is 0.12 cm² s⁻² yr⁻¹, and statistically significant, which is slightly higher than the trend of two-sat-glo of 0.06 cm² s⁻² yr⁻¹.

To identify the regions that are more likely to be affected by the impact of an increasing number of satellites, we computed the Pearson correlation coefficient between the all-sat-glo and two-sat-glo differences and the number of altimetry missions at each grid point (Fig. ). High correlations are especially pronounced in low-energy areas (Fig. ) such as the Adriatic Sea, Tyrrhenian Sea, eastern Levantine basin, and along the Libyan coast. These areas coincide with zones where all-sat-glo trends are statistically significant while two-sat-glo trends are not (Fig. ). This spatial coherence highlights that, in these low-energy regions, the trends observed in the all-sat products are most likely driven by the increasing number of merged satellite altimeters. In such regions, the signal-to-noise ratio is lower, making it more sensitive to observational changes. Moreover, the Alboran Sea, although a high-energy area, also presents a high correlation. High correlation values are found mainly in the western Alboran gyre, while the eastern gyre exhibits lower correlations. This contrast is likely linked to the orientation of the two-satellite tracks relative to the surface currents. In the eastern gyre, one of the two-sat tracks crosses close to the gyre center, providing better sampling of the geostrophic velocities. In the western gyre, however, the tracks mostly sample the edges of the gyre, leading to limited representation of its surface geostrophic velocities. This uneven sampling is further exacerbated by the highly dynamic nature of the western Alboran gyre, located at the entry point of the Atlantic Jet , where the circulation evolves rapidly and is not fully captured by the two-sat configuration. Consequently, the additional spatial and temporal coverage provided by multiple satellites substantially improves the reconstruction of the dynamics in this region, thereby increasing the observed correlation.

These spatial correlations between EKE differences and satellite coverage offer insight into where artificial amplification of EKE trends is most likely to occur. To place these findings in a broader spatial context, we now examine the full two-dimensional distribution of EKE trends across the Mediterranean Sea. These maps (Fig. ) reveal that most of the Mediterranean Sea exhibits positive EKE trends along the main southern surface currents, especially pronounced in regions such as the Alboran Sea, along the north African coast, and in the vicinity of the Mersa–Matruh area (southeast of the Ierapetra eddy). In contrast, the Ierapetra eddy is the only region showing strong statistically significant negative EKE trends. The northern Ionian Sea also shows negative trends, whereas the central Ionian displays positive trends, a pattern that may be linked to shifts in the basin's circulation between anticyclonic and cyclonic states . All products present a similar pattern with stronger trends for all-sat products. What differs substantially is the spatial coverage of statistically significant trends between the all-sat and two-sat products (Fig. ). Although 72.40 % and 71.27 % of the Mediterranean have significant trends for all-sat-euro and all-sat-glo, respectively, only 48.70 % is significant for two-sat-glo.

To understand how these regional differences influence basin-scale assessments of EKE variability, we compare trends as the spatial average of local trends with trends derived from the area-weighted mean EKE time series (Table ). “Mean of all trends” corresponds to the spatial average of all grid-point trends. “Trend of the mean” represents the linear trend computed from the area-weighted EKE time series averaged over the whole Mediterranean basin, the one shown in Fig. . This comparison highlights the influence of high-energy regions on basin-wide trends. In all-sat products, the trend of the area-weighted mean EKE time series is higher than the mean of the trends because high-energy regions, though spatially limited, dominate the average time series due to their strong EKE values and pronounced trends. In contrast, this behavior is not observed for two-sat because of weaker trends in energetic regions. These energetic areas are better resolved in all-sat, leading to greater influence on the overall trend. We also distinguish these trends in two categories with different reference periods to compute the anomalies: the entire period (1993–2023) that are the results presented in this manuscript (first raw in Table ) and 1993–2012 which is the reference period used in the CMEMS products (second raw in Table ). We systematically observe higher trends for the shorter reference period, suggesting an increase in kinetic energy in recent years and highlighting the need to compute geostrophic anomalies with respect to the full time period when computing eddy kinetic energy.

To evaluate the fidelity of the gridded products in capturing mesoscale variability, we compared the L3 and L4 data sets in two high-energy regions of the Mediterranean Sea (Fig. ), the Alboran Sea and the Ierapetra area, assessing how well the gridded fields reproduce the observed variability.

In addition to exhibiting high EKE values, the Alboran Sea region displays strong, positive, and statistically significant trends in EKE over the studied period (Fig. a and b). Notably, one of the reference satellite altimetry mission tracks nearly crosses the center of the eastern Alboran gyre (track 96; Fig. a). In this configuration, the geostrophic velocity component along the track is close to zero, while the cross-track component dominates the total geostrophic velocity. Therefore, the cross-track EKE along this track (EKE_⟂) is expected to be very close to the total EKE, and thus enhances the reliability of the observed trend. In fact, we note the similarity of the trends of EKE_⟂ between the L3-ref and all-sat-glo products for this track (Fig. b). This result indicates that the processing used to produce gridded L4 products does not introduce significant errors in EKE estimates along-tracks, supporting the suitability of L4 data for studying mesoscale variability in regions with dense track coverage. A closer examination reveals that the strongest positive and significant trends are located at the edges of the gyre rather than at its core (Fig. a and b), where geostrophic velocity variability is typically greater , leading here to enhanced positive EKE_⟂ trends.

In contrast, the Ierapetra eddy area, crossed by track 185, shows a negative trend in EKE (Fig. c and d). As observed in the Alboran region, the EKE_⟂ trends are consistent between the L3 and L4 products. Southeast of the Ierapetra eddy lies the Mersa–Matruh area, characterized by episodic rather than persistent eddy activity and reveals an increase in mesoscale activity (track 7, Fig. c and e). The agreement in EKE_⟂ trends across products further confirms the robustness of the results.

EKE captures the overall intensity of oceanic mesoscale variability, integrating contributions from a variety of dynamical processes, including frontal instabilities, current meandering, and coherent mesoscale eddies. Among these, ocean coherent eddies are of particularly interest as they are long-lived, rotating water masses that trap and transport heat, salt, and nutrients across vast distances .

In this section, we use the META Atlas to analyze the statistics of these mesoscale eddies derived from the all-sat-glo and two-sat-glo satellite products, distinguishing cyclonic from anticyclonic eddies, including their number, size, spatial extent, and rotational speed (Fig. ). A representative snapshot of detected eddies from each dataset on 4 January 2020 illustrates clear differences with the all-sat-glo product capturing a larger number of eddies and finer-scale structures compared to two-sat-glo (Fig. a).

Over the altimetric era, the total number of eddies detected per year shows a stronger increasing trend in all-sat-glo (27 812 per year on average) than in two-sat-glo (23 744 per year). Figure b further differentiates between anticyclonic and cyclonic eddies. For both products, a higher number of cyclones than anticyclones is detected with 57 % of the eddies being cyclonic, in agreement with . In parallel, the decreasing trend in the eddy radius size in the all-sat-glo product for both cyclonic and anticyclonic structures (Fig. c) reflects its improved ability to detect smaller-scale features . In contrast, in the two-sat-glo product only anticyclones show a decrease in radius, while cyclones do not display any detectable trend (Fig. c).

Despite these negative trends for the mean radius, when eddies are not separated by polarity, the total eddy area remains approximately constant and similar between the two products. However, separating cyclones from anticyclones reveals opposite tendencies in the two-sat-glo product (Fig. d): the total area covered by anticyclones decreases, while the area associated with cyclones increases. No marked polarity-dependent trends are observed in the all-sat product.

Finally, the average eddy rotational velocity (Fig. e), highly relevant to local biogeochemical variability, shows significant increasing trends in the all-sat-glo data, with a stronger increase for anticyclones, while no such trend is evident in two-sat.

In general, these results indicate that the progressive increase in the number of altimetric satellites has enhanced the detection of mesoscale activity , leading to more numerous, smaller, and faster eddies in the all-sat-glo product, while the total eddy-covered area remains unchanged. These discrepancies also mirror the EKE results and further highlights the limitations of the two-satellite configuration in resolving the diversity of mesoscale processes in the Mediterranean Sea, consistent with earlier work by .

The Alboran gyres and the Mersa–Matruh area confirm a general intensification of EKE in high-energy areas, consistent with patterns observed in global high-energy regions . However, trends over the Mediterranean Sea are not spatially uniform and the Ierapetra area is a high-energy region showing negative EKE trends.

The literature describing Alboran gyres is abundant . Although transient events have been observed , they are mainly persistent gyres with strong seasonal variability (stronger in summer) and are one of the most important signals observed in the Mediterranean Sea. The strong positive trends highlighted at their edges (Figs.  and a) reflect their interannual modulation . As seen in Fig. b the variability in EKE increases during the altimetric record (standard deviation of 44.24 cm² s⁻² for the period 1993–2008 and 59.04 cm² s⁻² for the period 2013–2023 for the Alboran Sea). There is still a gap in understanding the drivers of this variability in the circulation of the Alboran Sea, which include variations in the flow through the Strait of Gibraltar, local winds, and background stratification . For example, previous studies have indicated that the eastern gyre is less stable and may disappear or shift under specific conditions, such as during winter and spring when its kinetic energy decreases to nearly zero and it is replaced by the central cyclonic gyre under moderate Atlantic inflow, or through interactions with the western gyre . In terms of EKE, the most energetic scenario in the area is the eastward migration of the western Alboran gyre before its fusion with the eastern Alboran gyre . noted that these migration events were associated with instabilities of the Atlantic Jet, more precisely south shifts of the Atlantic Jet along the Moroccan coasts, which appear to be related to the establishment of easterlies .

In the Levantine basin, the Ierapetra eddy is the only clear seasonal signal , although it does not appear every year . Studies have highlighted the role of the interaction of Etesian winds (persistent northerly summer winds over the Aegean Sea) with Crete's orography in its formation and intensification . The Ierapetra eddy region is associated with a strong negative EKE trend (Fig. ). Figure d shows a net decrease also in the intensity of the EKE peaks. The analysis of the formation of anticyclonic eddies in this area reveals smaller and shorter-lived eddies (Fig. S2). In fact, have observed a decline in Etesian wind episodes across the eastern Mediterranean and we surmise that this decrease in wind episodes could damp the development of the Ierapetra eddy.

Concerning the Mersa–Matruh area, this region acts as an attractor of transient eddies without clear boundaries, but characterized by a relatively high EKE activity (Fig.  and ). Ierapetra eddies tend to go west along the Libyo-Egyptian current . Previous studies have suggested that some Ierapetra eddies may deviate south-eastward toward the Mersa–Matruh area , potentially contributing to the regional eddy activity. We investigated this behavior using the META atlas (see Supplement), which shows that such deviations are relatively isolated events. This indicates that the EKE trend dipole observed in the area (Fig. c) does not reflect a systematic shift of the same mesoscale structure.

In addition to the regional patterns discussed above, the Ionian Sea is known to exhibit pronounced decadal oscillations in its surface and intermediate circulation, commonly referred to as the BiOS (Bimodal Oscillating System; ). These recurrent transitions between cyclonic and anticyclonic circulation regimes can substantially modulate the intensity and position of the Mid-Mediterranean Jet (MMJ). During cyclonic phases, the MMJ tends to strengthen and shift southward, enhancing mesoscale activity along its path, whereas anticyclonic phases are associated with a weakening and northward shift of the jet . Such regime shifts can therefore imprint decadal variability on EKE in the Ionian Sea. This is consistent with the spatially heterogeneous trend patterns apparent in Fig. . In all three products, the Ionian region displays strong spatial contrasts, with negative EKE trends south of the Peloponnese and west of Greece, while positive trends emerge in the central and southern Ionian. The gray-hatched regions, indicating non-significant trends, are also more prevalent in the central Ionian, reinforcing the idea that this area is subject to decadal variability rather than a monotonic long-term change (Fig. S1). Taken together, these patterns are compatible with a transition from anticyclonic to cyclonic circulation in the Ionian Sea. In particular, reported a marked regime shift in 1997, when the circulation flipped from anticyclonic to cyclonic.

A central finding of this study is the discrepancy in the magnitude of the EKE trend between altimetry products based on different satellite constellations (Fig. ). In high-energy regions we showed that the intensification in the Alboran Sea and the decline in the Ierapetra area are supported by robust and statistically significant trends (Figs.  and ). However, when extending the analysis to the entire Mediterranean basin, the trend obtained with two-sat-glo is statistically not significant (0.06±0.04 cm² s⁻² yr⁻¹). While the all-sat products exhibit positive trends over the whole basin, the inconsistency between products and the limited significance of two-sat-glo trend make it difficult to confidently assert a basin-scale increase in EKE. Moreover, the EKE values derived from all-sat products show a strong correlation with the increasing number of merged satellite altimeters (Figs.  and ). This increase in merged satellites improves spatial and temporal resolution with time, introducing an artificial trend, especially in low-energy areas.

Eddy characteristics shown in Fig.  have also highlighted this artificial trend: the number of detected eddies increases with time for all-sat-glo, whereas the total area they occupy remains approximately constant. This suggests that the enhanced temporal and spatial sampling of all-sat products, arising from the growing number of satellites, enables the detection of smaller eddies that were previously unresolved.

Moreover, the distinction between cyclonic and anticyclonic eddies reveals several polarity-dependent differences in the long-term evolution of eddy statistics (Fig. ), pointing out intrinsic differences in their detectability and dynamical behavior. Overall, more cyclones are detected (57 % Fig.  and ). This asymmetry is fully consistent with the known differences in the reliability of eddy detection in gridded altimetric products. have shown that anticyclones in the Mediterranean are detected with high positional accuracy and moderate radius biases, while cyclones, particularly large ones, are substantially less reliable, with larger position errors, stronger overestimation of radius (only category without negative trend in Fig. c), and greater sensitivity to interpolation artifacts. These detection biases arise from fundamental dynamical differences: large anticyclones tend to be more coherent and longer-lived, whereas cyclones are more unstable and prone to splitting into smaller, fast-evolving sub-mesoscale structures that are poorly resolved by altimetric gridding. The two-sat product, which maintains a stable configuration over time, further supports this interpretation. While it shows no significant trends in eddy number or speed, separating polarities reveals a positive trend in cyclones total eddy area. This pattern is consistent with the “coarsening artifact” described in , where small structures are smoothed into larger, spurious cyclonic features during interpolation.

These polarity-dependent behaviors have direct implications for the interpretation of EKE trends in gridded altimetric products. The stronger increase in mean rotational speed for anticyclones than for cyclones in all-sat suggests that the positive EKE trends reported in multi-mission products are driven by anticyclonic intensification, as observed in the semi-permanent Alboran gyres.

As a result, long-term EKE trends in all-sat configurations should be interpreted with caution, as they may reflect improved sampling of anticyclones rather than energetic changes in the mesoscale field. Nonetheless, part of the differences in EKE trends can also be attributed to the two-sat product being derived from a two-satellite constellation and, as demonstrated by , a minimum of three concurrent altimeter missions are required to adequately monitor mesoscale variability in the Mediterranean Sea.