The Magellan Strait, situated in the southern Patagonia fjord region, connects the Atlantic and the Pacific oceans. Its internal dynamics govern the hydrography of the surrounding inland sea. Despite a sparse population, southern Patagonia possesses considerable economic significance for Chile. The growth of industries such as aquaculture, oil, green hydrogen, and tourism has altered land use, transport, and increased the ecological pressure on the Magellan Strait, Tierra del Fuego, and the Beagle Channel (Ariztía and Undurraga, 2025; Giesecke et al., 2024).

Large freshwater inputs from riverine and glacial sources significantly affect the region (Sassi and Palma, 2006). The low salinity of the Magellan Strait results from a north-eastward flow from the Patagonian Current (Brun et al., 2020). The distribution of these waters is governed by the combined effects of wind, tides, and the Antarctic Circumpolar Current (Palma and Matano, 2012). Previous research indicates that waters from the Atlantic Ocean extend their effect westward through the Magellan Strait as far as Punta Arenas. In contrast, the effect of the Pacific Ocean extends eastward to Carlos III Island, where a shallow sill limits its influence on surface waters (Antezana, 1999; Brun et al., 2020).

An important geographical feature is the Darwin Cordillera (DC) mountain range which extends 200 km from west to east along the southwestern peninsula of Tierra del Fuego Island (Fig. 1). The DC is the core of the southern westerlies and experiences low precipitation. The region is characterized by a strong W–E temperature gradient (Carrasco et al., 2002; Garreaud et al., 2013; Meier et al., 2018; Izagirre et al., 2024). Near the study region (Fig. 1c), the glaciers of Parry fjord have been relatively stable since 1986, but the Darwin glacier retreated 3 km between 1986 and 2014. In this period, the region's glaciers have lost 10 % of their area (Rivera et al., 2017).

The Almirantazgo Fjord (AF) is 75 km long, with a width of 16 km in the northern part and 5 km near its head with an averaged width of 12 km. The AF axis is 116° from the true north, an approximately East-Southeast (ESE) direction. It contains three major glacial connections: Brookes fjord (BF), the Ainsworth bay (AB), and Parry fjord (PF). At the head of the fjord, in Maria Cove, the fjord is shallow, with a mean depth of approximately 30 m for the first 7 km. The bottom then slopes steeply until the mouth of PF, deepening from 30 to 200 m in less than 3 km. From PF to BF, the fjord depth is approximately 220 m. The deepest part of this basin is in the Whiteside channel (WC), where the bottom depth is approximately 450 m. This is one of the deepest parts of the inner sea of Tierra del Fuego (Fig. 1).

Meteorological data for the region were obtained from the Dirección General de Aeronáutica Civil (https://climatologia.meteochile.gob.cl/, last access: 16 June 2026), sourced from airports at Punta Arenas (53.00° S, 70.84° W), Porvenir (53.25° S, 70.33° W), and Puerto Williams (54.93° S, 67.62° W). The dataset included wind magnitude and direction (meteorological convention), recorded at 1 h time intervals between 1960 and 2025.

Additionally, to characterize the wind patterns, this study uses wind reanalysis data from ERA 5. The wind components at 10 m high (u10, v10) were downloaded in spatial grids of 0.25° × 0.25°. The ERA5 is the fifth-generation ECMWF reanalysis of global climate and weather data. Reanalysis integrates model data worldwide observations into a globally complete and consistent dataset on physical laws (Hersbach et al., 2023).

Wind-stress (τ) was calculated using the bulk formula by τ=racd|U10|U10, where ra is the air density (1.2 kg m-3), U10 is the wind velocity vector (with components u10, v10), |U10| is wind speed (the scalar magnitude of U10) and cd is dimensionless wind-drag coefficient. The coefficient cd was calculated for each wind component value in the time-series following Yelland and Taylor (1996), for |U10|<6 m s⁻¹, cd=(0.29+3.1|U10|-1+7.7|U10|-2)×10-3, whereas, for 6 m s⁻¹ ≤ |U10| ≤ 26 m s⁻¹, cd=(0.60+0.07U10)×10-3.

Sea level data were acquired from four tide gauge stations located within the inner sea of the Magellan Strait and the Almirantazgo Fjord (Fig. 1): Gregorio Cove, Punta Arenas, Puerto Williams, and Ushuaia (Table 1). The dataset was obtained from the Sea Level Station Monitoring Facility (https://www.ioc-sealevelmonitoring.org/, last access: 16 June 2026) for the period of January to June 2024. At each station, measurements are recorded at a one-minute interval by a pressure sensor positioned near the seabed.

Additionally, a barotropic tidal global model TPXO9 (Egbert and Erofeeva, 2002) was applied to determine the spatial distribution of amplitude and tidal current ellipses in the region (Fig. 3). The model provides complex tidal heights (h) and transport coefficients for 15 constituents at 1/30° resolution. The principal harmonic amplitudes (M2, N2, S2, O1, and K1) were used to estimate the form factor (e.g. Pan et al., 2024) defined as F=(O1+K1)/(M2+S2) was useful to characterize the regimes of the barotropic tide in the inner sea of Tierra del Fuego.

The hydrography data were obtained using two different complementary methods: conventional CTDOF (AML Metrec X) deployments and CTDF (CTD-SRLD) casts from sensors affixed to Southern elephant seals.

The conventional instrumentation consisted of a CTD AML–Metrec X equipped with Temperature, Conductivity, Pressure, Fluorescence/Chlorophyll, and pH sensors, supplemented by an additional Aanderaa Optode 4831 infrared Dissolved Oxygen sensor. The CTDOF was configured for freshwater settings to permit measurements under low-conductivity conditions, with profiles acquired at a rate of four scans per second. At each oceanographic station, the instrument was activated on deck and subsequently lowered to 1 m below the surface, where it was held for one minute to ensure sensor stabilization. Data acquisition was then performed using an electrical winch to a depth up 200 m. To describe the oceanographic conditions, sampling was conducted at five cross-fjord oceanographic stations (Fig. 1). Data were collected during the austral summer (15 and 31 January 2024) and the austral winter (26 and 30 June 2024). A total of 60 cross-fjord profiles were acquired to characterize the oceanographic conditions near the head of the Almirantazgo Fjord. Post-processing of the AML-CTD data followed standard quality control procedures, consistent with other CTD instruments. This process included: visual inspection to remove spikes, exclusion of sensor stabilization time, selection of downward profiles, and vertical binning at 0.5 m intervals. Subsequently, mean profiles for January and June were calculated for each station and are presented in Fig. 6.

Oceanographic conditions in the region were also assessed using CTD-SRLD (Boehme et al., 2009) and CTD-Fluoro tags (Guinet et al., 2013). The CTD tags possess high accuracy (±0.005 °C for temperature, ±0.01 mS cm⁻¹ for conductivity, and 2 dBar ± (0.3 + 0.035 % reading) K⁻¹ for depth) and were calibrated by the manufacturer prior to deployment. These tags were affixed to Southern elephant seals (Mirounga leonina) from the colony located at Jackson Bay, located at the head of the Almirantazgo Fjord (Fig. 1c). In-situ temperature and salinity measurements were converted to Absolute Salinity (SA), Conservative Temperature (Θ) following the TEOS-10 (McDougall and Barker, 2011).

Water column stratification was characterized by the buoyancy frequency (N2), calculated as N2=-g/ρ0∂ρ/∂z, where g is the gravitational acceleration, ρ0 is the reference density, and ∂ρ/∂z is the vertical density gradient. Additionally, CTD data were used to estimate the energy required to homogenize the water column, expressed as the Potential Energy Anomaly (PEA), PEA=1/h∫-d0gz(ρp-ρ)dz, where h is the thickness for the calculation of average density (ρp). A high PEA value corresponds to high stratification (Simpson et al., 1979). The Freshwater Content (FWC), defined as the equivalent thickness of freshwater in the water column, was also calculated, FWC=∫-d0(s0-s)/s0dz, where s0 is the reference salinity at depth d (Blanton and Atkinson, 1983), here the study used s0=31 g kg⁻¹ which was the maximum salinity in the deeper zone of the Almirantazgo fjord. In all the profiles FWC was calculated by integrating over the entire water column down to the bottom depth.

To describe the hydrographic along-fjord patterns for January and June 2024, this study utilizes CTD-SRDL data collected between Inútil Bay (IB) and María Cove (MC). Although the elephant seals provided numerous hydrographic profiles across the region, the study defined an along-fjord route along the central axis between IB and MC (see Fig. 5a). This route consisted of points spaced at 0.1 km intervals over a 150 km span. All profiles within 2 km of each point, in a temporal window of 3-weeks centred at the CTD-AML dates (see Table 1), on the route were selected and averaged to obtain a single representative profile per point. For the vertical dimension, standard depths were established at every 1 m (0–20 m), 5 m (20–50 m), 10 m (60–100 m), and 50 m (150–500 m). Vertical interpolation was performed only at these standard depths, avoiding spatial (distance) along-fjord interpolation. This procedure was applied to the January and June 2024 data shown in Fig. 5c, d.

To assess the influence of different water masses, an Absolute Salinity of 31 g kg⁻¹ was used as an index defining the upper boundary of the Modified Subantarctic Water (MSAAW). This water mass is produced by the combination of surface freshwaters with Subantarctic Water (SAAW), the dominant water mass within the region (Silva and Vargas, 2014).

Time series data of the water column properties were acquired using two distinct but complementary ways. The first method comprised a six-month (January–June 2024) near-bottom (30 m depth) record of sea level, salinity, temperature and dissolved oxygen. The second method involved short-term (5 d) acquisitions of a fine vertical-scale temperature, salinity and dissolved oxygen during January 2025 (Fig. 2).

The six-month time series was collected using a CTD WiSens NKE Instruments (https://nke-instrumentation.com/, last access: 16 June 2026). This instrument was secured within a stainless steel, pyramid-shaped frame (1 m × 1 m; 0.7 m height) positioned on the seabed at 30 m depth near to 54.44° S, 69.05° W. The instrument specifications include a pressure range up to 300 m (0.1 % accuracy), a temperature range of the -2 to 35 °C (±0.02 °C), and salinity range 2 to 42 PSU (±0.1 PSU). Dissolved oxygen (DO) concentration was recorded by a PMEL MiniDOT (https://www.pme.com/products/minidot, last access: 16 June 2026), which functions via an optical sensor. This sensor has an accuracy of ±0.2 mL L⁻¹ and a resolution of 7×10-4 mL L⁻¹. The MiniDOT also contains temperature sensors with a range of 0 and 35 °C ±0.1 °C and its housing is rated for 300 m depth. Both the Wisens NKE and MiniDOT sensors were affixed in individual stainless-steel frames attached to the main pyramid structure. This entire assembly was lowered to the bottom and remained moored until retrieval at the end of June 2024.

The 5 d time series was acquired from a mooring at A2 station (Fig. 2, Table 1) between 22 and 26 January 2025. This instrument array was designed to measure the vertical gradient of temperature, salinity, DO, and pressure. The array included CTD NKE WiSens and a miniDOT positioned at 0.3 m depth and at 30 m depth (near the bottom) to determine the vertical gradients of salinity (∂S/∂z), temperature (∂T/∂z), and dissolved Oxygen (∂DO/∂z).

To characterize the temporal variability and dominant oscillations observed in the data, a Morlet wavelet analysis (Torrence and Compo, 1998; Grinsted et al., 2004) was applied to the time series of wind stress, sea level, Conservative Temperature, Absolute Salinity, and Dissolved Oxygen. This method permits the computations of both the dominant modes of variability and their temporal variations (Torrence and Compo, 1998). To quantify the relationship between the wind stress and the other variables, wavelet coherence was estimated following the methodology of Grinsted et al. (2004).

The potential for the wind stress to perturb the pycnocline was assessed using the dimensionless Wedderburn number (Wb) following Geyer (1997), Thorpe (2005) and Inall et al. (2015). The number is defined as: Wb = (τL)(ρg)-1h1-2, where L is the length of the fjord (or the horizontal scale of the wind), g is gravitational acceleration, and h1 is the depth of the upper layer directly influenced by the wind stress τ. In a simplified two-layer model, the density difference ρ=ρ2-ρ1 where ρ1 and ρ2 are the densities of the upper and deeper layer, respectively. According to this formulation, a value of Wb < 1 indicates that the density gradient dominates. When Wb = 1, wind forcing and buoyancy effects are comparable, and horizontal mixing becomes significant When Wb > 1, wind-driven effects are dominant (Inall et al., 2015).

The 54-years hourly reanalysis dataset was used to establish the wind climatology for the study region (Fig. 2). The results indicate that seasonal winds exhibit high directional consistency. Westerlies are the dominant pattern; south of 54° S in the Pacific Ocean, these winds acquire a slight southerly component, with the airflow generally following the coast. Wind magnitudes intensify near the coast, particularly during the warmer seasons (austral spring and summer), when speed exceeds 8 m s⁻¹ (Fig. 2a, b). During the colder seasons (austral autumn and winter), regional winds are weaker (<6 m s⁻¹), particularly offshore in the Pacific Ocean. Within the inner sea of the Magellan Strait and Almirantazgo Fjord, wind magnitude was low (<4 m s⁻¹), comparable to those observed in the open ocean. Regarding direction, winds within the Almirantazgo Fjord appear aligned with the fjord's axis and generally flow toward the fjord head throughout all the seasons (Fig. 2).

The climatology from the meteorological stations indicates a high consistency with the regional winds described previously (see Fig. A1 in the Appendix). At Punta Arenas (PA), Porvenir (PV) and Puerto Williams (PW), eastward winds (meteorological westerlies) were predominant (>25 %) during all the seasons. The highest wind speeds (ca. 15 m s⁻¹) were observed at PA and PV during summer and spring. PV was an exception, as northward/southward winds were also recorded there during autumn and winter. For the Almirantazgo Fjord (AF) wind data were derived from a time series of an ERA5 reanalysis pixel (Fig. 3). At this location, the distribution of magnitude and direction was highly consistent with the regional patterns, showing a dominance of eastward winds in all the seasons. The directional distribution indicates that up-fjord winds (directed towards the fjord's head) are prevailing conditions in AF (see Fig. A1).

Tidal amplitudes recorded by the region's tide gauges (Fig. 1, Table 2) were primarily controlled by the M2 semi-diurnal harmonics. The Gregorio station measured an M2 amplitude of 132.7 cm, a magnitude at least double that observed at the other stations (which ranged from 48 to 56 cm) located south of the Magellan Strait. At the Punta Arenas station, the M2 amplitude was 48.1 cm. In the Beagle Channel, south of the Almirantazgo Fjord, the Puerto Williams and Ushuaia stations recorded comparable M2 amplitudes of approximately 55 cm. The lowest F value (0.33) was observed at Gregorio, whereas at the other stations, F>0.6. The form factor (F) indicated that all the stations are characterized by a mixed, mainly semi-diurnal tidal regime.

To quantify the tidal influence in the study region, the barotropic tidal Model by Egbert and Erofeeva (2002) indicated that the northern part of the Magellan Strait and its connection with the Atlantic Ocean showed the highest M2 amplitude (>3 m). Amplitudes decreased southward, falling below 1 m off Punta Arenas and reaching approximately 0.5 m inside the Almirantazgo Fjord (Fig. 3). A marked change in amplitude (from high to lower) was observed at the Second Narrow, phase tends to increase from the Atlantic (10°) to the inner-sea reaching 90° into the Inutil Bay. The entire basin from the Inutil Bay to Maria Cove seems to be in phase at 90°. Other harmonic constituents were substantially smaller than M2 accounting for <10 % of the variability explained by the M2 component (Fig. 4).

The whole CTD-SRLD dataset (showed in Fig. A2) acquired from Southern elephant seals encompassed extensive areas, including the Pacific and Atlantic oceans and the inner sea of the Magellan Strait and Tierra del Fuego. To specifically address the dynamics of the Almirantazgo Fjord, the present analysis utilized only the CTD-SRLD data obtained from within the fjord. The hydrographic data reveal two distinct regimes for the waters of southern Patagonia (see Fig. A1d). Within the inner channels of Tierra del Fuego, low absolute salinity (ca. 31 g kg⁻¹) was observed in deeper waters (depths >100 m). Conversely, waters in the outer channels (in the Pacific and Atlantic oceans) were characterized by salinities exceeding 34 g kg⁻¹. These findings indicate that the inner channel waters are more stable (stratified) compared to the mixed conditions prevailing in the outer channels.

To compare and describe the seasonal patterns of the Almirantazgo Fjord, this study utilized CTD-SRLD data acquired from Southern elephant between January and June of 2024. The analysis was restricted to Conservative and Absolute Salinity measurements collected within the region extending from the Whiteside channel and Almirantazgo Fjord (Fig. 5).

The Conservative Temperature/Absolute Salinity (CT/SA) diagram (Fig. 5) acquired by elephant seals, illustrates the wide hydrographic variability within Almirantazgo Fjord. SA ranged from a minimum of 26.28 g kg⁻¹ near the surface (4 m depth) to a maximum of 31.13 g kg⁻¹ near the fjord bottom. CT varied from 4.24 to 10.94° C with a mean of 7.20±1.12 °C. The conservative density anomaly (segmented lines, Fig. 4b) extended from 1020.0 kg m⁻³ in the surface waters during January to a maximum of 1024.5 kg m⁻³ in near-bottom waters during June. The mean CT and SA profiles (Fig. 5c, d) show pronounced upper-water-column stratification in summer, contrasting with the well-mixed conditions observed in winter.

The cross-fjord transect data were acquired using a CTD-AML Metrec during field measurements in January and July 2024. During the austral summer, SA ranged from a minimum of 26.30 g kg⁻¹ near the surface to a maximum of 30.99 g kg⁻¹ near the bottom, with a seasonal average of 30.21±0.10 g kg⁻¹. Water temperature varies between 6.132 and 10.81 °C, with a mean of 7.68±0.03 °C.

During winter (June 2024), conditions in the Almirantazgo Fjord (AF) shifted. SA ranged from 29.01 to 31.01 g kg⁻¹ with a mean of 30.78±0.04 g kg⁻¹. The water column was colder, with CT between 2.97 and 7.74 °C (mean: 7.01±0.06 °C). A thermal inversion was observed, with colder upper waters overlying warmer deeper waters. Density range was 1018.8 to 1024.5 kg m⁻³. Mean cross-fjord profiles indicated that stratification was salinity-dominated. In summer, continuously stratified conditions were present in the upper 30 m depth layer. In winter, however, the SA profile was nearly homogeneous throughout the upper 100 m of the water column.

Based on the along-fjord and cross-fjord CTD sections and mean profiles (Figs. 5c, d and 6c, d), the study determined that during January, the thickness of the upper brackish layer near the head of the fjord (at PF) was h1=20 m, compared to h1=5 m during June. At the mouth of Almirantazgo Fjord, the upper layer depth was estimated to be h1=5 m during both months. These observations are consistent with a salt-wedge structure in the upper layer during January that was absent during June (Figs. 5 and 6).

Both sets of hydrographic data (along-fjord and cross-fjord CT and SA profiles) were used to estimate the Freshwater Content (FWC) and Potential Energy Anomaly (PEA) for January (summer) and June (winter) (Fig. 7). During January, the FWC exhibited a gradient, decreasing from the fjord head toward the Magellan Strait (MS). At 3.9 km from the head, which is near Parry Fjord (PF), the FWC was 1.37 m then FWC decrease trough the fjord's mouth presenting peaks events (ca. 0.6 m) near of Brookes Fjord (BF) at 71 km from the head reaching a minimum FWC of 0.14 m near the connection with the Magellan Strait (Fig. 7a). The along-fjord structure of PEA closely resembled that of the FWC, displaying a maximum of 157.2 J m⁻³ near PF and secondary maxima at 14 km (ca. 152.5 J m⁻³) and BF (ca. 71.3 J m⁻³). In June (winter), both FWC and PEA were lowers and noisy with lower values (FWC ∼ 0.2 m, PEA ∼ 7 J m⁻³) around BF location (70 km from the head) (Fig. 7b).

The cross-fjord FWC and PEA values were consistent with the along-fjord estimations, as summer values exceeded those recorded in winter. In summer, FWC ranged from 1.8 m (north side) to 2.1 m (south side), displaying a slight tilt toward the southern portion of the transect near PF (Fig. 6c). PEA values reached 200 J m⁻³ and exhibited a cross-fjord tilt analogous to that of the FWC. During the winter, FWC was less than 1.0 m and showed only a minimal cross-fjord tilt to the south. The winter cross-fjord PEA was <40 J m⁻³ a value approximately five times lower than observed in summer (Fig. 7d).

The wind-stress magnitude exhibited high spectral amplitude for the oscillation with periods greater than 24 h. This diurnal band showed its greatest amplitude from late January until the beginning of May, during the winter, this band was virtually absent (Fig. 10a). The subsequent period was observed in the 3 d band, which displayed high amplitude at the beginning of the time series (25 January to 24 February) and again around 20 March and mid-May. The 10 d band was the most significant period for the wind-stress magnitude. The first two weeks of February was the only time this band showed lower amplitudes; during that period, the variability was centred at 3 d. Furthermore, distinct events with high amplitudes spanning a broad range of periods (12 to 256 h) were centred on specific dates: 5 February, 30 March, 20 April, 10 May, 3 June, and 14 June (Fig. 10a).

The sea level (SL) data were dominated by tidal variability, with the highest spectral amplitudes occurring in the semi-diurnal (12 h) and diurnal (24 h) bands. Both bands exhibited a pronounced fortnightly modulation; this modulation became less distinct in the 12 h band from mid-May until the end of the time series. Other evident variability was contained at 4 h band with low amplitude and bi-weekly modulation and the low-frequency (longer than 1 d periods) which showed marked high amplitudes at 3 d (around 72 h) at the beginning of the time-series until middle of February and beginning of April. The most important (and significant) period band was centred at 10 d (256 h) into the synoptic band. This band persists along the entire time-series but diminishes its amplitude between May and June (Fig. 10b).

The CT data indicated the significance of the 3 d band. High amplitudes were observed in this band during January and February; its amplitude subsequently decreased until 3 and 14 June, when two distinct high-amplitude pulses were recorded (Fig. 9c). The global spectrum confirmed the relevance of the synoptic band, centred at 10 d (approx. 256 h). The wavelet power spectrum shows that this 10 d band was less prominent during March and April (Fig. 10c).

The DO wavelet exhibited a pattern analogous to that of CT at the start of the time series, characterized by high variability within the 3 d band. However, unlike CT, the DO data displayed high amplitudes across the 3 to 10 d bands between March and April; during this interval, the 5 d band appeared to intensify around 20 March and 24 April. Throughout the record, multiple high-amplitude events occurred across various periods (Fig. 10d).

For SA, the time series from 23 January to 5 March was characterized by low spectral amplitudes across all periods resolved by the wavelet analysis. Following this, from March to June, high amplitudes developed at periods greater than 3 d. Within the 5-to-10 d, SA variability initially exhibited high amplitudes centred at 5 d (late April). From mid-May until the end of the record, the dominant variability shifted to the 10 d band. High-amplitude events spanning all periods >24 h were observed centred on 3 and 13 June.

The CTD-SRDL data was used in the estimations the upper outflows exchange and to determine flushing times of the same layer. The mean profiles from the along-fjord CTD-SRDL (Fig. 5c, d) and the mean profiles of the CTD-AML (Fig. 6c, d) were highly consistent between the seasons; highly stratified in January and mixed during June. Despite the fact, the study has several profiles in the region: inner the fjords and in the open ocean (see Fig. A2). In the Almirantazgo fjord, only few CTD-SRDL profiles were nearest to the cross-fjord stations (see Fig. 6a). Thus, the study was able to make a comparison for data of the 27 of January and between 6 and 9 June 2024. During those dates, CTD-SRDL data was taken nearest to CTD2 and CTD3 stations (Figs. 2, 6a). Considering that the study used a h1=20 m during January, the comparison between upper layer (CTD depths ≤ 20 m) and deeper layer (CTD depths > 20 m) were used to quantify the agreement between CTD-SRDL and CTD-AML. The comparisons showed the best agreement for January temperatures (r>0.96). During June, the relation was lower with r>0.80 in the upper layer. The summer Salinities presented RSMD > 2.61 g kg⁻¹ with r=0.67 at CTD2 in the upper layer. In June, RMSD was lower than January (RSMD < 1.10 g kg⁻¹) with r between 0.85 (CTD2) and 0.94 (CTD3). In oceanic water, the performance change, the lowest errors and highest correlations were concentrated for profiles with depths > 50 m with RMSD < 0.3 and r > 0.8 for both temperature and salinity, indicating a good overall representation. Although the values in the surface layer were slightly lower, they remained consistently good for both variables. The low correlation in Salinity could be taken with the consideration that, (1) few profiles were able to make the comparison, (2) the dates selected were near under variable conditions of Salinity (see Fig. 8e), (3) the processing of the traditional CTD take the downcast to be analyzed, whereas CTD-SRDL record data when elephant seals are coming to surface.

Utilizing the hydrographic data acquired from two complementary sources (CTD-SRLD and CTD-AML), within the Tierra del Fuego fjord system, and considering the basin geometry, a first-order estimation of the flushing time for the upper layer (TF) of the Almirantazgo Fjord was estimated (from MC to BF). This approximation, based on the Knudsen theorem, this requires determining the upper Volume of the brackish layer. The study shows that during January the amount of freshwater was considerably higher than June (Fig. 7), additionally cross-fjord transect shows that during January the upper brackish layer (h1) could be as deeper as 20 m but during June was shallower, h1=5 m (Fig. 6c, d). Similarly, the along-fjord patterns showed that during January the fjord showed a salt-wedge pattern deeper near the head and shallower (ca. 5 m) near the mouth of the fjord (Fig. 5c) in contrary, during June the upper brackish layer has a constant and shallower (ca. 5 m) along the entire fjord. These cause geometry differences for the Volume between January (right trapezoidal prism) and June (right rectangular prism). Considering the average width of AF was nearly 12 km the upper Volume (V1) of the brackish layer was 11.25×109 m³ in January whereas in June was 4.5×109 m³ yields that the upper brackish layer of January was 2.5 times of June.

At the mouth of the AF (Fig. 1) the study estimates the along-fjord exchange of the upper layer, Q1=A1⋅v1, here A1 was the transversal area at the mouth and v1 is the advective along-fjord current. Expanding, A1=b⋅h1, taking an averaged width of 12 km and h1=5 m the transversal area in both January and June was A1=6×104 m². This study do not take currents measurements but the Fisheries Chilean Institute (IFOP) is working to modelling the Magellan strait, they use a MIKE 3D model for modelling (http://chonos.ifop.cl/, last access: 16 June 2026) the northern Patagonia (between Puerto Montt and Golfo de Penas, see Fig. 1) and they using to estimate different operational results for decision makers (e.g. Reche et al., 2021; Ruiz et al., 2021). For the study region, the model lack in circulation validation (Cristian Ruiz, personal communication, 1 May 2026) but the IFOP give us a weekly averages circulation patterns for January and June 2024 (see Fig. A3). Using that model output, the study obtained typical v1 to determine the upper exchange Q1. The results showed that during, January v1=0.3 m s⁻¹ whereas in June v1=0.1 m s⁻¹ considering these values the output exchange in January was Q1=1.8×104 m³ s⁻¹ and during June Q1=6×103 m³ s⁻¹. Using that estimations to determine the flushing time (TF=V1⋅Q1-1) of the upper layer, using this approximation TF for January and June were similar order of magnitude of one week (Table 3). In this study high RMSD for surface Salinity could limited the precision of the Knudsen-based estimations. Despite that, the TF quantification requires a second-order studies focusing into determine the water column exchanges using a validated circulation model. As an example, Pinilla et al. (2020) using a circulation validated model, determine the water age into the Puyuhuapi fjord obtaining longer than 1 year for intermediate and deeper waters of the fjord which could promote low DO in that region.

The calculated upper layer flushing times were relatively longer than those reported for other highly stratified systems in northern Patagonia. For instance, Castillo et al. (2016) determined an upper-layer flushing time of 3 d for the Relocavi Fjord, and Calvete and Sobarzo (2011) found a similar 5 d flushing time for waters between the Guafo mouth and Elefantes Fjord. These results indicate that the upper layer of the Almirantazgo Fjord has flushing times of approximately one week, which may promote the concentration of materials within the basin, but the limitations of the Knudsen theorem must be taken into account and to serve as a tool that supports management and decision-makers, this first-order estimation of exchange times will require validation by a second-order estimation in future works.

Traditional CTD measurements acquired near Jackson Bay during both January and June recorded salinities below 31 g kg⁻¹. Data from the time series mooring (recovered in June) further indicated that near-bottom (ca. 31 m depth) salinities reached a minimum of 27 g kg⁻¹ during the austral summer (January). This was followed by a seasonal increase, with values reaching a maximum of 30.5 g kg⁻¹ near the recovery dates in June 2024. The low salinities characteristic of the inner sea of Tierra del Fuego is attributed to the input of glacial meltwater, which exhibits strong seasonality. During summer, the along-fjord Freshwater Content (FWC) was higher in the first 20 km from the head of the fjord in the region where Parry fjord (PF) and the Azopardo river (AR) had the major influences on the freshwater input (Fig. 7a, b). Minor peaks of FWC and PEA were located at Brookes fjord (BF) suggesting the glacial melting influences on the FWC increase. Although a lack of comprehensive winter data prevents full confirmation, cross-fjord FWC measurements show that January values are at least twice as high as those observed during the June (Fig. 7a, b). This distribution is consistent with a surface slope toward the fjord head and an upper salt-wedge along-fjord structure (Fig. 5c). This typical structure of strong stratified estuaries, like fjords (Geyer and Ralston, 2011) has been documented previously in Chilean fjords like, the Reloncaví fjord (e.g. Castillo et al., 2012) and Puyuhuapi fjord (e.g. Schneider et al., 2014). The FWC has been useful to quantify the amount of freshwater input in Chilean fjords system between 43.5 to 46.5° S by Calvete and Sobarzo (2011) in this study the authors report FWC ∼ 7 m near a region with tidal-glacial influence, here FWC was < 2 m suggesting a low amount of freshwater in the Almirantazgo fjord system. The summer upper-layer salinity from the seal-borne data was less well constrained than temperature, this uncertainty propagates into the along-fjord Freshwater Content estimations.

Wavelet analysis of the time series indicated a dominance of low-frequency (periods > 24 h) variability in wind-stress magnitude, Conservative Temperature (CT), Dissolved Oxygen (DO) and Absolute Salinity (SA). The sea level (SL) was the exception (Fig. 9b), being dominated by the semi-diurnal (12 h) and diurnal (24 h) signals. This tidal dominance is consistent with the harmonics analysis and the calculated form factor (F=0.593).

In stratified systems, internal (baroclinic) seiches can be excited in the pycnocline, contributing to entrainment and enhancing heat and salt exchange between layers e.g. in the Gulmar fjord (Arneborg and Liljebladh, 2001), the Gullmaren fjord (Djurfeldt, 1987) and the Reloncavi fjord (Castillo et al., 2017). Up-fjord winds generate a surface slope (∂η/∂x>0), highest at the fjord head. In stratified conditions (ρ≫1), this results in a deepening of the pycnocline at the head, creating an opposing pycnocline tilt (∂h1/∂x<0) (Farmer, 1976). In this study, the effect of the wind over the dynamics of the fjord is the main focus thus to determine the baroclinic adjustment scale (ts), in the Almirantazgo fjord here the study consider the length of the AF which is 75 km to determine the time required for this pycnocline tilt to be established, is defined by ts∼L/c1 (e.g. Klymak et al., 2025). Here, for January conditions ts=3.34 d whereas for June ts=22 d. Comparing ts with TF (Table 3), the results indicates that during January the baroclinic adjustment is lower than the flushing time but during June, ts>TF suggesting that the wind-setup do not reach the quasi-steady state, probably due to the transient nature of the strong winds (Fig. 8) and the weak stratification of the water column during this month (Figs. 5 and 6) in the region.

Based on the established importance of the Earth's rotation, the 6-month mooring data, the short-time mooring data, and the hydrographic surveys, a schematic fjord dynamic is proposed. Winds in the region (Maria Cove, Fig. 2) are typically persistent and directed up-fjord (see Fig. A1). These forcing tilts the surface layer (∂η/∂x) and enhances the along-fjord barotropic pressure gradient. The response is a deepening of the pycnocline at the head and shallowing at the mouth (Fig. 5c). This salinity structure is concurrently modulated by summer glacial meltwater (Fig. 7a). These freshwater forms a buoyant plume. Under weak winds, this plume would likely be deflected by the Coriolis force and exit the fjord. However, moderate to-intense up-fjord winds – acting within the baroclinic adjustment time (ts=3.34 d) – appear to overcome this deflection, trapping the plume at the fjord head. During this process, the plume gains heat. Consequently, January up-fjord wind events are associated with pulses of warmer waters at 30 m depth (Fig. 8d). This is consistent to the observations made by Aravena-Yáñez et al. (2025) in Punta Santa Ana at the south of the Magellan Strait. The associated turbulence also increases DO and promotes mixing of the upper freshwater, which reduces the salinity of the deeper waters (Fig. 8e). This process is markedly seasonal. A regime occurred in mid-March 2024, after which wind intensity decreased, and strong events became primarily synoptic. In this early autumn period, up-fjord wind events began to be associated with colder waters. In winter, the up-fjord wind events were related to colder waters. During June, surface water resides longer (see Table 3) at the surface than in January and thus loses heat to colder overlying air. These colder events are associated with high DO, resulting from enhanced turbulence (oxygen ingress) and the higher solubility of oxygen in colder water (Fig. 8d). Observations in the short-term are consistent with the synthesis. Following a strong up-fjord wind pulse (Wb > 1), the DO gradient approached zero (∂DO/∂z∼0), implying oxygenation of the layer at 30 m (Fig. 8e, f). The time-domain analysis (Fig. 8) retained the linear trends to reflect the seasonality. The decreasing river discharge trend and increasing salinity trend reflect the freshwater impact in the study region. However, the FWC estimations (Fig. 7a) suggest that freshwater inputs from glacial locations are more significant than riverine inputs. In the region, Izagirre et al. (2025) using aerial imagery between 1945–2024 showed that Darwin's Cordillera increase the number of glacial lakes and thus the freshwater input by Glacial Lake Outburst Floods (GLOFs) due to a warming progress. Although the specific volume of glacial inputs was not estimated, this study concludes that the combination of this substantial freshwater input with tidal and wind-driven mixing maintains the low-salinity conditions of the inner-sea of Tierra del Fuego.

While ERA5 reanalysis data are valuable for assessing the spatio-temporal variability of the wind forcing in southern Patagonia, comparisons with in-situ meteorological stations (e.g. Punta Arenas and Porvenir) reveal that ERA5 magnitudes are typically 20 % lower than observations. This remarks that the wind-based estimations of the study were likely conservative and thus supporting the high influence of wind-driven on the dynamics of the Almirantazgo fjord.

To assess the potential for wind-driven mixing, the dimensionless Wedderburn number (Wb) was calculated. During January, intense wind-stress in AF could be as high as τ=0.10 N m⁻² taken a mean density difference ρ=1.5 kg m⁻³ (see Table 3) an upper layer h1=20 m and considering the AF length L=75 km the study obtained a Wb = 1.28. This value indicates that wind events can perturb the pycnocline during the stronger stratified condition of January. The June conditions were highly mixed with a ρ=0.1 kg m⁻³ and h1=5 m, wind stress was able to perturb the pycnocline even in weakest winds conditions. For other hand, the observational data from the short-term deployment (Fig. 9e) confirm this: during highly stratified conditions, Wb repeatedly exceeded 1 and reached values as high as 3.7, demonstrating that wind is sufficient to drive dynamics at the fjord head. The h1 depths for January and June were estimated by the cross-fjord and along-fjord maximum N2 depths at the sites near of the CTD cross-fjord transects (Fig. 2). The sensitivity of the Wb parameter was assessed by calculating the wind-influenced depth h1W=τLΔρg. In January, h1W=22.6 m closely aligns with the h1=20 m used on the estimations. In June, however, h1W=27.7 m exceeding the seasonal depth for Wb by a factor of six. Although direct validation via ADCP data was not available for this study – unlike the work of Wan et al. (2017) in Douglas Channel – the fact that Wb meets or exceeds the prescribed layer depth indicates that wind-driven forcing is a major driver of the surface layer in this system.

The time series (Fig. 9) also reveals a feedback mechanism. Strong winds, mix the upper water column, which diminishes the density gradient (ρ) (Fig. 9d). This reduction in stratification, in turn, increases the Wb (as ρ is in the denominator), enhancing the wind's mixing efficiency (Fig. 9e). Conversely, when winds weaken (due to synoptic and daily oscillations), freshwater discharge strengthens the density gradient. This re-stratification requires greater wind energy to overcome, and in the absence of strong wind, Wb returns to near-zero values (Fig. 9e). Under Wb > 1 the water column DO yields to mixing conditions (DO ∼ 0) showing a rapid response of the 30 m water column to strong winds (Fig. 9f).

To further quantify the wind's role, the power per unit area generated by the wind (dφW/dt=ϑρacdW103) available for mixing was calculated, following methodologies from Denman and Miyake (1973) and Bowden (1981). Here, ρa is air density (1.23 kg m⁻³), cd is drag coefficient (1.1×10-3) and W10 is the wind magnitude at 10 m height, here the study uses the mean magnitude (5.3 m s⁻¹) and percentile 90 (p90) of the magnitude (10 m s⁻¹) for the estimations. Additionally, the ϑ coefficient was determined by ϑ=ρacdρ where ρ is the water density. The study determines for mean magnitude dφW/dt=0.2×10-3 W m⁻² whereas for p90, dφW/dt=1.5×10-3 W m⁻².

To determine if this wind power is sufficient for mixing, it must be compared to the power required to maintain stratification dφE/dt by the estuarine circulation (Simpson et al., 1990). This stratification power is given by: dφE/dt=(1/320)(g2h15/Azρ)(∂ρ/∂x)2. Following Osborn (1980), the vertical eddy diffusivity is parameterized as Kz=0.2(ε/N2). Based on microstructure measurements from winter 2024 (Rojas-Celis et al., 2025), we assume a representative dissipation rate of ε∼O(10-6) W kg⁻¹ and typical buoyancy frequency of N2=0.0028 s⁻² which yields Kz∼7×10-5 m² s⁻¹.

The relationship between the vertical mixing of momentum and scalars is governed by the turbulent Prandtl number (Prt=Az/Kz; Thorpe, 2005). Assuming the standard Reynolds analogy for neutral or weakly stratified flows (Prt=1; Mellor and Yamada, 1982; Rodi, 1987), the resulting eddy viscosity (Az=Kz) yields a stratification power of dφE/dt=21×10-3 W m⁻². This value is one order of magnitude higher than wind power (dφW/dt), which would imply that wind forcing is insufficient for mixing. However, the Prt=1 assumption breaks down under stratified environments where buoyancy forces supress scalar mixing more efficiently than downward momentum transport (e.g. Munk and Anderson, 1948; Venayagamoorthy and Stretch, 2010). Consequently, in highly stable flows, Prt increases significantly, often reaching values O(10) (e.g. Pacanowski and Philander, 1981; Peters et al., 1988). Applying a scaling of Prt=10 to the summer conditions in the Almirantazgo fjord yields an effective eddy viscosity Az∼7×10-4 m² s⁻¹. This reduces the estuarine stratification power to dφE/dt=2.1×10-3 W m⁻².

Under these physically constrained parameterizations, first-order approximations indicate that estuarine stratification power becomes comparable to the wind power during intense episodic winds events of up to 9 m s⁻¹ (dφW/dt=1.2×10-3 W m⁻²). This scaling suggests that despite the high static stability of the water column, wind momentum can be sufficiently transferred downward to overcome stratification and induce mixing. Nonetheless, these steady-state assumptions warrant further validation through future observational and numerical studies to better constrain the spatiotemporal variability of mixing in the fjord.

Despite the dominant tidal forcing in the Magellan Strait (Fig. 3) and the strong salinity-driven stratification from glacial melt (Figs. 4c, 5c, 6a, c), the wind is a critical mixing agent. Both the Wedderburn number analysis and the mixing power comparison confirm the wind's capacity to mix the upper water column. This study provides the first quantification of the wind-driven effect in one of the southernmost fjord of Chile (Fig. 10).

Wind stress in the study area showed significant squared coherence with CT, SA, and DO at periods greater than 16 h, exhibiting marked synoptic variability at periods exceeding 70 h (Fig. 11). The most intense events (>0.07 N m⁻²; Fig. 8) were highly coherent within the 24–72 h band, suggesting that synoptic wind forcing drives a uniform response in hydrographic properties at 30 m near the fjord head. Interestingly, CT coherence for periods >64 h showed a distinct dip centered on 4 April 2024, despite remaining high during the surrounding periods (Fig. 11a). This indicates a temporary shift in the CT response to wind forcing between 30 March and 9 April, which resulted in the observed breakdown in coherence between the variables.

In the context of climate change and the projected intensification of the Southern Hemisphere westerlies, these findings are critical. The Almirantazgo Fjord ecosystem, with a seasonal stratification of the upper layer and mixing processes dependent on both tides and wind, will be profoundly influenced by future atmospheric changes. Therefore, while tidal and buoyancy forces establish the baseline conditions, the answer to the fjord's physical and ecological future is unequivocally linked to the wind.