Seals are capable of navigation and orientation during long distance movements, even in absence of apparent landmarks, in open seas, and at night (e.g., Lowry et al. 1998, McConnell et al. 1999, Gjertz et al. 2000, Lesage et al. 2004). Several ideas have been put forwards about marine animals' ability to orientate and navigate at sea (Mills Flemming et al. 2006, Lohmann et al. 2008, Chapman et al. 2011). However, little work has been carried out on seals (but see Matsumura et al. 2011). A number of experiments have been conducted on captive seals in order to test their sensory systems and orientation capacities (e.g., Kowalewsky et al. 2006, Mauck et al. 2008), but such experiments are difficult to conduct on free-ranging seals.

Modeling the animals' movements at sea in relation to environmental variables may elucidate the cues they use to orient and navigate. However, such free-ranging animal movements are always subject to the influence of local currents (Lohmann et al. 2008). Thus the incorporation of current data is necessary to reveal underlying navigational capabilities and strategies (Willis 2011). In this study, we model the observed sea surface tracks of two gray seals (Halichoerus grypus) that had crossed the English Channel in September 2011 (Fig. 1, 2). The seals (referred to as B23 and B24), were tracked by Fastloc GPS/GSM telemetry techniques. Their surface positions were drawn from the series of Fastloc GPS locations transmitted by GPS phone tags developed by the Sea Mammal Research Unit.1 These were glued to the animal's fur on the neck behind the head with quick-setting epoxy. The tags were configured to attempt Fastloc GPS locations every 10 min provided the seal was at the sea surface. Both seals were captured and tagged in the Molène archipelago, western Brittany, France.


Figure 1. Track plot of seal B24 from the Isles of Scilly, United Kingdom, to the Isle of Molène, France (trip duration 43.6 h). The real GPS track of seal B24 is shown by the gray line. Gray dots are locations obtained from the first model assuming a constant direction (α  =  165º) and swimming speed (Vseal = 2.1 m/s); white circles are identical to the gray dots for the first 18 h and then obtained from the adjusted first model (changing parameters to α  =  139º and Vseal = 1.96 m/s) from 19 h onwards. Black dots were obtained from the second model correcting swimming direction at each time step in order to head exactly to the destination point. Modeled plots present one location per hour.

Download figure to PowerPoint


Figure 2. Track plots of seal B23 from Porthleven (United Kingdom) to Les Sept Iles (France). A–E: Consecutive adjustments of the modeled seal track in heading and velocity (see Table 1 for details).

Download figure to PowerPoint

During their whole track duration (172 and 204 d, respectively), both seals crossed the English Channel, back and forth, moving directly from one colony to another. We consider here two such transits of the English Channel from the United Kingdom to known seal haul-out sites. These movements occurred outside the breeding or molting season of gray seals in the French colonies, but there could be breeding in the Isles of Scilly at that time of the year (September). Seal B24 departed from the Isles of Scilly (UK) and arrived at the Isle of Molène (France) after 43.5 h (Fig. 1). B23 crossed the English Channel (Fig. 2) in 48 h between Porthleven (UK) to the Nature Reserve of Les Sept Iles (France). At total, 158 Fastloc GPS locations were obtained for seal B23 over its crossing the English Channel (mean = 3.3 locations per hour) and 148 for seal B24 (mean = 3.4 locations per hour). Hourly ground track locations were then determined using linear interpolation of the raw track data.

In the English Channel, tidal currents dominate current patterns due to wind, wave, and thermohaline effects (Sentchev et al. 2009). For this reason, we estimated the currents along the seals' pathways using a tidal model. We used a 2-D model that estimates currents averaged over the whole water column (TELEMAC software, Hervouet 2007), which has been shown to be very effective for modeling tidal propagation in coastal waters (Nicolle et al. 2009, Davies et al. 2011). The model was initially developed for numerical simulations of tides and storms surges (Chevaillier 2011) and it was validated through extensive comparisons with the sea level and sea current measurements in the Bay of Biscay and in the English Channel.

In this study we define: “Ground Track” (GT) as a series of movement vectors built from the GPS time-stamped location fixes; “bearing” as the direction of a Ground Track vector; and “heading” as the direction a seal is pointing. A seal's ground track, D, can be represented as a sum of drifting and swimming vectors: D  =  Dd + Ds. We calculated the drift, Dd, as a Lagrangian transport displacement experienced by a passive particle in tidal flows (TELEMAC, Hervouet 2007). The measured surface velocities of the tracked seals were not used for the numerical simulations. These are the result of both the swimming speed of the seal and the current's speed, and we aimed at modeling the animals' movements from a data set completely distinct from the “real” seal data before comparison of the model vs. the track of the seals.

We compared two navigation rules. Rule 1 is that the seal keeps a constant heading along the whole transit (or at least a significant part). Rule 2 is that the seal continuously adjusts its ground track heading towards the destination.

To test Rule 1, an iterative process was performed where we chose values of seal swimming speed and heading (with a 0.1 m/s and 1º resolution) that gave a modeled track closest to the real seal trajectory. Rule 1 Ground Track (R1GT) locations were calculated by adding the correspondent time step current vector to this constant heading. If large difference developed between the R1GT and the GT a new heading was set. For Rule 2, the direction (α) to the destination was recalculated at every time step (every 10 s) from the seal position. Rule 2 ground track locations (R2GT) were calculated by adding the correspondent time step current vector to the calculated (variable) heading. Note that at this geographical scale we did not distinguish between the rhumb line (constant ground track bearing) and the great circle (variable ground track bearing providing the shortest path between two points). All references to time relate to UTC. All bearings and headings are with reference to true north.

Seal B24

For R1GT a heading of 165º at a constant speed (Vseal) of 2.1 m/s provided closest agreement with the GT during the first 19 h of swimming (Fig. 1). At 1500 on 17 September, the R1GT and GT began to diverge. Therefore we reset B24's heading to 139º and Vseal to 1.86 m/s until its arrival at the Isle of Molène. With this single reset, R1GT continued to closely match the GT. To match the last four seal positions (within 10 km of the Isle of Molène) we assumed that B24 had followed the R2GT there. Using R2GT over the entire transit predicted a route that was very different from the GT (Fig. 1).

Seal B23

The best-fitting ground track for Rule 1 resulted from resetting heading and Vseal three times (Table 1), thus dividing the transit into four legs. Each time the model drifted away from the GT, swimming speed and/or heading were modified to realign with the GT. The resulting R1GT track is shown in Figure 2. The last 7 h of the transit when B23 approached Les Sept Iles, the best match between the modeled and the GT was achieved by using the Rule 2 (α  =  variable, Vseal = 1.76 m/s; Fig. 2E). Therefore, most of the seal B23's trajectory modeled here was obtained by using the R1GT algorithm, considering constant speed and heading. However, in order to fit to the GT, the R1GT had to be adjusted several times all along the journey, and the final approach of the destination was better modeled by using the R2GT algorithm.

Table 1. Successive adjustments made to the R1GT parameters used to model the 48 h transit of seal B23
Time after departure (h)Heading (degrees from true north)Speed (m/s)

These best models allowed the prediction of the seals' movements with an accuracy of 2.4 km (SD 1.3, maximum 6.0) over the 214 km long trip for B24, and an accuracy of 4.4 km (SD 2.1, maximum 8.8) over the 223 km long trip for B23 (Fig. 3). Variations in the distance between the model and the GT did not decrease suddenly with the adjustment of model parameters, for both seals.


Figure 3. Distance between every hourly modeled location and the corresponding seal's GPS tracking location at the same date and time. Arrows indicate moments when model parameters were modified to better fit the real movements of the seals.

Download figure to PowerPoint

The close match of modeled and actual GT suggests that gray seals have the ability to maintain a constant heading at sea. Indeed the best-fitting ground tracks were obtained when assuming the seals kept a constant heading (R1GT) for a significant leg of the trip. The two navigational rules R1GT and R2GT discussed above are not mutually exclusive: changing a heading at every time step to keep the right bearing converts the first mode of navigation to the second one. Nevertheless, distinguishing between the navigational modes was useful in this study as it pointed to different seal behavior in coastal waters and in deep sea. Near the coasts, the locations of the animals suggest they are continually adjusting their course to arrive at a specific destination. In our two examples, the seals reached known gray seal colonies and probably had a good knowledge of the local habitat when getting close to these haul-out sites. At least in the Molène archipelago, where these seals were initially captured, gray seals' foraging areas shown by these telemetry tracks are located in the close vicinity of the haul-out sites, which means they spend a lot of time in the area. We suggest that the seals have a good knowledge of this habitat close to this destination point, which allows them to switch from one navigation strategy to another at the end of their trajectory. In addition to the local bathymetry and sea-floor shape and habitat, seals could use chemosensory cues such as gradients in salinity as sources of orientation (Sticken and Dehnhardt 2000).

At sea, the seals followed the “keep constant bearing” navigation (R1GT) for prolonged periods. Their behavior in the middle of the Channel looks like that of a ship's navigator (Brillinger and Stewart 1995) who determines the ship's new position at the start of a day and then corrects the heading of the course. Contrary to this example, however, there was no correlation between route adjustment and time of day and no feature or cues could be identified at the location of the change in direction and velocity in the middle of the course.

The successful modeling of the observed trajectories above implies that these seals have an ability to maintain heading along long legs of their travel. In this note we used a purely deterministic model assuming two plausible and nonexclusive, navigation strategies.

This approach is very different from the statistical modeling developed by Kendall (1974), Mills Flemming (2010), or by Brillinger and Stuart (1998). They used a geolocation system of poorer time resolution (one to four approximate locations per day), while in this study we obtained 80–90 GPS-quality locations/day. Instead of supposing an ability of seals to determine their position outside the “circle of confusion” (Kendall 1974) we supposed a kind of perfect seal that is able to maintain a heading and a speed in the absence of navigating cues. The question we asked was whether a simple navigation strategy could be found in order to match the observed seal trajectories. A simple navigation rule of keeping a constant heading over several hours resulted in a close fitting of the observed seal trajectories. It would be interesting to include random perturbations of seal movements in order to estimate the circle of confusion of seals navigation and to compare it to predictions of purely stochastic models (Mills Flemming 2010). This would be particularly important in modeling seals' swimming in three dimensions when the seal's diving depth is not known as accurately as its horizontal position.

Our deterministic model matched the real trajectories well. A series of trials with various values for heading and seal speed resulted in very different trajectories (not shown here) beginning with an orbital trajectory near seal's departure point when seal's speed is too low and going to a straight line when the speed is much higher than that of the tidal flow.

We propose to develop this model in two ways. First, we will extend it to three dimensions to incorporate the depth dependence of sea currents. Second, we will include stochastic perturbations of seals' locations, their heading and speed in order to evaluate the corresponding “circles of confusion.” We also propose to test what temporal or environmental cues (e.g., time of day, undersea features, navigational buoys) may be linked to course readjustment.


  1. Top of page
  2. Acknowledgments
  3. Literature Cited

We wish to thank everyone from the Marine National Park of Iroise, the Sea Mammal Research Unit, the University of La Rochelle, the Office National de la Chasse et de la Faune Sauvage, Oceanopolis, and the Zoo de La Fleche who helped with seal captures in the field. Seals were captured under license 10/102/DEROG delivered by the French ministries of Ecology and Fisheries, respectively. This project was funded by the Regional Council of Poitou-Charentes and by the Marine National Park of Iroise (France).

Literature Cited

  1. Top of page
  2. Acknowledgments
  3. Literature Cited
  • Brillinger, D. R., and B. S. Stewart. 1995. Elephant-seal movements: Modeling migration. The Canadian Journal of Statistics 26:431443.
  • Chapman, J. W., R. H. G. Klaassen, V. A. Drake, et al. 2011. Animal orientation strategies for movement in flows. Current Biology 21:R861R870.
  • Chevaillier, D. 2011. Réalisation et calibration d'un modèle de marées dans le Golfe de Gascogne [Construction and calibration of a tidal model in the Gulf of Biscay]. M.Sc. thesis, University of La Rochelle, La Rochelle, France. 40 pp.
  • Davies, A., J. Xing and E. Jones. 2011. A model study of tidal distributions in the Celtic and Irish Sea regions determined with finite volume and finite element methods. Ocean Dynamics 61:16451667.
  • Gjertz, I., K. Kovacs, C. Lydersen and Ø. Wiig. 2000. Movements and diving of adult ringed seals (Phoca hispida) in Svalbard. Polar Biology 23:651656.
  • Hervouet, J. M. 2007. Hydrodynamics of free surface flows: Modelling with the finite element method. John Wiley & Sons, Ltd., Chichester, England.
  • Kendall, D. G. 1974. Pole-seeking Brownian motion and bird navigation. Journal of the Royal Statistical Society, Series B 36:365417.
  • Kowalewsky, S., M. Dambach, B. Mauck and G. Dehnhardt. 2006. High olfactory sensitivity for dimethil sulphide in harbour seals. Biology Letters 2:106109.
  • Lesage, V., M. O. Hammill and K. M. Kovacs. 2004. Long-distance movements of harbor seals (Phoca vitulina) from a seasonally ice-covered area, the St. Lawrence River estuary, Canada. Canadian Journal of Zoology 82:10701081.
  • Lohmann, K. J., C. M. F. Lohmann and C. S. Endres. 2008. The sensory ecology of ocean navigation. Journal of Experimental Biology 211:17191728.
  • Lowry, L. F., K. J. Frost, R. Davis, D. P. DeMaster and R. S. Suydam. 1998. Movements and behavior of satellite-tagged spotted seals (Phoca largha) in the Bering and Chukchi seas. Polar Biology 19:221230.
  • Matsumura, M., Y.Y. Watanabe, P. W. Robinson, P. J. O. Miller, D. P. Costa and N. Miyazaki. 2011. Underwater and surface behavior of homing juvenile northern elephant seals. Journal of Experimental Biology 214:629636.
  • Mauck, B., N. Gläser, W. Schlosser and G. Dehnhardt. 2008. Harbour seals (Phoca vitulina) can steer by the stars. Animal Cognition 11:715718.
  • McConnell, B. J., M. A. Fedak, P. Lovell and P. S. Hammond. 1999. Movements and foraging areas of grey seals in the North Sea. Journal of Applied Ecology 36:573590.
  • Mills Flemming, J. E., C. A. Field, M. C. James, I. D. Jonsen and R. A. Myers. 2006. How well can animals navigate? Estimating the circle of confusion from tracking data. Environmetrics 17:351362.
  • Mills Flemming, J. E., I. D. Jonsen, R. A. Myers and C. A. Field. 2010. Hierarchical state-space estimation of leatherback turtle navigation ability. PLoS ONE 5(12):e14245.
  • Nicolle, A., M. Karpytchev and M. Benoit. 2009. Amplification of the storm surges in shallow waters of the Pertuis Charentais (Bay of Biscay, France). Ocean Dynamics 59:921935.
  • Sentchev, A., P. Forget and Y. Barbin. 2009. Residual and tidal circulation revealed by VHF radar surface current measurements in the southern Channel Isles region (English Channel). Estuarine, Coastal and Shelf Science 82:180192.
  • Sticken, J., and G. Dehnhardt. 2000. Salinity discrimination in harbor seals: A sensory basis for spatial orientation in the marine environment? Naturwissenschaften 87:499502.
  • Willis, J. 2011. Modelling swimming aquatic animals in hydrodynamic models. Ecological Modelling 222:38693887.