The Southern Ross Sea is characterized by the presence of a polynya, large area of open water surrounded by sea ice, that is generated by strong katabatic winds (a downslope wind caused by the flow of an elevated, high-density air mass into a lower-density air mass below under the force of gravity), that transport and control the dynamics and temporal coverage of sea ice (Zwally et al., 1983; Jacobs and Comiso, 1989). This physical phenomenon plays a major role in driving variability in the functioning of the ecosystem on a short time scale (Hecq et al., 1993). Typically, during late winter and early spring in the western Ross Sea, the ice-free surface spreads from south to north starting from the Ross Ice Shelf Polynya (RISP), causing the ice to retreat and melt and expose the water column to sunlight earlier than in other areas (Hecq et al., 1992). This characteristic extension of the marginal ice edge seems to be responsible for higher productivity, and an increase in the production season due to the shift in the length of the ice edge from south to north (Saggiomo et al., 1998; Lazzara et al., 2000).

To characterize sea-ice variability within the study area during the cruise period, monthly percent sea-ice concentration maps, derived from satellite observations (www.nsidc.org), were examined for the two months prior to the winter surveys, then during and up to one month after completion of the survey (Figure 1). Due to the very bad meteo-oceanographic conditions that were encountered between stations 1 and 2, and the very extensive ice cover between latitudes 65°S and 70°S, the sheduled distribution of the stations was changed to accommodate the conditions. The zone near the Sub-Antarctic Front (SAF), south of New Zealand, included only station 1, located at 50.9222°S and 172.0011°E (Figure 2), while stations 2–4 were located in the Polar Front (PF) free ice area, north (station 2) and south (stations 3 and 4) of the Antarctic Convergence (AC). South of the AC, stations 5 and 6 were located near the Balleny Islands in an area nearly continuously covered by pack-ice and the presence of icebergs, (e.g. station 7, which was NE of Cape Adare). The oceanic zone, which was covered with drifting ice between 66°S and 70°S, included stations 8 and 9. The pack-ice-free zone between 71° and 73°S included the northern stations 10–12 and the southern transects (stations 13-16), which cut transversely through the Scott Canyon that separates the Mawson Bank from the Iselin Bank, and connects the Ross Sea with the ocean. Stations 17–19 represented a north-south transect, from the Campbell Plateau to the Ross Sea shelf, thus including the area before the Convergence, the Antarctic Convergence and Divergence and the Continental Slope. Stations 19–25 represented a transverse transect that cut the Ross Sea from east to west, along the latitude of 75°S.

An interdisciplinary study of the Ross Sea ecosystem was undertaken during the V^th ItaliaAntartide Expedition on board the R/V Cariboo (Guglielmo et al., 1992). Between 25 November 1989 and 12 January 1990, 27 stations were sampled in the Pacific sector of the Southern Ocean, from New Zealand to the Western Ross Sea, across the Antarctic Convergence (Figure 2). Zooplankton was collected over several depth-layers, from the surface to near the seabed (Table 1), by BIONESS (Bedford Institute of Oceanography Net Environmental Sampling System), equipped with ten black nets of 500 μm mesh size (0.25 m² mouth area; Sameoto et al., 1980). An altimeter PSA-900 (Benthos Inc.) provided information on the real-time distance to the bottom. At each sampling station, a multiparametric probe SBE 911plus (Sea-Bird Scientific, Bellevue, WA USA), which was mounted on the BIONESS frame, recorded vertical profiles of temperature (°C), salinity, and fluorescence (µg L⁻¹ Chl a). The BIONESS was deployed at low speed along an oblique path to the maximum selected depth to be investigated, then towed at a speed of 1.5–2 m s^-1 until closing, with the simultaneous opening of a new net. At almost all stations, two vertical profiles were undertaken, one from the maximum depth up to 100–200 m and the other over the depth of the euphotic layer. Depths for the euphotic layer samples were chosen according to the thermohaline structure and the fluorescence profiles. The depths of the sampled strata depended on the bottom depth (Table 1). Fishing upward, the nets were opened and closed on command, between 10–20 and 200 m intervals. Due to differences in depth at the sampling sites it was not always possible to collect the same number of samples. The volume of water filtered for each layer varied between 43 and 372 m³, according to the thickness of the sampled layer. A total of 410 samples was collected, with a total volume of 32684 m³ of water filtered. The zooplankton samples collected at each station from the surface to the maximum sampled depth, during the downward deployment of the BIONESS, were not used for this work but observed in the laboratory for the detection of rare species. On board, all the zooplankton samples were preserved in a 4% buffered formaldehyde-seawater solution.

In the laboratory all amphipods were counted and removed, fixed in 70% ethyl alcohol and identified to species level based on the taxonomic keys provided by Bowman (1973, 1978), Bowman and Gruner (1973); Shih (1991); Vinogradov et al. (1996), and Zeidler (1999, 2003, 2004a, 2004b, 2009, 2012a, 2012b, 2015, 2016) and classified in accordance with Lowry and Myers (2017). Taxonomic names were verified against the World Register of Marine Species (WoRMS) (Horton et al., 2019). All occurrence records at the subspecies level, synonyms, and misspellings were corrected to the valid species name and included.

Subsamples (1 to 10 mL) were extracted by Stempel pipette and observed under a Leica Wild M10 stereomicroscope for the determination of mesozooplankton abundance. The abundance of amphipods in each stratum was calculated by dividing the total number (N) by the volume of filtered water and expressed as individuals 100 m^-3. For the entire water column, the total abundance of amphipods at each station was expressed as a weighted mean, summing all counted amphipods and dividing by the total of seawater filtered volume (m³), then multiplied by the water column thickness and expressed as individuals m^-2. The mean total abundance of a given species is represented by the geometric mean and expressed as individuals m^-2, to minimize the difference in water column depth.

Using the Seabird CTD profiles potential temperature (θ, °C) and potential density anomaly (σ₀, kg m^-3) were calculated and plotted with Ocean Data View software (Schlitzer, 2023). Flow velocity and filtration efficiency were monitored by internal and external TSK flowmeters.

Species richness R and genus richness D for amphipods were calculated for each station and used to summarize their diversity over the sampling region (Burridge et al., 2017).

To gain insight into the underlying causes of the latitudinal trends in species richness, a partitioning clustering method named k-means algorithm (Jain and Dubes, 1988) was used to classify observations and distinguish the spatial distribution of different groups or clusters. It classifies objects into multiple groups, so that objects within the same cluster are as similar as possible. Since the number of clusters (k) was not known a priori, a preliminary analysis was performed to evaluate the partitioning of the observations between the different clusters, considering a number of clusters between 2 and 10. For each k value, 10 iterations were performed, starting from random initial positions for the k centers and taking into account the relative average values obtained. To further evaluate how many clusters to consider when determining an optimal solution, internal validity indexes were applied. In particular, the indices used by Giacalone et al. (2022) were considered (Table 2), using the NBClust package (Charrad et al., 2014) in the R statistical environment (R Core Team, 2020).

The factoextra package in R was used for carrying out the cluster analysis through k-means. Once the number of clusters was identified, the main species that were more abundant in the study area were analyzed and boxplots were generated to illustrate their abundance and distribution within the identified clusters.

Relationships between the amphipod communities and environmental variables at the sampling stations were explored using a combination of non-metric multidimensional scaling (NMDS), an ordination method that can summarize relationships into 2-dimensional space and an Envfit function that fits environmental vectors or factors onto an ordination. For this a R’s Vegan package was used (Oksanen et al., 2017). To perform the NMDS analysis, the environmental data and species abundances were grouped in three sub-layers, that were chosen based on the physical features of the stations. The distance matrix used in the NMDS was the Bray-Curtis similarity index. The dataset was square-root transformed and standardized by means of the Wisconsin method before computing the distance matrix. The square-root transformation was used to downweight the importance of very abundant species, while the Wisconsin standardization (performing a double standardization in terms of species maxima and layer abundance) enabled the highlighting of the differences between layers. The performance of the ordination was evaluated by looking at the stress coefficient, where stress values higher than 0.2 highlight poor NMDS performance (Clarke, 1993) and results should be interpreted with caution.

The analysis of the hydrological characteristics of the 27 CTD stations provided evidence for the presence of several water masses. The potential temperature (θ in °C) and salinity (S) measurements obtained from CTD casts were used to construct the θ/S diagrams (Figures 3A, B), where the dashed line indicates the surface freezing point (Tf, Figure 3A). All waters with σ₀ less than 27.55 kg m^-3 can be generically encompassed under the definition of Antarctic Surface Water (AASW), in agreement with Bergamasco et al. (2002). This water mass occupies the upper ocean and is significantly affected by surface processes such as cooling/heating, air temperature, wind and precipitation (Wang et al., 2023).

Temperatures higher than 1.0 °C are typical of the stations dominated by Circumpolar Deep Water (CDW; Bergamasco et al., 2002). The CDW signal was identified at some CTD stations (4, 5, 8, 9, 10, 11, 14 and 15 in Figure 1) by the typical bell shape with temperatures higher than 1.0 °C between 200 m and 850 m (Figures 4 and 5). According to Orsi and Wiederwohl (2009), some CDW continues poleward and enters the cyclonic circulation of the Ross Gyre, entrenched between the colder AASW above and the Antarctic Bottom Water below (Orsi et al., 1999). During the survey this structure in the water masses was observed in the stations 8, 9, 10 and 11. The Modified Circumpolar Deep Water (MCDW) originates from CDW and is modified by cooling and mixing with shelf waters (Wang et al., 2023). It is defined by temperatures up to -1.0 °C at the shelf-break (Jacobs et al., 1985) and was observed at CTD stations 12, 13, 17, 19, 20 and 21. The waters with temperatures below the sea surface freezing point (Tf) define the Ice Shelf Water (ISW; Jacobs et al., 1985). ISW found on the continental shelf (stations 25, 26 and 27) had salinities around 34.7 and minimum temperatures reaching down to -2.08 °C, values which agree well with previous observations (Gouretskvi, 1999).

Fluorescence profiles show the presence of DCM (Deep Chlorophyll Maxima) at depths between 10 m and 28 m. The comparison of chlorophyll values (as a proxy of primary production), recorded in the S-N (Figure 4) and W-E (Figure 5) sections, shows that the innermost area of the Ross Sea is characterized by higher productivity.

Mean total zooplankton abundance across the 27 sampled stations was in the range 1392.27 ± 2457.23 ind. 100 m^-3. Copepoda was the most abundant taxon, representing 79.3% of total zooplankton (total weighted mean abundance 1103.76 ± 1943.43 ind. 100 m^-3), followed by Chetognatha (65.49 ± 85.27 ind. 100 m^-3, 4.7%), Euphausiacea (44 ± 152.46 ind. 100 m^-3, 3.2%), Pteropoda (32.93 ± 49.83 ind. 100 m^-3, 2.4%), Ostracoda (32.98 ± 36.34 ind. 100 m^-3, 2.4%), Polychaeta (34.46 ± 185.74 ind. 100 m^-3, 2.5%), Salpida (20.69 ± 120.17 ind. 100 m^-3, 1.5%) and fish larvae (13.96 ± 67.2 ind. 100 m^-3, 1%). All the other taxa represented less than 1% each and together accounted for 5.4%.

A total of 2058 pelagic amphipods were observed for the study region, representing 0.4% of the total zooplanktonic community abundance. Overall, 24.8% samples (106 out of 410) did not contain amphipod specimens, regardless of the filtered water volume. The total amphipod abundance was 925.6 ind. m^-2 and the mean abundance among the 27 sampling stations was 34.3 ind. m^-2 (± 33.6 ind. m^-2). Forty-three taxa were recovered in the investigated area (Table 3), that belonged to 18 families of amphipods and to 2 suborders. Twenty-seven species (belonging to 15 families), twelve genera (belonging to 8 families) and one taxon at family level (Phoxocephalidae) were identified.

Hyperiella dilatata (Hyperiidae) was the most abundant species, representing 44.9% (1033 total counted specimens, 15.4 ind. m^-2) of the total collected amphipods. Primno macropa and Pseudorchomene plebs represented 12% (169 total counted specimens, 4.11 ind. m^-2) and 11.8% (289 total counted specimens, 4.05 ind. m^-2), respectively. In decreasing order the amphipod community included Hyperiella macronyx 8.1% (191 total counted specimens, 2.77 ind. m^-2), Scina sp. 3.4% (44 total counted specimens, 1.16 ind. m^-2), Vibilia armata 3.3% (61 total counted specimens, 1.12 ind. m^-2), Themisto sp. 3% (59 total counted specimens, 1.02 ind. m^-2), Themisto gaudichaudii 2.6% (36 total counted specimens, 0.9 ind. m^-2), Epimeria (Epimeriella) macronyx 2.5% (47 total counted specimens, 0.84 ind. m^-2) and Pseudorchomene rossi 1.2% (32 total counted specimens, 0.42 ind. m^-2). The remaining 33 identified taxa contributed less than 1% to the total amphipod abundance, although many contributed to high biodiversity. Together these rarer taxa represented 7% of the amphipod community.

Amphipods were observed at all sampling stations, though not in every sample from each station. The horizontal distribution of total amphipod abundance across the 27 stations is shown in Figure 6. Highest abundance (119.4 ind. m^-2) was observed at station 25 within Terra Nova Bay and the lowest (6.73 ind. m^-2) at station 14. A high abundance was also observed at station 1 at the Sub-Antarctic sampling site south of New Zealand. Amphipods were particularly abundant in the Polar Frontal Zone, both at station 2, north of the Antarctic Convergence (six identified species and four genera of unidentified species) and at the southern station 3 (seven identified species and three genera of unidentified species). The lowest amphipod richness (three species) was found at stations 12 and 18 (Figure 7).

Hyperiella dilatata and P. macropa were distributed widely (20 out of 27 sampling stations), with a percentage frequency of occurrence (PFO) of 74% (Table 3). Hyperiella macronyx (PFO = 67%) was detected at 18 sampling stations. Pseudorchomene plebs was present at 48% of the stations, followed by P. rossi (37%, 10 stations), E. (Epimeriella) macronyx and Scina sp. (7 stations), Scina antarctica (5 stations) and V. armata (4 stations). All the other identified 25 species or taxa were detected occasionally, being present at one to three stations.

The horizontal distribution of the four most abundant species is shown in Figure 8. Hyperiella dilatata had the highest abundances in the Ross Sea, particularly inside Terra Nova Bay near the coast at station 26 (90.3 ind. m^-2). Primno macropa had a more pelagic distribution, starting from the eastern part of Cape Adare and the southern Pacific sector north of the Balleny Islands, with the highest abundance observed at station 7 (18.56 ind. m^-2). Pseudorchomene plebs was not detected off Cape Adare going north towards New Zealand. The highest abundance was at station 20 (23.3 ind. m^-2), where the juveniles (17.8 ind. m^-2) were more abundant than the adults. Hyperiella macronyx showed a similar horizontal distribution to P. plebs, but with the highest abundance inside Terra Nova Bay, especially at station 27 (14.6 ind. m^-2). The other six species that followed in decreasing order of abundance (see Table 3), present each one with a percent contribution between 3.4% and 1.2%, showed distinct narrow horizontal distributions. Scina sp. was absent in the Ross Sea and present only off Cape Adare going northwards; V. armata was detected at stations 1 to 5; Themisto sp. was collected at station 1, the sampling site closest to New Zealand; T. gaudichaudii was present at stations 1 to 3; E. (Epimeriella) macronyx and P. rossi were detected only in the Ross Sea and inside Terra Nova Bay.

Over the vertical plane the amphipod community mainly (75%) occupied the 20–100 m layer. The 7.3% of the community was collected in the surface layer, while 15.9% occurred between 100 and 500 m. Only 1.6% were detected in the deepest layers between 500 and 1000 m. The patterns of vertical distribution for the four most abundant species are shown in Figure 9. For H. dilatata, P. macropa and H. macronyx the abundance includes both adults and a very few juvenile specimens, while for P. plebs the two developmental stages are shown separately.

Hyperiella dilatata occupied the depth from 800 m to the surface, although the bulk of the population (88%) was found in the 0–100 m layer.

Primno macropa showed a more uniform distribution through the water column than H. dilatata and was collected from the surface to 700 m. The main part of the population (80%) showed a wide distribution, occupying the 40-300m depth layer. No individuals were found in the 400-500 m layer. Only six P. macropa juvenile specimens were collected, in the 300-400 m and 100-200 m layers.

Adult Pseudorchomene plebs occurred between 500 m and the surface. The main part of the adult population (87%) occupied the 20–100 m layer, while only 2% were present in the most superficial layer (0–20 m). Juvenile P. plebs were almost as abundant as adults and showed similar vertical distribution.

Hyperiella macronyx were distributed from 400 m to the surface. Approximately 80% occupied the 20–100 m layer, with ~4% in the more superficial layer and the 16% below 100 m depth. Only two juveniles were collected in the 200–300 m layer.

The spatial distribution boundaries of the amphipod taxa collected during this study, are shown in Table 4. Only two species (Phronima atlantica and Themisto compressa) and one genus (Themisto sp.) were found exclusively in the Sub-Antarctic zone (station 1) at latitude 50.9222°S, while three species (Halice tenella, Hyperiella antarctica and Parandania boecki) and one family (Lanceolidae) were found only in the area north of ​​the Polar Front (station 2, 61.9956°S). Cyllopus magellanus and T. gaudichaudii were recovered only between the Sub-Antarctic zone and Polar Front (stations 1 and 3), while Cyphocaris richardi and two genera (Cyllopus sp. and Vibilia sp.) were present only in ​​the Polar Front (stations 2 and 3), between 61.9956 and 63.0764°S. Four genera (Archaeoscina sp., Ctenoscina sp., Lanceola sp. and Parandania sp.) were found in the area south of ​​the Polar Front (station 4, 63.9489°S). One species (Scopelocheiropsis abyssalis), one genus (Acantoscina sp.) and one family (Archaeoscinidae) were found in the ice-covered stations 6 and/or 7. Hyperiella dilatata and H. macronyx were present throughout all the study area, including Terra Nova Bay, with a small difference in the northern limit of their detection: the Polar Front (61.9956°S) and the Balleny Islands (64.8675°S), respectively. Primno macropa was collected over a large area, from the Sub-Antarctic Front to the mid Ross Sea Region (at station 21). Some species were found only along the two transects at the northern entrance of the Ross Sea: Cleonardo longipes, Lanceola clausii pirloti and the family Phoxocephalidae were found exclusively in station 10 (71.2194°S), while Ctenoscina brevicaudata, Cyllopus lucasii and Temisto abyssorum were recorded at stations 13,14,15. Eusirus microps and Abyssorchomene scotianensis are exclusive species of Terra Nova Bay (about 75°S).

The k-means algorithm, applied to species abundances estimated for the 27 stations, enabled the evaluation of the spatial distribution of different groups. To evaluate the optimal number of clusters, 15 indices were estimated (Table 2). The results of this approach revealed the presence of three clusters (Figure 10). Cluster 1 was mainly composed of the southernmost stations, Cluster 2 included the three northernmost stations, while the stations in the central part of the study area were grouped in Cluster 3 (Figure 10). The boxplots of the main species in the three clusters are shown in Figure 11.

CTD profiles belonging to Cluster 1 showed the presence of a strong chlorophyll signal, with the DCM located at around 28 m depth. Profiles belonging to cluster 2 were characterized by an almost homogeneous upper layer (0–200 m) in terms of temperature, salinity and fluorescence; for this cluster no clear DCM signal was evident. Most of the stations belonging to cluster 3 were influenced by the presence of CDW and MCDW. At these stations the presence of a DCM was less evident than in stations of cluster 1 and was located around 50 m depth.

NMDS was used to investigate the relationships between amphipod taxa and the sampling stations. The NMDS stress value was 0.11, indicating acceptable fit. The NMDS plot (Figure 12) showed a clear separation between northern and southern stations along the first NMDS axis (NMDS1). In particular, the upper layers (0–50 m and 50–200 m) of the southern stations, belonging to Cluster 1 and affected by higher fluorescence and lower temperatures, were characterized by higher proportions of H. dilatata, H. macronyx and P. plebs, while the species Epimeria (Epimeriella) macronyx was more common in the middle layer (50–200 m) in the south-western side of the Ross Sea. The deepest layer (>200 m) in these stations was characterized by a species assemblage composed of Hyperiella dilatata, H. macronyx, Pseudorchomene plebs and P. rossi. The NMDS plot also showed that the upper layers (0–50 m and 50–200 m) of the stations belonging to Cluster 3, influenced by relatively higher temperature and lower fluorescence values, were mainly characterized by higher proportions of P. macropa. Lower proportions of H. dilatata, H. macronyx and Scina sp. were also observed at a few stations in this cluster. The deepest layer (>200 m), where higher salinity was observed, was mainly characterized by the presence of P. macropa and Scina sp. Higher proportions of T. gaudichaudii, V. armata, and Vibilia sp. characterized the three northernmost stations belonging to Cluster 2. Finally, juvenile specimens of Themisto sp. were found in the upper and saltier layers of station 1.