The ocean around Iceland is a key region where major water masses and currents interact, shaping the North Atlantic circulation, and play a crucial role on the Atlantic Meridional Overturning Circulation (AMOC). The Nordic Seas are among the few places on the globe where the formation of deep waters (1000–3000 m depth) occurs during winter deep convection (Petit et al., 2020). The southern end of the Nordic Seas is bounded by the Greenland-Iceland-Scotland Ridge (GISR). The North Atlantic Current (NAC) brings the warm and salty Atlantic Water (AW) northward into the Nordic Seas (Hátún and Chafik, 2018; Østerhus et al., 2019; Hátún et al., 2021). The AW crosses the ridge in three ways (Fig. 1): (i) between Greenland and Iceland, where the Irminger Current (IC) forms the North Icelandic Irminger Current (NIIC) bringing AW that flows clockwise around Iceland (Jónsson and Briem, 2003; Jónsson and Valdimarsson, 2012); (ii) between Iceland and the Faroe Islands (Mauritzen, 1996); and (iii) through the Faroe Shetland Channel (Hansen and Østerhus, 2000; Hansen et al., 2023), contributing up to 48 % of the total AW transport. The AW undergoes strong cooling and densification in the Nordic Seas and the Arctic Ocean (Mauritzen, 1996; Pérez-Hernández et al., 2019; Athanase et al., 2020; Huang et al., 2023). This modified AW is referred to as Atlantic-origin Overflow Water (AtOW; e.g., Håvik et al., 2017; Casanova-Masjoan et al., 2020) and is one of the two sources of Denmark Strait Overflow Water (DSOW; Semper et al. 2019). AtOW travels southward as a mid-depth water mass in the East Greenland Current (EGC; Håvik et al., 2017), from where, part of it diverts east and merges with the NIIC northeast of Iceland (Casanova-Masjoan et al., 2020).

The transformation of AW into AtOW takes place in different areas of the Nordic Seas: along the Norwegian Current (Håvik et al., 2017), in the Iceland Sea Gyre (Våge et al., 2013), on the eastern side of Greenland, or even – due to its proximity – in the Arctic Basin (Pérez-Hernández et al., 2019). This transformation has different driving mechanisms impacting mixing and convective processes. Wind-stress, sea-ice retreat, and high heat loss due to cold-air outbreaks drive the transformation east of Greenland (Våge et al., 2018). Sea-ice retreat, and heat exchange dominate north of Svalbard (Pérez-Hernández et al., 2019; Athanase et al., 2020), and heat fluxes are the main drivers in the center of Iceland Sea (Våge et al., 2013). Thus, the Nordic Seas region has been previously described as a “mixing pot” (Renfrew et al., 2019), largely responsible for the overall formation of deep overflow water (Lozier et al., 2019). The Nordic Seas are also a large repository of freshwater, primarily originated from glacier melt and river discharge. This water mass increases buoyancy and is carried southward by the East Greenland Current (EGC). Therefore, it is crucial to fully understand the variability of the upper ocean, where mixed layers (ML) develop and transform these water masses.

The Arctic Ocean is warming much faster than the global average, a process known as “Arctic Amplification”, which is also associated with the “Atlantification” of the Arctic (Polyakov et al., 2017; Dai et al., 2019). While the causes are still debated, Arctic Amplification has evident consequences, such as a decrease in seasonal sea-ice extent and a weakening of the cold halocline (Polyakov et al., 2020; Dai et al., 2019). Although these changes are less pronounced in the central Iceland Sea, similar processes have been observed in the central Greenland Sea and the northeastern shelf (Gjelstrup et al., 2022; Strehl et al., 2024), suggesting that Atlantification may also influence the Iceland Sea. Changes in temperature and salinity in the upper ocean modify upper-ocean stratification, which partially controls the mixed layer depth (MLD).

The depth and structure of the ML is primarily controlled by local buoyancy forcing, i.e., surface heat loss and freshwater fluxes, which modifies the water density (Kohler et al., 2018). For instance, within the Iceland Basin, wintertime buoyancy loss drives deep convection, shaping the thermohaline properties that influence the lower limb of the AMOC and its variability in the subpolar North Atlantic (Petit et al., 2021). The pre-existing stratification of the water column is responsible for controlling the effect of the surface forcing. Strongly stratified upper layers resist mixing, while weak stratification allows deeper penetration of turbulence and convection mixing (Pierce et al., 1986). Over shorter timescales, on the order of days, the MLD can significantly deepen as a result of the strong wind events with significant wind stress and associated large wave heights (Skyllingstad et al., 2023). MLD and stratification are strongly influenced by atmospheric forcing, including variability associated with the North Atlantic Oscillation (NAO), which modulates wind stress, surface heat flux, and freshwater input in the Iceland region (Hurrell, 1995; Dickson et al., 1996)

The Intergovernmental Panel on Climate Change (IPCC), Special Report on the Ocean and Cryosphere in a Changing Climate (IPCC-SROCC 2019), indicates with high confidence that roughly 40 % of the global ocean mean upper ocean stratification has increased about 3.3 %–6.1 % since 1960 due to both oceans warming and high-latitude freshening (Tesdal et al., 2018; Yamaguchi and Suga, 2019; Bindoff et al., 2019; Li et al., 2020; Liu et al., 2020; Sallée et al., 2021). Increased stratification is associated with less efficient diapycnal mixing, reducing the exchanges of heat and tracers from the mixed layer into the ocean interior. It has also been observed, with high confidence, that the ML is undergoing changes (Bindoff et al., 2019; on Climate Change, IPCC). Particularly, the shallow summertime ML, which is more likely to be affected by global warming, is deepening at a rate of 5–10 m per decade (Sallée et al., 2021). Despite the reported global patterns, it has been also acknowledged that regional changes might differ from the global estimates (Fox-Kemper et al., 2021).

The warming of the ML and the associated increase in stratification have an impact in biogeochemical processes like phytoplankton blooms and carbon or oxygen sequestration, key components for the Earth's climate (Ólafsson, 2003; Pérez et al., 2021). In the waters surrounding Iceland, the phytoplankton community is closely linked with the water mass properties and hence, an “Atlantification” will replace Polar communities with more Atlantic communities (Cerfonteyn et al., 2023). In the Arctic Ocean, north and northwest of Iceland, the early onset of stratification in spring gives rise to rapid shallowing of the mixed layer and triggers early spring phytoplankton blooms, whereas the weakly stratified water-column in the Atlantic water and the associated deep ML delay the spring bloom south of Iceland (Zhai et al., 2012). This also has strong consequences in carbon uptake, vertical nutrient supply, and biological processes (Yamaguchi and Suga, 2019). Other indirect impacts of the increased stratification include changes in upwelling, deep-water formation rates, biological production, and remineralization rates (Holt et al., 2016), and deoxygenation (Shepherd et al., 2017).

Thus, the overall goal of this study is to characterize the spatial and temporal variability of the mixed layer and stratification around Iceland, where Atlantic Water inflow, Arctic waters, and local surface fluxes shape upper-ocean properties. Using a 29-year hydrographic time series, we investigate the variability of water-mass properties, mixed-layer depth (MLD), density, and thermohaline structure across seasonal and interannual timescales. We examine correlations with atmospheric circulation patterns such as the North Atlantic Oscillation (NAO) and, to complement the observations, use a 1D PWP model to simulate the mixed-layer response to local forcing, helping to identify the mechanisms driving MLD variability.

We use Conductivity-Temperature-Depth (CTD) data from the repeated hydrographic observational program of the Icelandic Marine and Freshwater Research Institute between 1990 and 2019. The oceanographic surveys took place quarterly, mainly in February, May, August, and November with little coverage during the intermediate months. Observations are made at standard repeated sections. The profiles are obtained with a Seabird 911plus CTD mounted on a rosette with Niskin bottles. The conductivity data are calibrated with salinity samples taken at the bottom of each station. All sensors underwent regular calibrations by the manufacturer.

In our analyses we considered only the deepest stations in each section (red dots in Fig. 1), including nearby stations within an area defined by 1° × 0.5° in longitude and latitude (red boxes). The selected stations are located outside of the Icelandic shelf (about 500 m depth). This criterion was chosen to avoid topographic effects, such as across shelf processes on the stratification of the water column and to avoid MLDs limited by shallow bathymetry. Thus, the stations in gray, HB and IH in Fig. 1 were not considered as they fall on the shelf. For the sake of simplicity stations will be named with the acronym of the standard section, first two letters and the station number. The station full name (section and station number) can be found in Table 1.

In this study we analyze the inter-annual variability and linear trends of the ML over a 29-year period as well as the seasonal variability using the seasonal extremes (summer and winter), when there is more data coverage. From the CTD stations we estimated the MLD using the density threshold method with a criterion of σθ=0.01 kg m⁻³ (as, for instance, in Piron et al. (2016) in the Irminger Sea) and a reference depth of 10 m. We chose this criterion instead of the usual 0.03 kg m⁻³ (de Boyer Montégut, 2004) as the latter overestimated the MLD in more than 500 visually inspected profiles (not shown). For comparison and robustness of our chosen method, we also estimated the MLD using other criteria (de Boyer Montégut et al., 2004; Holte et al., 2017). We found the density threshold method appropriate for our region as it proves to be effective even for cases where the variations of salinity and temperature were large. Those variations usually compensate in density making this method more suitable. We have validated our method against previous work by Våge et al. (2018), where a glider data were available, and the results were satisfactory. However, automatic detection methods have limitations, as they may miss stacked mixed layers and other non-canonical representation MLs.

For each profile we computed the Brunt-Väisälä frequency (N2), defined as: 1ρN2=g1ρo∂σθ∂z, where g is the acceleration due to gravity, ρo is a reference density, σθ is the potential density and z is depth. N2 can be decomposed to show the relative contribution of salinity and temperature to the observed stratification as follows: 2N2=NT2+Ns2 where NT2 and Ns2 are the components representing the stratification set by the temperature and salinity, respectively and are defined as: 3NT2=gα∂T∂z4Ns2=gβ∂S∂z, where α is the thermal expansion coefficient and β is the haline contraction coefficient at constant pressure. This decomposition has also been made to classify the oceans by their stratification contribution into α-ocean, β-ocean, and transition zone, where in α-oceans stratification is permanently dominated by temperature, in β-oceans by salinity and the transition regions are either intermittently or seasonally dominated by temperature or salinity (Carmack, 2007; Stewart and Haine, 2016). For the water column to be statically stable, N2 must be positive. However, the contributions may not be positive; when either of its components, NT2 or Ns2 are negative, temperature or salinity respectively has a destabilizing effect on the resulting stratification that must be compensated by the other variable to maintain a stable water column. Small values of N2 indicate that the water column is weakly stratified, which favors mixing due to winter convection and deeper MLD.

To investigate further the driving mechanism of the MLD, we used the Price-Weller-Pinkel (PWP) model (Pierce et al., 1986). The PWP model was chosen to test the hypothesis that β- and transition oceans do not develop deep mixed layers, which is shown in Fig. 8. The PWP model is a one-dimensional vertical model used to simulate the evolution of the ocean mixed layer in response to atmospheric forcing, including wind stress, heat fluxes, and freshwater fluxes. The model evolves vertical profiles of temperature, salinity, density, and horizontal velocity based on surface forcing and a set of physical stability criteria. The first criterion is convective overturning. If surface cooling increases the density such that the water column becomes gravitationally unstable (i.e., denser water overlies lighter water), the model applies vertical mixing until static stability is restored. The second criterion is based on the Bulk Richardson number, which represents wind-driven mixing. The mixed layer deepens until the Bulk Richardson number (Rib=(gΔρh)/(ρ0(ΔU)2) reaches or exceeds the critical values 0.6. The final criterion is the Gradient Richardson Number Rig=N2/(∂U/∂z)2 which accounts for shear instability. When Rig<0.25, local vertical mixing is applied. The PWP model is initialized with the ERA-5 12 h dataset of wind stress, heat, and freshwater fluxes (Hersbach et al., 2020) and the summer/winter averaged vertical profiles of temperature and salinity from the observations presented here (Fig. S1–3). The 1D model allows us to address the relative contributions from diurnal heating/cooling, freshwater fluxes, and wind mixing.

In addition, we broaden the impact of our findings by using the hydrographic database published in Brakstad (2023) that includes, in addition to the dataset from the Marine and Freshwater Research Institute of Iceland, other multiplatform observations like Argo floats or cruise data between 1950 and 2019. For this objective, a larger oceanic region is used and classified into α-ocean and β-ocean using the spice frequency, K2, (Carmack, 2007; Strehl et al., 2024), defined as: 5K2=NT2-Ns2. K2 is positive in α-oceans and negative for β-oceans.

The stratification around Iceland in summer is roughly an order of magnitude higher than in winter, largely due to positive buoyancy forcing, resulting in shallower and more variable MLDs. Therefore, we study the atmospheric effect on these α- and β-ocean regions by implementing the Price et al. (1986) one-dimensional model. The model is forced starting in the fall before the deep MLD develops. The model results of the MLDs shown in Fig. 8 are within the range of the observed average ± standard deviations (thick black dots and lines in Fig. 8). For most stations a spring shoaling of the MLD is driven by reduced heat fluxes, while the MLD remains relatively deep due to the wind-stress.

In the model, the stations embedded within the α-ocean with AW (Fig. 8: FX9, SB5 and ST5) present the largest MLDs exceeding 300 m depth, which is consistent with the observations. In this α-ocean region, the development of a deep ML is driven mainly by heat. However, within these stations, the wind-stress steadily contributes to the development of the ML. During the summer, shoaling of the mixed layer is likely influenced by the changes of both heat and freshwater fluxes, with their effects on the MLD partially offset by wind stress (Fig. 8: FX9, SB5 and ST5).

The station LB8, despite being in Denmark Strait and presenting a large contribution of PSWw and PSW in the upper layers (driving a β-ocean stratification), shows that the development of MLDs can be influenced by both, heat flux and/or wind stress (Fig. 8: LB8). However, the contribution of wind-stress and freshwater cannot lead to MLDs deeper than 100 m (Fig. S4: LB8). Beneath the PSW and PSWw at LB8 we find AtOW. Hence as the wind-stress develops, the MLD evolution erodes the PSWw strata reaching the AtOW layer, allowing reduced heat fluxes to contribute to the MLD development. This erosion is not visible on the stations embedded within the EIC.

Wind stress becomes the leading forcing mechanism northeast of LB8 at stations KG6, and SI8, coinciding with the shift from α- to β-ocean stratification (Fig. 8). This region has a lower convective potential than regions with pure AW and therefore does not produce large MLDs (Figs. 4 and 8). The MLD there results from roughly equal contributions of convection and wind-driven mixing. At these stations, the best performance of the PWP model is obtained when both heat flux and wind-stress are included (Fig. S4). Notably, the summer MLD remains shallow and is roughly the same order of magnitude across all stations around Iceland (Figs. 8 and S4).

To complement the understanding of the stratification of the Arctic and Subarctic waters around Iceland, their connection with water masses, currents, and their variability, we used the spice frequency averaged in the first 200 m, estimated following the methods described in Strehl et al. (2024) implementing Eq. (5). For this analysis we used the hydrographic dataset in Brakstad et al. (2023). The spiciness distributions shown in Fig. 9 reveal that temperature dominates on the southern side of Iceland, marked by an α-ocean regime, while salinity dominates the northern side, associated with β-ocean. These areas largely correspond with the distribution of AW versus PSW/PSWw (See Fig. 2 for T-S definitions).

The seasonal distribution of spice frequency north of Iceland agrees with the seasonal behavior of the NIIC described in Casanova-Masjoan et al. (2020) where the NIIC surface extension in winter and spring remains constrained to the northwestern side of Iceland due to the cold southeast surface imprint of the EIC. In summer, the NIIC expands northeastward, reaching all the way to the eastern side of Iceland and the northernmost station at the SI transect and constraining the polar water between the northern end of the Kolbeinsey Ridge and the Greenland shelf. In fall, the NIIC northern extension narrows but is still able to surround Iceland. It is noteworthy that a clear increase in coverage of the α-ocean, mainly in summer, is observed by comparing the 1980–200 average and the 2000–2010 average.

In this study, we discussed the seasonal and interannual variability of the mixed layer characteristics and the stratification regimes around Iceland by using a long time series of CTD data. Based on our results, we propose the regionalization of the waters around Iceland into three dynamical regions α-ocean, β-ocean, and transition-ocean.

The southwestern region is dominated by AW both in winter and summer. Within this region, the winter MLD is the deepest and most variable of the whole study area, with ML's occupied by AW over the whole sampling period. This region is influenced by the dynamics of the Irminger Sea and the Subpolar gyre and has favorable conditions (α -ocean) for winter deep convection driven by heat fluxes to develop deep mixed layers, which agrees with previous studies (Carmack, 2007; Våge et al., 2008; Piron et al., 2016; Stewart and Haine, 2016; Petit et al., 2020). In the southern region, the ML salinity anomaly was negative over the last 5 years, which is consistent with previous numerical models and Argo observations showing a freshening trend of the North Atlantic (Tesdal et al., 2018; Holliday et al., 2020; Liu et al., 2020). However, since the south of Iceland is an α-ocean, these recent changes have not yet reached or affected the MLD.

The northwestern region, which includes the Denmark Strait, is a medley of all the water masses described in this study. In this region there is a confluence of Greenland shelf and slope waters and Iceland Sea origin waters (Harden et al., 2016; Foukal et al., 2020). This includes the NIIC generating an important variability in the water properties due to complex interactions of the regional currents (Lin et al., 2020; Mastropole et al., 2017). The MLD variability over time at the KG6 and LB8 stations is moderate (<100 m), except for the years 2000, 2007 and 2016 when the ML was anomalously deep. In this region, the stratification is notably year-round dominated by salinity (β-ocean), which is explained by the strong Polar influence of cold and fresh waters transported by the EGC. A broader look into the northwestern side of the basin reveals that this area can be divided at the center of the Denmark Strait into an α-ocean near the Icelandic shelf where the NIIC flows and a β-ocean as we progress towards Greenland (where the LB8 and KG6 stations are located). Near the Icelandic shelf, MLDs are driven mainly by wind-stress, with a secondary contribution of heat fluxes. The T-S properties as well as the MLT anomalies in the northwest region near the Icelandic shelf show that the NIIC waters there are getting warmer and saltier. This agrees with previous studies showing the transformation of the NIIC also accompanied by an increase in its transport with time (Casanova-Masjoan et al., 2020). Even if the α-ocean area is warming, it is not expanding northwestward. This suggests that the EGC may act as a barrier, bringing PSW into the region and maintaining the β-ocean state on the northwesternmost side of the strait. This β-ocean, where the MLD is shallow all year round has dynamical implications as the strong shallow stratification inhibits baroclinic instability and eddy generation (de Marez et al., 2025).

Northeast of Iceland, the ML exhibits intrusions of SW in the winter, while AW is present during the summer. In this region, the stratification changes from β- to α-ocean seasonally. The Kolbeinsey Ridge acts as a barrier where we find the eastward penetration of the EIC bringing fresh waters (PSWw) from the East Greenland Current (Macrander et al., 2014; Casanova-Masjoan et al., 2020). In summer, the NIIC expands northeastward, bringing AW into the area and changing the stratification regime to α. Hence, the mixed layer waters show important seasonal variability. They range from maximum temperature below 2 °C in winter to over 10 °C in summer. This is also the only region where a significant warming decadal trend emerges over the interannual variability and progressively results in a stronger α-ocean. This agrees with the AW warming observed in Casanova-Masjoan et al. (2020) and with the northward progression of AW named as “Atlantification” described by Polyakov et al. (2017). This shift to α-ocean or “Atlantification” may lead to deeper ML's (Moore et al., 2015; Våge et al., 2022), and the associated deeper convection may increase the potential of this area to contribute to the dense flow carried by the NIJ (Semper et al., 2019).

The regionalization proposed in this work, based on hydrographic properties, matches the recently proposed distribution of primary production around Iceland (Richardson and Bendtsen, 2021; Cerfonteyn et al., 2023), supporting the importance of MLD properties for the primary production (Ólafsson, 2003). The induced alterations on primary production can lead to ecosystem changes. For example, Iceland has witnessed a rapid increase in the population of mackerel, a relatively warm-water fish, since 2006 (Astthorsson et al., 2012; Valdimarson et al., 2012; Campana et al., 2020) starting from the southeast towards the north, and recently they have been reported almost all around the country. This migration is consistent with the increase in surface temperatures, i.e., northward shift of warmer isotherms over the Iceland Faroe Ridge (de Marez et al., 2025) and the increase of temperatures within the ML in the same regions over the last decade, which may establish new pathways for entire ecosystems.

The long time series investigated here revealed important interannual oscillations of the ML properties. Five main features are to be highlighted: (i) We do not observe any linear trend in the MLD, which is rather subject to strong interannual variability. (ii) Except for the southern stations, influenced by the subpolar gyre, the interannual variability was not correlated with the NAO. For example, FX9 shows a significant negative of MLT with the NAO (R=-0.41 and p-value <0.03). (iii) The southern stations, FX9, SB5, and ST5, within the α-ocean region, show a clear decrease in ML density, with statistically significant values and experience a total decrease during the 29-year period of -0.08 kg m⁻³. (iv) The linear fit indicates significant (at 95 %) warming trends in the MLT of most of the stations in winter, with the maximum trend of 0.08 °C yr⁻¹ at SI8 resulting in approximately 2.2 °C. This agrees with previous studies (Sarafanov, 2009) showing that the northern part of the North Atlantic (south of Iceland) is strongly dominated by atmospheric interannual to decadal variability, particularly, where AW is present. The exception here is the northeastern region of Iceland where we observe a clear warming trend of the ML (2010–2020). (v) We observe an “Atlantification” expressed as a northeastward progression of the α-ocean state. This progression will highlight the role of the northeastern area of Iceland as a convective zone where deep water could be formed and contribute to the NIJ.