A comparison of statistical techniques for evaluating body condition in New Zealand leopard seals (Hydrurga leptonyx) using citizen science data
Unlike other southern continents where leopard seals (Hydrurga leptonyx) are considered vagrant visitors from Antarctica, sightings of leopard seals in New Zealand have been increasing, resulting in a re-classification from a vagrant to resident species in 2019. However, their ability to adapt to no...
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2021
|
| Subjects: | |
| Online Access: | https://etheses.whiterose.ac.uk/28944/ https://etheses.whiterose.ac.uk/28944/2/WARREN_206060038_THESIS.pdf |
| Summary: | Unlike other southern continents where leopard seals (Hydrurga leptonyx) are considered vagrant visitors from Antarctica, sightings of leopard seals in New Zealand have been increasing, resulting in a re-classification from a vagrant to resident species in 2019. However, their ability to adapt to northerly environments still remains relatively unknown. Assessments of leopard seal body condition within this northern part of their range is one way in which scientists can understand their health status and thereby adapt strategies to protect populations. Preceding this study research investigating body condition of leopard seals often-employed invasive and costly techniques that carried risk to seals and researchers. This thesis presents four non-invasive procedures to examine body condition of leopard seals found in New Zealand waters, designed to assist with understanding their health status following such a substantial habitat shift. A body condition scoring system allocated leopard seal sighting records into body condition groups based upon presence/absence of bony protrusions and identified that sighting records of New Zealand leopard seals (n=80) were predominantly in Good condition (71.25%). Using these body condition groups, machine learning classifiers were successful in predicting sighting records of New Zealand leopard seals into Good, Moderate and Poor body condition using differences in body shape defined by photogrammetry. Whilst highest classification accuracy was obtained using photographic measurements of body width and Linear Discriminant Analysis (87.5%), an Artificial Neural Network based on leopard seal silhouettes (81.25%) was also identified as being suitable for examining body condition New Zealand leopard seals due to its ability to utilise large, complex datasets and flexibility with lower quality images. Methodologies developed here were enabled by a large photograph library collated by citizen scientists and volunteer researchers and can potentially be applied to assess the body ... |
|---|