Analysis of Laysan Albatross GPS tracking data
1. Introduction
The purpose of this article is to examine the flight path of an albatross to see whether anything can be learned about how albatrosses use dynamic soaring as part of their foraging behaviour. The Windward-Turn theory of dynamic soaring referred-to later is explained in the website
In 2012, when this website was first published, the sole purpose of it was to explain how dynamic soaring worked as a flight manoeuvre practised by albatrosses. At that time, I had only a little amount of fine-resolution GPS tracking data with height information, to analyse the dynamic soaring manoeuvre as performed in nature by the experts. That information came from a research paper concerned with high resolution GPS tracking rather than with albatrosses per se.
For scientists studying albatross behaviour, high resolution data or even GPS data is not necessarily needed. Tracking can be achieved with a simple light-detector and an accurate clock to determine times of local sunrise and sunset. These times can be processed to give latitude; while the time of local midday relative to GMT gives longitude. A simple immersion switch will detect whether the bird is airborne or on the surface of the water. Some birds have been persuaded to ingest a recording thermometer which then tells the scientist when the bird has swallowed cold seawater and therefore when it has been feeding! This information need only be collected every few minutes or hours rather than every second. That is enough to determine where the bird goes and what it is doing but it is not high enough resolution to analyse dynamic soaring. For that, resolution of one second or less is needed and, ideally height information as well.
2. The Data
From time to time, high resolution data of albatross tracking becomes available and then we can get some insight into how dynamic soaring fits-in with the albatrosses general foraging behaviour.
The dataset I am using here is one of several sets of GPS tracking of albatrosses and shearwaters, which was published on the website datadryad.org and comes from a research paper entitled: Flight paths of seabirds soaring over the ocean surface enable measurement of fine-scale wind speed and direction by Yonehara et al, University of Tokyo 2017. The paper was concerned with the calculation of wind-velocity by analysing seabird tracking data. The thinking was that this method could help to verify and fill-in the gaps in wind-velocity data derived from other sources such as ships, buoys and satellites. The data comprises approximately 46 hours of latitude and longitude coordinates recorded at one second intervals but with no height information. The data was downloaded to an Excell spreadsheet for processing and the results are shown graphically.
3. Calculating wind-velocity
The method of calculating the wind-velocity works like this: Using the difference between successive GPS latitude and longitude positions at one second intervals, the ground-speed and the track direction relative to True North are calculated. On a scatter-graph, plot ground-speed on the y-axis against each possible track from 0 to 360 degrees on the x-axis. When airspeed and wind-speed are each uniform, the variation of ground-speed versus track is an approximate sine-curve. The lowest and highest values of ground-speed occur at the exactly into-wind and down-wind tracks respectively therefore this gives the direction of the wind. The ground-speed varies by double the wind-speed during a full 360degree circle therefore the wind speed is half the difference between the greatest and least ground-speeds.
Since, in the natural world, neither a bird’s airspeed nor the wind-speed are exactly uniform and typically they are not flying circles, the distribution of the ground-speed values is a ‘smeared’, approximate and partial sine curve, which will only be a complete curve if data points are obtained for all tracks from 0 to 360. Furthermore, the data distribution is slightly ‘pinched’ at the point where the groundspeeds are greatest because the rate of turn and the rate of change of speed, is also not uniform. Despite appearances, the graph is not a pictorial view of the dynamic soaring manoeuvre. For instance, the section from the least speed to the greatest speed is the leeward turn. In practice, a ‘partial, smeared’ curve is obtained by plotting data from a given time period, say 300 data points from a 5-minute section, to which a clean sine curve can be ‘fitted’ manually, and that curve is used to get the average wind-velocity.
This is an example of such a graph comprising, in blue, 300 data-points at one-second intervals; and, in orange, a sine curve fitted to the data. The sine curve is fitted by eye; by moving it left and right, up and down and by adjusting its amplitude. The gap in the data means that there were few tracks between about 270 and 310. The numbers in the top line refer to the number of seconds into the data set.
The lowest point of the curve is at a track of about 130 degrees and 9 m/s. In other words, the bird was tracking South-East when heading directly into the wind. The highest point is at 310 and 17 m/s. The difference between the highest and lowest groundspeeds is 8 m/s. Therefore, the wind-velocity is 130 degrees True at 4 m/s. The wind-direction, from the South-East, is the direction FROM which the wind is blowing, which is the meteorological and navigational convention. This exercise is repeated for any particular location in the total data set.
The problem with using albatross dynamic soaring for the specific purpose of calculating the wind, is the fact that average dynamic soaring tracks on a scale of kilometers are relatively straight and a single dynamic soaring manoeuvre, comprising a single windward turn and a single leeward turn, on a scale of hundreds of metres, has a relatively narrow range of tracks: typically, only sixty degrees. Nevertheless, with a large enough data set and presuming relatively uniform winds, it is quite easy to manually fit the sine curve to the data.
4. Raw data
Before looking at particular points of interest, what can we do with the raw position and speed data? First, the overall track can be plotted on a rectangular grid to give a true scale representation of the track, as shown here, titled Ground plot. Second, the wind-speed can be plotted against time as shown next. The y axis is time in hours repeated in seconds, from the time the tracking device is switched on. Third, the ground-speed can be plotted against time.
What we can see is 46 hours of data comprising 42 hours in the life of a Laysan albatross plus four hours of data from before the tracking device was attached to the bird. The wind-velocity is indicated by the arrows and in degrees True and meters per second, at one hundred kilometre intervals. The red circles are the locations of points of interest that we will look at later.
We don’t know whether the bird is male or female or whether an experienced adult or a juvenile, so let’s call him Phil. Hi Phil. The journey begins in February 2014 at Ka’ena Point on the North West corner of Oahu in the Hawaiian Islands.
Once the GPS tracker is attached, Phil sets-off in a determinedly Northerly direction. In 42 hours Phil will travel 1000 kilometers to the North. This is surprising because the wind is a light Northerly for the first 36 hours.
Phil’s main soaring technique - dynamic soaring - is theoretically most energy-efficient when flying on crosswind headings but clearly upwind dynamic soaring in light winds is also a practical proposition.
To achieve an overall journey of just over 1000km, the total ground distance flown is 1747.403km. This is the simple sum of all of the one-second legs and is therefore probably an over-estimation. Nevertheless, it does reflect the combined effect of the sinuous nature of dynamic soaring at the small scale of a few hundred metres and the meandering flight path of the bird at the larger scale of hundreds of kilometres. This apparent inefficiency is offset by the energy-saving of dynamic soaring at relatively high speeds using only the energy of the wind.
Why is Phil heading North? If not led by the wind, then it may be that Phil is heading for a preferred feeding ground. We will see.
5. The Wind
In the next diagram, the wind has been calculated at one hour intervals and plotted against time in hours and in seconds as a bar-graph. The first six hours of non-flying data are excluded from this. To see whether there are any diurnal effects, the two orange markers are spaced 24 hours apart but arbitrarily placed. We might expect that the wind strength at low levels would reduce at night with cooler temperatures and less atmospheric turbulence and mixing. The wind data appears to show two reduced wind-speed periods but it is difficult to fit the two 24 hour spaced markers to demonstrate any diurnal effect. The variation of wind is probably a combination of geographical, weather and diurnal effects rather than purely an indication of night fall.
6. Ground-speed
If the reduction of wind-speed is an indication of night fall, there is no sign that Phil is sitting it out. When the wind reduces to near calm at about 23 hours, Phil alights more frequently but never for long periods. Is he feeding or resting or sleeping?
On the Ground-speed graph below, the first four hours (14400 seconds) are mostly to do with scientists switching-on the tracking device and carrying it by car and on foot to the roosting site prior to attaching it to Phil. The rest of the data is pure albatross flight-ing or alighting.
Whilst airborne the ground-speed varies to different degrees. The variation of ground-speed is either due to a variation of airspeed or a variation of wind-speed or a wider or narrower range of wind-angles used, in other words, more or less turning. The greater amplitudes of ground-speed do seem to correspond to the increasing wind-speeds during the latter part of the journey, while the smaller amplitudes of ground-speed appear to correlate to the lesser winds during the presumed night-time periods.
The ground-speed graph indicates when Phil is flying and when he has alighted. Albatross flying speeds are approximately 10 to 20 m/s therefore, a ground-speed of less than about 4 m/s shows when he has alighted. Phil was airborne for no less than 36.89 hours out of 41.96 hours of his journey; that is 88% of the time. During the first period of 37.2 hours Phil was airborne for 36 hours; that is 96.75% of the time. During that first period he alighted 20 times for an average of approximately 2.5 mins each time. For the second period of 4.76 hours, Phil was airborne for 0.91 hours, which is 19.1% of the time. In other words, during the second period he was on the surface for 81% of the time.
After 32 hours, 600km North of Oahu, the wind gets stronger from the North East and the variation of ground-speed becomes greater. Then at 37 hours and 700km North, there is a change in the ground-speed pattern. Phil starts making periodic 90 degree turns upwind or downwind. This could be a sign that Phil’s track is intersecting a scent trail and he is turning to follow it. Finally, at the Northernmost point of the track, it seems that Phil has found what he is looking for and he spends most of the time on the surface.
7. Foraging strategy
Maybe we can conclude that Phil has reached his feeding ground but how did he achieve that? Did he go to a known location or did he follow a scent trail upwind or something else? A valid foraging strategy would be to fly upwind when a scent trail is detected and fly crosswind if the scent trail is lost. Turn upwind again when the scent trail is detected again. We can look for evidence of this by zooming-in on selected sections of the data.
8. Ka’ena Point
Here is a Google Earth picture of Ka’ena point on the North West tip of Oahu. Note the 100m scale bar. The pre-flight action is taking place somewhere in the middle of this area.
9. Taking a closer look – wrangling and departure
Now let’s look at things in more detail. The first four hours of the data are nothing to do with Phil the albatross. The GPS tracker is switched on and then transported several kilometres by car and on foot along the coast road, past the airport, along a dirt road to a car park and then by foot to the roosting site. The next diagram takes up the story and covers about two hours of time but is only 150m from one side to the other. This is about in the middle of the photo of Ka’ena Point but note the different scale.
We start at point A. The line of data points, always at one second intervals, has a particular character, steady and ordered, which is maintained to point B and on to point C. This surely depicts the steady plod of the scientists carrying the tracker, entering the roosting area looking for suitable candidates for tracking, including Phil. There is some kind of action at B and another kerfuffle at C. Then, the character of the plot changes to a slower more chaotic or random track from C to D. Did the scientists tag their first bird at B and then find and wrangle Phil at C? Then, do we see Phil, his feathers slightly ruffled, wander off from C to D? 
10. Taxying, pre-flight checks and take-off
In the next diagram of 46 minutes, the meandering continues from D via E to F and then there is another change of gait. After shuffling around for a bit at point F, Phil lurches forward, going 15 meters over the ground in 3 seconds, maybe 20 meters through the air with a headwind. The ground is flat here and the wind relatively light; Phil is surely flapping! He lands at G and shuffles to H, 10 meters to his left. Is he practising flapping; getting ever lighter on his feet? Is Phil a juvenile practising for his first solo? An experienced bird would surely not be messing about like this.
Finally, not with a lurch but a launch, facing the Northerly breeze, striding out and flapping gamely, in seven seconds from H to I, Phil accelerates to 10 m/s; airborne and free!
11. Take-off
Next, we see the take-off again as Phil achieves full flying speed from point H to point I. The wind is from the North-north-east, 030 degrees 5 m/s, about 11 miles per hour. The scale bar is 50m. Phil immediately bears off to the left to achieve a cross-wind heading. An experienced bird would immediately go into dynamic soaring mode but Phil does not. If he is a juvenile, he has never done dynamic soaring before.
12. The first 5 minutes of flight school
The next diagram is a larger scale, note the scale bar of 500m but is only 5 minutes of data. The wind analysis chart shows a wind of 030 degrees 5m/s. Phil’s track is relatively straight from H to I, although still meandering +/- 10 to 15 degrees. Ground-speed is 13m/s +/-2m/s. The variation of ground-speed corresponds to the variation of track but obviously we cannot see if there is any variation of height. The wavelength of the speed- and direction-changes is about 7 seconds which is close to dynamic soaring. This is surely flapping flight, although there is no direct evidence of that; no indication of a flapping frequency.
Suddenly at point I, after only a minute of flight, something clicks in Phil’s brain. He locks his out-stretched wings and changes gait to a more sinuous motion. The wavelength of the motion increases to about 12 seconds and ground-speed increases to 15m/s +/-5m/s. This is easier than flapping and he can sense through his tube nostrils that he is sustaining his airspeed. This is dynamic soaring!
The serpentine flight path from I to J gives a greater variation of ground-speed due to the greater variation of track direction. Where the data points are more widely spaced indicates greater ground-speed and therefore more downwind headings. From I to J, Phil practises curves and circles. The pattern is not well-ordered although it is crosswind which is the most efficient dynamic soaring orientation. He will keep this up for an hour before alighting briefly, maybe to feed, maybe to rest.
Certainly, if Phil is a beginner he is wise to avoid splashing down near the roost in the fledging season. You never know what might be lurking below the surface!
13. The first touch-down at sea
Moving on, Phil is 13.5 kilometers North of Oahu, after an hour of flying upwind into a North-easterly breeze; here we can see seven and a half minutes of Phil’s track on a grid of 100m squares.
Although most of the distance of this particular plot is in flight, most of the time of this plot, 7.5 minutes, is spent on the surface. At point A Phil alights for about 5 minutes with a brief hop at point B before setting off again at point C.
14. Getting the hang of it
The next glimpse of Phil’s progress, after nearly five hours, sees him 70 kilometers North of Oahu, doggedly pushing-on into a light North-easterly. His flight-path is undulating but still somewhat irregular, his average ground-speed is approximately 13m/s. His track is mainly crosswind but he is making some distance upwind.
15. Night flying
At just under 11 hours of flight time, Phil is 230 kilometers North of home and is tracking North-East at around 14m/s. (below)
The wind has died away to a negligible westerly or light and variable. If we assume Phil was tagged around midday on the first day, then it is probably night but possibly moon-lit or star-lit.
Phil’s ground speed has increased slightly because he now has a slight tailwind. This 5 minute section is 3043m as the crow flies but is 4200m as the albatross flies. The crow would be expending energy while flapping but the albatross is dynamic soaring and the energy saving makes it worthwhile travelling 38% further and taking longer to achieve the same distance.
16. Climbing and diving
The next diagram shows a curious episode. 12 hours into his flight, it is probably still night time and Phil is heading North-North-West against a light Westerly. There is a strange episode of convoluted flying between A and B with the average ground speed about 14 m/s. Between A and B the overall speed is about 2.5 m/s or about 5 knots but there is no indication of a touch-down. Is he soaring in the updraft on the upwind side of a ship moving slowly Northward? We will never know. Phil abandons this game and sets off Northwards. After another 300m there is about 200m of relatively straight flight at reduced speed, presumably a flapping climb, and then at point C a sharp reversal of direction and an increase of speed to 22m/s, about 49miles per hour, presumably a dive. Is Phil just having fun? He turns right and proceeds to the North-West.
17. Cruising flight
Here, Phil is 17 hours and 380 kilometers into his journey, making steady progress to the North-north-west. Still the wind is light and variable. Well-named is the Pacific but, to an airman with distance to make, a headwind, however light, is not benign.
This is routine commuting. No sign of anything happening.
This diagram and the next are either side of a 40degree turn across the wind. Phil clearly wants to go North but appears to not want to fly directly into the wind. The closer is the average dynamic soaring heading to cross-wind, the more efficient is the manoeuvre due to the lesser load-factor. Phil appears to be compromising between heading directly into the North wind and heading cross-wind.
18. Pausing
At 19.5 hour’s journey time, 445 kilometers from home, Phil has turned across the wind and is now tracking North-north East. He takes another couple of short breaks on the surface. The second break, of 2mins 20 secs, is shown as A-B in these two diagrams. Note the different scale bars.
In the diagram on the right, Phil is on the surface and moves 50 m North against the wind, possibly drifting with the current, possibly paddling maybe even pursuing a meal.
Even though the track on the left is mostly flying and that on the right is on the surface, the track is curiously fractal at scales of 10m 100m and 1000m.
19. Turning upwind
Another 10 hours flying and we find Phil at 773 kilometers from home. The wind is 4m/s from the North-east, Phil makes a sudden turn into wind and increases speed. Has the scent trail increased or has he intersected a scent trail? After 600m he has found nothing or at least he does not alight and he turns crosswind to the North again.
20. Foraging
At 880 kilometers North of Oahu, the wind has backed to the South-East and Phil is tracking North East with a steady 5m/s crosswind. During the 3.3 hours of this extract, covering about 55 kilometers, clearly something different is happening and it is worth taking a closer look.
The first two points of interest A and B do not look very different in the Ground plot view but in the speed chart, the first is a flight sequence and the second is a surface event.
21. Dynamic soaring
The first event A shown here as A1 to A2, illustrates 2 minutes of the relaxed and easy grace of classic dynamic soaring; an average cross-wind heading with long, windward turns of 5 to 10 seconds to the right and quick leeward turns to the left at an average groundspeed of 15 to 20m/s.
This looks much more orderly, smoother and more efficient compared with Phils efforts a day and half before.
Is this an illustration of how a juvenile albatross masters the art of dynamic soaring during its first two days on the wing?
No alighting here.
22. Alighting
At the second event B shown here as B1 to B2, Phil alights briefly a couple of times. During this 4 minutes and 10 seconds of convoluted flying, concentrating on a relatively small area, Phil alights for about 30 seconds, presumably for a quick meal. This is rather different to the calm drifting shown earlier.
Phil gets airborne and heads off to the North East again.
23. Tacking upwind - gybing downwind
Then suddenly at point C1, Phil takes a turn to the right and heads upwind for a couple of kilometres towards the South East. With apparently nothing found or at least no reason to alight, at point C2 he then doubles back downwind to the North-West and picks up the trail again at point C3. But note he does not fly directly downwind. Rather, he zig-zags or gybes downwind like a sailboat, suggesting that while dynamic soaring works crosswind as A1 to A2 with a downwind drift element or upwind as from C1 to C2, it does not work so well for tracking directly downwind.
24. Tacking upwind
Here is a 5 minute section from C1 to C2. The average ground-speed is slightly less than seen on the crosswind leg A1 toA2 and on the downwind leg C2 to C3. The track and the ground-speed looks more chaotic and less well ordered.
The Windward turn theory of dynamic soaring suggests that, while crosswind dynamic soaring is possible at little more than 1G load factor, upwind dynamic soaring requires a greater load-factor and therefore more effort on the part of the bird. Furthermore, the wind-gradient becomes more important when going upwind but is of course invisible and the out-come of any particular windward or leeward turn is more uncertain.
25. Gybing downwind
Next is an expanded version of the section from C2 to C3. In the Windward Turn theory of dynamic soaring, it is indeed difficult, if not impossible, to find a solution that allows a direct downwind flight path. Upwind or crosswind flight path solutions are relatively easy to find but downwind tracks are best achieved with a variation of the crosswind technique which can be flown at approximately 1G load factor but at about 135degrees to the wind direction. In this case, the 5 minute zig-zag section from C2 to C3 comprises five legs, each with about five dynamic soaring manoeuvres on each leg and yields an average ground-speed of about 15m/s with a direct downwind ground-speed of 10 m/s which is more efficient than simply meandering around at random and drifting downwind at the speed of the wind, as seen later. Looking at the downwind leg again, from C to E, we can see the zig-zag pattern repeated in a fractal manner at scales of 100m 1000m and 5000m.
26. Homing-in on a meal
At point C3, Phil turns upwind again and alights at point C4. Has he found something?
At C4 the downwind path reverts to a sinuous but somewhat chaotic flight path as seen below. It looks like Phil is drawn to investigate this area and alight a couple of times for about two minutes. This could be feeding behaviour.
27. Meandering downwind
Next is the section at D showing a different downwind flight pattern. Instead of the well-ordered zig-zag track from C2 to C3, the flight path here looks more chaotic with a ground-speed varying from 10 to 20 m/s; covering a direct line distance of 3200m in 16.2 minutes, at an average of only 3.3m/s which is less than the speed of the wind. Phil appears to be interested in this particular patch of sea, oriented in the downwind direction but does not appear to alight. Clearly, this is a less efficient way of travelling downwind compared to the zig-zag method. After about 16 minutes Phil reverts to the more efficient gybing pattern.
During 1.8 hours it appears that Phil has investigated a definite line from C to E, about 20 kilometers long, aligned approximately with the wind.
Finally, at point E he breaks off and heads North again.
28. Darkness
Moving on, we are 34.7 hours into the journey and 930 kilometers from home. After a day and a half since departure it could be dark again. The wind is a light South-Easterly
Phil is initially heading North East. Again, he takes a sudden turn to the right and flies upwind towards the South-East for about 2 kilometers. At point F there appears to be some investigations taking place but Phil does not alight for long. He turns crosswind again for another 2 kilometers and we join him again at point G1. From G1 to G2 is about 10 minutes but about 9 minutes is spent on the surface.
We can see how Phil turns into wind at the points of landing at H1 and take-off 100m further North. The section from H1 to H2 is shown next.
The 25m section from H1 to H2 appears to show variable speeds with darting movements but following a line rather than dwelling at a particular spot. Is this the difference between scavenging and active predation of live prey?
Phil is shown full scale, although he is on the surface here, not in flight.
29. Journeys end
Finally, at 1000 kilometers North of Oahu, Phil seems to have mastered his craft. The wind remains at about 4 m/s from the South East; Phil is tracking North-North-East. From J to K we see classic dynamic soaring flight. The flight path is undulating on a scale of a hundred metres but is fairly straight on a scale of 1000m. Maintaining a sinuous cross-wind heading, the resulting track is slightly downwind, according to the drift angle. An average triangle of velocities is added to illustrate the three velocity vectors and the angle of drift.
At K, Phil takes a sharp right turn through 90degrees and begins to head upwind to the South-East. Has his cross-wind track once again, intercepted a scent-trail drifting downwind?
From K to L, the character of the dynamic soaring undulations changes subtly, becoming more compressed and the ground-speed reduces with the headwind. On the ground-speed graph, the ground-speed from K to L is visibly less than from J to K. The dynamic soaring technique is slightly different when going upwind compared to crosswind. According to the Windward Turn Theory, the angle of bank and load factor are greater during upwind dynamic soaring, requiring greater effort from the bird.
At point L, Phil alights and now spends most of his time on the surface with brief flights. The circled area from L to S is expanded below.
The wind speed has picked up a bit to 8 m/s but still from the South-East.
Finally, a two minute, 20 m section of surface activity between O and P is expanded next. Note the 5m scale bar and the average speed only 0.5 m/s. This time the line of interest is normal to the wind direction. Phil is not drifting downwind; he is making deliberate changes of direction and speed.
Also, there are distinctive loops in his track as if he focusing on some particular item in the water. This behaviour has been noted before by others.
Phil makes a final appearance to scale although he is not flying here; he is on the surface and presumably he has found his feast.
Between L and S, Phil spends a lot of time on the surface with brief linking flights. From Q to R is such a linking flight. We can see the classic crosswind dynamic soaring flight; long windward turns at a small rate of turn linked with short, quick leeward turns. Phil alights at R and pursues another line of interest to S, normal to the wind.
Bon appetite and Bon voyage!
30. Conclusions
During his 42 hour journey, it appears that Phil’s dynamic soaring technique improved rapidly so that it might be inferred that he was indeed a juvenile on his first flight.
Phil was airborne for over 88% of the time but due to the sinuous nature of dynamic soaring the distance flown was up to 75% greater than the straight line distance. It is impossible to know how much flapping was involved but anecdotally we can estimate there was very little.
There is strong evidence that the albatross foraging technique is to travel upwind or crosswind until some trace of prey is detected, presumably through scent trails drifting downwind and then to turn upwind to home-in on the target.
