We combine new observations from two drifting profilers and a hydrographic field campaign with existing data from the monitoring station in Disko Bay to construct a hydrographic time series for 2022–2024 (Fig. ). Table  gives an overview of all hydrographic profiles included in the time series, and profile locations are shown in Fig. b. Additionally, we utilise profiles from five hydrographic surveys between 2022 and 2024 for a spatial analysis.

To estimate atmospheric forcing on local ocean variability in Disko Bay and across the shelf region, we use the European Centre for Medium-Range Weather Forecasts ERA5 reanalysis product . The product has 0.25° spatial and hourly temporal resolution. We obtained sea-ice concentration and 10 m u and v wind components over a region covering EDS (Fig. a).

For sea-ice concentration inside Disko Bay, we use data from the merged MODIS-AMSR2 satellite product . Daily sea-ice concentration (SIC) with 1 km resolution was downloaded for the Disko Bay region (68° 42^′ N–69° 31^′ N, 51° 24^′ W–53° W) and averaged spatially to obtain a time series of weekly mean SIC.

The two principal water masses are evident in the Temperature–Salinity (Θ–SA) diagrams from the Monitoring Station and Floats (Fig. a–c). At the surface, PW varies strongly over the seasonal cycle, with temperatures ranging from the in situ freezing point in winter to ∼8 °C in summer (Fig. a–c, d) and salinity ranging from ∼31 g kg⁻¹ in summer to 33–33.4 g kg⁻¹ in winter and early spring (Fig. a–c, e).

Between 50 m depth and the WGIW boundary, PW maintains a broad temperature range (∼0 to 4 °C), with the coldest subsurface temperatures in winter and early spring (Fig. b–c, d). After sea-ice melt, surface waters re-stratify and warm, while subsurface PW retains a characteristic temperature minimum. This minimum is most pronounced in summer and persists at greater depths in autumn (Fig. d).

Mixing with other water sources accounts for the wide property range of PW. Fresh meltwater pulls the near-surface observations along the runoff line in spring (Fig. a–c). Similarly, early autumn observations align with the runoff line, while later in autumn they shift towards the mixing line with submarine meltwater (Fig. a–c).

In contrast, WGIW (σ0>27.2 kg m⁻³) shows less temporal variability, with temperatures of 2–4 °C and salinity near ∼34.6 g kg⁻¹ (Fig. a–e).

The five near-synoptic GEM cruises (summers 2022–2024), together with overlapping Float 1 and 2 profiles, reveal significant spatial variability in PW properties across Disko Bay (Fig. ). This variability is especially pronounced in August, with differences of up to 3 °C along isopycnals in the σ0<27.1 kg m⁻³ range (Fig. b, d, f). In all August cruises, the coldest upper layer consistently occurred north of Ilulissat Icefjord, while the southwestern region near Aasiaat exhibited a notably warmer subsurface PW relative to all other areas. The spatial contrasts were most distinct in August 2022 (Fig. b), when the southern arm of the trough (near Aasiaat) contained the warmest subsurface PW, the middle arm was slightly cooler, and progressively colder conditions were observed towards the station nearest to the trough, then north of Ilulissat Icefjord, and finally at the Monitoring Station south of Disko Island. The central basin was the coldest area overall, aside from a sharp subsurface temperature minimum at the Monitoring Station.

In addition to along-isopycnal variability, PW also exhibits substantial variability along depth levels, particularly in August. The pattern is consistent with isopycnals sloping downward towards the coast, as previously documented north of Ilulissat Icefjord . In the spatial cruises, the same tendency is evident at the near-coastal stations both north of Ilulissat Icefjord and near Aasiaat, where, for example, the σ0=26.7 kg m⁻³ isopycnal was more than 60 m deeper than in the central parts of the bay.

In contrast, WGIW exhibits much weaker spatial variability. Near-synoptic observations within WGIW fall along a narrow line in the Θ–SA space (Fig. b–f), indicating minimal variability along isopycnals. Differences are somewhat greater along depth levels, but remain small: the PW–WGIW boundary depth varied by 20–30 m across the bay, and standard deviations in density at fixed depths were relatively small and decreased with depth. The average within-cruise standard deviation was ∼0.023 kg m⁻³ at 300 m and ∼0.014 kg m⁻³ at 400 m, with slightly larger variability in 2024 (both May and August), compared to 2022 and 2023 (Table ).

Because floats sampled the deeper parts of the bay, analysis of WGIW seasonality relies largely on float data. These observations can be used to assess WGIW variability along isopycnals with confidence, as well as at depth levels, provided that the variability in time exceeds the background spatial variability noted above. The wide spatial coverage of Float 1 in 2022–2023 is considered acceptable for studying WGIW properties, as that period showed limited spatial variability (Fig. b, c; Table ). In contrast, PW seasonality is better assessed from fixed-point observations at the Monitoring Station, since the pronounced spatial variability in PW would obscure the seasonal patterns if Float data are used.

The MLD, mixed-layer salinity, and mixed-layer temperature all exhibit a consistent annual cycle over the observation period (Fig. ). During winter, mixed-layer temperatures remain at the in situ freezing point (Fig. b), while sea-ice formation increases mixed-layer salinity through brine rejection (Fig. c). The maximum MLD occurs in winter, reaching 61 m in February 2023 and 48 m in January 2024. Continued sea-ice formation increases mixed-layer salinity to a peak of 30–33.4 g kg⁻¹ in March (Fig. c), although enhanced stratification and elevated salinity below the mixed layer limit further deepening (Fig. e–f).

As sea ice starts to melt in late April, the mixed layer warms and freshens. This timing is consistent for both 2023 and 2024 and agrees well with satellite-derived SIC, which drops below 10 % at that time (Fig. b–c). Sea-ice melt establishes a shallow mixed-layer (MLD < 20 m), which warms rapidly due to solar insolation (Fig. a–b). Through summer, increasing freshwater input lowers mixed-layer salinity by ∼2 g kg⁻¹, resulting in either a shallow or absent mixed layer (empty markers in Fig. a). By August–September, mixed-layer salinity reaches its minimum (31–31.4 g kg⁻¹), while temperature peaks at 8–10 °C (Fig. b–c).

In autumn, the mixed layer cools and deepens, accompanied by a gradual increase in salinity, returning the system to winter conditions and completing the annual cycle.

At the onset of summer, near-surface salinity reduces in conjunction with mixed-layer freshening (Fig. c–d), establishing a stratified layer with a core that is bound approximately by the σ0=26.5 kg m⁻³ isopycnal (Fig. e–f). As freshwater input and surface warming intensify, the stratified layer thickens and extends to greater depths, reaching 50 m by late summer, and rapidly extending to >100 m during autumn (Fig. e–f). Although both the Floats and the Monitoring Station capture this seasonal evolution, we assess PW seasonality primarily from the three-year Monitoring Station observations to avoid conflating the substantial spatial variability in PW that Floats would capture (Sect. 4.2).

The mean salinity of the upper 120 m, encompassing the depth range of the stratified fresh layer (σ0<26.5 kg m⁻³), shows a recurring seasonal reduction beginning in June (2023) or July (2022, 2024) and continuing until Monitoring Station observations end in November (Fig. a). The magnitude of the summer–autumn freshening varies interannually, reaching ∼0.5 g kg⁻¹ in 2022, ∼1 g kg⁻¹ in 2023, and ∼ 0.6 g kg⁻¹ in 2024. Within the σ0<26.5 kg m⁻³ layer, temperature rises through summer, peaks in August–September, and subsequently cools in autumn while salinity continues to decline (Fig. a–b).

Beneath this layer, denser PW (σ0≈26.5–27.1 kg m⁻³) maintains the summer temperature minimum characteristic of PW (Fig. d). In contrast to the lighter PW above (Fig. b), along-isopycnal temperatures within this denser PW layer continue to increase steadily through autumn – rising by 1.6 °C in 2022, 1.1 °C in 2023, and 0.4 °C in 2024 between August and November (Fig. c).

In Θ–SA space, this steady autumn warming of the denser PW (σ0≈26.5–27.1 kg m⁻³) appears as a gradual erosion of its sharp summer temperature minimum, with Θ–SA properties shifting towards warmer values during autumn (Fig. a–c). Given the strong spatial variability in August PW properties (Sect. 4.2; Fig. b, d, f), the warming at the Monitoring Station likely reflects advection from upstream. In August, the station near Aasiaat, in the bay's southwestern corner, consistently had the warmest PW, whereas the Monitoring Station had the coldest PW temperature minimum. When the August Θ–SA properties from Aasiaat are overlaid onto the Monitoring Station data, the latter's October–November properties closely resemble those observed upstream two to three months earlier (Fig. a–c), consistent with cyclonic circulation and advection of water masses. The ∼170–200 km distance between Aasiaat and Monitoring Station implies mean velocities of ∼3.3–3.9 and ∼2.2–2.6 cm s⁻¹ for a two- and three-month lag, respectively.

The magnitude of autumn warming at the Monitoring Station along a given isopycnal matches the temperature difference between the Monitoring Station and the Aasiaat station along that same isopycnal in August. For example, along σ0=26.8 kg m⁻³, warming from August to November was 1.65 °C in 2022, 1.1 °C in 2023, and 0.48 °C in 2024 along σ0=26.8 kg m⁻³ (Fig. c). These values are comparable to the corresponding along-isopycnal temperature offsets between the Monitoring Station and Aasiaat station in August: 1.89 °C in 2022, 1.1 °C in 2023, and 0.56 °C in 2024 (Fig. ).

A similar relationship was found only once between the Monitoring Station and the station north of Ilulissat Icefjord. In 2022, denser PW properties at the Monitoring Station in September resembled those north of Ilulissat Icefjord in August. The 1.2 °C warming along σ0=26.8 kg m⁻³ between August and September at the Monitoring Station (Fig. c) was comparable to the 1.04 °C temperature difference between the two locations in August 2022 survey (Fig. a). The distance of ∼90 km and a 1-month lag yield mean velocities of ∼3.5 cm s⁻¹.

At depths larger than 300 m, WGIW is isolated within Disko Bay by shallow bathymetric barriers (245 m at the entrance to Vaigat Strait, 245 m at Ilulissat Icefjord, 300 m at EDS; Fig. b). Regarded as basin water, WGIW can be renewed if equally dense or denser waters pass over the topographic constraints , with the deepest and most relevant being the EDS . The onset of a WGIW renewal is characterised by an increase in density below 300 m.

In early November 2022, basin density at 400 m increased from a September–October mean of σ0(400m)=27.31 to 27.36 kg m⁻³ by early December (grey shading in November–December in Fig. e). Basin temperatures rose simultaneously by ∼0.3 °C (Fig. f). Around two weeks after the onset of density increase, the upper WGIW boundary (σ0=27.2 kg m⁻³) shoaled rapidly by > 100 m within a span of two weeks (Fig. d), reflecting the uplift of lighter WGIW that previously resided in the basin. This renewal coincided with a seasonal shift in prevailing wind direction, whereby the along-coast (north–south) winds switched from being predominantly southerly in summer to northerly in autumn to spring (from positive wind stress values to negative in Fig. b). In the absence of strong sea-ice cover, northerly winds (negative wind stress) drive upwelling over the Egedesminde Dyb and its sill (Egedesminde Dyb Sill, EDS). Using hourly data, we define “strong upwelling” as WE≥0.45 m d⁻¹ (upper quartile of the hourly data distribution), typically derived under τy≤-0.06 N m⁻². November 2022 exceeded these thresholds with τ‾y=-0.1 N m⁻² and W‾E≥0.7 m d⁻¹ (Fig. b, c); WE frequently exceeded 1 m d⁻¹ (upper quartile), with episodic peaks approaching 3 m d⁻¹. The strength and persistence of this forcing appear to have lifted the dense waters to the west over EDS, initiating the observed basin renewal in November–December 2022 (Fig. d–f).

Through winter 2022–2023, basin density continued to rise gradually from σ0(400m)=27.36 to 27.39 kg m⁻³ between February and late April (Fig. e), accompanied by a further ∼0.5 °C increase in basin temperature which reached the annual peak at the end of April 2023 (Fig. f). This dense renewal lifted the overlying WGIW, which continued to rise until early June, when the upper WGIW boundary (σ0=27.2 kg m⁻³) rose to a minimum depth of 120 m (Fig. d). Northerly winds (τ‾y=-0.04 to -0.07 N m⁻², upper quartile ≤-0.12 N m⁻²), Ekman divergence, and positive vertical velocities (W‾E=0.47–0.82 m d⁻¹, upper quartile ≥1.5 m d⁻¹) persisted through February–April until the prevailing winds reversed in mid-May, ending the upwelling-favourable conditions (Fig. b–c).

In autumn 2023, the upper WGIW boundary (σ0=27.2 kg m⁻³) shoaled by ∼50 m in early November (Fig. d). There were no Float observations deep enough in the basin to document subsequent changes until early winter. However, as neither density nor temperature increased from October 2023 until January 2024, basin renewal likely did not occur. The average wind stress was near zero from September to December, with brief episodes of negative τy in October and November (Fig. b). Calculated upwelling velocities were W‾E=0.3 m d⁻¹ in October, with short-lived episodes of WE>0.6 m d⁻¹ (upper quartile), but overall, the forcing was weaker and less persistent than in November 2022 (Fig. c).

There are signs of a renewal in March–April 2024, when σ0(400m) increased by 0.01 kg m⁻³ and the temperature rose by ∼0.3 °C within one month (orange shadings in Fig. e, f). Lack of observations between 23 January and 11 March hinders the ability to determine when this renewal began; however, density and temperature in the basin were already higher by 0.02 kg m⁻³ and ∼0.4 °C by 11 March, suggesting it was already underway. Winds during January–April were upwelling-favourable (τ‾y=-0.01 to 0.05 N m⁻²), with mean vertical velocities of 0.1–0.6 m d⁻¹. The strongest forcing (upper quartile) occurred in March with τy≤-0.11 N m⁻² and WE≥1.1 m d⁻¹. By late May–early June 2024, the WGIW boundary rose to a minimum depth of ∼150 m (Fig. d).

In both years, the upper WGIW boundary was shallowest in May–June, shoaling to 120 m in 2023 and 150 m in 2024 (Fig. d). This uplift brought WGIW waters in Disko Bay above 240 m depth, corresponding to the Ilulissat Icefjord sill depth. At 240 m depth, density showed a pronounced seasonal cycle, peaking at σ0=27.3–27.35 kg m⁻³ in May and declining to a minimum of σ0=27.1 kg m⁻³ in August–November (Fig. a). Temperature at 240 m varied in parallel, with an amplitude of ∼2 °C and maxima above 3.5 °C in May–June (Fig. b). A renewal event in November–December 2022 resulted in ∼1 °C warming at 240 m (Fig. b).

Over two annual cycles (2022–2023, 2023–2024), renewal of WGIW in Disko Bay occurred primarily in spring, when the densest WGIW filled the basin, resulting in annual maxima in density, temperature, salinity, and vertical extent of WGIW (Fig. d–f). The springtime renewal began in February–March and lasted until May–June. This provides direct observational evidence supporting the hypothesis of , who proposed that exchange over the EDS likely occurs in spring. This timing also matches observations further north in Uummannaq fjord system , and south in Godthåbsfjord . At 240 m in Disko Bay (Ilulissat Icefjord sill depth equivalent), WGIW density peaks at σ0=27.3–27.35 kg m⁻³ in May (Fig. a), matching the density range observed within Ilulissat Icefjord basin . This correspondence suggests that spring renewal in Disko Bay delivers the densest waters entering the fjord basin.

The repeated springtime renewal in Disko Bay is consistent with regional-scale hydrographic variability. At Davis Strait, isopycnals tilt up towards the Greenland shelf in spring, with ∼140 m vertical displacement of WGIW at the shelfbreak coincident with densification along the shelf . Increasingly dense waters thus become available over the EDS, eventually exceeding the density of resident basin waters in Disko Bay and driving the springtime renewal (Fig. b). Persistent upwelling-favourable winds locally through spring likely enhance the renewal process (Fig. b–c).

In addition to the repeated spring renewal, we observed a distinct renewal in November–December 2022, marked by rapid increases in basin density and temperature and a more than 100 m shoaling of the WGIW boundary (Fig. d–f). This autumn/winter renewal was unique to the 2022–2023 annual cycle and coincided with particularly strong and persistent upwelling-favourable winds. These conditions likely induced upwelling over the EDS, lifting the dense waters over the sill and into Disko Bay (Fig. a). Estimated vertical velocities frequently exceeded 1 m d⁻¹, with episodic peaks approaching 3 m d⁻¹, implying uplift on the order of 20 m during this period. The actual uplift may have been greater, as our estimates only quantify the effect of wind stress curl, and not coastal upwelling, which could also contribute, given the proximity of the coastline east of EDS.

While we cannot determine the precise magnitude of upwelling or the properties of upwelled waters over the EDS without hydrographic observations on the shelf, the conditions observed are consistent with other studies linking upwelling-favourable winds to WGIW intrusions along the west Greenland shelf. For example, upwelling-favourable winds were linked to the observed increased WGIW presence in the Uummannaq trough (300 km north of Disko Bay) during December–January , and to modelled areas of frequent upwelling along the coast both north and south of Disko Bay . Near Cape Farewell, wind-driven upwelling events have been shown to draw Atlantic-origin waters from ∼250 m depth onto the shelf (∼150 m), raising both temperature and salinity . In southeast Greenland, at Sermilik Fjord, along-shelf winds were also shown to play an important role in uplift and onshore intrusion of dense Atlantic Waters .

Our results highlight that a pronounced autumn/winter WGIW renewal, such as in 2022, can shoal the WGIW boundary enough to increase the temperatures at Ilulissat Icefjord sill depth by ∼1 °C (Fig. b). Ilulissat Icefjord renewal is thought to be primarily driven by subglacial discharge during the melt season . However, our observations show that episodic uplift of dense waters in Disko Bay during autumn–winter can also bring dense waters above the fjord sill and possibly initiate fjord basin renewal earlier in the year than previously expected , even during periods of limited subglacial discharge forcing . Observations in Disko Bay reported by indicate that late autumn–early winter renewal may have taken place in 2021 as well, as temperatures at 240 m increased rapidly by ∼0.5 °C.

We therefore extend the schematic of to include a wind-driven autumn/winter renewal pathway across the EDS into Disko Bay, while also acknowledging the possibility that such events may enable dense inflow into Ilulissat Icefjord basin (Fig. a), although direct evidence for this process is not yet available.