Movement tracking and bycatch risk in the Southern Ocean
This project analyses GPS tracking data from 24 grey-headed albatrosses at Campbell Island, New Zealand, to understand how and where they feed — and where they face the greatest risk from longline fishing operations.
Using a technique called a Hidden Markov Model, the analysis classifies what each bird is doing at every GPS fix across a full year: resting on the ocean surface, actively searching for food, or travelling between areas. This behavioural classification is then linked to satellite ocean data (sea surface temperature, chlorophyll concentrations) and international fisheries records to map bycatch risk at both population and individual level.
Key Question: Where do grey-headed albatrosses forage, and where are they most exposed to longline fishing gear?
Headline Finding: 99.1% of foraging occurs in New Zealand's Exclusive Economic Zone and adjacent subantarctic waters — not in the CCAMLR-regulated Southern Ocean. Addressing bycatch risk for this population is primarily a matter for New Zealand fisheries management, not international Antarctic agreements.
The most striking finding is the degree of variation between individuals from the same colony, in the same year. Some birds spend 59% of their time actively foraging; others just 33%. Some range 2,700 km from Campbell Island; others rarely exceed 750 km. This is not random variation — published research on seabirds suggests individual foraging strategies are stable across years, meaning these differences reflect genuine specialisation.
Conservation implication: Risk is not evenly distributed across the population. Three birds appear to be chronically exposed to fishing areas by virtue of their long-range foraging strategy — not because they happened to be unlucky in 2013.
A statistical model found that where a bird is located (distance from colony, latitude, longitude) is a better predictor of foraging behaviour than local ocean conditions (temperature, productivity). This suggests these birds rely partly on learned knowledge of where food tends to be found, rather than purely following environmental gradients in real time.
Climate implication: If productive ocean zones shift under climate change, birds with strong spatial fidelity may not track those shifts — continuing to return to historically productive areas even if conditions have deteriorated.
CCAMLR (the international body managing Southern Ocean fisheries) records show 58 million longline hooks were set in the convention area in 2013. However, because 99.1% of foraging by these Campbell Island birds occurs north of the CCAMLR boundary — in New Zealand's EEZ — that data tells only part of the story. A complete risk assessment would require access to New Zealand domestic fisheries observer records, which are managed separately by the Ministry for Primary Industries.
The temporal picture shows moderate synchrony (index 0.31) between bird foraging peaks (October–November) and CCAMLR longline effort (January peak, with secondary activity in May and December). The December risk window — when both fishing effort and late-season bird activity coincide — is the most directly actionable finding from the available data.
Single year of data (2013): All findings are from one tracking season. The identification of high-risk individual birds, and the claim that their strategies are consistent across years, relies on published literature rather than multi-year data from this population.
NZ EEZ fishing data gap: The most consequential limitation for conservation application. Longline fishing effort within the birds' primary foraging zone is regulated nationally by New Zealand and was not available for this analysis. The risk surface presented here uses CCAMLR Southern Ocean records, which cover only 0.9% of the birds' foraging area.
Environmental resolution: Satellite sea temperature and chlorophyll data are monthly averages at 10 km resolution. Fine-scale features — upwelling filaments, mesoscale eddies, tidal fronts — that may drive moment-to-moment foraging decisions are not captured at this resolution.
CCAMLR spatial approximation: Fishing effort data in the CCAMLR bulletin contains area codes but no spatial coordinates. Effort is visualised using approximate area centroids rather than precise fishing ground locations.
Ground-truthing: Behavioural state classification (resting, foraging, transiting) is based on movement patterns, not direct observation. While the HMM results are internally consistent and ecologically plausible, they have not been validated against direct behavioural observation data for this specific population.
To move from exploratory analysis to operational conservation application, the following would be most valuable:
The full dataset, interactive map, and analysis code are publicly available:
Open to collaboration on environmental data science projects and actively seeking opportunities in geospatial analysis and conservation science