More than just counting: the pack-ice seal census

Estimating the number of animals in a population is a tricky business. Unlike a human census where most subjects make themselves available for counting, animals may be overlooked even though visible, or they may be present but hidden by obstacles. The science in an animal census is not in the counting but in estimating how many weren’t counted.

Such problems made a recent survey of seals in the pack-ice off East Antarctica one of the most difficult animal population studies ever undertaken. To address them, researchers utilised recent advancements in survey methodology and technology.

The survey aimed to estimate the number of crabeater, leopard and Ross seals in the pack-ice between 60oE and 150oE, a vast area covering over a million square kilometres. To do this we used Sikorsky S76 helicopters flying at 80 knots on north-south tracks to count seals hauled out on the ice in 1200m wide strips (600m each side of the aircraft). It was impossible to cover the entire area so the strips represented only a sample, but provided the strips are representative then extrapolating to the entire area is straightforward.

Looking down from an aircraft would seem an ideal position for counting seals, but not so. Ridges and hummocks in the pack-ice can hide more distant seals from the observer, and the more uneven the ice, the more animals missed. A method that addresses this common animal survey problem is called distance sampling, where the distance between sighted animals and the flight path is measured. Typically fewer animals are sighted as distance from the flight path increases. Provided we can be certain that all animals are seen within a specified distance of the flight path, we can use the relative frequency of sightings at that ‘certain’ distance to correct for animals missed at other distances.

Distance sampling was developed for ground surveys where seeing an animal on the survey track could be considered a certainty. Counting from the air gives no certainty of seeing an animal, even close to the aircraft, because there is not time to search all the strip thoroughly. It is possible to slow down in a helicopter, but then there would not be enough time to gather a representative sample of the million square kilometres. Instead we maintained sufficient speed to sample enough area, and developed methods to estimate how many seals were missed in the zone of best visibility.

To do this we adapted another procedure for estimating animal numbers, called mark-recapture, by which a sample of animals is caught, given an identifying feature (usually a tag) and released. Some time later another independent sample of animals is caught and the proportion of tagged animals (recaptures) in the sample used to estimate the total number of animals in the population.

So how do we mark animals from a fast-moving helicopter? The ‘tag’ can be simply a unique identifier for a particular seal. We have two observers (front and back) on each side of the helicopter, searching the same area independently. On sighting an animal, the front observer ‘tags’ it by recording unique features
identifying it from other sightings, such as the exact time the seal passes abeam of the aircraft, the seal’s distance from the flight path, its species and group size.

In combination, these features can form a unique identifier for each sighting. If the back observer sees the same animal independently of the front observer, as indicated by the tag identity, it is considered ‘recaptured’. From then on the sums are similar to a normal mark-recapture exercise.

If each observer records both the exact time, distance, species and groups size for each sighting, we have the potential to use both distance sampling and the adaptation of mark- recapture in combination to estimate all the animals that were present but not sighted in the survey strip, regardless of distance from the flight path. But before the survey began there was no easy way of recording these data conventionally in the short time available to record data for each sighting.

We put the problem to Australian Antarctic Division engineers, who designed an electronic data logging system in which a laptop computer logs data from several sources. Altitude and position are logged at 10-second intervals using a radar altimeter and GPS unit. Each observer has a sighting ‘gun’ and a keypad to record data for each sighting.

As the seal passes abeam of the helicopter, the observer points the gun at the seal and presses a record button, then enters species and group size data via the keypad. The angle of declination of the gun, in combination with the altitude, is used to calculate the distance of the seal from the flight path, which together with the exact time from the keypad entry, forms the ‘tag’ for that observer.

After the survey, distance/time ‘tags’ for sightings by front and back observers are compared to determine the number of ‘captures’ and ‘recaptures’. From there
the number of missed seals in the strip can be estimated, which combined with the number counted gives the total number of seals on the ice in the strip.

But there are still some missed animals to account for. Seals are marine mammals, spending much of their time foraging in the ocean. The above methods have been aimed only at seals that are visible, out on the ice.

To estimate the proportion of seals in the  water,  we  need  to study the haulout behaviour of the seals. This is now possible with the development of small electronic dive recorders glued  to  the  fur  of  a sample of captured animals (the recorder later falls off with the moult). These dive recorders have a conductivity sensor that measures when the seal is in and out of the water, data which can then be transmitted to the researcher by satellite.

Catching a seal in the pack-ice is very difficult, and re-catching the same animal some time later nearly impossible. But a satellite link means that once a dive recorder is attached to an animal, it does not have to be recaptured to retrieve the data. The transmitted data can be used to estimate the proportion of
time seals spend on the ice and in the water at any time of the day, and so provide a correction for seals in the water.

No survey will provide a completely accurate result. Surveying over so large an area of remote ocean will always be less accurate than surveys undertaken in smaller areas or in more favourable conditions elsewhere. But the various techniques described above have given us data that is several times more accurate than anything we have had previously, and which we can apply with confidence to larger Southern Ocean ecosystem questions.

Colin Southwell
Antarctic Marine Living Resources Program,
Australian Antarctic Division