Also in the Article


Likely encounters and fishing effort
This protocol is extracted from research article:
Global hot spots of transshipment of fish catch at sea
Sci Adv, Jul 25, 2018; DOI: 10.1126/sciadv.aat7159

Likely transshipment events (encounters) were detected using satellite and tower-based AIS data between 2012 and 2017, as described by (17). AIS was designed as a tool of maritime safety to avoid ship collisions. Transponders installed aboard vessels send position and vessel identification messages to receivers on other ships, land, and satellites every few seconds. These messages can be used to reconstruct vessel tracks with high precision and allowed us to analyze their activity on the basis of an automated analysis of movement patterns.

Likely encounters were identified by Global Fishing Watch as locations where two vessels remained within 500 m of each other for longer than 2 hours, traveling at less than 2 knots while at least 10 km from an anchorage (including ports). These parameters balance the need to detect vessel pairs in close proximity while recognizing our ability to identify long periods in which vessels are in immediate contact is limited by satellite coverage and inconsistent AIS transmission rates. Some vessels are known to transship within ports, but these events are more likely to be subject to surveillance, and therefore, we focused on events that do not occur within the vicinity of port and the accompanying oversight. Here, we used a subset of the data analyzed by (16), only including encounters where AIS data are available for both the reefer and the fishing vessel engaged in the encounter.

To exclude vessel meetings that occur within port, encounters were filtered to be more than 10 km from an anchorage (defined as docking in port or anchoring close by) by using a global anchorage data set developed by Global Fishing Watch and made publicly available at http://globalfishingwatch.org/datasets-and-code/anchorages/. Briefly, the anchorage data set was developed by applying an approximately 0.5-km grid to the globe using S2 grid cells (level 14) (http://s2geometry.io/). Using AIS messages from 2012 to 2016 from all vessel types, those grid cells where at least 20 vessels remained stationary for at least 48 hours were identified. For each grid cell, the mean location of the stationary periods was calculated, and this point was labeled as an anchorage. This method identified 102,974 anchorages, and the mean location of an encounter was required to be at least 10 km from any anchorage.

A maximum encounter duration of 3 days was chosen to exclude encounters too short to offload catch and encounters that significantly exceed expected catch offload durations. These events likely represent vessels meeting for other reasons, such as repairs. This upper bound resulted in the removal of 97 events, representing less than 1% of the identified encounters.

Fishing vessels, refrigerated cargo vessels, fish carriers, and fish tender vessels were identified using vessel lists from the International Telecommunications Union and major RFMO fleet registries. Additional vessels were identified by a vessel classification neural network developed by Global Fishing Watch to predict vessel types based on movement patterns. Vessels that were identified as likely reefers by this neural network were manually reviewed through web searches and national, as well as RFMO registries. We do not expect that this list includes all vessels capable of receiving catch at sea, but it likely includes a majority of large specialized reefers that transport fishing for much of the offshore fishing fleet. Of the 641 refrigerated vessels identified in this manner (17), 501 were involved in likely transshipment events with AIS-tracked fishing vessels.

Fishing vessels included in this study were cross-checked for gear types through web searches using fleet registries and other reliable sources such as fishing company websites. To estimate the amount of catch landed directly by a fishing vessel versus catch brought to port via a reefer, we identified encounters and port/anchorage visits longer than 24 hours for each fishing vessel. For this analysis, a vessel was not considered to have “visited” a port or anchorage if it did not remain for longer than 24 hours to avoid assigning fishing effort to a port where a vessel was not present long enough to offload significant catch. For reefers, we identified the port visited following an encounter and the hours of fishing per fishing vessel that took place between events (the hours of fishing since the previous encounter or port visit). The fishing that preceded a port visit was assumed to have been landed in that port. Fishing hours that preceded an encounter were assumed to have been transferred from the fishing vessel to the reefer and offloaded in the next port that the reefer visited. The total fishing hours were aggregated by gear and attributed accordingly to ports (Russia considered separately from Asia and Europe).

Fishing activity and vessel gear type were classified following the methods described by (15). Briefly, two convolutional neural networks were trained on data from fleet registries, logbooks, and data labeled by experts to identify vessel types and classify their behavior (transiting and fishing) based on movement characteristics as seen in the AIS data.

Note: The content above has been extracted from a research article, so it may not display correctly.



Also in the Article

Q&A
Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.



We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.