“We’ve received another big shipment!” a dock official’s unkempt voice barked over the din of the shore port. Shouted orders, the deafening sound of flies, the roar of boat horns and engines… the excitement was non-stop. Such was the life of law enforcement and customs officials whose job it was to protect endangered animals in both legal and illegal trade. Horns, scales, jaws, fins – the list of things to look out for and identify was seemingly endless!
More than two decades after the passage of the International Plan of Action for Conservation and Management of Sharks (IPOA-SHARKS) in 1999, the globalized market for shark products is largely unknown. However, one thing that always stands out is the shark fins. Depending on what species they come from, they can come at a hell of a price. And although many countries and regional fisheries organizations have banned finning (the practice of removing a shark’s fins and discarding its carcass at sea) since 2004, shark products can still be sold…both legally and illegally.
Legally, the fins and snout are separated when the whole animal is landed. Because of this, there is a constant need to identify these shark products along the supply chain, particularly when they are listed in the Appendices of the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES). Identification of fins in fisheries landings before they enter international trade is a key constraint for CITES compliance. And although it seems like it has a simple solution – “Just identify the damn thing!” as one anonymous commenter joked on Twitter – identifying shark fin species is complex, especially in a global context.
“Being able to identify fins of species listed in the CITES Appendices, if they have already been taken at the landing sites, is a necessary first step to establish the sustainability and legality of the catch, which is necessary to to provide the certification required for their shipment and full traceability,” explained Monica Barone, adviser on fisheries resources and shark specialist at the Food and Agriculture Organization of the United Nations (FAO). Carlotta Mazzoldi, Professor at the Department of Biology at the University of Padua, agrees: “Any conservation action requires a thorough understanding of the actual threat to the species. Inclusion of shark species in CITES does not ensure effective enforcement of these regulations unless effective controls are implemented.”
Official CITES statistics on import and export of shark goods largely understate species and quantities traded based on catch data and trade content analysis. Enter iSharkFin. FAO and the University of Vigo began collaborating in 2014 to develop the system, which is based on machine learning technology that can learn from various data inputs. “In the case of iSharkFin, the algorithm learns by adding sets of fins with new morphological features, thereby increasing detection performance. iSharkFin V1.0 software, introduced in 2015, was designed as a user interface to both predict an identification for an unknown dorsal fin based on current inputs and to generate the new data inputs to further train the learning algorithm. Successive versions have been released with refinements, the latest (V1.4) allowing users to select qualitative variables such as fin tip color to refine subsets of morphometric variables,” illustrate the authors of this new study.
Barone and her team reported on the current power of the iSharkFin system, which aims to allow users to identify the species of wet shark dorsal fin from its image. Using photographs of over a thousand wet dorsal fins from 39 shark species collected in 12 countries, they trained the algorithm over a four-year period. “CITES enforcement tools like iSharkFin need to achieve a high level of reliability. This is important for the credibility of the tool and to convince frontline customs officials and others around the world to work with it,” said co-author Frederik Mollen, Authorized CITES Expert on Sharks and Rays. “For us as developers of the tool, we wanted to make its performance transparent. It also gave us the opportunity to review the strengths and weaknesses of subsequent versions of the tool and make it even more powerful in the future.”
The hope is that with continuous training, the iSharkFin system will be able to correctly predict new cases. As the researchers fed the program more and more images, the accuracy iSharkFin showed varied but increased, reaching 85.3% and 59.1% at the genus and species levels, respectively. The accuracy in predicting CITES listed sharks compared to unlisted sharks was 94.0% based on the 39 species currently represented in the baseline.
Promising… but iSharkFin won’t be enough to identify shark fin traded legally and illegally. So authors and shark fin identification experts Lindsay Marshall and Jenny Giles say, “It’s a problem that needs to be addressed from many angles and with many resources. Expertise is key. We were able to identify CITES species compared to non-CITES species within our data set with a good level of confidence. [For] CITES, this will be a handy screening tool when developed for specific settings. In terms of enforcement, samples must still undergo expert forensic identification using morphology or DNA.”
While this latest study is quite technical and intended for those who work on wet dorsal fin fisheries landings, the authors hope it can raise public awareness of the origins of the food they eat. “I think this paper can further raise public awareness about shark conservation and our responsibility as consumers. Our choices can support and even drive conservation efforts or, on the contrary, encourage trade in endangered species,” concludes Mazzoldi.