Fighting overfishing with AI: Satellites Expose Illegal Fishing in Marine Protected Areas

0:00 illegal fishing vessels thought that they can hide, but satellites and AR are catching them in the act in marine protected areas. Today I'll show you how this new technology is turning the tide for ocean conservation. When I read a recent study, I was shocked to see how many fishing vessels were detected inside marine protected areas, especially the more remote ones. These are places we imagine as untouched. But in reality, they're often the hardest to enforce and the easiest for illegal fishers to exploit. That hit me personally because MPAs are supposed to be our ocean's havens. but here's the good news. Satellites and AI are changing the game. We're finally catching vessels that thought that they can hide, and enforcement is starting to get smarter and stronger.

0:45 In this episode of the How to Protect the Ocean Podcast, I'm gonna share real world examples of how these tools are already protecting the ocean and why it gives me hope for the future. So let's start the show. Hey everybody. Welcome back to another exciting episode of the How to Protect the Ocean Podcast. I'm your host, Andrew Lewin, and this is the podcast where you find out what's happening with the ocean, how you can speak up for the ocean, and what you can do to live for a better ocean by taking action. Today we're gonna be talking about a lot of different things, all combined into one terrific story. We got AI, we got satellite imagery, we got synthetic aperture radar, which is like a really cool satellite. We've got machine learning. We got AIS, which is automatic identification system.

1:27 we got marine protected areas. We got illegal fishing, fishing boats. There's all in one enforcement. It's a very exciting story, all packed in one research paper in the research journal science. Now we're gonna be going a little beyond that because this research paper showed an increase in illegal fishing activity in marine protected areas, especially those that were further away from the coast, so more remote, and they spent longer within those time, four hours. Per vessel in marine protectors per year. Not a really great story in terms of optimism, but then I wanted to go a little bit further and I found some examples of opportunities where satellites, AI and machine learning were used to detect an enforce, of course, and prevent further illegal fishing in marine protected areas.

2:12 So we're gonna talk all about that on today's episode, which I can't wait to do that. But first, before we get into that, if you are interested in marine stuff, you wanna learn about the ocean, you wanna learn more about these conservation stories, you can get more information each and every weekday with our newsletter. Just go to speak up for That's speak up for blue.com, all one word slash newsletter to get access to our newsletter and get information on ocean conservation in your inbox five days a week. It has all of our podcast stuff, news articles that are coming out, something that's very interesting each and every day, check it out. Speak up for blue.com/newsletter. Let's get back into the story.

2:52 This is really interesting. So this is how they use it. They use Synthetic Aperture, Aperture Radar SAR, satellite imagery, combined with AIS data, which is Automatic Identification System and AI and machine learning to detect industrial fishing vessels. So essentially what they did is they took the satellite imagery, which is really cool. It's not like your regular satellite. It basically takes the surface of the water and makes it really like an impenetrable marker. it looks like concrete and then it takes fishing vessels and makes them dark. So you can actually detect fishing vessels each and every way. It's really cool. They use it to monitor a lot, anything at the surface, they know where the boundaries of the marine protected areas are. They can find out where these fishing vessels are, so they overlay a map or a polygon of where these boundaries are.

3:34 'cause they're invisible normally. We don't see an actual physical barrier in the ocean that goes down the water column to prevent people from coming in and fishing in marine protected areas. Now, before we go further, marine protected areas are areas where they protect any kind of extractive activity, including illegal fishing. It's usually to protect fish, allows fish to grow bigger, allows fish to grow more in terms of number and bigger so that they spill over the boundaries of the marine protected areas. And they get fished by fishermen legally, but they're not allowed. The fishermen are not allowed inside the boundaries of the marine protected areas because that is where they're allowed to grow. They're special areas, and so we need to protect them. These fishing vessels will come in because you know why?

4:16 Nobody's there to watch them. Because the ocean is massive and you can't just keep, you know, boats there all the time. It's expensive. It's almost impossible to do that. So we have satellite imagery like sar, right? the satellite imagery allows, this picture to be taken at some point in time. To find out if there's any illegal boats that are in there. AIS data, Automatic Identification System, is placed on some of these boats that allows it and detects the GPS coordinates of where they are at any point in time. Well, within a certain interval. I think sometimes it's seconds. I've worked with this data before. You could really see, like if you look at worldwide and how this AIS has done for those boats, which are still a lot. Those boats that use it, you can actually see tracks all over the world and where these boats have gone, what flags they're flying.

5:00 It's really cool data. Hard to get ahold of. It's usually managed and kept by the Coast Guard, but I've worked with it when I worked in government. It's really cool data to have. So you have these boats. So you know where they are. We know which boats are where, and then we have the satellite imagery to detect which boats are within the marine protected areas. And then you can find out basically on speed. So if they're trolling, you can see how they slow down. All this different type of stuff, you can find out all this with this data, and of course, AI and machine learning can quicken that processing speed so you can catch 'em almost in the act. So if they're coming to a border, you know where they're going. You can like basically deploy Coast Guard vessels or whatever the vessels are for each country to get out and to enforce this illegal fishing activity to prevent it from happening.

5:41 So the faster the authorities can get on it, the better. So AI and machine learning help it. So you got satellites and AIS to detect the illegal activity. And you have AIS and machine learning to deploy the enforcement, to go out and catch them. So that's what the story was essentially. They focused on coastal marine protected areas globally between 2022 and 2024. Only coastal MPAs that are larger than one kilometer squared were considered since Ramona MPAs often lack consistent satellite data coverage. Okay. SAR satellite data coverage. So prevalance of fishing in MPAs, they detected industrial fishing happened in 47% of coastal MPAs worldwide during the study period. So that's only within two years.

6:21 So many of the vessels detected 67% were untracked by public monitoring system. So no AIS so the satellites can detect all the boats that are there. The AIS can identify which boats have AIS and which boats don't. If you don't have AIS and you just see a boat there. Why aren't we being able to see this? This is consistent with other studies that show 75% of the fishing vessels around the world do not have AIS, they are untracked. This is very scary in terms of what is happening with fishing and illegal fishing because we can't track these boats. More effort needs to be done to track every single fishing vessel that is around anywhere. So having untracked vessels, not good. So surprisingly, the most restrictive MPAs where extractive activity is supposed to be prohibited at all costs, still had a lot of untracked vessels.

7:08 About 80% of detections in those MPAs were on track. So that makes sense because they're not supposed to be there. So why would you have a tracked vessel in there? So that makes sense but still pretty scary. So the effectiveness or the lack thereof was not just tied to how restrictive the MPA is, like legally from by IUCN standards, but more strongly to remoteness and size of the MPA. So large and remote MPAs had more phishing vessel detections. Now, what's interesting to me is I wonder if we actually talk to these fishers if they even care about the boundaries or they are just going where the fish are now. Ideally in the marine protective areas because they work so well, you will see bigger and more fish. And so that's why they're probably going within those areas.

7:48 so the average fishing vessel inside the MPAs was around four hours per square kilometer per year when both tracked and untracked vessels were counted. That's a lot of time. I know that sounds weird, but that's a lot of time thinking about how often these fishing vessels go into these areas where there's marine protected areas. So that is something that is not good, right? That's a lot of time in these marine protected areas. So the density and presence of fishing vessels, so tracked and untraced in MPAs correlate more strongly with MPA size and remoteness rather than the formal protection category. So it's not as if it's just, hey, it's protection, it's actually this size and the remoteness of the MPA. So here's the thing. We do have gaps in monitoring, right?

8:29 That is one of the limitations that this study identifies. So public tracking systems, AIS miss a large portion of the phishing activity. Untracked vessels make up a large and major invisible pressure. That's something that we can't see through AIS. If you can't see the data on a map, it's really difficult, even though there's so much data, there's so many more vessels that are go on tracked. We also have to worry about paper parks. It's a risk when MPAs are designated, which takes a long time to designate them when they're designated, they need to be enforced. If they're not enforced, they're just considered paper parks, meaning that they're on paper. We see them on maps, but we don't actually enforce them. And so without the lack of sufficient enforcement and monitoring, we may see considerable phishing in these areas, which kind of nullifies the point of the MPA.

9:14 We also need to integrate OnTrack surveillance. So using things like SAR, like the satellite machine learning and combining multiple data sources improves understanding of true phishing pressor. So a lot of the times when we do these studies, we have to understand how much pressure there is in phishing within these MPAs or just around the world. When you don't have the ability to identify who's fishing and who's not, or what's tracked and what's not tracked like what fishing vessels aren't tracked, we lose the ability to understand the true phishing pressure. So if we only look at tracked vessels, we only get a snapshot of what's actually happening. It's only a small representation when we have more vessels that are untracked, this can't be done.

9:54 This is up to the fishing fleets or even the countries that allow fishing vessels to fly their flags. That's a whole thing. If you look at my recent interview with Dan Skerritt, you'll understand what these flags do and how easy it is to manipulate these flags and to fly under a different flag. To be able to do pretty much whatever you want. When these countries are not enforcing any of their laws, or if they even have laws, they may not have as many laws. So in integrating this information, surveillance, so SAR machine learning, AI, combining multiple data sources improves the understanding of true phishing pressures. Then you have model predictions. So because of limitations, not all vessels are greater than 15 meters. Variable satellite coverage, which is hard to get total satellite coverage whenever you want because they're in orbit and they only fly around certain areas once or twice a day.

10:43 Distinguishing transit versus fishing vessels. So some are just in transit, some are actually actively fishing. And then the metrics that are estimates, actual fishing effort in some marine protectors may be higher, especially in places with weaker AIS coverage. So the fact that there's so many variables, we're not there physically all the time, these enforcement people, and we are seeing it from like, you know, orbit and we're seeing things like from a digital landscape, like on a GIS or a mapping tool and only seeing data points and maybe missing data points from untracked vessels. There are a lot of underestimations that are happening right now, and we have to assume that there's more fishing pressure around the world, especially within these marine protected areas. So now that's the bad news.

11:24 I know this a lot and I appreciate you. You hand me the good news is coming. There are a lot of things that we can do and there are a lot of things that are being done, and I think that's something really interesting. Detecting dark vessels with SAR and AI. So here's how it works. SAR satellites can detect ships. I just kind of mentioned it, regardless of the weather or lightning, so that's good. AI models match SAR detections with AIS data to flag quote unquote dark vessels. Those that are not transmitting AIS. So, like I said before, the satellite imagery can actually detect what's in the marine protected area boundary, even though the AIS, like the DOT doesn't come up on a map. So an example of this is a 2025 study found that 67% of vessels inside the MPAs were untracked.

12:05 This is the study that we just talked about. So now that we can detect untracked vessels, we're gonna be able to understand how many of these untracked vessels are gonna be there and maybe be able to deploy these vessels through like AI prediction models or machine learning prediction models. So authorities can prioritize patrols towards areas with high track vessel activity, closing a major enforcement gap. That's something that's really important. So here's another thing that they can do. Guiding patrols with real-time alerts. So AI driven analytics turn raw satellite detection into real-time alerts for coast guards and fisheries officers. So an example is the country. Gavin partnered with NGOs and Global Fishing Watch to use satellite data to direct naval patrols leading to the arrest of industrial trawlers operating illegally in MPAs.

12:49 So instead of patrolling blindly, authorities can intercept in real time saving fuel and resources. That's a big, big win and a great example. Then you have the legal and court backed evidence. So satellite imagery combined with vessel tracks creates verifiable evidence admissible in legal or policy disputes. So an example, in Europe, court rulings upheld trawl bans and MPA strengthened by satellite based evidence showing incursions. So we see the actual stuff in court saying, Hey look, we may not have caught these vessels right away, but we know this is happening. We know the vessels. We're able to track them through AIS and satellites. Now we can give them the business, right? We can give them those court decisions where they're not gonna be very happy and they're gonna be fine.

13:32 Now, those court decisions can be tougher if we put the right laws in place, and that's another issue, but that's something that we need to worry about in the future. So that provides a lot of like a black box record that regulators and cords can use to uphold restrictions. Now you've got persistent oversight of remote MPAs using that satellite imagery. and AI large scale surveillance. Remote MPAs are too vast for constant petroleum. We can't do it, but the UK Blue Belt program uses daily satellite imagery or monitoring for Tristan Una's, I think that's how you pronounce it, remote MPA to detect potential illegal activity with alerts directing patrols to areas to be avoided. It prevents remote MPAs becoming paper parks by providing consistent low cost oversight.

14:16 So again, directing patrols through quick processing of this data to say, Hey, we actually found something when we're able to, and we're gonna direct these patrols to enforce. Over time, that's going to decrease what's happening. In Palau in 2025, using the combined satellite surveillance with long range drones to enforce its National Marine Sanctuary covering one of the world's largest no-take zones. So they use satellites to detect suspicious vessels, drones to verify and collect high resolution imagery and AI models confirming phishing activity, so the gear types, the patterns of behavior, and et cetera. So it creates a layered monitoring system, Increasing deterrence while reducing cost of ship based patrol.

14:58 So think about this again. It's not just blindly patrolling the water where you'll never be able to find this. If you ever played battleship, if you ever played the game battleship, before you hit your first boat, you're blindly trying to hit like a 10 or B this, or B five or something like that, and then you finally hit something. Well, that's doing it blindly. But imagine if you had somebody behind you to just kind of be like, Hey. You know what? A 10 is actually gonna miss. Let's go B five. You're gonna hit the big battleship, right? And it's having something there to direct you. And now I know for the game that's cheating and don't cheat in battleship. The person you're playing against is not gonna like that. But here we have to cheat a little bit to get into it. And just by using data, we can better inform our patrols and save on a lot of money for patrolling.

15:40 So then you got the promotion of transparency and deterrent. So publicly sharing satellite detected vessel activity, pressures fleets to comply. So just that public pressure allows us to say, Hey. This fishing vessel's awful. Where do they work? Oh, they work for such and such. That's who sells to such and such place. Don't buy from them. That public pressure, as we've seen, like in cancellation society that we live in today, that could help deter boats from doing such a thing. So Chile publishes industrial and artisanal fishing vessel tracks on global fishing watch and uses AI analysis to plan patrols. They just publicly just put it out there. So the transparency itself becomes a deterrent. Knowing your vessels, movements may be visible to NGOs, journalists, and the public reduces rule breaking.

16:25 So think about that. Just think about these things that are actually working in countries where we can do better. We've been doing better. It is a very difficult challenge to manage fisheries across the globe, especially in areas in the high seas where countries don't really have a lot of ability to enforce, but in their own boundaries, in the exclusive economic zone where they can enforce things, it becomes a bigger picture. And if you can use things like AI, you can use things like satellite imagery, AIS, any kind of GPS technology 'cause there's others than AIS, you can use, drones even to direct patrols to say, Hey, you know what? We found the vessel. Boom, let's go get 'em and let's enforce what we're trying to do and let's protect this marine protected area.

17:09 Truly and really decrease the amount of paper parks that we actually have. We know marine protected areas work. But we need to help them by enforcing them from extractive activities like illegal phishing, and we can do that using the new technology. It's being done. Although we can detect a lot of stuff, have bad stuff happening now we can really focus on remote areas, larger remote areas to say, Hey, we need to invest more of this technology or more money into this technology to detect things that are illegal activities such as phishing, and be able to enforce it better. That's the episode for today. I thought I would do a little bit of a deep dive. I know this is a longer episode than we normally do, but I thought this was really interesting and it's something that we need to continue to invest.

17:50 So what you can do specifically as an individual is support places like the global Phish watch, right? Support organizations like the Ocean Conservancy, Oceana, World Wildlife Fund. These are all organizations, large organizations, international organizations that have global partners, not only from a conservation side, but also on the fishing side because you know, there's a lot of fishing partners that want to make sure that fishing is sustainable and that they're following the rules and be able to stomp out illegal activity. Let's reward the fishers that are doing things legally and let's, you know, enforce the rules for fishers who are doing things illegally. You can do that by supporting these organizations who are out there actively doing this type of stuff.

18:33 So that's something I wanted to talk about today. I wanted you to know about more and more and more. We're gonna be covering these things in the future, so I'm really happy to do that. If you have any questions or comments, leave a comment If you're watching this on YouTube, in the comment section below. And of course, if you're listening to the audio version, which I love the fact that you're listening to the audio version, start off on audio. Always will do audio. Please get ahold of me by going to the website. Speak up for blue.com/contact. Just fill out the form or you can just hit me up on Instagram at How to Protect the Ocean. Just DM me. I would love to chat with you. I'd love to hear your questions and comments. And that's it for today's episode of the How to Protect the Ocean Podcast. I'm your host, Andrew Lewin from the True Nord Strong and Free. Have a great day. We'll talk to you next time and happy conservation.