Hydrography and trawl catches
In November and December, the temperature was 8.5 and 4.5 °C in surface waters, respectively, increasing to nearly 10 °C at around 30 m. From 60 m and below, the values stabilized at around 8 °C (Fig. 2a). The salinity at the surface was 30 in November and 24 in December, the salinity increased to nearly 32 at 30 m (in both months), and in the lower half of the water column, the salinity was stable at 33 (Fig. 2a). The oxygen content in November and December 2009 ranged between 2 and 3 ml O2 L−1 from 15 to 60 m. The values decreased rapidly below 70 m, and deeper than 80 m, the values were close to 0 (Fig. 2b). A water renewal in mid-February brought more oxygenated water into the fjord, and in April, the oxygen values had increased to >4 ml L−1 from 30 m and deeper (Fig. 2b). The fjord became ice covered between 06 January and 08 January, after which no sampling was carried out until after the ice melted (between 01 April and 04 April).
A total of 7,345 sprat were caught during the sampling in December 2009 and April 2010, making sprat the most abundant species in the trawl catches. In December, the highest catches of sprat were obtained in the range of 60–80 m during the day (~720 sprat per 10 min of trawling) and 40–70 m at night (~85 sprat per 10 min of trawling). There were no catches of sprat below 80 m in the beginning of the winter. During daytime in April, most of the sprat were collected from 70 to 95 m (~520 sprat per 10 min) with nearly none distributed deeper than 100 m or shallower than 60 m (<5 sprat per 10 min). The catches contained furthermore a total of 162 herring (Clupea harengus), 58 gobiids (Gobiidae spp.), 12 whitings (Merlangius merlangus), 4 anchovies (Engraulis encrasicolus), 1 pollack (Pollachius pollachius), and some gelatinous plankton. Catches of other invertebrates like krill and shrimps were negligible this winter.
Vertical distribution of sprat
The vertical distribution of the sprat population changed throughout the winter. In the beginning of November, a major part of the population was schooling in upper waters during the day, while schools dispersed and individuals descended to deeper waters at night (i.e., their behavior can be classified as inverse diel vertical migration; IDVM). Most fish were distributed in a layer ranging from around 30–60 m at night (Fig. 3a). Fish joined this layer after their excursions to the surface (Fig. 1a).
The nocturnal layer became more dispersed and extended to shallower depths by the end of the month, and by December, a part of the population carried out normal DVM and schooled in deep waters during the day. A bimodal distribution was displayed at night (Fig. 3b). A similar pattern was observed in January (when the fjord became ice covered), but the two nocturnal modes were separated over a larger depth range as one layer was distributed close to the surface and the other below 50 m (Fig. 3c). The sprat inhabited waters down to ~75 m from November until end of January. In the following months, the nocturnal distribution extended deeper, but the majority of the sprat were still distributed in the upper half of the water column (Fig. 3d). During daytime, the sprat schooled at various depths from 0 to ~90 m from February until the end of the study in April.
Abundance of sprat
The surface integrated values, sa (interpreted as acoustic biomass), were highest in the period before the fjord froze over with a monthly average of >300 in November and ~200 in December (Fig. 4a). The monthly average values declined to <~100 during the period with ice-covered waters (January, February, and March; Fig. 5). The sa increased in April (when the ice had melted) with daily values fluctuating between 100 and 400 (Fig. 4c).
Surfacing and gas release in ice-free waters
Surfacing sprat was detected every date during ice-free conditions (November 11–December 12, 2009 and April 5–12, 2010; Fig. 4a, c). Recorded surfacing events ranged from 1 per night to nearly 50 per night, with an average daily estimated “surfacing rate” of ~3.5 times fish−1. There was a significant relation between the total area backscattering coefficient (sa) and number of surfacing fish per day (Linear regression, r
2 = 0.38, F
1,31 = 20.9, P < 0.0000). There were no records of surfacing sprat during daytime.
Sprat surfaced after sunset, particularly early at night, and released gas in the following hours during nighttime (Fig. 4b, d). Release of gas was prominent in the morning hours, especially in November when most of the gas bubbles were detected within 2 hours prior to sunrise (Fig. 4b). The estimated “rate of gas release” was 72 times fish−1 day−1 in the period before ice cover (November and December) and 35 times fish−1 day−1 in the period after the ice had melted, i.e., the rate of gas release was an order of magnitude higher than the rate of surfacing (Fig. 6). The relation between number of surfacing events and amount of gas release per day was significant for all months combined, but the percentage of explained variation was low (Linear regression, r
2 = 0.31, F
1,30 = 14.7, P = 0.0006). The low relationship may to some extent be explained by fish releasing air also below the depths of our records with the 200 kHz echosounder as data from the 120 kHz echosounder showed that the fish released air between 0 and 40 m depths. However, the majority of gas release appeared to take place in the 30 m interval covered by the 200 kHz echosounder.
Surfacing and gas release in ice-covered waters
The fjord became ice covered between 06 January and 08 January. There were few surfacing fish during the first 2 weeks of January (<10). This was followed by a conspicuous increase in recorded surface events during the third week of January ranging from 39 to 101 per day (Fig. 5a) giving an average daily surfacing rate of ~24 times fish−1 for that period. The high number of surfacing fish was only apparent for around 1 week followed by a period with a few detections until the second half of February (Fig. 5b). In March, surface activity was detected every day (Fig. 5c). Combining all 3 months with ice cover, the daily estimated rate of surfacing was ~12.5 times fish−1. Several of the fish just under the ice repeatedly ascended to the surface before descending (see echogram example Fig. 7), which was a behavior that was not observed in ice-free waters. There was a significant relation between the total area backscattering coefficient (sa) and number of surfacing fish during the months with ice cover (January, February, and March; Linear regression, r
2 = 0.08, F
1,71 = 7.3, P = 0.0085), but the coefficient of determination was very low.
Gas release was strongly reduced when the fjord was ice covered, and the pattern of nocturnal gas release that was observed in ice-free waters was not apparent. Gas bubbles were sporadically detected, but only at a rate of 2 times fish−1 day−1 (Fig. 6). More than 60 % of the gas releases recorded during this period (194 in total) were detected during the last 2 days of March (a couple of days before the ice melted; not shown). The reduced amount of gas bubbles was confirmed by the 120 kHz echosounder which covered nearly the whole range of the sprat population (0–80 m).
Cloud cover and moon phase
The weather was relatively cloudy during the whole winter with an average cloud index mostly fluctuating between 5 and 8 (8 is maximum) (not shown). No consistent pattern between frequency of surfacing events and cloud index or moon phase was evident.
Swimming speed and TS measurements
During ice-free conditions, the average downward swimming speed after surfacing was approximately twice as high as the average upward speed, ~28 and ~13 cm s−1, respectively (Fig. 8). Vertical swimming associated with surface events was considerably slower when the fjord was ice covered, with the descent and ascent rates being ~11 and ~ 8 cm s−1, respectively (Fig. 8). The swimming speeds differed significantly by both swimming direction and environmental conditions (before ice, during ice, and after ice cover), and there was a significant interaction between the two factors (two-way ANOVA, F
1,2633 = 784.6, F
2,2633 = 760.1, F
2,2633 = 342.4, P < 0.0000, for all tests). A post hoc Tukey’s test (P value of <0.05) revealed that among fifteen possible comparisons, only three were not significantly affected by interaction: downward swimming before ice versus downward swimming after ice (P = 0.18), upward swimming before ice versus upward swimming after ice (P = 0.99), and downward swimming during ice versus upward swimming after ice (P = 0.86). The two factors are difficult to interpret because of the significant interaction; however, the results show that the swimming speed depended on environmental conditions and that the difference between upward and downward swimming speed was greater in ice-free waters (Fig. 8).
The TS values were weaker in ice-covered waters than in ice-free waters. The average TS values of ascending and descending sprat during surfacing in ice-free waters were −54.2 and −53.5 dB, respectively, and −57.8 and −54.3 dB in ice-covered waters, respectively. The TS values differed significantly by both swimming direction and ice conditions (ice-free waters versus ice-covered waters) (two-way ANOVA, F
1,2635 = 100.5, P < 0.0000, F
1,2635 = 85.3, P < 0.0000), and a significant interaction between the factors was also present here (F
1,2635 = 21.1, P < 0.0000). TS of upward swimming sprat in ice-free waters versus TS of downward swimming sprat in ice-covered waters was the one comparison (among six) that was not affected by interaction (P = 0.98, post hoc Tukey’s test, P < 0.05).