Realizing the potential of population information from seal photo ID studies

Demographic information can be collected from individual-based studies. Female grey seals can be identified by their unique coat patterns which remain recognisable during their adult lives. Photographs of seals which show these patterns in sufficient detail can be used in analyses of seal longevity, breeding success and movement patterns. The project aimed to provide a semi-automated program to help extract patterns from photos, store these in a database and compare new and existing records. Tests of the ability of the system to make matches and deal with some of the biases associated with mark-recapture data were proposed. This system was to be made available to groups throughout the UK who could use such a product.

Demographic information can be collected from individual-based studies. Female grey seals can be identified by their unique coat patterns which remain recognisable during their adult lives. Photographs of seals which show these patterns in sufficient detail can be used in analyses of seal longevity, breeding success and movement patterns. The project aimed to provide a protocol for taking pictures and associated data in a standardised way, and build a semi-automated program to help extract patterns from photos, store these in a database and compare new and existing records. Tests of the ability of the system to make matches and deal with some of the biases associated with mark-recapture data were proposed. This system was to be made available to groups throughout the UK who could use such a product.

1. The SEAL system was developed to run as a stand-alone package. It uses ExtractCompare (Conservation Research Ltd) as the main processing engine for extracting patterns and comparing them for matches.
2. Necks, heads, flanks, chest and abdomen were identified as suitable areas to have extracts taken: of these most photographs show heads, necks and flanks, depending on whether they were taken at sea or on haulouts.
3. Miss rates (failure to match 2 extracts of the same female) could be as high as 33% if no restictions were placed on photo quality etc. Miss rates fell quickly s more images of the same animal were available, reaching almost zero at 5 or more images.
4. Apparent adult mortality at the breeding colony of N Rona, Scotland was estimated at 0.84 using an open population model, and was found to be similar to that estimated by Smout et al (2011), and likely to be a contributory cause of the falling pup production at the colony.
5. The ability of the program to provide almost real time matches in the field provides a huge improvement in the power of individual recognition for all other individual-based research. For example, the demographic estimates can now be derived from a sample of around 300+ pelage identified females rather than the 100 or so artificially animals marked previously and potentially biased by handling.
6. Provisional tests on a small number of animals have shown that there is some retention of weaned pup pelage patterns through to adulthood, offering a useful avenue for early survival studies.