The study area was located north of 80° N, between the Nares Strait and the Lincoln Sea with the primary sampling program focused on PF and SOF (Figure 1; Supplementary Table 1; Jakobsson et al., 2020a). The oceanography and geomorphology of these two fjords are described by Jakobsson et al. (2020a) and Stranne et al. (2021). In brief, both fjords reach maximum depths exceeding 800 m and are home to the only two marine terminating glaciers in northwestern Greenland with floating ice tongues. Surface and subglacial meltwater contribute to the outflowing current above 200 m and remnants of circulated Atlantic water from the Central Arctic Ocean flow inward at depths below approximately 300 m in each fjord. Ryder Glacier in SOF drains into the Lincoln Sea where multiyear sea ice accumulates. In contrast, Petermann Glacier drains into the Nares Strait, a conduit for glacial ice export in the region.

Sampling stations (Figure 1) were located in SOF and PF during the Ryder Glacier 2019 expedition (August 5 – September 10^th, 2019) on the icebreaker Oden led by Stockholm University and University of New Hampshire (Jakobsson et al., 2020b). At each sampling station, optical (particle imaging), acoustic, and biological (zooplankton) data were collected. In this study we only compared the data and results from SOF and PF, but the images from an additional 13 nearby locations in Kennedy Channel, Hall Basin and the Lincoln Sea were used to better train the marine snow image clustering algorithm in categories of particles (Supplementary Table 1). The seasonal progression of ice conditions in the region was assessed using satellite imagery (MODIS Characterization Support Team (MCST), 2017). Using these satellite images, we describe surface sea-ice extent at approximately 2-week intervals during the cruise period.

Particle and zooplankton images were collected using an Underwater Vision Profiler (UVP5, Hydroptic ^©) mounted to the water sampler rosette. The UVP5 system detects and counts all objects larger than ~100 μm in a ~1 L volume illuminated with a digital camera (Picheral et al., 2010). The system automatically detects objects and stores vignettes of objects >80 pixels (from approximately 0.5 to 200 mm). For this study, the UVP5 was mounted onto the rosette water sampler that also carried conductivity, temperature, and depth sensors (CTD; SBE 911plus, Seabird Scientific, Bellevue, WA, USA). The acquisition frequency reached up to 8 Hz, which was necessary for the resolution of the profile. The rosette was lowered with an average descent speed of up to 0.5 ms⁻¹; however, descent speeds varied considerably during challenging ice conditions. Due to the high levels of suspended sediment particle entrainment in subglacial plumes, the UVP5 was unable to consistently quantify particle distributions below 100 m; therefore, our analyses are restricted to the upper 100 m of the water column. The CTD-Rosette with UVP was deployed at 25 stations in SOF and 10 stations in PF (Figure 1).

Acoustic data (volume backscattering strength S_v in dB re 1 m^-1) for the measurement of mesozooplankton distribution were collected using a rosette-mounted wideband autonomous transceiver (WBAT, Simrad ^©), equipped with an ES333 split-beam transducer that was mounted sideways facing relative to the rosette frame. Data were collected to 50 m range using a 0.5 s sampling interval. Pulses were transmitted in broadband (also known as frequency modulated, FM) mode using a 320–420 kHz bandwidth pulse and a transmitted pulse duration of 1.024 ms and 200 W power. Acoustic data containing biological backscatter were mostly limited to 25–100 m depth due to inherent challenges in the sampling environment: above 25 m, measurements were contaminated by backscatter from nearby icebergs and the air-water interface. Below 100 m, much of the mesozooplankton data were masked by strongly scattering particulates entrained in subglacial turbidity plumes. Ping-depth intervals varied between 0.2 to 0.8 m, depending on the rosette speed during lowering.

Zooplankton samples were collected using a multinet (Hydrobios ^©) deployed at 10 sampling stations across SOF and PF to collect zooplankton at discrete depth intervals. The sampling intervals in the top 100 m were divided as 0–50 m and 50–100 m layers, except at one station where the top layers were divided in 25 m intervals. (Supplementary Table 2). The multinet comprised five nets, each with an opening of 0.25 m² and 200 μm mesh that terminated in a PVC cod end. Following a non-sampling descent, zooplankton were sampled during retrieval, with each net opened/closed at the desired depth by an arrangement of spring-loaded levers which are triggered by a motor unit. An integrated pressure sensor allows for continuous supervision of the operating depth indicated on the display of the deck command unit. After retrieval, the net was returned to the deck, rinsed with fresh seawater, and samples were collected from each cod end and preserved in borax-buffered 4% formalin-seawater solution and later taxonomically identified using microscopy. The multinet was deployed at 7 stations in SOF and 3 stations in PF (Figure 1).

During the 2019 sampling season, sea-ice break up in SOF and PF occurred on similar timescales (Figure 2). Beginning in mid-July, fissures began to form at the glacial tongues and localized ocean circulation and wind drove ice outward toward the opening of each fjord. In PF, ice was eventually exported into the Kennedy Channel – Hall Basin region by August 14^th. In contrast, the large icebergs and fractured sea ice in SOF created ice damming and remained entrapped within the fjord, staying in circulation throughout the short open-water season (Figure 2).

Based on these ice conditions, it was apparent that the waters of PF were exposed to ice-free, open water conditions for a longer period, and thus a higher level of continuous solar irradiance and increased susceptibility to wind-driven mixing. By late August the fjord was open to the surrounding waters, allowing for increased full-depth horizontal exchange with the Nares Strait. While SOF remained nearly free of sea ice in its inner parts throughout the study period, SOF’s mouth was completely sea-ice covered.

In response to these differences in ice regime and fjord circulation, PF and SOF displayed strongly contrasting oceanographic conditions in the upper water column (0–100 m; Figure 3). Vertical temperature and salinity profiles in SOF revealed a stronger surface stratification than in PF, whereby warmer meltwater, reaching conservative temperatures of 3.5 °C and absolute salinity of 15 in the top 15 m. This surface layer overlaid cold intermediate waters with a maximum temperature of 0 °C and higher salinity of 30 (Jakobsson et al., 2020a). Fluorescence and oxygen profiles, which can be indicative of both phytoplankton biomass and the level of aerobic respiration, also show contrasting conditions between the two fjords. The magnitude of fluorescence was higher in PF with a peak observed higher in the water column, on average at ~20 m, compared to ~30 m in SOF. Oxygen saturation was consistently higher in SOF, peaking at a depth of ~15 m with values of 115% (Figure 3).

Marine snow was classified into four categories: Large aggregates (large and fluffy with a median ESD of ca. 2 mm, low circularity, high and relatively uniform brightness); Small spheres (small and circular with a median ESD of 1 mm with high and relatively uniform brightness); Flakes (medium size with a median ESD 1.5 mm, irregularly shaped particles with high and relatively uniform brightness); Dark mix (dark and round with a median ESD of ca. 1.5 mm) (Figure 4). We observed clear differences in marine snow particle structure across the two fjords (ANOSM Global R: 0.78; P = 0.002; Figure 5), which were driven by a greater abundance of all particle types in PF (Figure 5; Supplementary Table 4). Depth-specific concentrations of Small Sphere particles were notably higher in the upper stratum of the water column in SOF and were the most abundant particle type above the pycnocline (Figure 5, Supplementary Figure S2). In contrast, Dark mix particles were most prevalent below the pycnocline (~20 m depth). Depth-specific trends in particle densities were only obvious in the unstratified PF for Dark mix particles that were less abundant in the upper water column (Figure 5). In PF, Flakes were the most abundant particles above and below the pycnocline depth in SOF (20m).

Based on multinet sampling, copepods accounted for 97.7% of the total zooplankton abundance in the upper 100 m. The remaining 2.3% of zooplankton consisted of appendicularia, cnidaria, ctenophores, amphipods, polychaetes, ostracods, and euphausiids. We restricted our quantitative analysis to copepods due to low numbers of the remaining taxa. Large copepods included the three major Calanoid copepods Calanus glacialis, Calanus hyberboreus, and Calanus finmarchicus as well as Metridia spp. and Euchaeta spp. Small copepods included Oithona spp., Microcalanus spp., and Pseudocalanus spp. The nauplii group contained nauplii of all copepod species listed above.

An analysis of similarity (ANOSIM) showed statistically significant differences in the community structure of copepod size classes in the upper 100 m between fjords (R = 0.60, p < 0.005). SOF was characterized by consistently low nauplii abundance (< 10%) and high numbers of small copepods (~50%) (Figure 6). In contrast, PF had a more even distribution of size classes, with nauplii comprising approximately 30% of the total copepod abundance. This resulted in higher total abundance of copepods in PF than in SOF. The lower oxygen saturation levels between 15–75 m in SOF compared to PF are likely related to higher total respiration by zooplankton.

Comparison of broadband acoustic profiles and zooplankton images revealed distinct vertical distribution patterns among fjords (Figure 7). Overall, acoustic backscatter profiles showed a trend of attenuation of copepod backscatter with depth, with highest magnitudes found near the surface (Figure 7A). Functional Data Analysis (FDA) showed strong differences between acoustic profiles in each fjord (p <0.001, F-type statistic = 293). Profiles across SOF showed a more consistent structure, with lower variation among sites compared to PF. Within profiles, SOF also showed a less even distribution of backscatter (Equivalent Area = 59.3 m vs 71.0 in PF) and a shallower center of mass (COM) of backscatter (COM = 50.7 m vs 54.8 m in PF). In PF there was a consistently higher proportion of copepod backscatter below 75 m.

When comparing UVP zooplankton imagery of copepods among fjords (n=310), SOF also showed a shallower distribution of copepods, concentrated in the upper 50 m, whereas PF showed a deeper distribution of copepods, matching patterns in acoustic backscatter (Figure 7B).

In seasonally open-water fjords, such as PF, strong winds can enhance nutrient upwellings and productivity (Castro de la Guardia et al., 2019). Buoyant subglacial plumes can also enhance the nutrient load, creating upwelled regions of high productivity within fjords (Meire et al., 2017). In contrast, in semi-enclosed fjord systems such as SOF, constrained meltwater discharge can produce particularly strong, near-surface stratification, resulting in a shallow nutrient-depleted layer which thereby limits primary production and phytoplankton biomass in the euphotic zone (Hopwood et al., 2020). Increasing meltwater volume from sea ice and glacier melt trapped by ice damming can further enhance this effect, creating thermohaline gradients requiring strong energetic processes such as higher wind stress and turbulence to break down (Cottier et al., 2010).

Although not a direct relationship, the lower fluorescence values in SOF suggest lower density of phytoplankton and is likely a consequence of a reduction in mixing and nutrient resupply in the surface layer (Stranne et al., 2021). This observation was consistent with satellite imagery, which shows a shorter open water period and consistently higher ice cover in SOF than PF (Figure 2). While these observations are not reflective of phytoplankton bloom stage (e.g., pre- or post-bloom), the ice breakup around the same period (mid-July) in both fjords and the peak in chlorophyll below the pycnocline suggest similar post-bloom periods at the time of sampling, even though the higher ice concentration in SOF might have delayed the onset of the bloom to a later period in SOF than PF. In any case, the difference in what is interpreted as lower productivity in SOF is likely related to lower bloom intensity rather than pre-bloom conditions.

In association with the oceanographic differences across our study fjords, we observed strong differences in the overall abundance of marine snow particles and their relative abundance (Figure 5). In the less productive SOF, marine snow particles were at less than a quarter of the densities observed in PF. We believe this to be largely related to the different volumes of senescent phytoplankton building blocks of marine snow in SOF.

Stratification can affect the vertical distribution of marine snow communities in other ways. For example, MacIntyre et al. (1995) documented porous flocs increasing from 38.7 to 59.3 aggregates per liter and settling at ≥20% slower rates near density gradients. Diercks et al. (2019) noted that persistent pycnoclines reduce settling speeds by 2.2–3.5 times and increase particle volume two- to threefold. Their results are explained by the resuspension of the particles at the pycnocline. While we did not observe strong evidence of elevated densities of any marine snow particles at SOF’s pycnocline, we did observe higher densities between 10 and 20 m (Supplementary Figure 2) and evidence of other stratification effects. Small sphere particles in SOF were the most abundant type above the pycnocline whereas Dark mix and Flakes were the most abundant below. Similar changes in dominant particle types were not apparent in the unstratified PF where Flakes were most abundant throughout the water column. Small spheres have attributes that resemble fecal pellets (Poulsen and Kiørboe, 2006) and with relatively low surface area, would sink faster compared to other particle types. We therefore did not anticipate their high abundance in the upper water column. It is likely that ice-related stratification indirectly influences their distribution, not by entrainment, but rather by the increased production of these particles by the SOF zooplankton that concentrated higher in the upper water column to feed. This is supported by observations in PF where copepod distributions were deeper than in SOF and Small sphere particles were least abundant at the surface.

Our data suggest secondary production (i.e. larger zooplankton and nauplii community, abundance, and vertical distribution) is also related to stratification, productivity, and marine snow, at least indirectly. In both SOF and PF, most copepods were mainly distributed below or just above the pycnocline. However, the higher center of mass of zooplankton in SOF could be related to the delayed bloom period (Daase et al., 2021), higher competition for scarce resources, or advection of different zooplankton communities.

Copepods are intrinsically linked to marine snow through their consumption, production, and breakdown of biogenic particles (Long et al., 2007; Jackson and Checkley, 2011; Toullec et al., 2019; Svensen et al., 2024). Differences in marine snow morphology can also influence palatability for copepods. While zooplankton are responsible for creating marine snow, they are also capable of reshaping marine snow communities and energy export patterns through selective grazing (Svensen et al., 2024). Lower-latitude studies using sediment traps have suggested that particles in the upper 100 m are typically composed of algal-derived aggregates, amorphous aggregates and fecal pellets (Riser et al., 2001; Goldthwait and Alldredge, 2006). Large aggregate and Flakes were both the largest and most irregular in shape (associated with previously described qualitative descriptors such as ‘fluffy’ and ‘agglomerated’; Lombard et al., 2013; Trudnowska et al., 2021) and provide more surface area for microorganisms, potentially making them attractive food items for zooplankton. While some copepod species, for example Oithona similis, feed on fecal pellets (Gonzalez and Smetacek, 1994), others such as Calanus spp. and Pseudocalanus spp. selectively feed on, fragment and repack large aggregates (van der Jagt et al., 2020).

The distribution of high-latitude copepods is often shaped by biotic factors, particularly food availability (Herman, 1983; Longhurst et al., 1984; Basedow et al., 2010). In PF, relatively high nauplii abundance suggests a match between the appearance of zooplankton and peak in primary productivity (Cushing, 1990). In contrast, in SOF the lower intensity and potentially delayed bloom likely resulted in low nauplii survival or production due to a mismatch situation typical of a late ice breakup (Daase et al., 2021). Nauplii and adults of copepod species can switch to particle feeding when phytoplankton is scarce (Søreide et al., 2010; Trudnowska et al., 2020; Svensen et al., 2024). Moreover, larger copepods as those observed here can detect and follow the chemical trails left by sinking marine snow aggregates, which can influence their vertical positioning (Lombard et al., 2013). Zooplankton feeding on marine snow typically position themselves at or near the base of the euphotic zone, a strategy that balances optimal feeding opportunities with reduced predation risk in surface waters (Aksnes and Giske, 1993). We observed median depth to occur below the fluorescence maximum in each fjord but this association was deeper and much more decoupled (higher fluorescence maximum and lower median copepod depth) in the more productive and mixed PF. This suggests that in SOF, where phytoplankton abundance was lower, copepods needed to be closer to the food accumulating near the pycnocline due to scarcity elsewhere. This would explain the higher fecal pellet production (i.e., occurrence of Small Spheres) above the pycnocline. Small spheres in the surface of SOF could also be related, in part, to the accumulation and aggregation of small particles transported with fresh water runoff from Ryder Glacier.

The low density of our Large aggregate and Flake fluffy particle types in SOF suggests poor feeding conditions for zooplankton in SOF. However, fully understanding the origins and quality of our marine snow morphotypes as food for zooplankton would require knowledge of the particle composition (e.g. Alldredge, 1998). Unfortunately, we were not equipped to collect samples during our expedition.

Our study compares the cascading effects of contrasting sea ice conditions in two unstudied Arctic fjords. The remote location and harsh environmental conditions provide, however, a small window to access these areas. While our study captures strongly contrasting oceanographic conditions across a small spatio-temporal gradient, we provide only a snapshot of a seasonally dynamic ecosystem. It is therefore difficult to fully disentangle local conditions from differences in seasonal production, succession, and utilization across fjords. We argue however, that seasonal succession is also influenced by the local ice conditions we compare. We also acknowledge that there were likely some spatial gradients in succession within fjords. Unfortunately, our limited sampling window and number of stations required pooling data and precluded in-depth spatial analyses within fjords. Nevertheless, the vertical distribution patterns of both marine snow particles and copepods we observed were stable throughout several weeks of sampling.

Our study also features complementary technologies that include hydroacoustics, optical sensors and traditional nets. These combined methods filled holes caused by technological limitations (e.g. UVP captured near surface zooplankton data unavailable to the acoustic instruments) and provided inference across multiple ecosystem components. Even so, deeper parts of the water column remained inaccessible to our suite of instruments and full coverage of all instruments would have improved our understanding of important processes that occurred in surface strata.

Future studies would benefit from in-situ light measurements and high-resolution particle characterization to better distinguish between biological and physical drivers of copepod vertical distribution. Moreover, collecting and analyzing marine snow particle composition would have provided a better indication of marine snow quality in these environments and their value to zooplankton and benthic ecosystems.