Overlap of the distribution of black petrel (Procellaria parkinsoni) with New Zealand trawl and longline fisheries

Citation

Abraham, E. R., Richard, Y., Bell, E., & Landers, T. J. (2015). Overlap of the distribution of black petrel (Procellaria parkinsoni) with New Zealand trawl and longline fisheries. New Zealand Aquatic Environment and Biodiversity Report No. 161. Retrieved from https://mpi.govt.nz/document-vault/10034

Summary

Black petrel, Procellaria parkinsoni, has been identified as the New Zealand endemic seabird species that is most at risk from incidental captures in commercial fisheries. The goal of this project was to use available data to estimate the overlap of New Zealand black petrel with commercial trawl and longline fisheries.

The main black petrel colony is on Great Barrier (Aotea) Island in Hauraki Gulf. Black petrel attend the island from October to May, migrating to the coast of South America during the austral winter. During the breeding season, they forage on the shelf edge, principally along the north-east coast between East Cape and North Cape. Data on black petrel abundance at sea are available from counts made by government fisheries observers of seabirds behind fishing vessels between 2004–05 and 2012–13; from sightings made by a tour operator during birdwatching trips in Hauraki Gulf and to Kermadec Islands between 2004 and 2012; and from high-resolution Global Positioning System (GPS) tracking of 15 birds carried out during the 2012–13 breeding season. All of these data sets have limitations: the observer data contain mistaken identifications, the sightings data from the birdwatching tours are mainly focused inside Hauraki Gulf, and there are only limited high-resolution GPS tracks available.

To generalise from these data to obtain an overlap with fishing effort, boosted regression trees were used to model the data. Boosted regression trees are insensitive to outliers, and are able to represent complex interactions between variables, due to the hierarchical structure of the trees. Two models were fitted, a simple “lat-long” model that related black petrel abundance to latitude and longitude (as well as the count type, method, and breeding season), and a habitat model that related abundance to variables that may be relevant to black petrel (distance from Great Barrier Island, water depth, sea-surface height, sea- surface temperature anomaly, and chlorophyll-a concentration) instead of latitude and longitude. When tested using a ten-fold cross-validation procedure, the lat-long model explained more of the underlying variation in the data. This model was selected as the best model, with results from the habitat model available for sensitivity analysis.

Both models showed a peak in the black petrel abundance on the shelf edge (peaked at around 1000 m water depth), to the north of Great Barrier Island. This finding is consistent with previously published tracking data that were not used in this study. When compared with the observations, the lat-long model was able to explain much of the variation in black petrel abundance between regions and fishing methods. The overlap with trawl and longline fisheries was calculated from the black petrel abundance estimated at each fishing event. When compared with the overlap derived from the distribution used in a recent risk assessment of commercial fisheries to New Zealand seabirds, overlap with bottom-longline fisheries decreased in Hauraki Gulf, and increased towards the shelf break. Overlap with inshore trawl fisheries decreased, while overlap with bluenose bottom-longline fisheries (which are mainly on the shelf break) increased. When compared with observed captures, there appeared to be low overlap on the East Coast North Island, south of East Cape. There were no seabird abundance observations made by fisheries observers on bottom-longline vessels in this area. Both models (the lat-long model and the habitat model) gave similar overlaps for the main fisheries in the region.

Although there were limitations with the source data sets, the boosted regression tree models provided a method for estimating overlap between black petrel and commercial fisheries. This overlap will be used in a revision of the risk assessment to allow a better estimate of the capture of black petrel in poorly observed small-vessel trawl and longline fisheries.