Coastal Summer Freshening and Meltwater Input off West Greenland from Satellite Observations

Abstract

Coastal waters off west Greenland are strongly influenced by the input of low salinity water from the Arctic and from meltwater from the Greenland Ice Sheet. Changes in freshwater content in the region can play an important role in stratification, circulation, and primary production; however, investigating salinity variability in the region is challenging because in situ observations are sparse. Here, we used satellite observations of sea surface salinity (SSS) from the Soil Moisture and Ocean Salinity mission produced by LOCEAN and by the Barcelona Expert Center (SMOS LOCEAN and SMOS BEC) and from the Soil Moisture Active Passive mission produced by the Jet Propulsion Laboratory (SMAP JPL) as well as by Remote Sensing Systems (SMAP RSS) to investigate how variability in a narrow coastal band off west Greenland is captured by these different products. Our analyses revealed that the various satellite SSS products capture the seasonal freshening off west Greenland from late spring to early fall. The magnitudes of the freshening and of coastal salinity gradients vary between the products however, being attenuated compared to historical in situ observations in most cases. The seasonal freshening off southwest Greenland is intensified in SMAP JPL and SMOS LOCEAN near the mouth of fjords characterized by large inputs of meltwater near the surface, which suggests an influence of meltwater from the Greenland Ice Sheet. Synoptic observations from 2012 following large ice sheet melting revealed good agreement with the spatial scale of freshening observed with in situ and SMOS LOCEAN data. Our analyses indicate that satellite SSS can capture the influence of meltwater input and associated freshwater plumes off coastal west Greenland, but those representations differ between products.

1. Introduction

Variability in ocean salinity plays a key role in a variety of processes, including water mass formation, stratification and mixing, sea ice dynamics, and ocean biogeochemistry (e.g., [1,2,3,4,5]). Changes in freshwater content in high latitudes, in particular, can have large implications for global ocean circulation and climate [6,7,8]. Geochemical fingerprints attained in polar regions can be exported into the neighboring subpolar systems and beyond [4]. However, few in situ observations are available in high latitudes in comparison to mid and low latitude regions [9,10,11], which significantly limits our ability to investigate salinity variability in polar regions [12].

The advance of satellite sea surface salinity (SSS) observations over the last decade has been changing this picture, providing new insights on the linkages between surface salinity and climate variability, the global water cycle, and biogeochemical processes in the ocean [13]. They have also provided a new ability to monitor mesoscale and synoptic-scale salinity features [13]. Despite these advances, investigations of salinity variability in high latitude regions remain challenging. This is, in part, because current missions such as the National Aeronautics and Space Administration (NASA) Soil Moisture Active Passive (SMAP) and the European Space Agency (ESA) Soil Moisture and Ocean Salinity (SMOS) are based on L-band radiometry, which are characterized by low sensitivity to salinity in cold waters [14]. Another important source of bias for satellite SSS in high latitudes is sea ice [11]. The presence of sea ice makes it difficult to obtain reliable observations of salinity using L-band radiometry because the brightness temperature of sea ice is much higher than that of seawater [15]. Therefore, even a small area of sea ice increases the radiometric signal, which can be interpreted as an erroneous decrease in salinity or freshening [15]. Comparisons among several satellite SSS products in polar regions have identified discrepancies due to the different treatments of sea ice [16].

Much effort has been spent recently characterizing errors and uncertainties in salinity measurements in the Arctic Ocean (e.g., [17,18,19]). Detailed comparisons of satellite SSS with in situ surface salinity data have revealed that, although significant biases are often observed (e.g., [19]), satellite measurements can capture salinity variations away from regions of significant sea ice concentration and can nicely capture seasonal and interannual variations in surface salinity averaged over the entire Arctic Ocean [18]. SSS variations at monthly time scales are often consistent with effects of sea ice seasonal cycles [16,20]. Satellite SSS has also been shown to present a consistent response to discharge anomalies from Arctic rivers [21], allowing for assessing the impact of river runoff in areas where no in situ observations are available [17]. Collectively, these studies have demonstrated that spatial and temporal variations of freshwater content often exceed the remote sensed SSS uncertainty in many polar/subpolar regions (e.g., [22]), indicating that those observations can be useful for investigating the freshwater cycle and demonstrating their potential for coastal studies even in high latitude systems.

Here, we focused on investigating salinity variability along the west coast of Greenland, including the eastern Labrador Sea and eastern Baffin Bay. Baffin Bay is a seasonally ice-covered basin linked to the Arctic Ocean through the many small and narrow straits that make up the Canadian Arctic Archipelago and to the Labrador Sea through the Davis Strait. The Labrador Sea is directly connected to the North Atlantic Ocean. Along these shelves and shelf breaks flow the West Greenland Current, the Baffin Island Current, and the Labrador Current, linking the large scale cyclonic circulation in the region [4]. Freshwater from the Arctic is transported to the region through two marine export pathways (see Figure 2 in [4] for a schematic of major ocean currents). First is the pathway east of Greenland through Fram Strait [23], where freshwater is carried southward by the East Greenland Current (e.g., [24]) and then northward into west Greenland by the West Greenland Current. The second pathway is on the west side of Greenland, where arctic outflow enters Baffin Bay through the channels of the Canadian Arctic Archipelago [25,26], before being transported southward though the Davis Strait [27,28]. The southward flow on Arctic waters in Baffin Bay has been shown to sometimes interact with the northward flow off west Greenland (e.g., [29]). Recent studies have suggested that southward transport on the west Greenland shelf may occur under some conditions [30]. A second important source of freshwater to the region is melting from the Greenland Ice Sheet. Greenland Ice Sheet mass losses have accelerated during recent decades [31,32,33,34,35]. Meltwater input from the Greenland Ice Sheet can occur locally in the Labrador Sea off southwest Greenland (e.g., [33,36], but also farther north in Baffin Bay [37], which can then be transported southward into the Labrador Sea through the Davis Strait [28]. The largest land ice freshwater flux anomalies from the Greenland Ice Sheet over the past two decades have been found in Baffin Bay and off southwest Greenland [34].

Several recent studies have evaluated different satellite products in the Arctic by comparing the remote-sensed observations with in situ data (e.g., [16,19,20,21]). Off west Greenland, in particular, Fournier et al. [18] used in situ thermosalinograph data to show that satellite SSS products are generally able to capture large spatial gradients in the region, with correlation coefficients between in situ and satellite salinity measurements along ship transects ranging between 0.53 and 0.97, and root mean square differences ranging between 0.2 and 2.7 psu. To expand on that previous work, we focused on identifying how different satellite products capture seasonal salinity variability off west Greenland, including their ability to capture the influence of meltwater input from the Greenland Ice Sheet. Salinity variability off west Greenland is important because the Labrador Sea has long been known as a site of intermediate water formation [38,39,40], and its potential contribution to the global meridional overturning circulation is an active topic of research (e.g., [41,42]). Understanding how freshwater inputs affect circulation in the region [36,43,44,45] is of fundamental importance to understand the future of Labrador Sea Water formation [46,47]. Freshwater distribution can also exert a strong influence on coastal stratification and primary production off west Greenland [48,49], highlighting the importance of characterizing salinity variability in the region as well as identifying if satellite products are able to capture the influence of meltwater input from the Greenland Ice Sheet. We focused our analyses on summer/fall conditions when sea ice coverage in Baffin Bay is at a seasonal minimum [26,50], which allows for better spatial coverage. Focusing on the time of the year when surface waters are the warmest is also beneficial, given that the root-mean square differences between in situ and satellite salinity measurements in high latitude regions tend to improve with increasing temperature [18].

2. Methods

Sea surface salinity (SSS) data were obtained from multiple satellite products. For SMOS, we used the Level 3 debiased version 7 product distributed by LOCEAN/ACRI-ST Expertise Center (SMOS LOCEAN; [51]). The 9-day running mean maps were made available on a 25 km grid every 4 days (https://www.catds.fr/Products/Available-products-from-CEC-OS/CEC-Locean-L3-Debiased-v7 (accessed on 28 August 2022)) and cover the period from 2010 to 2021. We also used the Barcelona Expert Center (BEC) SMOS Arctic Sea Surface Salinity product v3.1, which was available from 2011 to 2019 (SMOS BEC; [52]). The product consisted of 3-day average maps in a 25 km grid, and it was distributed at a daily interval (https://bec.icm.csic.es/ocean-arctic-sss (accessed on 29 July 2021)). The main difference from the previous version of the data set was an improvement in the effective spatial resolution, which allowed for better identifying mesoscale features [52]. A product consisting of 9-day averages is also available. We chose to use the product with highest temporal resolution (i.e., 3-day average maps) so that short-term events could be potentially captured; however, using the 9-day product yields results quantitatively similar to those reported here. We also used the Level 3 version 5.0 SMAP SSS product distributed by the Jet Propulsion Laboratory (JPL; [53]), which consisted of 8-day running means on a 0.25 × 0.25° grid at a daily interval from 2015 to 2021 (SMAP JPL; https://podaac.jpl.nasa.gov/dataset/SMAP_JPL_L3_SSS_CAP_8DAY-RUNNINGMEAN_V5 (accessed on 18 June 2022)). Lastly, we used the Level 3 version 5.0 SMAP SSS product distributed by Remote Sensing Systems (SMAP RSS; [54]), with a 70 km spatial resolution and an 8-day running mean on a 0.25 × 0.25° grid (https://www.remss.com/missions/smap/salinity (accessed on 10 August 2022)). We did not used observations from Aquarius because of its lower spatial resolution [55], which makes it difficult for the smaller scale variability that is typical of coastal processes (e.g., [56]) to be fully captured.

A detailed description of how each of these products handle the presence of ice can be found in [18]. We used a less permissive approach with regard to the presence of ice to minimize errors, even though that meant that some features near ice edges would be removed [16]. For SMOS BEC, we only used observations flagged as being of good quality. For SMAP JPL, we only used observations characterized by ice fractions smaller than 0.1%. Previous comparisons of satellite SSS with in situ measurements along thermosalinograph transects suggested that satellite data were able to capture salinity gradients in regions with ice fractions smaller than 0.5% [18]. For SMAP RSS, only data flagged as not being contaminated by sea ice were retained. A detailed evaluation of each product in the Arctic based on extensive comparisons with in situ observations is presented in [18].

In situ observations were obtained from the National Oceanic and Atmospheric Administration (NOAA) National Centers for Environmental Information (NCEI) data base (www.ncei.noaa.gov/access/world-ocean-database-select/dbsearch.html (accessed on 25 June 2022)). Profiles were obtained using high-resolution Conductivity–Temperature–Depth (CTD) probes, moored buoys, profiling floats, gliders, undulating oceanographic recorders, ocean station data, and autonomous pinniped data [57]. Temperature and salinity profiles were processed following [58,59,60] to eliminate nonacceptable observations. Given the interest in investigating surface salinity variability, we only retained profiles that contained at least one observation in the top 6 m. After processing, a total of 20,327 profiles were available in the Labrador Sea, while 6493 were available in Baffin Bay. Of these, 10,294 and 5406 were available between July and October in the Labrador Sea and Baffin Bay, respectively. For each profile, salinity observations within the top 6 m were averaged and are referred to as surface salinity. We also used termosalinograph data from the Labrador Sea and Baffin Bay that were collected and validated by the French Sea Surface Salinity Observation Service [61] (http://sss.sedoo.fr/# (accessed on 4 April 2022)).

We used the algorithm developed by [62] for the Arctic Ocean using Moderate Resolution Imaging Spectroradiometer (MODIS) remote-sensing reflectance data (https://oceancolor.gsfc.nasa.gov (accessed on 18 March 2022)) to determine the chromophoric dissolved organic matter (CDOM) spectral slope coefficient S_275–295 from 2010 to 2021. S_275–295 has been shown to be highly correlated with dissolved organic carbon-normalized yields of lignin phenols in multiple systems [63], including the Arctic [62], and it is commonly used as a tracer of riverine input into the ocean (e.g., [64,65,66,67]). Because of its higher resolution compared to SMAP and SMOS, MODIS observations can provide complementary information in river-influenced waters, including the characterization of sharp fronts [68]. Riverine organic matter of Arctic origin has been identified in the East Greenland Current and North Atlantic Deep Water [69,70,71], and, as such, it is a possible source of terrigenous material to coastal waters off west Greenland. Terrigenous material may also be introduced into the coastal ocean with meltwater input. Subglacial streams exiting the base of land-terminating glaciers in west Greenland have been shown to contain a large fraction of lignin-like material [72], which may be derived from overridden soil and vegetation during glacial advance [73] or from deposition of atmospheric organic matter [74]. Proglacial ponds can also accumulate terrestrial-derived materials from surrounding terrestrial soils and vegetation [72]. Lastly, time series of meltwater input around Greenland were determined from an ice sheet surface elevation model [75]. Discharge rates are available from 1966 to 2012 [36,45,48,49].

The various products used here are available for different time periods. The in situ and various satellite-derived data are available simultaneously for only a few years (spring 2015 to 2019), which would result in too few observations for robust estimates of the seasonal evolution of surface salinity. This is especially true for in situ observations, which are particularly sparse in high latitudes [10]. Thus, when computing averages (e.g., Figure 1) or estimating the seasonal evolution of fields (e.g., Figure 2), we used the entire time series available for each product. Using the longest possible time series for each product presumably yields a better estimate of seasonal variability in each case, although that also means that some of the differences observed between the various products may be related to interannual variability (e.g., anomalous conditions during a given year may be included in one product but not in others).

3. Results

Surface waters off western Greenland are characterized by large spatial and temporal salinity gradients during summer (Figure 1). In situ observations from the historical record (Figure 1a) reveal a strong meridional gradient, with salinities in Baffin Bay being substantially lower than in the central Labrador Sea. Baffin Bay is also characterized by strong cross-basin gradients, with the western portion of the basin presenting lower salinity due to the influence of the inflow of low-salinity waters from the Arctic via the Canadian Arctic Archipelago (e.g., [26,76]). A thin band of low-salinity water is observed off southwest Greenland in the eastern Labrador Sea from July to October, consistent with previous studies (e.g., [77]). Low-salinity waters are also observed in the western Labrador Sea, as freshwater from Baffin Bay is transported southward through the Davis Strait [28] and into the Labrador Current.

Climatological averages of surface salinity from the four satellite products analyzed here reveal a picture that is largely consistent with the in situ observations, although differences between the products are observed (Figure 1b–e). This is especially true for western Baffin Bay, where salinity values are lower, and the pool of low salinity water spans a larger area in SMAP JPL compared to in situ observations. This is likely related to the influence of sea ice, as SMAP JPL observations are generally not available in western Baffin Bay in July if a more stringent threshold for the ice fraction is used to identify valid observations (e.g., ice fraction < 10⁻⁴%, Figure S1). The averaged magnitude and extent of the freshening in western Baffin Bay are closer to in situ observations in SMOS LOCEAN, although few in situ data are available in July and August. SMAP JPL and SMAP RSS are generally characterized by higher salinity in eastern Baffin Bay compared to the other products and to the in situ data. The coastal freshening off southwest Greenland detected with the in situ data is also observed in most satellite products, although its magnitude is reduced. We note that the coastal gap is larger for SMAP RSS observations, and that SMOS BEC is characterized by smaller spatial variations compared to in situ data and to the other satellite products.

Given the recent interest in the effects of inputs of meltwater from the Greenland Ice Sheet on ocean dynamics in the region (e.g., [46], we plotted Hovmöller diagrams of climatological averages of surface salinity within 60 km from the west Greenland coast to reveal the seasonal evolution of coastal salinity (Figure 2). Although ideally one would compute the average salinity over the shelf (or inshore of a given bathymetric contour), the width of the shelf varies substantially along the coast and can be quite narrow in some regions; therefore, considering the relatively low resolution of satellite salinity products, that would result in several locations where no satellite observations are available. Given our focus on coastal salinity variability, SMAP RSS data were not used for the remaining analyses because of its larger coastal gap. Observations in Baffin Bay are generally only available from June or July to October, given that the region is characterized by extensive sea ice cover during the rest of the year [26]. Off southwest Greenland to the south of the Davis Strait, in situ salinity generally peaks around April/May, decreasing during summer and early fall (Figure 2b). The summer freshening is possibly influenced by the local input of meltwater along the coast (e.g., [36,45]), but also by the transport of low salinity waters of Arctic origin [23]. A Hovmöller diagram of the CDOM spectral slope coefficient, S_275–295, along the coast from high-resolution MODIS observations indeed reveals a progressive decrease from spring to summer/early fall (Figure 3). Low S_275–295 values are indicative of increased terrigenous dissolved organic matter content [62], which could be related to meltwater input containing terrigenous material [72] or to the transport of water from the Arctic in the East and West Greenland Currents. Riverine organic matter of Arctic origin has been previously identified in the East Greenland Current [78] and in the North Atlantic Deep Water [69,70].

The seasonal progression of surface salinity off southwest Greenland based on observations from SMOS BEC and SMOS LOCEAN (Figure 2c,d) is similar to that based on in situ observations (Figure 2b), although the summer freshening is reduced in SMOS BEC. The large freshening observed in northern Baffin Bay from July–September (Figure 2b) is also nicely captured by SMOS LOCEAN (Figure 2d), but only to a much lesser extent by SMOS BEC (Figure 2c). Even though salinities are substantially higher in SMAP JPL (Figure 2e and Figure S2), the product does capture the summer coastal freshening off southwest Greenland, especially during September. The summer coastal freshening in northern Baffin Bay is also attenuated in SMAP JPL data.

We also investigated the seasonal evolution of coastal salinity gradients by tracking the salinity difference, ΔS, between observations collected within 60 km from the coast (shown in Figure 2) and those obtained between 60 and 110 km offshore (Figure 4). An increase in negative salinity differences near the coast associated with the coastal freshening off southwest Greenland is clearly observed in the in situ data from July to December (negative ΔS in Figure 4b). An increase in negative salinity differences is also seen in SMOS LOCEAN and SMAP JPL observations, particularly in August and September for SMOS LOCEAN and from July to September for SMAP JPL (Figure 4d,e), although its magnitude is reduced compared to in situ observations. Although salinity differences off southwest Greenland are slightly negative in SMOS BEC data, indicating some coastal freshening, no clear intensification during summer is observed (Figure 4c). The salinity difference (ΔS) changes sign in Baffin Bay during summer (Figure 4b), likely due to the influence of the inflow of low salinity waters from Arctic origin via the Canadian Arctic Archipelago [25,26] and sea ice melting, which results in offshore freshening (Figure 1a). This is mostly missed by the satellite products, with the exception of SMOS LOCEAN, which presents weakly positive ΔS in Baffin Bay, especially from August to early October (Figure 4d).

Although the previous analyses indicate that the various satellite products are able to capture the coastal freshening during summer off west Greenland to some degree, the details of that representation vary considerably, as demonstrated by an EOF decomposition of the observations. We focused on the period from July to October, when sea ice extent in Baffin Bay is at a seasonal minimum [26]. Furthermore, we excluded observations from western Baffin Bay from the analyses. If western Baffin Bay is included, salinity variance is dominated by the large freshening observed in that region, especially in July and August (Figure 1), which may be partially related to ice contamination in at least some of the products. Fournier et al. [18] showed that the standard deviation of surface salinity in Baffin Bay is strongly enhanced in the western sector (see their Figure 3). Thus, to focus on the potential impact of meltwater input from the Greenland Ice Sheet on salinity variability, we only retained observations from eastern Baffin Bay in the analyses (Figure 5). Only observations not contaminated by ice (as defined in the Methods section) are used, and EOFs are computed following [79]. For each product, we show the EOF mode that explains most of the local variance near the coast off west Greenland. The amplitude time series are shown as the average plus or minus the standard deviation between the multiple years to emphasize how each product captures the summer freshening. Note that the color scale varies between the plots to reveal as much of the spatial variability as possible in each case.

The EOF for all salinity products is characterized by a band of negative values around the coast, extending approximately from the southern tip of Greenland to Disko Bay (~68°N; Figure 5, left panels). The amplitude time series in all cases progressively increase from July to September (Figure 5, middle panels), indicating that during that time, coastal salinities off west Greenland decrease in all products, which is consistent with results shown in Figure 2. The freshening is largest in SMAP JPL and SMOS LOCEAN, as revealed by larger negative EOF values near the coast and by the larger increase in the amplitude time series from July to September. For those products, the mode explains more than 50% of the local variance off southwest Greenland (Figure 5, right panels). For SMOS BEC, on the other hand, which is characterized by a weaker summer freshening, the mode explains approximately 20% of the local variance along the coast.

The analysis also reveals that there is little alongshore variability in EOF values along the coast up to Disko Bay for SMOS BEC (Figure 5a, left panel), which suggests uniform freshening. This is in contrast to the pattern observed with SMOS LOCEAN and SMAP JPL observations (Figure 5b,c, left panels). Although those products also reveal broad freshening along the coast, especially up to Disko Bay, they are also characterized by larger negative EOF values near the mouth of some fjords (areas shown in dark blue). The largest negative intensification for SMOS LOCEAN occurs around 64–65°N, while for SMAP JPL, large negative intensifications occur around 64–65°N and also around 66–67°N. Climatological averages of meltwater input integrated into 1° latitudinal bands reveal increased inputs of meltwater from glaciers at 64–65°N and 66–68°N (Figure 6). Note that most glaciers off southwest Greenland are land-terminating [48], where meltwater derived from surface melting is discharged as a surface plume. Thus, the increased freshening near the mouth of fjords captured by SMOS LOCEAN and SMAP JPL may be related to local inputs of freshwater associated with melting of the Greenland Ice Sheet. Freshwater outflow associated with melting glaciers can be associated with ice transport [80], however, and thus the freshening signals identified in the satellite data near the mouths of fjords could be influenced by contamination due to the presence of ice.

To the north of Disko Bay, meltwater input is reduced (Figure 6). Additionally, most glaciers at and to the north of Disko Bay are marine-terminating [37,81], where the fjord circulation supplies heat to melt the ice at subsurface [82,83]. Meltwater is thus delivered to fjords deep below the surface, and the diluted meltwater is then exported to the coastal ocean at subsurface as glacially modified water [84,85,86]. This is consistent with the weaker surface freshening observed near the coast in that region in the EOF analyses (Figure 5).

To further evaluate the potential impact of the input of meltwater from the Greenland Ice Sheet on coastal salinity variability, we plotted time series of surface salinity within 120 km from the coast around 64–65°N (i.e., in one of the regions identified as potentially influenced by meltwater input; area in dark blue in Figure 5b,c, left panels, marked by a black square in Figure 7a). We also plotted time series for regions immediately to the north and to the south of that region (green and red squares in Figure 7a, respectively). The climatological time series of in situ surface salinity in the central area around 64–65°N reveals a freshening of ~2.5 psu from June to September (black line in Figure 7b). Freshening is also observed to the north and to the south of the central region (green and red lines in Figure 7b), suggesting that the freshening extends throughout coastal southwest Greenland. The freshening in the northern and southern areas is smaller, however, at ~1.7–1.8 psu. The additional freshening at the central area could be related to the increased input of meltwater that occurs locally at 64–65°N (Figure 6).

Climatological averages of surface salinity from the various satellite products are also shown in Figure 7c–e, which reveal differences among the products. Salinity in June, before the summer freshening occurs, is approximately 1 psu higher in SMAP JPL observations than in the other products, for example. Despite these differences, coastal salinity in the central area (black lines) decreases progressively for all products reaching a minimum in late August/early September, which is consistent with the in situ observations (Figure 7b). That is also consistent with the seasonal evolution of S_275–295 (Figure 7f), which remains approximately constant during June but then progressively decreases from July to September indicating an increase in the terrigenous signature during that time. Comparing the magnitude of the summer freshening among the various satellite products and the in situ data indicates that it is lower in SMOS BEC at ~1.2 psu. For SMOS LOCEAN and SMAP JPL, freshening of 2.9 and 2.5, respectively, are observed in the central region, which is comparable to the magnitude observed with the in situ data. Interestingly, the summer freshening in the southern area (red lines) is lower in all satellite products compared to the central and northern areas. Given that the mean flow off southwest Greenland is to the northwest due to the presence of the West Greenland Current [87,88], this could be related to an input of meltwater at the central region at 64–65°N (Figure 6), which is then transported northward by the large-scale circulation. As a result, the summer freshening in the southern region (i.e., in the red box to the south of 64°N, upstream of the location of large meltwater input) is reduced. We note that this is not observed in the in situ data, however. Although this could indicate that the reduced freshening observed in the southern area (red lines) in the satellite data is an artifact, this difference could also be related to the sparsity of the in situ observations (i.e., the in situ data are not synoptic and represent averages of different locations in different years; thus, it is possible that synoptic events such as the transport of meltwater plumes by the mean northward flow could be missed). Despite the differences among the satellite products and the fact that in some cases biases in excess of 1 psu can be observed in the climatological averages, this analysis suggests that the various satellite-derived salinity products can capture the influence of meltwater input off southwest Greenland, although to different degrees. Note that the ice fraction in these coastal regions is virtually zero from July to September when the largest freshening is observed, although it begins to increase reaching 0.1% in October (Figure 7g).

While climatological averages were used in Figure 7, the influence of meltwater input from the Greenland Ice Sheet on coastal salinity can also be seen at synoptic time scales, especially following large events. Greenland Ice Sheet melting reached record levels during summer 2012 [89], resulting in large inputs of meltwater into the coastal ocean [36]. In situ observations collected along a ship transect using a thermosalinograph from 20–27 August 2012 off southwest Greenland reveal lenses of low salinity water near the coast, especially from ~62°N to ~66°N (Figure 8a). Substantial coastal freshening was also observed in SMOS LOCEAN observations on 25 August 2012 (Figure 8b) from 62.5°N to 66°N, approximately coinciding with the location where maximum freshening was observed in the in situ data. SMOS LOCEAN salinities were 2–3 psu higher than in situ measurements on average however (note different color bar used in Figure 8a,b), and missed the extreme freshening observed in some locations in the in situ data (Figure 8d). The coastal freshening was much less significant in SMOS BEC data (Figure 8c). Note that observations are not available as close to the coast in that case. Note also that SMAP JPL was not used in this comparison, since those observations first became available in 2015. Satellite SSS can still capture some coastal freshening during less extreme events, but the comparison is often less favorable. For example, in situ observations reveal coastal freshening around 62.5–64°N in August 2018 (Figure S3a). SMOS LOCEAN also captures some freshening in that region, although the freshest water is shifted slightly to the north at 63.5–65°N (Figure S3b). SMAP JPL also indicates freshening at 62.5–64°N, but salinities are significantly higher than observed using in situ measurements (Figure S3d; note that a different scale was used in panel d to highlight the freshening; see also Figure S3e). The core of low salinity water near the coast is less visible in SMOS BEC (Figure S3c).

4. Conclusions

In situ oceanographic data in high latitude regions are sparse in both time and space [9], which have traditionally made it difficult to investigate salinity variability in those systems from an observational point of view. The recent availability of sea surface salinity (SSS) observations from satellite measurements have provided a new ability to monitor changes in those under-sampled regions [13], and several recent studies have evaluated satellite products against the few in situ observations that are available (e.g., [17,18,19]). This is critically important, given that changes in high latitude systems can have significant implications for a variety of processes with global significance. For example, changes in freshwater content in the Arctic Ocean, as those reported by [90], can potentially impact convection in the North Atlantic and the meridional overturning circulation [91].

In this study, we investigated surface salinity variability off west Greenland and how that variability is captured by various satellite products. We were particularly interested in identifying if and to what extent these products can capture the influence of meltwater input from the Greenland Ice Sheet, given that the large increases observed recently in that freshwater input [35] can also potentially play a significant role in ocean circulation [46]. Comparing in situ observations and satellite measurements is often difficult due to their different sampling characteristics, however [18,92,93]. In situ data represents a single location and time, while satellites measure salinity over the satellite footprint size from ~40–70 km. Additionally, satellite observations represent the top few centimeters of the ocean [93], while in situ data capture salinity in the top few meters of the water column. Off coastal west Greenland, the salinity difference in the top 6 m of the water column was lower than 0.3 psu in 83% and lower than 0.1 psu in 77% of the profiles for which in situ observations were available very near the surface. In 14% of the profiles that difference was higher than 1 psu. Lastly, the in situ measurements and the various satellite products used here cover different time periods. Despite these differences, our results indicate that the various satellite products analyzed here can capture the seasonal evolution of coastal salinity and the summer freshening off west Greenland that peaks around September, although some large biases are observed. SMOS LOCEAN, in particular, is characterized by summer freshening off southwest Greenland and in northeastern Baffin Bay that are comparable to that detected using in situ observations. However, all products seem to underestimate coastal salinity gradients.

EOF decompositions of the SSS fields have revealed small variability during summer and early fall to the north of Disko Bay in eastern Baffin Bay. Most of the meltwater from Greenland is introduced into Baffin Bay’s fjords at the subsurface at the grounding line [81], both as subglacial discharge and as basal melt from the glacier face by the warmer ocean [94]. The input of meltwater into fjords below the surface results in buoyancy-driven upwelling and entrainment of deeper water rich in nutrients to the surface, where it has the potential to support the growth of phytoplankton [95,96,97,98]. The depth at which the plume becomes neutrally buoyant depends on several factors, including the depth of the glacier grounding line and the magnitude of the subglacial discharge volume flux [98,99], as well as on the cumulative turbulent entrainment and ambient stratification at the glacier face [100,101]. Noble gas observations (e.g., [84]) have revealed that neutral buoyancy in Baffin Bay’s fjords is often achieved at subsurface, however [37]. This means that the diluted meltwater is generally exported to the coastal ocean below the surface [84,85,86] and, thus, a large surface signature along eastern Baffin Bay is not expected, which is consistent with the observations. The relation between surface salinity and top-to-bottom freshwater content has been shown to be weaker in Baffin Bay than in other Arctic areas [12].

To the south of Disko Bay, on the other hand, significant coastal freshening is observed from July to September at a time when the local ice fraction is virtually zero. That freshening is observed using three independent sources of information: the in situ data, all satellite salinity products, and MODIS-derived CDOM spectral slope data (the latter reveals an increase in the terrigenous signature, which can be associated with freshwater input either due to local inputs of meltwater from the Greenland Ice Sheet [72] or due to alongshore advection of low-salinity waters of Arctic origin [23]). The coastal freshening captured by the in situ data and by the satellite products is in excess of 2 psu in some locations, except for SMOS BEC, which captures a weaker freshening [77]. Off southwest Greenland, most of that meltwater input occurs at the surface at land-terminating glaciers, which behaves similarly to a river plume [102,103,104,105] with the potential to create large surface salinity anomalies. The few marine-terminating glaciers off southwest Greenland have a shallow grounding line, and neutral buoyancy is achieved in those fjords very near the surface [37]. We can obtain a crude estimate of the possible salinity anomaly due to meltwater input if we assume that the meltwater introduced to the coastal ocean from June to August to the south of Disko Bay is uniformly distributed in an area extending 100 km from the coast and 700 km in the alongshore direction. Assuming that the meltwater is uniformly distributed in the top 15 m of the water column (mixed layer depth is approximately 10–20 m off southwest Greenland from July to September [48]), the input would be enough to lower coastal salinities by ~3 psu (~4 psu if we consider only meltwater input since 2009). That is evidently an upper bound estimate, since part of this meltwater will be transported away from the region by coastal currents (e.g., [36,43,44]) during the melting season. In any case, this back-of-the-envelope calculation indicates that meltwater input can potentially make a substantial contribution to the coastal freshening observed between July and September. This is consistent with SMOS LOCEAN and SMAP JPL SSS observations, which are characterized by intensified freshening near the mouth of some fjords, which is consistent with local inputs. That is true not only for climatological averages but also on synoptic time scales, with significant freshening being observed in SMOS LOCEAN near the mouth of fjords following a period of high melting in Greenland, agreeing with in situ observations. The signature of meltwater plumes is not as clearly seen in SMOS BEC, however. This is possibly due to the objective analysis scheme used in the product, which results in smoothing that makes it more difficult for small-scale variability to be captured [18].

Collectively, our analyses suggest that even though clear biases are observed in satellite products (e.g., [16,17,19]), with average salinities being sometimes up to 1 psu higher than in situ observations, satellite SSS can provide useful information about the input of meltwater and the associated distribution of freshwater plumes in the coastal ocean off west Greenland. Given that the various satellite products use different retrieval techniques, geophysical filtering, and sea ice and land masks [16], it is probably best to investigate freshwater distribution in the region using all products available, since differences among the products can provide a crude measure of the potential uncertainties in the observations.