Marine farmers using AI to find out what’s happening under water

Source: Radio New Zealand

At the top of the South Island, in the centre of New Zealand’s aquaculture and fishing industries, artificial intelligence is increasingly being used to provide better information about what’s going on under water.

Scientists and engineers are working together to develop new tools that give marine farmers better metrics, in a bid to push growth in the sector.

Tim Rensen is one of the founders of UWAI Robotics, a Christchurch-based company that’s built an underwater drone that uses artificial intelligence and advanced imaging to scan mussel lines, so farmers can see their crops in real time.

“What we are working on at the moment is being able to tell farmers what the condition of their mussels is, because at the moment they have to shuck them open to look at the quality of the meat, but we are looking at that, looking at what other biofouling is growing on the lines and what have the conditions been.”

Once the drone is in place on a mussel farm using sensors and GPS, it scans the lines and provides a video stream to a laptop on the surface, capturing images of thousands of individual mussels.

The data was then instantly analysed using AI and Rensen said it had taken eight years of development to get to this point.

“It is about distilling all of this raw data into a few key metrics like count, size, the variability on the line, we could be looking at whether the mussels are clumping together or if they are well distributed,” Rensen said.

They can provide insights on expected yield, when to reseed and harvest, with the aim of providing even more information as the technology develops.

In Havelock, the Greenshell Mussel Capital of the World, Mills Bay Mussels operations manager Maegan Blom said checking the quality of the shellfish usually involved hauling lines out of the water and opening mussels to inspect them.

“Those quality decisions that you make when you are checking the crop for harvest are really crucial to maintain the value of the product. If you tell your customer what you have got for them and then it arrives at their factory and it is different, that is really not good.”

The company, which supplies fresh live mussels, mussel meat products and health supplements to the domestic market, has marine farms in the Marlborough Sounds and Golden Bay.

She said it had been fascinating to see Rensen use the underwater drone on a Mills Bay farm in Kenepuru Sound, and then almost instantly supply her with a graph showing the size of the mussels.

“It’s not necessarily quicker, but the accuracy of the information is just incredible so you can make better decisions.”

Blom said had imagined technology like what UWAI Robotics was developing and was excited to see what was already being used in other industries being adapted for use in marine farming.

She said there was lots that couldn’t be controlled when farming in the ocean, so it was important to focus on what could be. Quality data collected over time would enable better operations, previously learnt through years of experience.

“You can control things like when you seed your lines, the density, where you put your farms, what time of the year you harvest, but then there’s things you can’t control, how much food’s in the water, the weather, what happens in the natural environment with predators like when the snapper are going to come and eat the mussels.

“With good information you can actually start to draw conclusions, patterns, trends and become a more efficient mussel farmer.”

In Nelson, marine tech company Seaweave have developed camera systems, sensors and artificial intelligence for use in the aquaculture and fishing industries.

Chief executive Chris Rodley said not only was it hard to see what was happening in a marine farm, but access to those sites was weather dependant.

“If a big storm comes to Tasman Bay and the farmers they can’t go out on their vessels to check the farm, they can’t even go out the next day because the weather is so bad, they can’t see it from land and how do we help those guys put their head on the pillow and actually sleep at night?”

He said the benefits were two fold, visual data was available in real time, but could also be interpreted over time to identify patterns and trends.

“Number one, what’s going on in real time on the farm, what’s the water quality like, are there any issues with lack of food, or a huge weather event that dumps a bunch of freshwater into the harbour.”

He said AI was changing the way that data was accessed.

“What if you could talk to your farm, if you could ask it, ‘what is going on right now?’ and your farm could simply answer.

“In the past that was difficult and now we just put that data into a model and it can be queried and that’s exciting.”

Seaweave first started working with mussel farmers and had since moved into other areas, including wild caught fisheries.

Last year, Seaweave signed an agreement with the Bioeconomy Science Institute to further develop AI imaging applications for fish in aquaculture, which Rodley said it was like facial recognition for fish.

He said nine fish species were currently recognisable. It means individual fish can be identified, assessed for breeding programmes and monitored underwater.

Rodley said New Zealand had some of the best seafood in the world, but because of its distance from key markets, so there was a need to focus on quality and telling the story behind the product.

“We can identify individual [fish] as they travel through the supply chain at various points and that’s huge for population modelling and management.

“Linking that to the end consumer allows us to create a huge amount of value.”

It’s hoped the technology will help the aquaculture industry reach its goal of quadrupling its annual revenue to reach $3 billion by 2035.

Sign up for Ngā Pitopito Kōrero, a daily newsletter curated by our editors and delivered straight to your inbox every weekday.

– Published by EveningReport.nz and AsiaPacificReport.nz, see: MIL OSI in partnership with Radio New Zealand