Study area
The North Sea is a continental sea connected to the Atlantic Ocean through the English Channel in the southwest and between northern Shetland along the 61° latitude parallel to Norway in the north (Fig. 1). It is bordered by Norway, Denmark, Germany, the Netherlands, Belgium, France and the UK, and has a surface of 570,000 km2. The North Sea has an average depth of 95 m, yet maximum depths of ca. 700 m are found in the Norwegian Trench. The maximum tidal amplitude of the North Sea can reach up to 8 m, average winter sea surface temperatures are ca. 6 °C and average summer temperatures reach ca. 17°C33. The English Channel encompasses the marine strait between the UK and France. It covers 75,000 km2, has an average depth of 63 m, a maximum depth of 174 m and can reach a maximum tidal amplitude up to 12 m. The average winter and summer sea surface temperatures in the English Channel are ca. 5 and 20 °C, respectively54.
Tagging
In total, 320 silver eels were tagged with pop-off archival tags (Table 1; Supplementary Table S2). In Belgium, 238 eels were caught and tagged at a drainage system upstream of the Yser Estuary (hereafter referred to as the Belgian eels) in 2018–2020 via nets that were attached to gravitational discharge sluice gates (coordinates: 51.127 N, 2.761 E) in October, November and December (n2018 = 102, n2019 = 60 and n2020 = 76). In Germany, 82 eels were tagged in 2011 and 2012. In early December 2011, seven eels were caught at Lake Plön (coordinates: 54.137 N, 10.334 E) with fyke nets. During September, October and November 2012, eels were caught in the Rivers Eider (n = 30; coordinates: 54.190 N, 9.093 E) and Havel (n = 45; coordinates: 52.419 N, 12.571 E) with fyke and stow nets, respectively.
Upon capture, the eels were anaesthetized with 0.3 ml/L clove oil (Belgium), 0.4 ml/L ethylene glycol monophenyl ether (Germany 2011) or 120 mg/L MS-222 (Germany 2012), and various morphometric characteristics were measured to identify the life stage55: total length (to the nearest mm), weight (to the nearest g), horizontal and vertical eye diameter (to the nearest 0.01 mm in Belgium and to the nearest 0.1 mm in Germany) and pectoral fin length (to the nearest 0.01 mm and 0.1 mm in Belgium and Germany, respectively). Given that their total body length was > 450 mm, all eels were considered female55. According to the morphometrics, five Belgian eels could be considered in the premigratory stage (FIII); however, based on visual inspection, they were considered silver eels (i.e. silver-coloured abdomen, dark grey on the dorsal side, jaw hinge not proceeding beyond the eye, enlarged eyes and dark coloured pectoral fins). The other 315 eels identified as silver eels based on both morphometry and visual inspection (201 FIV stage and 114 FV stage).
Eels weighing ≥ 550 g were externally fitted with a G5 PDST (CEFAS Technology Ltd, UK), which log temperature and pressure (providing information on depth). They were attached applying the three-point Westerberg attachment method56. Two tag types were used: one with a separate tag and pop-off mechanism (Germany) and one where both mechanisms were integrated (Belgium). The flotation collar of the PDSTs was painted bright red, contained contact information and a cash reward to stimulate retrieval by the general public (e.g. beach combers and fishermen). The seven eels caught in 2011 in Germany (minimum 1220 g) were fitted with PSATs (X-Tag, Microwave Telemetry Inc., USA), also using the Westerberg-method56. Like the PDSTs, the PSATs record temperature and pressure. After release, they drift to the surface and transmit the data to the user via the ARGOS satellite system (www.argos-system.org). For the specifications of the different tags, we refer to Supplementary Table S3.
Upon recovery from the anaesthetic, eels tagged with PDSTs were released close to their capture locations in the rivers Eider (coordinates 2011: 54.381 N, 9.009 E; coordinates 2012: 54.379 N, 9.013 E), Elbe (coordinates 1: 53.793 N, 9.402 E; coordinates 2: 53.569 N, 9.700 E; coordinates 3: 53.396 N, 10.171 E) and Yser (coordinates: 51.135 N, 2.757 E) (Table 1). The seven eels captured for PSAT tagging in 2011 were held for several weeks in the Thünen Institute of Fisheries Ecology, then tagged and released the same day; others were tagged in the field.
Preprocessing
Once downloaded, the temperature and pressure data obtained from the PDSTs was subsampled to 1-min (Belgian eels) or 2-min (German eels) intervals to reduce the datasets and improve geolocation calculation time; this discrepancy is due to the minimum logging rate of the tags (Supplementary Table S3). Linear regression was applied to correct for pressure sensor drift over time. Indeed, pressure values increased over time even if the tag was kept at atmospheric pressure level. The regression was applied between 15 min before release and the moment the tag popped off and reached the surface, since the tag was then considered at sea level and hence to be under zero pressure.
The PSAT data were retrieved through the ARGOS satellite system as a subset with 15-min intervals and converted to values of pressure and temperature. Contemporaneous values of temperature and depth were not always transmitted due to the transmission method. As a consequence of the tag release programming, the transmission of the first position for one of the tags was only received five days after the tag reached the sea surface.
Geolocation
The daily movements of each electronically tagged European eel were reconstructed using an adapted version of the tidal geolocation model of Pedersen et al.57. The geolocation model uses a novel Fokker–Planck based method to combine the tidal location method of Metcalfe and Arnold58 with a hidden Markov model (HMM), such that an individual’s daily location d is modelled conditionally on its previous location (d − 1), its inferred behavioural state ds, where behaviour is defined by a single diffusivity parameter (i.e. the maximum amount of movement permitted in a given day), and the observations made between d and d − 1. In this case, observations consisted of the recorded depth (m; D1, …, Dn) and temperature (°C; T1, …, Tn), where n is the number of measurements made per day (the HMM down-samples to 10-min intervals, hence 144 measurements per day), and any hydrostatic (tidal) data which are derived from the sinusoidal pressure cycle recorded in the depth data when a fish is at rest on the seafloor. In addition to bathymetry and tidal amplitude with phase, the model was developed to include sea surface temperature (SST), which can provide additional validation when fish are swimming at or near the surface (i.e. depth ≤ 20 m)59,60, and temperature at depth, which can provide additional validation when fish remain at depths well below the sea surface61,62.
The model was run in three different configurations for each recovered dataset: (i) using the tidal location model only (as for Pedersen et al.57), hereafter termed TLM geolocation; (ii) using the TLM plus sea surface temperature (as for Wright et al.60), hereafter termed SST geolocation, and (iii) using temperature at the surface and sub-surface, hereafter termed 3D geolocation (Supplemental Fig. S3). The final trajectory output for the PDST Belgian eels and PSAT German eels was obtained via 3D geolocation, while SST geolocation was used for the PDST German eels. The reason for this discrepancy is that the German PDST eels stayed closer to the coast and in shallower water. Consequently, the 3D geolocation results were more prone to error due to coastal influences on water temperature. As a result, we used the SST geolocation method for these datasets to obtain more reliable results.
Data for the model were derived from publicly available resources. Gridded global bathymetry data were obtained from the general bathymetric chart of the oceans (Gebco; British Oceanographic Data Centre, Liverpool, United Kingdom, 2009). Tidal constituents were obtained from the Oregon State University Tidal Prediction model, as described in Egbert and Erofeeva63. Sea surface temperature data were sourced from OSTIA64, while temperature at depth data were sourced from the operational Mercator global ocean analysis and forecast system65. These datasets were downloaded from the Copernicus Marine Environmental Model Service (CMEMS: documented here http://resources.marine.copernicus.eu/documents/PUM/CMEMS-GLO-PUM-001-024.pdf). Data were sourced so as to fit the spatial scale of the model (30°N to 80°N and from 110°W to 60°E) and coarsened to reduce model run-time by modifying the spatial grid to a 1/10th of a degree resolution. The output of the model is a nonparametric probability distribution of the geographical position from which a most probable location, for each day at liberty, and a most probable movement path can be estimated.
Prior to running the model, a number of constraints and input parameters were defined to ensure that the model ran effectively. The recapture information was either set as (a) the latitude and longitude where the tag was recaptured, with a high confidence (< 5 km error) or (b) as an estimate of the location closest to that which matched the maximum depth and sea surface temperature on the day before the tag came to the surface, or the day before it was predated (low confidence: 50 km error). Model diffusivity (a proxy of the swimming speed of the fish) was set to low default values that matched the expected travelling speed of eels (ca. 20 km per day, maximum 50 km per day16). Two values were used; a low diffusivity value of 30 km d−2, corresponding to days on which vertical movement was low and infrequent, and a larger value of 100 km d−2, corresponding to days when vertical movement was high and frequent. These values broadly correspond to localized (resident) and migratory behaviours, respectively57. Finally, daily estimates of longitude were used to modify the weighting given to the likelihood areas generated from the SST-geolocation and the 3D geolocation. For eels that did not reach oceanic depths (i.e. depths > 200 m) and hence did not exhibit diel vertical migrations, the input estimates of longitude were based on a simple linear interpolation from release to estimated pop-up. However, for eels that did reach oceanic depths, the time of local noon was estimated (based on the timing of significant diel vertical migrations, as for Righton et al.16), and used to estimate longitude. Geolocation was conducted with MATLAB software66.
Migration routes
Only datasets containing ≥ 100 km of net tracking distance were included for further analysis, leading to 54 datasets from the 96 retrieved tags and 320 tagged eels. The net tracking distance was identified as the distance along the reconstructed trajectory between the release of the tagged eel and the pop-off event. When an eel was ingested by a predator, leading to the tag tracking the predator rather than the eel, the data were excluded from the day the eel was predated. The 100 km cut-off point was arbitrarily chosen to select migration paths of sufficient length for further analysis (e.g. migration direction); tracks had a minimum deployment duration of 4 days.
Migration speed
To exclude a size-effect, we first applied an independent two-sample t-test to confirm eel sizes (i.e. weight) did not differ between Belgian and German eels. The assumptions of normality (Shapiro–Wilk test), homogeneity of variances (F-test) and independence were met (weight measurements are individual-specific and therefore independent).
Next, an independent two-sample t-test was conducted to test if the total migration speeds (i.e. the ground speed along the reconstructed trajectory between the release of the tagged eel and the pop-off or predation event) differed between Belgian and German eels. The assumptions were tested and met as described above.
Finally, we tested if the daily migration speed (i.e. the ground speed along the reconstructed trajectory per day) differed according to the eel’s position (i.e. modelled latitude and longitude) via a linear mixed effects model. The tag IDs were implemented as a random effect to account for autocorrelation. Since the two-sample t-test showed a significant difference between Belgian and German total migration speeds, we performed a separate analysis on eels from both countries. Assumptions of normality, homogeneity of variances and independence were tested and met.
The migration speed analyses were conducted in R (version 3.6.3)67. The packages ‘lme4’ and ‘nlme’ were used to conduct the linear mixed effects model.
Ethical statement
Eels were tagged using approved protocols by trained and individually licensed scientists working under national project authority in accordance with institutional and national guides for the care and use of laboratory animals. These guidelines are consistent with Institutional Review Board/Institutional Animal Care and Use Committee guidelines. Tagging in Belgium was carried out in accordance with the Belgian national and regional regulations for animal welfare and treatment (Permit ID: EC INBO-011). Tagging in Germany followed German legislation concerning care and use of laboratory animals, and ethical permission for the experiments was given by the Ministry of Energy, Agriculture, the Environment, and Rural Areas of the federal state Schleswig–Holstein (reference numbers V312-72241.123-34 (90-8/11) and V311-7224.123.3 (93-6/12) for tagging in 2011 and 2012 respectively).
Source: Ecology - nature.com
