 Hello and thank you for tuning in for this talk. I am Cindy Javizio and I work for the U.S. National Marine Fisheries Service Alaska Fisheries Science Center. I am presenting on a project that we're just starting. This presentation will be focused on the background and project goals. Please feel free to contact myself or any of the project collaborators with questions. Federal fisheries in Alaskan waters are managed in two fishery management plan areas for FMPs, the Gulf of Alaska, the Goa in green, and the Bering Sea and Aleutian Islands with BSAI in blue. There is also an Arctic FMP, but commercial fishing is currently prohibited in that area and no stock assessments are conducted as of yet. Since this is an international audience, I'll put it in perspective. The entirety of Alaska stretches the same distance as the rest of the United States, particular United States from the east coast all the way to California on the west coast. For the BSAI and Goa shark stock assessments, there are four species. There are the spiny dogfish, the specific sleeper shark, salmon shark, and a group, a catch-all group called the other unidentified sharks, which consists mostly of blue sharks. In the Goa spiny dogfish, the small shark species is the most prominent species, while in the BSAI, specific sleeper shark are the dominant species caught. There is no direct fishing for sharks in either of the FMPs, and none of the species have market value. Therefore, almost all are entirely discarded. All sharks are considered nuisance species. Thus, it is an extra layer of challenge to promote improving the stock assessments for these highly undesirable species. Both assessments are considered data limited, and for most of the species, severely data limited. With the exception of spiny dogfish in the Goa, all harvest specifications are based on either the maximum or the mean historical catch. Unfortunately, there are some significant concerns regarding the accuracy of the catch. Unobserved and undocumented catch of sharks occurs in some state of Alaska-managed commercial fisheries, such as the gillnet fisheries for salmon. This picture is an example of how a gillnetter can have an entire net swamped by spiny dogfish, and due to safety concerns must abandon the net. These sources of removals are not important. Another source of unobserved catch for IFQ fisheries. Prior to 2013, the commercial IFQ fisheries for Pacific halibut and staple fish were exempted from observer coverage, as well as any vessels less than 60 feet for about 18 meters long. The observer program was restructured in 2013 to include the IFQ fleet as well as vessels down to 40 feet or 12 meters in length. The problem in observed fisheries is that observers may not have access to large species. This challenge applies to marine mammals, as well as sharks, and sometimes even birds. For the remainder of this talk, we will focus on the large shark species, Pacific super sharks, salmon sharks, and then the others caught within Alaska ground fish fisheries, and I'll simply be calling them just the sharks. The first challenge for observers trying to gain access to sharks in long run fisheries is getting them on board. Most long line vessels are not large enough to safely bring a large shark on board. Even large ships, as in this picture, cannot reasonably bring large sharks aboard without stopping hauling gear, rigging the crane, and devoting crew and man time to managing the shark. When sharks are brought on board, the observer is limited by their equipment, and observers only supply the 50 kilogram scale. If an observer has the time, they can section the animal and weight parts to get to some total weight, but that is nearly impossible given their other duties. They can take a length measurement and convert that to a weight, which is what is often done in trawl fisheries. In long line fisheries, that approach is nearly impossible due to limited deck space. Therefore, weights of sharks are often either not reported or limited to small sharks and result in an underestimate. In trawl fisheries, depending on the size of the vessel and the harvest sector, being either a catcher vessel or a catcher processor, the observer may have access to the shark and the ability to take measurements, which can be converted to weight. But in the case of catcher vessels, which deliver shore side, sharks are often discarded without an observer getting access to them. Electronic monitoring programs, or EM, can address the data gaps in both the long line fisheries, being the weight of the catch, and in the trawl fisheries, being identifying the presence of catch and getting data from it. There are currently two EM programs in the federal ground fisheries within Alaska, a fixed gear program, which is long line and caught fishing gear, and an experimental program on trawl catcher vessels. The fixed gear program began in 2018 and is designed for catch estimation in numbers of discarded species. The trawl EM program is different in that it is currently experimental, designed for compliance monitoring. This program launched in 2020 and is planned through 2023, with the expectation of it becoming a regulated program in 2024. All catch is required to be retained for shore side observers to sample, exception is that sharks are allowed to be discarded at sea. The catch of sharks is minimal in pot fisheries, and therefore for the rest of the talk, we will focus on long line portion of the fixed gear EM program. Vessels carrying EM systems are randomly selected to have their cameras on during a trip. Once they return and submit the hard drives, a random subset of 30% of the halls are reviewed in a complete census of all catches tally. There is no biological sampling on these vessels, nor on the catch when it is delivered shore side. Catch weights are calculated from, sorry, the catch weights of discarded fish are calculated from an average weight value borrowed from a nearest neighbor search pattern within the catch database. However, there is a video record, which is what prompted the development of this project. In the troll EM program, all halls on all trips are reviewed. However, only a random subset of the trips have biological samples taken when the vessels offload shore side. The video review process only notes discards or violations to the program goals. All sharks are to be measured by the crew and recorded in the vessel log book. Okay, now for the reason that I'm here talking to you today. EM Innovations has been going on in Alaskan fisheries since about the mid-20th, as the fixed gear EM program was in the development phase. Suzanne Romain will be giving a talk later about some of the tools they have already been working on. For this project, we wanted to utilize the great experience and knowledge of the existing EM Innovation team to develop tools to meet the unique needs of sharks. This project was originally intended to be a long line EM project, but we are expanding the project goals to integrate troll EM videos as well now. The project is set up around four primary goals. The first being to verify that EM systems see the same sharks that are being caught. To achieve this, we've become a voluntary shark reporting program. This is to report their shark catches to our Google form or the email to receive some fantastic logo wear. Number two is to correctly identify the species of sharks. This is actually well underway as shown in this image. The shark head is barely in the frame and hardly distinguishable from the surrounding water, but the ID system, the system can ID it. Measuring and assigning the shark to a size category is number three. We have observers currently doing this and labeling sharks as small, medium, or large based on estimated measures of whatever part of the body that they can see. Observers are trained to estimate sizes of fish based on known size items they can also see, such as the length of the gap. For sharks, they use this technique to estimate a size for one or more of the cues. Let's do the table here and convert that size to a category. For this project, I plan to add a fourth size category called Gigantor because machine learning should be able to have a higher precision than an observer 20 feet away from the roller bar. This work will build off of that published recently by Matt, my atoll in 2021 in the journal Computer Science, which Suzanne will be chatting about. The last component of this project is to integrate this into production for use in stock assessment. These are completely new data types and well outside the normal data streams that feed the total catch estimation process, something that I like to call the beast. And as any programmer knows, once you mark around with the data streams, the beast get grumpy. So the final objective is to turn these new data streams into inputs for the total catch accounting process and outputs for the stock assessments without breaking it for everyone else. For progress today. Well, we began this program and voluntary reporting program back in March of 2021 when the law history season began. We've had quite a few submissions already and had good feedback from industry participants. Unfortunately, the weather isn't always as nice as in these mixtures. Images and video clips for EM systems are slowly coming in. We are limited by the number of vessels that have agreed to allow us to use their videos, but so far so good. Clips are being annotated, which is a quite laborious process. Project staff are already working with the University of Washington collaborators of the EM innovation team to develop the algorithms for this project. All in all, this project is farther ahead than I expected it to be at this point. Well, this is a long line specific project. We are developing tools in conjunction with the troll EM program as well. The program programs are likely to expand to other fisheries within Alaskan waters and our project will result in tools that will be adaptable to the growing EM field. We're taking a unique approach to measuring large animals, at least within Alaska. The size data is almost entirely lacking to inform hedge estimates of sharks and long line fisheries and gaps in access to sharks has included sampling and others. Using size categories allows for measurement error, while still providing some size scale to the data. And then EM allows for data from fisheries where data were previously lacking. The EM innovation team is developing many great tools and check out Suzanne's talk later. The author list is any indication. This project is collaborative with many entities. The special thanks to all the vessels and industry representatives who have provided images and or given us permission to use video clips. All images shown here are either Noah or used with permission. Please do not distribute or copy without notifying the author. This project is funded by the Noah National Catch Shares Program.