Juvenile Atlantic herring (Clupea harengus) use a time-compensated sun compass for orientation

The Atlantic herring (Clupea harengus Linnaeus 1758), an ecologically and economically important species in the northern hemisphere, exhibits complex seasonal migration behaviour. In the western Baltic Sea, adult herring migrate hundreds of kilometers between the Sound (overwintering grounds), coastal areas of the western Baltic Sea (spawning grounds) and Kattegat, Skagerrak and eastern North Sea (summer feeding grounds) (see Fig. 1; Aro, 1989, Clausen et al., 2015, Miethe et al., 2014). Goal-oriented migration over mid-range distances is a difficult task and requires the right cues for orientation and navigation. Fish can use a range of sensory cues for orientation: sound (e.g. Radford et al., 2011), olfaction (e.g. Gerlach et al., 2007), Earth's magnetic field (e.g. Bottesch et al., 2016, Cresci et al., 2017) or celestial cues, such as the sun (e.g. Loyacano et al., 1977, Mouritsen et al., 2013, Quinn, 1980). While orientation based on olfactory and auditory cues is usually limited to a few kilometers (e.g. leaving a fjord), magnetic and sun compass orientation is mostly used for long-distance migration (e.g. herring migration between spawning and feeding grounds). To be able to effectively use the sun as a compass, animals must observe the path of the sun and link the sun's azimuth at different time points during the day to their circadian clock (reviewed in Mouritsen, 2018).

Western Baltic spring-spawning herring (WBSSH) migrate from their overwintering grounds to their coastal spawning grounds to lay benthic eggs in shallow, low-saline areas such as Greifswald Bay and Kiel Canal mainly between February and May (Fig. 1; Biester, 1989, Clausen et al., 2015, Paulsen et al., 2014, Polte et al., 2017). After spawning, adult herring leave the area. Yolk-sac larvae are found close to the spawning beds. Pre-flexion and flexion stage larvae are then observed in pelagic areas and post-flexion larvae return to shallow littoral areas where they remain until after their metamorphosis to the juvenile stage (Polte et al., 2017). It is assumed that juveniles initially stay near the coast and only start migrating to offshore nursery areas between Germany and Denmark in July/August (Oeberst et al., 2009; Payne et al., 2009; Polte et al., 2017). However, the exact path during the first migration of juvenile herring (age 0) is unknown.

To date, only Atlantic herring larvae have been tested in various orientation experiments between 14 and 28 days post-hatch (Cresci et al., 2020), weeks before they begin their first long-distance migration. Results of this study propose the potential use of a sun compass, but could not exclude the use of other orientational cues. To date, there are no published data on orientation mechanisms in juvenile herring as they embark on their first long-distance migration. In this study, we caught juvenile WBSSH in summer (July and August) 2021 in the Kiel Fjord and investigated whether they use a sun compass to orient towards offshore nursery areas and the Sound, where Baltic herring overwinter (Fig. 1; Payne et al., 2009).

All animal procedures were approved by the Animal Care and Use Committees of the Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit (LAVES, Oldenburg, Germany), Az.: 33.19-42502-04-17/2721 and performed in accordance with the relevant guidelines and regulations. Permission for scientific fishing using fishing rods, fish traps and sink nets was granted by the Schleswig–Holstein State office for agriculture, environment and rural areas to GEOMAR.

The experiments were performed at the University of Oldenburg, Germany, from August to October 2021. Juvenile herring were caught in the Kiel Fjord (54°19′46.0″N, 10°08′58.1″E) in July and August 2021 from the pier in front of the GEOMAR Helmholtz Centre for Ocean Research Kiel west shore building using large hand nets. Fish were kept in a flow-through tank (with Baltic Sea water) for 2 weeks and fed 2–3 times a day with Artemia sp. shrimp or natural plankton (caught in the Kiel Fjord) before being transported to Oldenburg (180 km SE) in plastic cooling boxes filled with continuously aerated Baltic Sea water (according to Kiel Fjord conditions: 15 PSU and 17°C).

After arriving in Oldenburg, fish were separated into two groups (sun compass and time-shift sun compass) and transferred into 200 l glass aquaria with water conditions matching those in Kiel Fjord. The tanks were filled with natural seawater, temperature was controlled by a TECO Aquarium Cooler TK 500 and for continuous filtration, an external filter was used (Eheim Professional). The light cycle was set to 14 h light:10 h dark with lights on between 06:00 h and 20:00 h for normal sun compass experiments and between 00:00 h and 14:00 h for time-shift sun compass experiments. The time-shifted tank was covered with opaque curtains to prevent external light from influencing the light:dark cycle. Once per day, when the light was on for both groups, the juveniles were fed Artemia sp. shrimp, enriched with Aqua Biotica Orange+ (Mrutzek Meeres-Aquaristik). The time-shift sun compass group was clock-shifted for 7 days before starting the experiments.

Subsequent to performing the orientation test, the body length of all fish in each experimental group was measured. In the normal sun compass group, the juveniles measured 5.8±1.7 cm (mean±s.d.) and in the time-shift sun compass group they were 5.2±0.8 cm. Animals tested in September (second cohort) had a standard length of 8.0±0.8 cm.

In total, 79 fish were tested, with 38 juvenile herring in the normal sun compass experiment and 41 juveniles in the time-shift sun compass experiment. A clock-shift is necessary to exclude the use of additional omnipresent cues like the earth's magnetic field. Each individual fish was tested at least twice (2–6 times) in each condition, respectively. This multiple-testing method is also used in migratory birds (Hein et al., 2011; Wiltschko et al., 1993) to gain reasonable accuracy on the intended mean orientation direction. Additionally, multiple testing of the same individuals can significantly reduce the total number of experimental animals needed, which complies with the 3Rs principle. All of the tests were performed between 08:00 h and 13:00 h for time-shifted fish and between 08:00 h and 16:00 h for normal sun compass experimental fish.

During the experiments, juvenile herring were transferred into testing bowls (Ø 30 cm) filled with 5 l of sea water (temperature and salinity similar to housing tank water) and placed on top of a 10 l bucket that was positioned outside in the field with a clear, unhindered view to the sky and no visible landmarks. After 5 min of acclimatization, each herring was tested individually; pictures were taken every 10 s for 20 min using a GoPro camera (GoPro Hero 4 silver) placed on the bottom of the bucket aligned to the north (see Fig. 2A). The position of the fish's head relative to the center of the testing bowls was assessed by uploading the recording in ImageJ (National Institutes of Health, Bethesda, Maryland, USA) and clicking on the fish's head every 30 s (in every 3rd frame). By choosing 30 s intervals, we made sure to have no correlation between the successive data points due to lag of time in between them (autocorrelation). The resulting 40 captured positions per tested fish were then translated into compass directions. In addition to the orientation direction, three different people assessed cloud cover in 5% steps based on the recorded sky during the experiments (cameras were placed under the testing bowls and pointed at the sky). We averaged these judgments of cloud cover.

After each experimental run, the juvenile herring were removed from the experimental setup and placed in the shade to rest for at least 30–45 min before being tested again.

It is not known yet if fish are able to exactly measure the sun's azimuth or if they just estimate the azimuth in accordance to their internal clock – assuming a steady shift throughout the day. Therefore, we used two different methods to correct for the sun position at different times of the day and consequently correct the raw data that have been collected during the time-shift sun compass experiments. One is the generalized correction and the other is the accurate correction.

Using the generalized correction, a 15 deg change in sun azimuth per hour is assumed (Lambrinos et al., 1997), leading to a 90 deg shift in 6 h. These 90 deg were added to every mean vector that had been calculated using the 40 positional data points of each significant experimental run to get a generalized corrected mean vector.

For the accurate correction, the exact position of the sun at the specific location, day and time of each experimental run (at a time point mid-experiment, meaning after 10 min of the 20 min trial) was determined using www.sunearthtools.com (October 2020, based on Michalsky, 1988, see Fig. 2B). Afterwards, the position of the sun 6 h after each experimental run was also determined. The deviance of the two angles is calculated by subtracting the sun position during the actual experiment from the sun position of the expected experimental time. This deviance is then added to the original mean vector to get the accurate corrected mean vector.

All data were analysed using the circular statistics program Oriana (Kovach, 2011) as well as R (version 4.1.3) with ggplot2 (Wickham, 2016). From the 40 observed directions in the respective experimental run of each individual, we calculated the mean vector of orientation. We analyzed whether orientation preference was statistically different from a random distribution using the Rayleigh Uniformity Test [P<0.05 in our case corresponds to a length of mean vector (r)>0.272]. The mean direction of a fish during one trial was averaged with the mean direction of the same fish during the following trials to gain only one direction per individual that was used to assess the groups' mean orientation direction.

To evaluate whether the fish were able to effectively use the sun as an orientation cue, the group result of the normal sun compass experiment was initially compared with that from the uncorrected time-shift experiment using a Mardia–Watson–Wheeler test. After proving that both data sets differed significantly, the data of the normal sun compass experiment were compared with those of the corrected time-shift data (generalized and accurate). We hypothesized that if the fish used a time-compensated sun compass, the sun compass direction and corrected time shift direction should be the same.

To analyze whether orientation depends on a clear view to the sky, the effect of relative ‘cloud cover’ on directionality [length of the mean vector (r) of every individual experimental run] was determined using logit transformation of the r vectors, Bonferroni correction for multiple comparisons and paired t-tests.

Herring expressed a significant sun compass orientation towards east-southeast (ESE) with a mean vector of 108 deg (see Fig. 3A) (for details, see Table S2). A 6 h clock-shift where the animals are tested before noon, when they think it is afternoon, should lead to a change in orientation when using a sun compass. Indeed, when clock-shifted 6 h backwards, fish showed a significant group orientation towards north-northeast (NNE), with a mean vector of 33 deg (see Fig. 3A) (for details, see Table S3). Comparison of these two datasets by performing a Mardia–Watson–Wheeler test showed a statistical difference (W=8.507, P=0.014). The change in mean vector was 75 deg.

To understand whether herring changed their directional swimming according to the time-shift, we used two different methods (generalized and accurate correction). First, we corrected by adding 90 deg (assuming 15 deg change per hour, see Fig. 2B) and calculated a group swimming direction of time-shifted juvenile herring towards ESE, with a mean vector of 123 deg. By performing correction according to the accurate sun position (see Fig. 3A), herring showed a SE direction with a mean vector of 137 deg. Both these mean swimming orientations were statistically similar to the normal sun compass orientation direction when compared using a Mardia–Watson–Wheeler test (generalized: W=0.309, P=0.857, accurate: W=1.9, P=0.387). The change in mean vector after correction amounts to only 15 deg (generalized correction) and 29 deg (accurate correction), respectively.

To determine if cloud cover influences directional swimming, we tested whether clouds had an impact on the orientation ability of the fish.

We categorized cloud coverage from 1 to 6 [(1) 0–20%, (2) >20–40%, (3) >40–60%, (4) >60–80%, (5) >80–100% and (6) 100% cloudy] and calculated the mean r vector±s.e.m. After logit transforming the r vectors and Bonferroni correction for multiple comparisons, the paired t-test showed that orientation was significantly higher at 0–20% cloudiness (r=0.439±0.04) than at >80–100% (r=0.286±0.01) or 100% cloud cover (r=0.246±0.05). No significant difference was found when the cloud cover was between >20 and 100% (see Fig. 3B and Table S1). It became apparent that the orientation ability of the fish decreased when the percentage of cloud coverage increased and the most successful tests were conducted when only 0–20% of the sky was covered with clouds.

Taken together, these facts demonstrate that juvenile Atlantic herring are able to use a time-compensated sun compass for orientation. This finding extends the results of Cresci et al. (2020), who used a transparent drifting in situ chamber (DISC) in which larvae at 14–16 and 25–28 days post-hatch were tested for their orientation ability in coastal areas of a Norwegian fjord. Cresci et al. (2020) found that both age groups display a significant orientation towards the southeast. However, fish larvae tested in a drifting DISC are exposed to many different environmental cues such as sound, olfactory cues or visual landmarks. Therefore, it is hard to determine the influence of one specific orientation cue. Cresci et al. (2020) suggest that larvae use the sun for orientation, since directionality decreased with increasing cloud cover. We performed experiments on juvenile herring during their first summer/autumn when they start to leave their shallow coastal nursery grounds and head towards offshore nursery or overwintering grounds; therefore, they should be in ‘migration mode’. We tested the effect of body length on the individual bearing, but fish of the second cohort (which were older and larger) showed no significantly different bearing than juveniles tested in August. This confirms that the smaller fish from the first cohort were indeed already in ‘migration mode’ when caught. Furthermore, they were tested under controlled conditions, giving them no visual cues for orientation other than the sun. Clock-shifting the fish, which has not been done in the larval experiments, confirmed that herring juveniles use a time-compensated sun compass for orientation. This conclusion is based on the fact that they misinterpreted the sun's actual position and changed their orientation direction accordingly. In addition, orientation was more precise under sunny conditions than when clouds prevented the direct view of the sun.

Even though Table S1 and Fig. 3B show data collected at ‘100% cloud cover’, which should lead to unoriented fish and therefore non-significant experimental runs, we still observed directional swimming. This could indicate that herring have further orientation capabilities, such as magnetic orientation or perhaps the use of a skylight polarization pattern (Hawryshyn, 1992; Waterman, 2006; Pomozi et al., 2001), but more analysis is needed to determine whether herring can indeed detect these cues. Even if celestial cues such as the sun were used as a primary compass sense in our experiment (as proven by the clock-shift), herring might use additional orientation mechanisms. The sun compass could be used collectively with other compass senses or be part of a cue cascade where one cue comes more into play if a higher ranked one is less available. Magnetic compass orientation would be plausible as this cue is omnipresent and has been demonstrated in a variety of fish species (Bottesch et al., 2016; Cresci et al., 2017; Putman et al., 2020). Whether herring can indeed use additional orientation mechanisms still needs to be determined. Even though herring larvae did not respond to rotations of the magnetic field (Cresci et al., 2020), magnetic orientation might develop in later life stages.

Juvenile Atlantic herring perform their first migration without the guidance of experienced adult herring. Similar difficulties are experienced by juvenile solitary migrating songbirds (e.g. the European blackcap Sylvia atricapilla), which use compass orientation during their first migration. When translocated, they do not find their way back to their migration route. However, as experienced adults, they have a map-based orientation and after translocation for several hundred kilometers, can adjust their direction accordingly (Merlin and Liedvogel, 2019). In a different study, we displaced juvenile coral reef fish (Ostorhinchus doederleini) by 180 km in a similar manner to the experiments on juvenile birds. We did not observe any change in orientation direction and conclude that they have no positional information and therefore no map-based orientation (L.S., F. Curdt, A. Bally, N. Janzen, P. Kraemer, B. Leberecht, M. J. Kingsford, H. Mouritsen, M. Winklhofer and G.G., unpublished results). Naive travelers, such as juvenile herring, coral reef fish or night-migratory songbirds seem to rely on simple inherited orientation mechanisms, such as a sun compass.

Juvenile Baltic herring start their journey at the end of summer, leaving the coastal areas to find areas of higher food availability (Kvamme et al., 2003). In our study, juveniles oriented to the east-southeast, which might indicate that they take the eastern route to reach the Sound (Fig. 1, solid blue line from the Sound to Kiel) or remain in shallower nursery areas at the German coast, potentially joining schools of similarly sized individuals from different spawning areas (e.g. Lübeck, Wismar, Greifswald) as suggested by Nøttestad et al. (1999). In future experiments, we plan to investigate additional compass- and map-based orientation mechanisms to further characterize the sensory orientation capabilities of Atlantic herring.

We thank Felix Mittermayer and Fabian Wendt for sample collection, Susanne Wallenstein, Anke Müller and Thilo Appeldorn for video analysis as well as Simon Käfer for creating circular data plots in R. We also thank GEOMAR Kiel for hospitality, help and logistic support.