2.1 MALINA sampling overview

The MALINA expedition was conducted from 31 July to 24 August 2009 in the southeastern Beaufort Sea on the research icebreaker CCGS Amundsen. A total of 167 CTD rosette casts were carried out during the expedition with water sampling conducted at 28 station locations (Fig. 1). The locations, sampling times, and bottom depths are provided in Table S1 in the Supplement. A small barge was launched to conduct coincident surface water sampling away from the ship's influence on 26 of these stations. In addition, the barge visited 12 additional stations in coastal waters too shallow for the ship (Fig. 1; Doxaran et al., 2012). The CTD rosette on board the icebreaker was equipped with 24 12 L Niskin bottles for water collection and various in situ instruments including an SBE 911plus CTD (Sea-Bird Electronics, Inc.), a C-Star 25 cm beam transmissometer (WET Labs, Inc.) for measuring the particulate beam attenuation coefficient at 660 nm, c_p(660), and a WETstar fluorometer (WET Labs, Inc.) for measuring fluorescence of chlorophyll a (chl a).

2.2 Determinations of SPM and POC

Niskin bottles were triggered during CTD rosette upcasts to collect water samples at three to four depths, which always included the near-surface water (1.5–3 m depth range) and subsurface chlorophyll-a maximum (SCM) if present. To ensure representative sampling of entire particle assemblages within Niskin bottles (including particles settled below the level of the spigot), the full content of the 12 L Niskin bottles was drained directly into 20 L HDPE carboys (Nalgene) by opening the bottom lid (Knap et al., 1996). Aliquots were then sampled from the carboys after mixing. If a sufficient volume of water was available, filtration for SPM and POC determinations was made in triplicate for each examined depth. However, this was not always possible in clear waters with low particle concentrations, in which case either duplicates or single samples were prepared. Water samples for SPM and POC on the barge were collected by directly submerging a 20 L HDPE carboy below the sea surface (Doxaran et al., 2012). Doxaran et al. (2012) reports on coefficients of variation for SPM and POC for these surface samples measured in triplicate. Additional near-surface water samples were occasionally collected by lowering a bucket from the side of the ship.

Water samples for SPM and POC were filtered through 25 mm diameter Whatman GF/F filters under low vacuum (≤5 psi). Prior to the cruise the filters for both SPM and POC determinations had been rinsed with Milli-Q water, combusted at 450 ^∘C for 1.5 h to remove organic material, and weighed using a Mettler Toledo MT5 balance (±0.001 mg precision) to obtain the blank measurement of the filter mass. Filters were stored individually in petri dishes until the time of sample filtration. The volume of filtered seawater was adjusted to optimize particle load on the filter but not to cause filters to clog. This volume ranged from 0.2 L for very turbid samples collected near the Mackenzie River mouth (station 697) to 5.8 L at station 780. Immediately following filtration, filters were rinsed with about 50 mL of Milli-Q water to remove salts, transferred back to the petri slides, and dried for 6–12 h at 55 ^∘C. The dried filters were stored at −80 ^∘C until processing. After the cruise, filters were again dried at 55 ^∘C in the laboratory for about 24 h before measuring their dry weight using the same Mettler Toledo MT5 balance. The SPM (in units of grams per cubic meter) was determined by subtracting the blank filter mass from the sample filter mass and dividing by the volume of water filtered. The relative humidity of the room was about or below 40 % during weighing of filters to minimize the effect of uptake of moisture by the filters during the measurements. The protocol used for SPM determinations is consistent with standard methodology (e.g., Babin et al., 2003a).

SPM and POC were determined on the same GF/F filters. After the weighing for SPM, POC content was determined with an organic elemental analyzer (PerkinElmer 2400 Series II CHNS/O) with a standard high-temperature combustion method as described in Doxaran et al. (2012). Prior to insertion of samples into the analyzer, the filters were acidified with 200–350 µL of 2N HCl to remove inorganic carbon and then dried at 60 ^∘C. Filters were compacted into small (∼5 mm diameter) rounded pellets within pre-combusted aluminum foil. Blank filters for POC determinations were treated and measured in the same way as sample filters. The combustion temperature was kept at 925 ^∘C. The final POC values (in units of grams per cubic meter) were calculated by dividing the mass of organic carbon measured (in units of micrograms) on the sample filter (corrected for blank filter) by the filtered volume. In these calculations, the correction for blank filters was made using the average mass concentration of organic carbon determined on nine blank filters, which was determined to be 21.2±8.1 µg (corresponding to a range of ∼2 to 50 % of measured signal for the sample filters).

2.3 Particle size distributions

The particle size distribution (PSD) of 54 discrete seawater samples collected with the CTD rosette or from the barge were measured using a Beckman Coulter Multisizer III analyzer following the method described by Reynolds et al. (2016). In 40 of these samples, data were collected using both the 30 and 200 µm aperture sizes and merged into a single PSD ranging from 0.7 to 120 µm. Seawater filtered through a 0.2 µm filter was used as the diluent and blank, and multiple replicate measurements were acquired for each sample. Each aperture was calibrated using microsphere standards following recommendations by the manufacturer. The average number of particles per unit volume within each size class, N(D) (per cubic meter), where D is the midpoint diameter of the volume-equivalent sphere in each size class, was obtained after subtracting the counts for the blank. The particle volume distribution, V(D) (dimensionless), was then calculated from N(D) by assuming spherical particles.

2.4 Beam attenuation measurements

C-Star transmissometer data were recorded at 24 Hz as raw voltages and merged with the depth recording from the CTD rosette. Downcasts were processed to 1 m vertical bins centered at integers by averaging the interquartile range of the voltages within bins. This method effectively removed spikes and noise from the data, if present. Time series of transmissometer data were also collected at selected depths and processed similarly to above, by taking the average of the interquartile range of the voltage values recorded over the periods when the rosette was stopped for water sampling during upcasts. These data were used for correlational analysis with SPM and POC data from discrete Niskin bottle water samples. The particulate beam attenuation coefficient at 660 nm, c_p(660) (in reciprocal meters), was then calculated from the binned voltage signal, V_signal, as

where x is the pathlength of 0.25 m, V_dark is the dark voltage offset, and V_ref is the reference voltage associated with particle-free pure seawater (see C-Star User's Guide, WET Labs, Inc.). For MALINA, V_ref was taken as the highest V_signal reading observed during the expedition, i.e., it was determined to be 4.7362 V (lower than the factory-supplied value of 4.8340 V) observed with the same instrument during the Geotraces cruise that immediately followed the MALINA cruise (cast 0903_26 on 4 September at depths between 1900 and 2500 m where water temperature and salinity averaged −0.40 ^∘C and 34.94 PSU, respectively). This V_ref was only marginally higher than maximum values observed during the MALINA expedition. The above method also assumes a negligible contribution by CDOM to c_p at 660 nm (Bricaud et al., 1981), which is a reasonable assumption based on data shown in Matsuoka et al. (2012). V_dark was found to be 0.0517 V when measured immediately after a deep cast when the temperature of the instrument was equilibrated to seawater temperature. The factory supplied value was 0.061 V. However, discrepancies in V_dark are of little significance compared to V_ref. For example, for relatively turbid conditions with V_signal as low as 3.7 V (representing a c_p(660) of 1 m⁻¹), the change from 0.0517 to 0.061 reduces the calculated c_p(660) by only 0.2 %.

In this study we also use the C-Star transmissometer data obtained during CASES (2004) and IPY-CFL (2008) expeditions on the CCGS Amundsen (Ingram et al., 2008; Barber et al., 2010) to compare and contrast to the MALINA observations. The data were processed in the same way as the MALINA 2009 downcast data. One exception was that the factory-supplied V_dark values were used exclusively as they had not been determined on board the vessels. The V_dark values were 0.0570 and 0.0586 V for the CASES and IPY-CFL expeditions, respectively. The highest V_signal readings were 4.6783 and 4.7902 V, respectively.

Four deep CTD casts were additionally collected in the Canada Basin during the Joint Ocean Ice Study (JOIS) on 21–23 September 2009 and the data were obtained from the Beaufort Gyre Exploration Program website hosted by the Woods Hole Oceanographic Institution. These transmissometer data were processed as described above with a V_dark value of 0.0633 V (factory calibration) and V_ref value of 4.9408 V (maximum recorded value at station CB-21 on 9 October 2009).

2.5 Determination of surface water mass distributions

During the MALINA expedition, water samples were collected at 51 stations on the Mackenzie Shelf by either the CTD rosette or from the barge. Oxygen isotope ratios (δ¹⁸O) were analyzed at the Light Stable Isotope Geochemistry Laboratory (GEOTOP-Université du Québec à Montréal) using a triple collector IRMS in dual-inlet mode with a precision of ±0.05 ‰. Total alkalinity (TA) was measured by open-cell potentiometric titration (TitraLab 865^® radiometer) with a combined pH electrode (pHC2001, Red Rod^®) and diluted HCl (0.03 M) as a titrant. Oxygen isotopes and TA collected during CASES 2004 are described, and partially published, in Lansard et al. (2012). We use salinity (S), δ¹⁸O, and TA data to estimate the fractional composition of sea ice meltwater (f_SIM) and meteoric water (f_MW) in the surface layer on the Mackenzie Shelf, following the protocol described in Lansard et al. (2012). The calculations follow Yamamoto-Kawai et al. (2008) and Lansard et al. (2012) with the sea ice melt (SIM) end members 4.7 PSU, −2.5 ‰, and 415 µmol kg⁻¹; the meteoric water (MW) end members 0 PSU, −19.5 ‰, and 1620 µmol kg⁻¹; and the saline Pacific Summer Water (f_PSW) end members 31.5 PSU, −3.0 ‰, and 2250 µmol kg⁻¹ for S, δ¹⁸O, and TA, respectively. The Mackenzie River represents the major source of meteoric water on the Mackenzie Shelf.

2.6 Additional environmental data

To describe ocean currents, temperature, and salinity near the shelf break, in addition to CTD casts we used data from a current meter (RCM11, Aanderaa Data Instruments) moored at station CA05 near the center of line 100 (Fig. 1). The locations where the mooring CA05 was deployed and the depth of the current meter varied slightly between years. During season 2003–2004, it was deployed in 250 m deep water (71.42^∘ N, 127.37^∘ W) at a depth of 202 m. In 2007–2008 and 2008–2009, the bottom depth was about 200 m (71.31^∘ N, 127.60^∘ W) and the instrument depth 178 m. In addition to current speed and direction, the instrument recorded water temperature, conductivity, turbidity, and dissolved oxygen content, all at 0.5 h intervals. The conductivity sensor did not function in 2007–2008.

Annual estimates of Mackenzie River discharge and ice concentrations on the Canadian Beaufort Sea shelf for the years 2004, 2008, and 2009 were obtained from publicly available data provided by Environment Canada. Daily discharge rates (in units of cubic meters per second) for the Mackenzie River at the Arctic Red River location (10LC014) were obtained from Water Survey of Canada (Environment Canada) hydrometric data online archives. Ice coverage with a 1-week resolution for the Mackenzie Shelf area was calculated using the IceGraph 2.0 program (region: cwa01_02) provided online by the Canadian Ice Service (Environment Canada).

Estimates of wind speed over the shelf were obtained by averaging 10 m elevation wind data over grid points located over the shelf break in the southeastern Beaufort Sea obtained from National Centers for Environmental Prediction (NCEP) (Fig. 1). As pointed out by Williams et al. (2006), NCEP data are readily available and may be preferable over observations made at coastal stations because the latter may be affected by the presence of land. We use the NCEP wind data in a qualitative sense to identify conditions that may have induced upwelling or downwelling of seawater within the shelf area (e.g., Kirillov et al., 2016).

Figure 2Fields of water salinity (a, c, e) and particulate beam attenuation coefficient at 660 nm (cp(660), a, c, e) for (a)–(b) sea surface, (c)–(d) 30 m, and (e)–(f) 80 m during the MALINA 2009 expedition. The presented results in (a) include both barge and only CTD rosette measurements from the CCGS Amundsen, but only CTD rosette measurements in (b)–(f) marked by black dots. Black contour lines in (a) mark salinity in the estuary, while the white dashed contour lines are the fraction of meteoric water (%) of the freshwater calculated using samples collected by both barge and CTD rosette.

Download

3.1 Water mass distributions and circulation during August 2009

During the MALINA cruise in August 2009, there was a distinct east–west gradient in the observed surface salinity on the shelf (Fig. 2a). To the west, surface salinities below 24 PSU were caused by the presence of the river plume that flowed along the coast and over the Mackenzie Trough in response to easterly winds during June 2009 (see Sect. 3.3.3). The river plume formed a near-surface layer of about 15–20 m in thickness, which covered the full extent of line 600 and line 700. To the east, water with salinity above 29 PSU was observed to reach the surface in the area north of Cape Bathurst. Williams and Carmack (2008) described such upwelling from within the Amundsen Gulf as topographically induced in response to easterly winds. Salinity values in excess of 32 PSU were measured near the shelf bottom at 30 m (Fig. 2c), which corresponds to Pacific waters in Amundsen Gulf at a depth of about 80 m (Fig. 2e). Generally, for the Arctic Ocean, salinity controls the vertical stratification such that higher salinity is found at greater depth. The water mass definition that ensues follows Carmack et al. (1989) and is consistent with descriptions in Lansard et al. (2012) and Matsuoka et al. (2012). The salinity range between 30.7 and 32.3 PSU corresponds to the Pacific Summer Water mass, which originates from waters flowing through Bering Strait during summer. Underneath, the Pacific Winter Water is characterized by salinity between 32.3 and 33.9 PSU and typically found at a depth from ∼180 to 220 m (e.g., Jackson et al., 2015). This is followed by a transition to waters of Atlantic origin with salinity >34.7 and temperature above 0 ^∘C typically found between ∼220 and 800 m. Cold and dense deep water is found at greater depths and down to the bottom.

The relative contributions (%) of the two sources to the freshwater content, i.e., meteoric water f_MW and sea ice meltwater f_SIM, in the surface layer are shown by the contours in Fig. 2a. The percent values are calculated as follows: $f_{\mathrm{MW}}/(f_{\mathrm{MW}}+f_{\mathrm{SIM}})\times \mathrm{100}$. Apart from the Mackenzie River mouth, the freshwater in the surface layer was a mixture between sea ice melt and river runoff. River water prevailed along the coastline, while sea ice melt had a larger contribution further offshore. A larger river water fraction also extended further along the west coast with the northwest-flowing river plume. In the upwelling region north of Cape Bathurst, river runoff and ice melt contributed about equal amounts to the relatively small freshwater content of ∼10 %. The high ice melt proportions in excess of 80 % were found in offshore waters with melting multi-year sea ice (Bélanger et al., 2013).

Figure 3Sections of geostrophic current velocity (colors and white contours) perpendicular to transect lines 100 (a), 200 (b), 300 (c), and 600 (d). Geopotential heights were referenced to 500 m. Positive current values are generally for the direction perpendicular to the transect lines (see Fig. 1) towards the northwest (a), west (b–c), or southwest (d). The green-yellow square in (a) indicates the location of the current meter on the CA05 mooring.

Download

Geostrophic currents for the cross-shelf sections 100, 200, 300, and 600 were calculated using temperature and salinity data from August 2009 CTD casts (Fig. 3). The reference depth, where the current velocity was assumed to be zero, was selected as 500 m, corresponding to a water mass originating in the Atlantic in which geopotential gradients are small (McPhee, 2013). The sections reveal a westward mean flow of up to 9 cm s⁻¹ in the Canada Basin (Fig. 3b, c), which is consistent with the anticyclonic circulation of the Beaufort Gyre. Similarly, currents over the shelf were typically westward with speeds on the order of a few centimeters per second. A notable feature was the presence of the eastward-flowing shelf break current centered between 100 and 150 m depth (Pickart, 2004). The shelf break current is an indicator for downwelling flow from the shelf to the basin (Dmitrenko et al., 2016). Both Dmitrenko et al. (2016) and Forest et al. (2015) present mooring data collected at the Mackenzie Shelf shelf break location showing events of wind-driven shelf break current intensifications (with flow up to 1.2 m s⁻¹ in January 2005) during downwelling favorable winds. However, to our knowledge, the current intensification along the Mackenzie Shelf shelf break during summer has not been shown in the literature to date. The mean easterly flow was around 3 cm s⁻¹ (Fig. 3a–c), which is consistent with the observations of Pickart (2004) for the summertime period along the Alaskan Beaufort shelf break. The section along line 600 in the Mackenzie Trough captured an anticyclonic mesoscale eddy (∼50 km diameter), which impacted the patterns of c_p(660) and chl-a fluorescence (see below).

3.2 Characteristics of particles suspended in seawater in August 2009

Empirical relationships between the beam attenuation coefficient and SPM are dependent on the composition and size distribution of particle assemblages (Kitchen et al., 1982; Bunt et al., 1999; Babin et al., 2003b; Reynolds et al., 2010; Woźniak et al., 2010; Hill et al., 2011). In this section we present several water characteristics encountered in August 2009 that help understand the origin of suspended particles and composition of particle assemblages in the Canadian Beaufort Sea. The absorption associated with organic and inorganic material is described elsewhere (Doxaran et al., 2012; Bélanger et al., 2013). However, the measured particulate absorption at 660 nm was found to be smaller by 1–4 orders of magnitude than c_p(660) and can thus be ignored (data not shown). Particle size distributions during MALINA and the relationship to backscattering are described in Reynolds et al. (2016). The environmental conditions encountered during MALINA showed large spatial variability; yet, a statistically significant and strong correlation was found between the particulate beam attenuation coefficient (c_p(660)) and SPM, as well as POC (see Sect. 3.3). Although we recognize the possibility of interannual and seasonal variability in particle characteristics, the wide range of particle characteristics observed during the MALINA expedition gives us confidence in the applicability of the derived statistical relationships to infer suspended particle concentration fields on the Mackenzie Shelf and southeastern Beaufort Sea.

Figure 4Voltage readings from the chlorophyll fluorometer (a, c, e) and CDOM fluorometer (b, d, f) for transects (a–b) 100, (c–d) 300, and (e–f) 600.

Download

Generally, c_p(660) in the near-surface layer decreased from >1 m⁻¹ in coastal waters to <0.02 m⁻¹ in offshore Canada Basin waters (Fig. 2), reflecting the riverine and coastal sources of particulate matter. To the west, the fresher surface layer influenced by the river plume featured relatively high c_p(660) ranging from 0.1 to 0.4 m⁻¹ (Fig. 2b) and high colored dissolved organic matter (CDOM) fluorescence (Fig. 4; Matsuoka et al., 2012). The highest ship-based observation of surface-water c_p(660) of ∼2.6 m⁻¹ was observed at station 394 in 13 m deep waters at the mouth of Kugmallit Bay; however, c_p(660) reached 8.8 m⁻¹ at a depth of 10 m and presumably higher values near the seabed. The surface waters in the area of upwelling just north of Cape Bathurst appear to also have been a hot spot in terms of particle concentration; c_p(660) at the surface of station 170 reached values over 1.2 m⁻¹ (Fig. 2b).

The high c_p(660) values near the shelf seafloor in August 2009 were accompanied by a strong chl-a fluorescence signal, both of which also extended from the shelf far into the Canada Basin as a subsurface chl-a maximum (SCM) layer (Fig. 4a, c, e). The SCM layer is a consistent feature in the southern Beaufort Sea during summer (Martin et al., 2010). The SCM was centered at depths between the 31.5 and 32.3 PSU isohalines, which corresponds to the lower portion of the Pacific Summer Water. The underlying Pacific Winter Water is characterized by maxima in both nutrients and CDOM (Fig. 4; Matsuoka et al., 2012). The nutrient maximum is typically found at the center of the Pacific Winter Water near the 33.1 PSU isohaline (Martin et al., 2010).

Table 1Salinity, saline end member, meteoric and sea ice melt fractions (%), and meteoric water percentage of freshwater content (MW% $=f_{\mathrm{MW}}/(f_{\mathrm{MW}}+f_{\mathrm{SIM}})\times \mathrm{100}$) for surface seawater samples obtained at matching station locations during CASES 2004 and MALINA 2009. A few matching stations were also sampled during IPY-CFL 2008. The samples were collected with Niskin bottles on a CTD rosette. The fraction of the saline end member (f_PSW) represents Pacific Summer Water with a salinity of 31.5 PSU.

^* Approximately 5 nmi south of station, which is halfway to the next station.

Download Print Version | Download XLSX

Following Woźniak et al. (2010), the data representing discrete seawater samples were partitioned into three composition-related groups based on the POC∕SPM ratio: (1) mineral dominated when POC∕SPM <6 %, (2) mixed when 6 % ≤ POC∕SPM ≤25 %, and (3) organic dominated when POC∕SPM >25 %. Only at station 394 (13 m bottom depth) near the entrance to Kugmallit Bay did the CTD rosette sampling from the CCGS Amundsen take place sufficiently close to the coast to reach the mineral-dominated water masses (Fig. 5a). However, the results from barge sampling in August 2009 show that mineral-dominated particle composition was mostly limited to shallow waters less than about 20 m deep near the two Mackenzie River mouths where f_MW contributed >90 % of the freshwater content (Fig. 5a). This agrees with past observations suggesting that most mineral-dominated particles transported by the Mackenzie River plume settle to the bottom within the delta or shortly after reaching the shelf where the plume speed decreases (Macdonald et al., 1998). For the rest of the shelf and basin surface waters the particle composition in our collected samples showed considerable variability within the organic-dominated and mixed types (Fig. 5). The one exception was, however, the surface sample at station 110 located furthest east in the Amundsen Gulf where the POC∕SPM was less than 1.8 % (SPM =3.56 g m⁻³). Although the possibility of contamination of the sample from station 110 cannot be excluded, the high SPM load could also have been caused by the release of ice-rafted sediments as the ice melted (Bélanger et al., 2013). Deteriorated multi-year ice was observed in the vicinity of station 110, which could have been the source of minerogenic material. Sea ice meltwater was found to have a slightly larger contribution at station 110 compared to other stations along line 100 (Table 1).

Figure 5(a) POC-to-SPM ratio for surface samples within the study area, (b) relationship between POC-to-SPM ratio and meteoric water fraction of freshwater in surface waters (see Fig. 2a), and relationship between volume fraction of particles less than 1 µm in diameter and (c) salinity, and (d) meteoric water fraction of freshwater. Values in (a), (b), and (d) are limited to surface water samples, while data points in (c) with salinities over 30 PSU represent subsurface samples.

Download

For a detailed description of the particle size distribution (PSD) data measured during MALINA, readers are referred to Reynolds et al. (2016). Here, we provide an overview of the spatial distribution of the PSD by calculating the volume fraction of particles less than 1 µm in diameter D to the total particle volume between 0.7 and 120 µm. A notable feature in the particle volume distribution, V(D), was the presence of high concentrations of <1 µm volume fractions in surface waters and their reduced abundance in subsurface waters (Fig. 5c). The highest increase in the abundance of submicron particles relative to larger particles was found in samples collected furthest to the west along lines 600 and 700 where surface water salinity associated with the river plume was less than 24 PSU. A similar observation also pertains to the surface water sample from station 380 located near the Mackenzie River's Kugmallit Bay channel, even though the salinity was ∼28 PSU (Fig. 5c). However, the fraction of meteoric water was similar to that at station 620 (Fig. 5d). The PSD measurements for low salinity, highly turbid samples nearest to the river mouth (stations 390, 394, and 690) were not possible due to limitations of the Coulter technique. Station 110 stands out among line 100 stations with <1 µm volume fractions of 0.29 at the surface (salinity of 29.1 PSU) and 0.09 at 60 m (31.6 PSU).

To conclude, from the data in Fig. 5 we find that (1) when f_MW increased in the surface waters of the southeast Beaufort Sea, POC∕SPM ratios decreased while the <1 µm particle fraction increased, and conversely (2) when the f_SIM influence increased, POC∕SPM increased while the <1 µm particle fraction decreased in surface waters.

3.3 Relationships between SPM, POC, and particulate beam attenuation

The SPM of the samples examined during the MALINA cruise ranged from 0.04 to 140 g m⁻³ with associated POC from 0.007 to 1.5 g m⁻³ (Doxaran et al., 2012). Organic-dominated and mixed particle assemblages were predominant in the portion of the dataset obtained from ship-based sampling, with SPM extending to 5.6 g m⁻³. The mineral-rich particle assemblages were more common in turbid estuarine waters located close to shore (Fig. 5a). These waters were sampled using a small barge with an optical package that included a WET Labs AC-9 meter (Doxaran et al., 2012) but no WET Labs C-Star 660 nm. The nearest wavelength band on the AC-9 was 676 nm. It thus provided c_p(676). Note that much higher sediment loads were observed in the region in the past. For example, Carmack and Macdonald (2002, their Fig. 10) reported near-bottom SPM values of 3000 g m⁻³ due to resuspension of bottom sediments during a storm in September 1987.

Data from all 28 stations with coincident measurements were used in the development of relationships between c_p(660) and SPM and between c_p(660) and POC. The particulate beam attenuation coefficient correlated well with both SPM and POC (Fig. 6a, b).

and

where SPM and POC are in grams per cubic meter and c_p(660) in per meter, with r² of 0.71 and 0.74, respectively. Further details on the evaluation of the regression fits are provided in the Supplement. In some instances, for example in biogeochemical modeling studies, the objective may be to estimate light transmission from SPM or POC that has been either measured or is available as model output. For this case the best-fit regression functions are c_p(660)=0.4267 PM^0.9068 and c_p(660)=3.088 POC^1.098, respectively.

The slopes of the best-fit lines (with intercepts set to zero) obtained through linear fitting to all pairs of c_p(660) vs. SPM and c_p(660) vs. POC data were 0.404 m² g⁻¹ (r²=0.70) and 3.39 m² g⁻¹ (r²=0.72), respectively. These slope values represent average SPM-specific and POC-specific particulate beam attenuation coefficients, respectively, for the examined dataset. Our average SPM-specific particulate beam attenuation coefficient at 660 nm is consistent with the range 0.2–0.6 m² g⁻¹ reported by Boss et al. (2009) and Hill et al. (2011) for a 12 m deep coastal site in the North Atlantic Ocean (Martha's Vineyard, MA, USA). Our average POC-specific value is near the middle of the range from 2.31 m² g⁻¹ at c_p(660)=0.45 m⁻¹ to 4.10 m² g⁻¹ at c_p(660)=0.07 m⁻¹ observed by Stramska and Stramski (2005) in the north polar Atlantic. Jackson et al. (2010) reported beam attenuation vs. SPM and POC correlations for measurements in the Arctic Ocean in 2006–2007, from which we estimate SPM-specific values of 0.34–0.50 m² g⁻¹ and POC-specific values of 3.4–3.7 m² g⁻¹ for the c_p(660) range from 0.07 to 0.45 m⁻¹. The slopes calculated from our data within this same c_p(660) range were 0.46 m² g⁻¹ (r²=0.57) for c_p(660) vs. SPM and 2.47 m² g⁻¹ (r²=0.69) for c_p(660) vs. POC, with the latter being consistent with other datasets (e.g., Cetinić et al., 2012) but notably smaller than the Jackson et al. (2010) value.

Figure 6SPM and POC as a function of particulate beam attenuation coefficient at 660 nm based on measurements from CCGS Amundsen during the MALINA expedition in 2009 (a–b) and as a function of the particulate beam attenuation coefficient at 676 nm measured with the AC-9 from the barge (c–d) (the latter data contain only surface samples). The dotted squares in (c) and (d) indicate axis limits in (a) and (b), respectively. The colors of the data points indicate POC∕SPM categories: mineral dominated (red), mixed (blue), and organic dominated (green).

Download

The data of SPM used in fitting the relationship of SPM vs. c_p(660) range from about 0.04 to 5.6 g m⁻³ (Fig. 6a). This corresponds to c_p(660) values up to about 3.1 m⁻¹; however, the highest measured c_p(660) where WET Labs C-Star measurements were made (but not accompanied by SPM sampling) was 8.8 m⁻¹ (at a depth of 10 m at station 394), which according to Eq. (2) would correspond to SPM of about 14.8 g m⁻³. For the purpose of examining SPM patterns we extend the use of Eq. (2) beyond the maximum measured SPM. A similar nonlinear least-squares regression analysis that included the highest observed SPM values and corresponding beam attenuation values measured at 676 nm using a WET Labs AC-9 resulted in a very good fit and a trend line approximating that of the extrapolation of Eq. (2) (Fig. 6c). This supports the assumption that the estimation of SPM from beam attenuation measurements can be reasonably well extended to cover the broader range of values measured with the WET Labs C-Star, thus being valid from the very clear open ocean to the highly turbid estuarine waters.

The situation is different for the POC vs. c_p(676) regression. Coincident observations of POC and c_p(676) reveal a tendency of POC to level off at the very high attenuation values (Fig. 6d). These high c_p(676) values were all observed from the barge in the shallow estuarine waters of the Mackenzie River mouth (Doxaran et al., 2012). As the particle assemblages within these coastal waters are dominated by mineral particles, a weak relationship between POC and c_p is expected. However, within the POC range up to about 0.45 g m⁻³ and c_p(660) ≤ 3 m⁻¹ covered by ship-based observations (Fig. 6b), which included only organic-dominated and mixed particle assemblages (POC∕SPM ≤25 %), both c_p(660) and c_p(676) are well represented by Eq. (3). This covers the range of c_p(660) observed along all the ship-based transects (Fig. 1).

3.4 SPM distributions on the shelf, slope, and beyond

The large range in concentration and composition of suspended particle assemblages (Figs. 5 and 6) collected as a part of the MALINA dataset allowed the determination of empirical relationships for estimating SPM and POC from c_p(660) (Eqs. 2–3) in Canadian Beaufort Sea waters. In the following, SPM distributions in the Canadian Beaufort Sea are investigated by applying the SPM algorithm to c_p(660) data collected during three cruises in the Canadian Beaufort Sea. These cruises include the 2-year-long projects CASES (2003–2004) and IPY-CFL (2007–2008) and the MALINA project in August 2009, which altogether cover a wide range of conditions encountered during the open-water season in the Canadian Beaufort Sea. Additionally, c_p(660) data from four deep casts in the Canada Basin collected during the JOIS expedition in September 2009 are examined to show conditions further away from the shelf break (Fig. 1).

Here, we focus on the cross-section plots for transect lines 100, 300, and 600 only (Fig. 1). These transect lines have also been repeatedly measured during other field campaigns (e.g., Carmack et al., 1989; Tremblay et al., 2011; Lansard et al., 2012; Mol et al., 2018). Line 100 crosses the Amundsen Gulf near its entrance from north of Cape Bathurst towards the southwestern point of Banks Island. Line 300 is a south-to-north transect located approximately along 134^∘ W and associated with Kugmallit Valley. Line 600 follows the Mackenzie Trough and provides the western border to the Mackenzie Shelf. The Mackenzie River delta is a maze of tributaries; however, the main discharge channel exits at Mackenzie Bay near the end of line 600, while the second largest channel exits at Kugmallit Bay near line 300.

Figure 7Concentration of suspended particulate matter, SPM, calculated from measurements of particulate beam attenuation coefficient at 660 nm, c_p(660), using Eq. (2) for lines 100, 300, and 600 during different field campaigns in 2004, 2008, and 2009, as indicated.

Download

Figure 7 shows the SPM fields from the three expeditions, derived from c_p(660) profiles using Eq. (2). Figure 8 provides the supporting temperature and salinity fields. Black contour lines show SPM values up to 10 g m⁻³ (Fig. 7f). We recall from Sect. 3.3 that both Eqs. (2) and (3) are derived from ship data and are strictly valid for c_p(660) values up to 3.1 m⁻¹ (Fig. 6). Thus, this excludes the most mineral-dominated waters on the shelf with SPM over 5.6 g m⁻³ and POC over about 0.5 g m⁻³. However, comparisons against nearshore data collected with the barge indicate that Eq. (2) for SPM is reasonably valid for a wider range (Fig. 6c). This is not the case for POC. Within the valid range (c_p(660)<3.1 m⁻¹) the presented SPM (g m⁻³) fields can be converted to POC (g m⁻³) according to POC =0.1279 SPM^0.7307, which is derived from the regression analysis of POC vs. SPM data.

Figure 8As Fig. 7 but for measurements of water temperature (colors) and salinity (contour lines).

Download

Elevated SPM values were generally present in shelf surface waters and associated with a lower salinity surface layer or plume. The highest values were seen nearest to the shore in shallow waters, indicating the riverine origin. SPM in the surface layer decreased past the shelf break, often reaching very low values, except within the northwest-flowing Mackenzie River plume during the 2004 CASES and 2009 MALINA expeditions (Fig. 7g, h). Clear waters with SPM ranging between 0.04 and 0.06 g m⁻³ were found offshore on line 300 in each of the three expeditions (Fig. 7d–f). The corresponding POC ranged from 0.01 to 0.02 g m⁻³. The low SPM values were especially widespread in August 2009, likely related to the high f_SIM content (Table 1).

Wedges of very clear water are seen extending far onto the shelf particularly during 2009. The extension of clear waters onto the shelf as a wedge between the surface plume and the turbid near-bottom layer has been described by Carmack et al. (1989). It appears that neither particle settling from the surface plume nor the upward mixing of bottom sediments was sufficient in August 2009 to increase these clear-water values of c_p(660) above those found in deep basin surface waters. The landward extension of the clear-water layer was particularly noticeable on line 600 (Fig. 7h), which corresponds to the Mackenzie Trough, the main river channel and the most distinct surface plume feature of the transects.

Figure 9Vertical profiles of (a) suspended particulate matter, SPM, calculated from particulate beam attenuation coefficient at 660 nm, c_p(660), and (b) temperature, T, and salinity, S, at selected “deep” stations. Inserts (c) and (d) show transmissometer data (converted to SPM using Eqs. 1–2) that were collected in the Canada Basin during 21–23 September 2009 and made available by the Beaufort Gyre Exploration Program based at the Woods Hole Oceanographic Institution in collaboration with researchers from Fisheries and Oceans Canada at the Institute of Ocean Sciences. Insert (e) shows a close up of the potential temperature, θ, and S for CB-23, CB-27, and CB-21 at the interface to the Canada Basin Bottom Water layer. Grey horizontal lines indicate bottom depths and are underlain by station numbers (see Fig. 1 for locations).

Download

3.5 Environmental forcing and oceanographic conditions

As is evident, SPM is not a conservative property of a water mass but undergoes settling or resuspension at rates that are dependent on particle composition and size and water dynamics. Consequently, in this section, the environmental forcing and oceanographic conditions during each of the three expeditions are first described and contrasted. Then, in Sect. 3.6, the observed patterns of the SPM fields are compared and discussed in the context of these oceanographic conditions, and in particular as these patterns relate to river runoff, sea ice melt, and wind.

Figure 10(a) Daily discharge for the Mackenzie River at the Arctic Red River location (10LC014). Data obtained from Environment Canada. (b) Weekly ice coverage for the Mackenzie Shelf area calculated using IceGraph 2.0 provided online by the Canadian Ice Service. The distributions of sea ice types are provided in Fig. S4. Time periods for the three expeditions considered in this study are also indicated in color shades. The x-axis labels are dates expressed in the format of month/day and are spaced 4 weeks apart.

Download

3.5.2 Wind forcing

The large freshwater inputs to the Mackenzie Shelf during summer result in strong vertical stratification and a vertically sheared two-layer circulation (Fig. 3) (Carmack and Macdonald, 2002; Carmack and Chapman, 2003; Mol et al., 2018). This estuarine circulation is reflected in the patterns of SPM across the shelf (Fig. 7). Sustained easterly along-shelf winds, particularly when strong, are known to cause offshore Ekman transport of shelf surface waters, thereby generating upwelling of deeper nutrient-rich water of Pacific origin onto the shelf (Carmack and Kulikov, 1998; Williams et al., 2006, 2008; Yang, 2009). The high salinity observed during the MALINA expedition in Kugmallit Valley (line 300), Mackenzie Trough (line 600), and near the coast west of 140^∘ W indicated the occurrence of upwelling (Fig. 2). During westerly winds, onshore Ekman transport will generate downwelling flow on the shelf (Dmitrenko et al., 2016). During westerly or weak winds, the river plume turns right to flow eastward along the coast of Tuktoyaktuk Peninsula (Carmack and Macdonald, 2002). Relaxation or reversal of either of these winds will cause return flow to occur towards or from the shelf. Furthermore, strong winds, and brine released from ice formation during late fall and winter, promote vertical mixing and may mix shallow shelf waters to the bottom (e.g., Carmack and Macdonald, 2002; Forest et al., 2007).

Figure 11Progressive vector plots of (a) daily average wind (NCEP, 10 m) and (b) currents from mooring CA05, with am inset for 2003–2004 CASES data. Colors in (a) indicate daily average wind speeds shown in the color bar. The black squares in (a) and (b) indicate the start of each month (the first letter of month followed by year, e.g., D03 stands for December 2003), while the blue circles show the approximate times of the ship-based transect sampling across the Mackenzie Shelf break used in this study. The black line shows the direction along the shelf break referenced to true north. In (b), the same vector plot for currents is shown three times, but the colors of each plot indicate current speed, salinity, or temperature as denoted next to the lines and shown in more detail in (c) and (d). For the inset in (b) showing 2003–2004, the colors indicate salinity as denoted. Time series for 2008–2009 for (c) current speed (colors are current directions), (d) salinity (red), and temperature (blue) at a depth of 178 m.

Download

The wind vectors reveal a predominance of easterly winds during our study periods in 2004 and 2008–2009 often with a southward component resulting in an along-shelf wind component (Fig. 11a).

The predominance of easterly winds is also a driving force behind the large-scale anticyclonic circulation of the Beaufort Gyre and its ice cover. The occasional reversals of the Beaufort Gyre are related to transient synoptic weather patterns (Asplin et al., 2012) that also affect the circulation on the shelf. Two notable periods dominated by westerly winds occurred during the month of October 2008 and during December 2008 to the end of January 2009. Typically the westerly wind episodes were characterized by relatively low wind speeds.

The wind conditions prior to the ship-based expeditions (marked by blue circles) are shown in Fig. 11a. During June–July 2004 (CASES) the wind speed ranged from 2 to 8 m s⁻¹ with a variable direction. IPY-CFL sampling (late June and early July 2008) overlapped with CASES in terms of time of year; however, winds were notably different with a month of easterly winds prior to the sampling. The conditions leading up to the MALINA expedition in August 2009 are characterized by <10 m s⁻¹ winds at directions inducing the upwelling in June and most of July, but with a turn to northerly winds during the first part of July, which probably were a contributing factor keeping sea ice on the shelf. Winds turned to easterly (upwelling inducing) for the last week of July with wind speed >8 m s⁻¹. Winds during the MALINA expedition were comparatively weak (<6 m s⁻¹) with variable direction.

3.6 Effects of river runoff and sea ice melt on SPM distributions on the shelf