By Dr. Leigh Torres, Director of the GEMM Lab
In our modern world we often share space with people, but never really interact with them. Like right now, I am on a train in France with a bunch of people but I’m not interacting with any of them (maybe because I don’t speak French…). This situation extends to our efforts to understand the bycatch of marine predators in fisheries.
Productivity in the ocean is patchy, so both fishing vessels and marine predators, like seabirds and dolphins, may target the same areas to get their prey. This scenario can be considered spatial overlap, but not necessarily interaction because the two entities (predator and vessel) can independently chose to be in the same place at the same time. Also, overlap can happen at larger spatial and temporal scales than interaction events, which typically must occur at small scales. Again, consider me on this train: all my fellow passengers and I are overlapping on a 500 m long train for 2.5 hours (larger scale) but I only interact with the passenger in the seat 1 m across from me for a minute (smaller scale) while I explain that I don’t understand what they are saying.
Distinguishing overlap from interaction between seabirds and fishing vessels is important to help managers determine how to best direct their efforts to reduce bycatch. Different management approaches can be applied depending on whether seabirds are using the same habitat as fishing vessels (overlap) or are attracted to vessels for feeding opportunities (interaction) and then incidentally caught/injured in the fishing gear. Furthermore, if we can describe discrete interaction events we may also be able to identify the individual fishing vessel, fishing gear used, country of origin, and other such specific information that can help direct bycatch reduction efforts.
However, studying the spatial and temporal relationships between seabirds and fishing vessels is challenging, and highly dependent on the quality of data we have, or can collect, about the movements of birds and boats at-sea. Tracking the movements of seabirds has evolved rapidly with the development of tagging technology and miniaturization, so that over the past 10 years seabird ecologists have collected a large amount of high-resolution GPS data of seabird foraging. While these data reveal fascinating patterns of seabird ecology, our ability to relate these seabird distribution data to fishing vessels has remained limited due to limited access to fishing vessel location data. Historically, fishermen have not wanted to divulge their fishing locations for fear of losing their ‘secret sweet spot’ or regulatory infractions. So, where fishing vessels fish has often been a mystery, at least fine scales. For a long time fishing effort data was only released at scales of 5 x 5 degree grid cells and monthly scales (Fig. 1) (Phillips et al. 2006), which is only broadly useful for assessment of overlap and not useful for assessing interaction events. The situation has improved in some countries where Vessel Monitoring Systems (VMS) data are available but even these GPS data are often too coarse to reveal interaction events (although it’s much better than what was previously available!). In fact, I wrote a paper about this topic in 2013 called “Scaling down the analysis of seabird-fisheries analysis” that called for higher resolution vessel position data to better evaluate and manage seabird and fishing vessel interactions (Torres et al. 2013).
Progress was made in 2016 with the release of Global Fishing Watch (globalfishingwatch.org) that has significantly increased transparency in the fishing industry and revolutionized our ability to monitor fishing vessel activities (Robards et al. 2016). Almost every fishing vessel in the world is required to use the Automated Identification System (AIS) that pings GPS quality position data to satellite and shore receiving stations around the world. AIS was originally developed to increase maritime safety by reducing collision risk, but Global Fishing Watch has developed methods to acquire these AIS data globally, distinguish fishing vessels (from cargo ships or sailing vessels), classify fishing vessels by fishing method, and disseminate these data in an accessible and visually understandable able format (de Souza et al. 2016; Kroodsma et al. 2018). When I saw the Global Fishing Watch website for the first time I actually let out a ‘Woohoo!’ because I knew this was the missing piece I needed to move from overlap to interaction.
So, I assembled a great team of collaborators including Dr. Rachael Orben – seabird movement ecologist extraordinaire – and colleagues who have collected GPS tracking data from three species of albatross in the North Pacific Ocean. Another important step was acquiring funding to support the research effort from the NOAA Bycatch Reduction Engineering Program, and to establish a collaboration with Global Fishing Watch. Fast forward a year and through many data analysis and R coding puzzles, and we have made the jump from overlap to interaction, with some preliminary results to share.
We compiled GPS tracks representing foraging trips conducted by Laysan (Phoebastria immutabilis) and black-footed (P. nigripes) albatrosses breeding in the Hawaiian islands, and juvenile short-tailed albatross (P. albatrus) from Japan. First we identified overlap between bird and boat at daily and 80 km scales. Next, we quantified encounter events at scales of 10 minutes and between 30 and 3 km, which was the assumed distance at which birds are able to perceive a boat. Finally, interaction events were identified when birds and boats were within 3 km and 10 minutes of each other.
At an absolute level, short-tailed albatross overlapped, encountered and interacted with many more fishing vessels than black-footed and Laysan albatross. However, it is important to point out that these results may be biased by the temporal sampling resolution of the GPS tracking data (how often a location was recorded), which we have not accounted for yet. Nevertheless, what is interesting is that when the percent of interaction events that derived from encounter events is assessed, black-footed and Laysan albatross demonstrate much higher rates of fisheries interactions. These results indicate that when a black-footed albatross encountered a fishing vessel engaged in fishing, nearly 50% of these opportunities turned into an interaction event. This rate was 39 and 26 percent for Laysan and short-tailed albatross respectively. This species-level difference between absolute and relative (percentage) interaction with fisheries may be due to the overall distribution patterns of the different albatross species, with short-tailed albatross using areas that overlap with fishing activity more frequently (coastal margins). Furthermore, these results indicate that short-tailed albatross may be more ‘vessel-shy’ than black-footed and Laysan albatross. The high black-footed albatross percent interaction rate aligns with the high by-catch rate of this species, and emphasizes the need to better understand and manage their interactions with fishing vessels.
While these results from our novel analysis are an interesting start to helping inform bycatch mitigation efforts, perhaps the most illustrative (and coolest!) output so far are the below animations that show the fine-scale movement tracks of an albatross and fishing vessel (Fig. 2 and 3). Both animations are a 24 hour period and show an albatross (red dot) and a fishing vessel (yellow dot). But, Figure 2 illustrates an overlap event, where the bird and boat clearly overlap spatially and temporally but do not interact. However, in Figure 3 we see how the albatross flies directly to the vessel and the bird and vessel remain spatially and temporally linked, demonstrating an interaction event. Our next steps are to improve our ability to distinguish these interaction events (assessment of duration and trajectory correspondence) and to describe the driving factors (e.g., albatross species, fishing vessel method and flag nation, environmental variables) that lead an albatross to move from overlap to interaction.
Figure 2. Fine-scale animation of overlap between the movement path of a Laysan albatross GPS track and the AIS track of a fishing vessel, overlaid on bathymetry. While the bird and boat overlap at this scale, the animation illustrates how the bird and boat do not interact with each other.
Figure 3. Fine-scale animation of overlap between the movement path of a Laysan albatross GPS track and the AIS track of a fishing vessel, overlaid on bathymetry. This animation illustrates how the bird and boat act independently at the start, and then the bird travels directly to the vessel’s location and the movements of the two entities corresponded spatially and temporally, demonstrating a clear interaction event.
de Souza, Erico N., Kristina Boerder, Stan Matwin, and Boris Worm. 2016. ‘Improving Fishing Pattern Detection from Satellite AIS Using Data Mining and Machine Learning’, PLoS ONE, 11: e0158248.
Kroodsma, David A., Juan Mayorga, Timothy Hochberg, Nathan A. Miller, Kristina Boerder, Francesco Ferretti, Alex Wilson, Bjorn Bergman, Timothy D. White, Barbara A. Block, Paul Woods, Brian Sullivan, Christopher Costello, and Boris Worm. 2018. ‘Tracking the global footprint of fisheries’, Science, 359: 904-08.
Phillips, R. A., J. R. D. Silk, J. P. Croxall, and V. Afanasyev. 2006. ‘Year-round distribution of white-chinned petrels from South Georgia: Relationships with oceanography and fisheries’, Biological Conservation, 129: 336-47.
Robards, MD, GK Silber, JD Adams, J Arroyo, D Lorenzini, K Schwehr, and J Amos. 2016. ‘Conservation science and policy applications of the marine vessel Automatic Identification System (AIS)—a review’, Bulletin of Marine Science, 92: 75-103.
Torres, Leigh G., P. M. Sagar, D. R. Thompson, and R. A. Phillips. 2013. ‘Scaling-down the analysis of seabird-fishery interactions’, Marine Ecology Progress Series, 473.