Fisheries Human impacts

Tracking Global Fisheries with the Help of Computer Neural Networks

Paper:  Kroodsma, D. A., Mayorga, J., Hochberg, T., Miller, N. A., Boerder, K., Ferretti, F., Wilson, A., Bergman, B., White, T., Block, B. A., Woods, P., Sullivan, B., Costello, C.,  Worm, B. (2018). Tracking the global footprint of fisheries. Science, 908(February), 904–908.

It is no surprise that we are pretty big fans of the ocean here at Oceanbites. Oceans cover over 70% of our planet, and are widely important for global economies. A report conducted by the World Wildlife Fund even estimated that our oceans are worth $24 trillion annually (yes, trillion with a “t”)! One of the reasons for this high price tag comes from our reliance on marine fisheries, which is estimated to be the primary protein source for about 3 billion people  – that is just about 20% of the global population.

Billions of people rely on fish like the ones at this market as a source of food. Image from: maxpixel.com 

While scientists are aware that fishing occurs on a large and widespread scale, it has been difficult to directly quantify just how far-reaching the industry is, how much effort is put into obtaining the fish we see in markets, and what drives the fishing effort on a global scale. With the advent of new satellite technology, scientists are beginning to find answers to some of these important questions and uncover drivers in global fisheries.

In a recent paper published in Science, a team of researchers explored the global tracks of fishing vessels from 2012 – 2016 to gain insight into worldwide fishing practices. The scientists made use of data generated by Automatic Identification Systems (AIS), which are required on many large, commercially operated, and internationally operating vessels worldwide. AIS report the identity and location of ships, assisting with vessel navigation, traffic control, and collision prevention.  Because AIS broadcasts vessels’ position, speed, and direction (i.e. if the ship is turning) to transponders on land or to satellites, the scientists were able to track the movement and nature of trips made by all ships using AIS. There are a lot of vessels traveling through the ocean at any given time. The researchers had to parse through 22 billion AIS positions from a range of vessels (ranging from cruise ships to cargo ships).

With this large dataset, the researchers had to find a way to pull out only the information on fishing vessels, which is not necessarily obvious just from the “identity” reported by AIS.

You can track fishing vessels around the globe at globalfishingwatch.org . Bright blue dots show the location of fishing vessels.

To do so, the researchers relied on the help of two different convolutional neural networks (CNNs), a computer system that is able to identify images or maps in a way similar to the human brain. The researchers trained one of the CNNs to identify different types of vessels based on images. Much like training a child to identify a dog vs. a cat, the researchers took images of 45,441 different vessels (both fishing and non-fishing vessels) and trained the CNN to recognize a fishing vessel. The CNN was able to discriminate vessels enough to classify them into six different categories of either non-fishing or fishing vessels respectively. The CNN was 95% effective at identifying vessels and could even predict the size, length, engine power, and weight of each vessel.

Researchers then looked at the AIS positions and tracks taken by fishing vessels to predict the type of fishing that was taking place on that vessel. For example, a long-line fishing vessel that is setting out lengths of baited hooks over several hours and returning to haul in their catch moves at a different pace and fashion and requires a different type of fishing vessel than a trawling vessel that tows a large net off the back of the boat. Much like using a signature to identify people, researchers used these known differences in vessel type and movement to train the CNN to spot different vessel tracks and identify the type of fishing that was occurring on the vessel with a greater than 90% efficacy.

The Global Fishing Footprint in Numbers

Using the data from the CNNs, the researchers were then able to analyze trends in fishing over the 4 year time frame. In the course of a single year, the researchers found that fishing activity around the globe equaled a whopping 40 million hours and that fishing vessels traveled over 460 million kilometers!  To put that in perspective, you could travel to the moon and back 600 times or to Mars and back in that same distance.

According to their data, the researchers estimate that fishing is occurring in over 55% of our oceans (about 77 million square miles), which is a far greater footprint than is taken up by the approximately 34% (around 19 million square miles) of land used by agriculture. Most of this fishing footprint (about 85%) was made up by fishing vessels from just five countries: China, Spain, Taiwan, Japan, and South Korea. The most common fishing practices were longline fishing, purse seining, and trawling.

The Economic and Cultural Drivers of Fisheries

Just as the productivity of different crops varies seasonally, the productivity of oceanic ecosystems is seasonal. Yet, the researchers found little evidence of fishing practices changing in response to the natural cycles of the fish they are chasing.

Top: Amount of non-fishing (left) and fishing (right) activity in 2016. Darker colors indicate increased fishing. Bottom: Fishing activity for non-Chinese fishing vessels between 2013 – 2017. Declines in fishing activity coincide with Christmas each year. Images from Kroodsma et al., 2018 – supplementary information.

The researchers compared global fishing effort to the net primary productivity (the abundance of primary food sources for fishes and a measure of ocean productivity), changes in climate (such as weather patterns driven by El Niño, which is known to impact fish stocks), and changes in fuel cost. None of these natural or economic features drastically influenced the level of fishing globally. Instead, the researchers found that fisheries appear to operate more like commercial businesses with little weekly or seasonal change in activity. In fact, the largest determinant of fishing effort was a more “cultural” driver: holidays. A visible drop in fishing activity occurs in Chinese fisheries around the Chinese New Year while a similar drop in activity occurs in the European and US fisheries around Christmas. Political influences also appeared to impact fisheries as mandated government moratoriums in China observably decreased fishing effort.

 

 

Implications of Monitoring Fishing Fleets

This eye-opening research proves how important marine fisheries are world-wide as they take up a larger portion of the globe than any other form of food production.  This study did not incorporate many small fishing operations (like recreational fishing or small hook and line vessels) that can still contribute to global fisheries stocks, so the estimates reported are still conservative. Even so, by using the techniques and data implemented in this study, it is possible that we could improve how we monitor fisheries on a large scale. The researchers have made the data publicly available on Global Fishing Watch where you can view fishing fleets worldwide.  By tracking individual fishing vessels, we can collectively monitor fishing activities, better determine if fishermen are accurately reporting catches, determine the effectiveness of fisheries management practices, and work to find under-fished areas that could be good candidates for areas worth protecting and conserving.

If you are interested in learning how the fish you eat are harvested, visit http://www.seafoodwatch.org/ or download the app to find out more about your seafood choices and the most sustainable options in your area. Even small changes like picking a “good alternative” to an over-harvested fish can help conserve marine fish that are so important to global economies and the health of our oceans.

Leave a Reply

Your email address will not be published.