The first validation data for the Aquadopp observations were obtained on 21 and 22. June 2010. The sound-scattering layers were identified at the raw echogram (Fig. 3) as mesozooplankton aggregations by comparison with the net sampling data (Fig. 4). The zooplankton net sampling data were consistent with the acoustic backscatter, indicating short-term variations in biomass and diel vertical migration of zooplankton.

The total mesozooplankton biomass in the entire water column varied from 0.99 to 3.57 gDWm-2. Zooplankton was dominated by the copepod Calanus euxinus, which made up the mean 58 % with the standard deviation (SD) ± 14 % of the total biomass (below for the sake of brevity, such estimates are denoted as mean ± SD%). The contribution of chaetognaths Parasagitta setosa was 21 ± 11 %, followed by copepods Pseudocalanus elongatus (13 ± 7 % of the total biomass). The sum share of other groups of mesozooplankton did not exceed 7 % of the total biomass.

The pattern of the vertical distribution of mesozooplankton biomass reveals a relatively uniform distribution over depth in the evening twilight (18:05–19:00) and at dawn (05:30–06:20) (Fig. 4, left column). At night (21:10–21:55 and 00:05–00:50), the highest concentration of zooplankton was observed in the thermocline layer, while in the daytime (09:00–09:50), the zooplankton maximum was in the layer between the density surfaces σΘ 15.7 and 15.4 (Fig. 4, left column), in accordance with the diurnal changes in the volume backscatter strength (Fig. 3). The deepest layer bounded by isopycnals σΘ 15.9 and 15.7 was inhabited by nonmigrating copepods, the fifth copepodite stage (CV) of C. euxinus (median prosome length 2.3 mm), persistently staying at this depth throughout the day (Figs. 3 and 4, middle column). Visual inspection of live samples revealed quiescent behavior of these specimens and large oil sac volume inside their body, suggesting a diapausing state in C. euxinus CV collected from the deepest layer (Vinogradov et al., 1992).

Three migrating species, copepodites CIV–CVI of C. euxinus (median prosome length 2.6 mm), CV–CVI P. elongatus (median prosome length 0.92 mm), and chaetognaths P. setosa (median length 19 mm), formed daytime zooplankton aggregations in the oxygen-deficient zone (Fig. 4, middle column). Ctenophore Pleurobrachia pileus, also inhabiting the deep layers in the daytime, contributed negligibly to the total biomass due to the low dry matter content in their gelatinous bodies and their low abundance (shown as Var in Fig. 4). At night, most of the migrating zooplankters were concentrated in the thermocline and did not ascend to the warm UML, which was inhabited by small copepods, cladocerans, and small (< 6 mm) chaetognaths.

The echograms based on the data from horizontal-beam transducers A1 and A2 often reveal aggregations of zooplankton at depths of 80–120 m in the daytime (Fig. 5). The aggregations begin to rise around sunset and descend before dawn. Thin, nearly vertical lines on the echogram indicate acoustic traces of the migrating mesozooplankton species. The echogram also shows patches that occupy the entire water column, from the upper to the lower measurement depth, penetrating below the surface σΘ 15.9 and then deeper into the hydrogen sulfide zone. These are clouds of suspended particles (see, for example, Klyuvitkin et al., 2016). Acoustic scattering by clouds of particles sinking through the water column can obscure zooplankton aggregations.

The layers of elevated acoustic backscatter amplitude due to deep zooplankton aggregations are accounted for using the R graphs (Fig. 5b) that were validated by net sampling on 6 October 2013 (Fig. 6), although sampling was not performed at night due to stormy weather. Since the depths of the isopycnals of 15.9 and 15.7 differed by only 3 m, the integrated zooplankton sample was taken in the layer between σΘ = 15.9 and σΘ = 15.4. In this layer, the contributions of Calanus euxinus, Parasagitta setosa, and Pseudocalanus elongatus to the total biomass were 60 %, 26 %, and 12 %, respectively (Fig. 6b). The extremely low zooplankton biomass (< 2 mgDWm-3) in the upper 50 m layer (Fig. 6a and b) is consistent with data on a 4-fold decrease in the annual average biomass of upper dwelling zooplankton in 2013 compared to previous years (Arashkevich et al., 2015).

Zooplankton diel vertical migration trajectories in the R graph are noticeably clear below 40 m (Fig. 5b). The explanation of these phenomena could be that the scattering area of the elongated bodies of the zooplankton species is larger in the horizontal projection than in the inclined projection at an angle of 45∘. Therefore, the orientation of the bodies of mesozooplankton species appears to be mainly vertical during migration. At night, these specimens are randomly oriented in the upper layer, where R≈1. In addition to the diel vertical migrations, intraday vertical fluctuations of zooplankton occur with an inertial period (Fig. 5b). The vertical displacements of the daytime deep mesozooplankton aggregations are coherent with the vertical displacements of both isopycnals and isooxylines. The displacements of isopycnals with amplitudes up to 20 m are mainly due to near-inertial waves.

In October 2016, persistent aggregation of diapausing C. euxinus was detected in the acoustic backscatter signal (Fig. 7), unlike October 2013 (Fig. 5). Zooplankton sampling was performed at midnight and midday on 4 and 5 October 2016 (Fig. 8). The pattern of zooplankton distribution was similar to that in June 2010 (Fig. 4), both in terms of the total biomass and composition of zooplankton and in terms of the day–night vertical distribution. The total mesozooplankton biomass of 1.8–2.3 gDWm-2 was dominated by three species: C. euxinus (59 %–76 %), P. setosa (9 %–22 %), and P. elongatus (5 %–10 %). At night, the maximum aggregation of migrating zooplankters was in the thermocline layer, and at midday, it was in the layer between isopycnals σΘ 15.7 and 15.4 (Fig. 8a). Daytime zooplankton aggregation consisted mainly of C. euxinus (92 % of total biomass) with a small contribution from chaetognaths (7 % of total biomass) (Fig. 5b). The layer between isopycnals σΘ 15.9 and 15.7 was persistently occupied by diapausing C. euxinus CVs. The chaetognaths found in this layer were represented by spent specimens and corpses (Fig. 8b). UML was inhabited by nonmigrating small copepods, cladocerans, and small chaetognaths.

The net sampling data on the day–night vertical distribution of mesozooplankton agreed broadly with the acoustic backscatter observations obtained during the next few days (Fig. 7). On the echogram, one can see a persistently existing backscattering layer associated with the isopycnal layer near σΘ = 15.9, as well as patches of the high-volume backscattering strength at depth during the daytime and their movement into shallower layers at night.

The two-layered structure was also observed at the end of June–early July 2014 (Fig. 9) and validated by day–night zooplankton sampling on 1 and 2 July (Fig. 10). Deeper zooplankton aggregation was monospecific, consisting only of diapausing C. euxinus CVs (Fig. 10b) and formed a thin layer (5–10 m thick). This layer was visible all day and night and was usually located above the isopycnal surface of 15.9. It is clearly distinguished by the value R > 1.1 (Fig. 9b). The daytime zooplankton aggregation consisted of three migrating species and their different developmental stages, C. euxinus, CIV–CVIs, P. elongatus, CV–CVIs, and P. setosa, 14–22 mm in size (Fig. 10b). Since the amplitude of vertical migration is different for different components of this assembly, the daytime deep aggregation reached 35 m in thickness. Before sunset, migrating zooplankters began to move upward and at night formed aggregations at depths above 40 m (Fig. 9a) and peaked in the thermocline at 17–25 m (Fig. 10a).

Since migrating zooplankton aggregations were observed in the deep layers only during the daytime, it is worth comparing the daytime average R profile with that for the nighttime (Fig. 11). Such a comparison clearly reveals the deep maximum of R at the daytime migration depths of mesozooplankton at 90–120 m, as well as the persistent maximum of the diapause layer within the deeper layer at 125–140 m. Notably, the depths of the persistent SSL change by approximately 5 m from night to the daytime, while they completely overlap when considered vs. the density. Such variations in the depth of the SSL might be linked to inertial oscillations (Ostrovskii et al., 2018).

In Sect. 3.2, it was shown that the mesozooplankton species float on isopycnals in the lower part of the oxycline and in the hypoxic zone. Both the diapausing aggregations and the daytime aggregations are displaced coherently by near-inertial waves. The deep aggregations of mesozooplankton are bounded by certain isopycnal surfaces and isooxylines.

Since the oxygen stratification strongly depends on the density stratification in the pycnocline (e.g., Vinogradov and Nalbandov, 1990; Codispoti, et al., 1991; Konovalov et al., 2005; see also example in Fig. 12), it becomes possible to switch from the depth profiles of the directional acoustic backscatter ratio R(z), where z is the depth, to the R([O2]) profiles to investigate the seasonal changes of the sound-scattering mesozooplankton layers in terms of R vs. [O2].

The average monthly profiles of R([O2]) were constructed from R(z) and [O2](z) data for every month when the data were available. To compute the averages, the daytime was defined as a period beginning 2 h after the local time of sunrise (at a given date) and ending 2 h before sunset. The nighttime was defined as a period beginning 1 h after sunset and ending 1 h before sunrise. Example plots of the average profiles 〈R([O2])〉 computed as arithmetic and bootstrap mean values along with 95 % bootstrap confidence intervals are shown for November 2019 in Fig. 13. In the hypoxic zone, the average values 〈R([O2])〉 for the daytime are significantly higher than those for the nighttime. The daytime averages 〈R([O2])〉 > 1.06 were in the range of [O2] = 9–40 µm in November 2019.

The average monthly 〈R([O2])〉 profiles show the seasonal evolution of the mesozooplankton distribution (Fig. 14). The SSLs are barely discernible in January. One can note some activity in the upper part of the oxycline in February. Although we unfortunately do not have data for March, in April, two peaks appear in the 〈R([O2])〉 profiles in the layers where the concentration of dissolved oxygen is 25–60 µm and 4–9 µm. These maxima correspond to the daytime mesozooplankton aggregations and the diapause layer, respectively. The upper maximum of 〈R([O2])〉, which corresponds to the daytime aggregations of mesozooplankton, may weaken in June–July. However, it becomes stronger again at the end of summer and in autumn. The largest value for this maximum over the entire observation period 〈R([O2])〉 = 1.18 is observed in October. At that time, the maximum shifts into the layer where [O2] is 10–25 µm. In December 2019, this peak was between the 10 and 30 µm isooxylines.

The maximum of diapause mesozooplankton was strongest in May and July 2020, reaching almost 1.2 at [O2] = 5–8 µm. In August, the diapause mesozooplankton layer shifts in the lower part of the suboxic zone where [O2] = 3–7 µm. It becomes substantially weaker in September. In October, this layer degrades further. In November, it tends to disappear.

Previously, acoustic measurements at a frequency of 2 MHz were not considered a tool for observations of the mesozooplankton SSLs in the sea due to the limited range of soundings. However, with the advent of ocean profilers with acoustic Doppler current meters, such as the Nortek Aquadopp, it has become possible to obtain the depth profiles of the volume scattering strength at 2 MHz frequency in the entire water column and to study the vertical distribution of zooplankton, such as those in the Black Sea (Ostrovskii and Zatsepin, 2011; Pezacki et al., 2017). Acoustic sounding of mesozooplankton at two angles is made possible using the side-looking head of the Nortek Aquadopp instrument. The combination of horizontal and tilted beam signals allows, on the one hand, the patches of particles to be eliminated and the background scattering level of the echogram to be equalized and, on the other hand, the preferred orientation of mesozooplankton species migrating through the oxycline to be determined. Earlier, Stanton and Chu (2000) reproduced the influence of the orientation of a 3 mm calanoid copepod (modeled as a high-resolution approximation of an animal profile) on the acoustic target strength at 2 MHz with respect to an incident sonar beam. The reduction was found to be 5 %–15 % when copepod orientation was shifted from 0∘ (broadside incidence) to 30–60∘. Benfield et al. (2000) carried out field observations using the Video Plankton Recorder on George Bank and showed that most Calanus finmarchicus (75 %) in the depth range of 10–70 m were within ± 30∘ of the prosome-up or prosome-down orientation. It was suggested that one reason for the behavior underlying the head-up orientation pattern might be due to the predator avoidance strategy aimed at reducing the conspicuousness of C. finmarchicus when viewed from above. Such individuals would present a significantly reduced cross-sectional area to an echo-sounder's transducer with correspondingly diminished target strength. It was concluded that it is necessary to know how the orientation of individuals changes with depth to correctly account for the biomass of mesozooplankton. Experiments using a multiple-angle acoustic receiver array on live copepods and mysids in a laboratory tank showed that it is possible to use the scattered acoustic signal to distinguish among zooplankton taxa (Roberts and Jaffe, 2008). Reflections in the frequency range from 1.5 to 2.5 MHz were recorded from untethered 1–4 mm calanoid copepods and 8–12 mm mysids over an angular range of 0–47∘. That study demonstrated the utility of a multiple-angle acoustic array for zooplankton identification.

To distinguish the SSLs against the background patterns of vertical flow of settling particles and to study the orientation of zooplankton species, we propose a simple method for the processing of ultrasound sensing data at three angles. This acoustic three-beam geometry provides a partial pragmatic solution for the quest towards the multiple-angle scatter measurements suggested by models (Stanton and Chu, 2000; Roberts and Jaffe, 2007) and laboratory experiments (Roberts and Jaffe, 2008). Since the late 1990s, researchers' efforts have been focused on creating multichannel instruments to measure acoustic backscatter (volume scattering strength) at several frequencies, which contain information about the size composition of the scatterers, since different frequencies bounce off objects of different sizes (Wiebe et al., 2002; Smeti et al., 2015). Multichannel instruments in conjunction with video cameras are fairly expensive systems that are used for the identification of mesozooplankton in its natural habitat. Plausibly, a multichannel three-angle system featuring several relatively cheap short-range three-beam acoustic units each operating at an individual frequency when installed on a vertically profiling carrier would be a very effective tool for visualizing zooplankton aggregations.

Comparison of the Nortek Aquadopp acoustic backscatter observations with the data obtained by stratified zooplankton sampling showed good agreement of the features of the diel vertical distribution of zooplankton. This was made possible by sampling narrow depth strata (10–15 m layers) targeting deep-water aggregations visualized on the echograms. The two-layered structure of the aggregations seen on echograms and R graphs in the daytime (Figs. 3, 7, and 9) reflected the species composition of zooplankton in these layers (Figs. 4, 8, and 10). The deepest layer bounded by isopycnals 15.9 and 15.7 was visible in the suboxic zone all day and night and was formed by diapausing CV Calanus euxinus. To some extent, this monospecific layer was contaminated by crustacean exuviae and carcasses, spent females, and zooplankters' remains sinking from the upper layers and apparently retained on the density gradient. The existence of a nonmigrating diapausing stock located in the suboxic layer from mid-spring to mid-autumn is confirmed by observations from submersible Argus (Vinogradov et al., 1985; Flint 1989), by high vertical resolution sampling with 150 L water bottles (Vinogradov et al., 1992), and by zooplankton net sampling (Arashkevich et al., 1998; Besiktepe, 2001; Svetlichny et al., 2009). However, for some unknown reasons, this nonmigrating layer was not detected by shipborne echo sounders at frequencies of 38–200 kHz (Erkan and Gücü, 1998; Mutlu, 2003, 2007; Stefanova and Marinova, 2015; Sakınan and Gücü, 2016), unlike our data obtained by Aquadopp at a frequency of 2 MHz.

The inclusion of diapause (or dormant stage) in the life cycle of all Calanidae species living in high-latitude and temperate environments is well known (e.g., see review in Baumgartner and Tarrant, 2017). Having accumulated a large amount of lipids, diapausing copepods descend into deeper ocean layers, where they can exist for several months at the expense of energy reserves. Decreased metabolic rate and developmental delay are characteristic features of diapausing copepods. In the Black Sea, a decrease in the metabolic rate in diapausing C. euxinus is caused not only by internal physiological reasons, but also by hypoxia in their dormant layer. The oxygen consumption rate in diapausing CV C. euxinus in hypoxia decreases by almost an order of magnitude, and the rate of ammonia excretion decreases 6 times compared with those in their active counterparts in normoxia (Svetlichny et al., 1998).

During the daytime, the upper SSL mostly located above σΘ = 15.7 consisted of four species, copepods C. euxinus and Pseudocalanus elongatus, chaetognaths Parasagitta setosa, and ctenophores Pleurobrachia pileus; the latter had a negligible contribution to dry biomass. This assembly had a wide range of body lengths from approximately 1 mm in P. elongatus to 22 mm in P. setosa. The different species had different swimming speeds. It was also possible that these species had different physiological tolerances to oxygen deficiency. This confirms earlier observations from the manned submersible, which showed that the daytime aggregation of migrating zooplankton had a layered structure: the lower layer was formed by chaetognaths, whereas the older stages of C. euxinus were located above, and ctenophores inhabited the upper part of the aggregation (Vinogradov et al., 1985; Flint, 1989). Furthermore, the different developmental stages of copepods C. euxinus and P. elongatus occupied different depths, deepening as their size increased (Morozov et al., 2019).

In the evening, approximately 2 h before sunset, zooplankters begin to ascend to the upper layers, where they spend all the dark hours concentrating in the thermocline layer and below it. In this layer, while feeding, they move in different directions and are oriented randomly (Kiørboe et al., 2009), so they cannot be discernible in the R graphs. According to our data, cold-water herbivorous C. euxinus and P. elongatus only occasionally ascend into the warm UML, mainly inhabiting colder layers rich in phytoplankton (see also the supplement to the paper by Morozov et al., 2019). Predator chaetognaths P. setosa move upward following copepods, their main prey (Drits and Utkina, 1988). The time of zooplankton migration clearly visible on the echograms is confirmed by the results of net sampling and is consistent with other published data (see for reference the Supplement of Morozov et al., 2019).

The vertical migration of zooplankton can increase the vertical flow of carbon and thus contribute to the functioning of the biological pump in the ocean (Tutasi and Escribano, 2020). The mesozooplankton that feed at the surface but metabolize and excrete at depth contribute to the transport of organic matter; more quantitatively, this contribution is estimated to be between approximately 10 %–50 % of the local sinking flux of organic particles (Bianchi et al., 2013, and references therein).

In the lower part of the oxic zone, the vertical displacements of SSLs coincide with the oscillations of isopycnal surfaces (Figs. 5 and 10). The dissolved oxygen concentration profile tightly hinges on the density stratification in the Black Sea since both are basically due to vertical mixing processes (e.g., Ostrovskii et al., 2018). Hence, displacements of the SSLs with regard to the oxy-isolines are much smaller than those vs. the depths. The vertical oscillations with a period of approximately 17 h near the mooring site are due to near-inertial waves (Ostrovskii et al., 2018). Irregular changes in isopycnal depths occur due to hydrodynamic events, such as individual internal waves, oceanic fronts, and jets. Occasionally, the isopycnal depth may change by 30–40 m within a day (Ostrovskii and Zastsepin, 2016).

It is unlikely that copepods maintain their positions on certain isopycnal surfaces by swimming, as displacements of such large amplitudes by tens of meters require an additional depletion of energy reserves. A more beneficial strategy would be to adjust their buoyancy to neutral. Having neutral buoyancy in the hypoxic zone, the copepods would not need to spend much additional energy floating up and down following crests and troughs of internal waves while avoiding entrainment into the suboxic layer. Indeed, direct observations from manned submersibles revealed a quiescent behavior of diapausing copepods and their slow response to light and noise produced by underwater vehicles, both in the Santa Barbara Basin (Alldredge et al., 1984) and in the Black Sea (Mikhail Flint, personal communication, 2020). Neutral buoyancy has been hypothesized to be regulated by changes in lipid composition (Visser and Jónasdóttir, 1999); however, Campbell and Dower (2003) argued that this buoyancy regulation mechanism is inherently unstable because wax esters are more compressible than seawater. An alternative mechanism for buoyancy regulation in diapausing copepods that involves the replacement of heavy ions with lighter ammonium ions in hemolymph has been proposed by Sartoris et al. (2010) by analogy with other invertebrates. Later, Schründer et al. (2013) found high concentrations of ammonium ions in the hemolymph of a diapausing species, Calanoides acutus, and suggested that these copepods could achieve neutral buoyancy through their biochemical body composition without swimming movements. This mechanism obviously would better explain the observed phenomenon of diapausing copepod movement synchronized with the displacements of the isopycnal surfaces in the Black Sea.

For most of the year, from April to October, the sound-scattering profiles in the deeper part of the oxic zone were bimodal during the day (Fig. 14), reflecting the vertical distributions of two different zooplankton cohorts, migrating and diapausing. The oxygen concentration at which the maximum backscattering signal from migrating zooplankton was observed decreased throughout the year, from ca. 70–90 µm in January–February to 30–40 µm in April–August and to 15–20 µm in September–December. Two explanations can be considered for the seasonal shift in the preferred oxygen concentration in these zooplankters. On the one hand, this shift can be attributed to the deepening of the suboxic layer in January–February, followed by gradual shallowing from April onwards (Fig. 15). If we assume that the depth of daytime zooplankton aggregation depends to a large extent on the species-specific migration amplitude, then with a deepened suboxic layer ([O2] < 10 µm), the migrating zooplankton will reach shallower depths, and its aggregation will be at a higher oxygen concentration. Conversely, with a shoaling suboxic layer, zooplankton localize at lower oxygen levels.

This assumption is consistent with the data of Vinogradov et al. (1992), who found the daytime aggregation maximum of migrating CV and female C. euxinus at an oxygen concentration of 18 µm when the suboxic layer was at a depth of 110 m and at [O2] = 36 µm when the suboxic layer was at a depth below 170 m. This suggests that the trade-off between the additional metabolic cost for extended swimming and metabolism reduction caused by low oxygen is in favor of a decrease in the diel migration amplitude of C. euxinus. This differs from observations (Wishner et al., 2020) on the migration of Lucicutia hulsemannae in the eastern tropical North Pacific, where this species changes its daytime location in response to changes in the depth of the oxygen minimum zones (OMZs). For example, at the lower oxycline, the depth of maximum abundance for L. hulsemannae shifted from ∼ 600 to ∼ 800 m in an expanded OMZ compared to a thinner OMZ but remained at similar low oxygen levels in both situations. L. hulsemannae is an example of a “hypoxiphilic” species (Wishner et al., 2020). However, unlike Calanus spp., L. hulsemannae is a strong swimmer, capable of diel vertical migration with amplitudes as large as approximately 1000 m.

Another explanation for the seasonal shift in the depth of the daytime aggregation in the Black Sea is the change in taxonomic and age composition of the migrating cohort. A decreasing/increasing share of strong/weak swimmers with different tolerances to oxygen deficiency may lead to a shift in the depth of daytime aggregation. The oxygen concentration was in the range of 15–60 µm in the layer of daytime aggregation of migrating species in the Black Sea. Similarly, the vertical distribution of migrating CV and adult Calanus chilensis off northern Peru was characterized by high abundance in hypoxic waters at oxygen concentrations between 5 and 50 µm (Hirche et al., 2014).

Based on our data, a diapausing cohort of C. euxinus appeared in April and terminated in November (Fig. 14). According to Vinogradov et al. (1985) and Svetlichny et al. (2009), the diapausing stage of C. euxinus was not found in the Black Sea in March. Hence, it is assumed that the diapausing stock is formed in April, when the offspring of the first generation of C. euxinus develop into the CV and accumulate sufficient lipid reserves. The energy reserve and a decrease in metabolic rate allow diapausing CV to exist without food for 7 months in the suboxic zone (Vinogradov et al., 1992), which agrees with our observation of the diapause duration. The suboxic zone provides a refuge from large visual predators, which generally need higher oxygen concentrations (e.g., Bianchi et al., 2013).

The diapausing layer is bound by 3 and 10 µm and peaks at 5–7 µm oxygen. Earlier, the oxygen survival threshold for diapausing C. euxinus was determined experimentally at [O2] = 1.8 µm (Vinogradov et al., 1992). Physiological tolerance for hypoxia in diapausing C. euxinus resembles that reported for diapausing stage CV Calanus pacificus, which formed narrow dense aggregations at an oxygen concentration of 6.25 µm in the Santa Barbara Basin (Alldredge et al., 1984). The diapausing C. pacificus at a depth of 450 m was characterized by quiescent behavior, low laminarinase activity (as a proxy of feeding), and high lipid storage.

Similar tolerance for hypoxia was reported for diapausing Eucalanus inermis. Based on lactate dehydrogenase activity in the deep-dwelling CV and female E. inermis, their oxygen tolerance threshold was defined at the level of 4.47 µm in the Pacific Ocean (Flint et al., 1991). Wishner et al. (2020) found a monospecific aggregation of E. inermis diapausing at extremely low oxygen, 1.0–5.7 µm, in the eastern tropical North Pacific. In the Black Sea, the diapause depth is directly associated with certain density surfaces and consequently with a specific concentration of oxygen. From April to November, the isopycnal surfaces bounded by the suboxic layer move upward from depths of 130–160 to 110–140 m. On this seasonal trend, the superimposed surfaces exhibit strong variations up to 60 m in amplitude at timescales from 17 h to several days (Fig. 15). Diapause layers varied in depth along with isopycnal oscillations, allowing the copepods to remain in a constant-low-oxygen habitat.