by Daire Carroll
To observe an animal in its undisturbed state is the wildlife ecologists dream. In the past this meant crouching in blinds and spending hours slowly creeping up on your study species, hoping that a chance change in the breeze wouldn’t give you away. Now, the advent of new technology is opening up wild spaces as never before and allowing researchers to observe without being observed. Our group at Gothenburg University focuses on studying seal ecology. In recent years, we have been taking advantage of drones and trail cameras to survey seal populations and quantify animal behaviour. In our latest publication , we present a new method to non-invasively estimate the seal mass from drone images.
An animal’s body condition is closely related to its health and welfare. For seals, mass is a particularly important factor. In the autumn of their first year, a seal pup’s mass is directly related to its chance of surviving through the winter. For decades, the blubber thickness and mass to length ratio of seals has been surveyed as an indicator of the health of marine ecosystems. These surveys have only every been able to assess small numbers of individuals due to the time and skill required to either capture a live seal or preform an autopsy on a dead seal.
Our new method involves flying a drone on a preprogrammed route over known seal haul outs and building maps called ‘orthomosaics.’ Within these maps, machine learning is used to detect and automatically outlined seals. This gives us a set of seal shapes which can be measured to estimate length and width. By collecting data from hundreds of seals, we have developed a set of equations to translate from these measurements to an estimated seal mass. We tested our method on harbour seals at Slottsskogen zoo in Gothenburg. On separate visits in February and April a pup experienced a 7.7 kg change in his true mass. We were able to detect this as a 6.9 kg change using the drone-based method. While this level of accuracy is extremely useful, the real benefit of our method is that it is semi-automated and can estimate the mass and length of thousands of seals within minutes of the data being collected.
Being able to track the body condition of close to the entire population of seals promises to enable us to explore how environmental variability impacts animal health and welfare. Already, we have begun to apply the method during the breeding season to track the health of harbour seal mothers and estimate the overwinter survival probability of pups. All this takes us a step closer to the dream of tracking a wild animals in an undisturbed state.
Remote Sensing in Ecology and Conservation Blog 