Non-Destructive Marine Monitoring in Fiordland
with Baited Remote Underwater Video (BRUV)

Dr. Adam N. H. Smith

Director, Sea Through Science Ltd.
Senior Lecturer, Massey University

Acknowledgements

• Supporters and funders:
  • Fiordland Marine Guardians
  • Department of Conservation
    especially Rich Kinsey
  • Ministry for Primary Industries
    Jean Davis, Mark Geytenbeek, Rob Win
• Skippers and crew of the Southern Winds
  Miriam MacFadgen, Pete Kirkman, Richard Kinsey, Ross Funnell, Millie Mannering
• Other helpers in the field
  Odette Howarth, C. Robert Priester, Jenni Stanley
• Annotators
  Odette Howarth, Alex Burton

Photo credit: C. Rob Priester

I’d like to begin by acknowledging a few people.

Thank you to the Fiordland Marine Guardians for inviting me here to speak today

To other supporters of the work I’m going to present, at DOC and MPI

And finally, thanks to the skippers and crew of the Southern Winds, and other helpers in the field and other aspects of this research

Protecting fish stocks

• Broad agreement that it is the right thing to do
• Requires appropriate management
• Need good information on:
  • the current status of populations
  • trends through time
• But… counting fish ain’t easy!
• Many methods are available

• Broad agreement

Now, this is a value statement, and I’m not here to convince anyone about what they should and shouldn’t value.

As a scientist, that’s not really my role.

Values are subjective, and they vary from person to person.

It’s the role of groups like the Guardians to represent the values of the citizens and stakeholders.

That said, I suspect that almost everyone can agree that we shouldn’t overfish to the point of fishery collapse. We want these resources to be available to future generations.

• Protecting fish stocks requires good management. And good management is a balancing act between protecting the ecosystem without being overly restrictive. If rules are too restrictive, people lose trust in our institutions and they might just flout the rules. Not restrictive enough, and we degrade our resources.

• To ensure management is justified and effective, we need good information on…

• the current state of populations, some measure of the abundance of individuals, and perhaps information on the size and sex structure of the population.

• Perhaps more importantly, we need to know something about population trends through time. This is true especially when there are changes to regulations.

To look at trends, we need regular monitoring. The more often we survey, the more information we have on trends. There’s simply no substitute for regular surveying if you want to assess trends.

• Monitoring fish is hard – trust me, I’ve spent most of my career counting fish using various methods and analysing the data. As my colleague Russel Millar once said: Counting fish is hard because you can’t see them, and they move.

• There are myriad methods out there for quantifying fish populations, and each method has it’s advantages and disadvantages that make it appropriate or inappropriate for any particular scenario.

For example,

• for broad-scale surveys across a continental shelf, research trawls might be a good idea. This sort of data can feed into bit fishery fish stock assessment models, which are typically done at large spatial and temporal scales, and require truckloads of historical catch data.
• in more sensitive, shallow, coastal environments, counts by scuba divers works
• Tag-recapture methods can be used for some populations. You catch a bunch of fish, tag them, come back another time and catch more fish, and we can get an idea of the population from the proportion of tagged fish in the second visit.
• And, of course, many many other methods

Which methods are appropriate?

Depends entirely on the context:

• Objectives
• Costs and budget
• Constraints of the physical environment and organisms’ biology
• Impacts of the method on the populations and broader ecosystem
• Quality and quantity of information provided by the method

Which methods are appropriate depends entirely on context

Firstly, the objectives are important. Everything stems from objectives. It’s no use just going out and counting stuff. We need to exactly what we want to measure, and why.

Then there are the constraints posed by costs, and by the physical environment, and biology in general

The impacts of the method can be very important – do we want to use a destructive or non-destructive method

And, of course, we need to consider the quantity and quality of the information provided, and whether that information meets our objectives.

Again, it all has to come back to the objectives, and this is one of the main reasons why I really wanted to make the trip down here and talk with you. I wanted to get a better idea of what we need to know, and whether the work that we have begun meets those needs. And, if so, how we can ensure that it continues.

New fishing regulations for blue cod and other species in Fiordland

How will we know if the new regulations are effective?

What is the current state of populations of blue cod and other key species?

How will these populations trend over the coming years?

Can the regulations can be relaxed in future? When?

Answering any of these questions requires ongoing monitoring.

With that in mind, here’s the broader context of the work.

As you know, earlier this year, new regulations were brought in for recreational fishing of blue cod and some other species in Fiordland.

Changes in regulations pose some important questions:

Answering any of these questions requires ongoing monitoring.

So, what is the current state of monitoring in Fiordland?

Blue cod potting

• In the 2014 survey, 1799 blue cod
  were captured (1237 kg; 96% killed)
• Non-neglible losses for a sensitive,
  low-productivity ecosystem
• No potting survey since 2014

?

The main method used to monitor blue cod fisheries is potting surveys.

• Potting surveys were done in Dusky Sound in 2002, 2008, and 2014. Here are the Catch Per Unit Effort numbers from these surveys, measured as kg of blue cod per pot. I’ve split them between inside the marine reserves vs outside the marine reserves.

No strong trend apparent there, but the last survey was a decade ago. Why?

• Potting is destructive method. It kills a lot of fish. For example, the 2014 survey captured over a tonne of blue cod, most of them killed.

Such losses may be ok for a large scale, productive fishery from which hundreds of tonnes are taken annually, but in a sensitive, enclosed environment like the fiords, many consider potting surveys to be inappropriate.

It is simply not fit for purpose.

• So, there has been no potting survey since 2014, a full decade ago, which means we’re relying largely on anectodal evidence when making potentially contentious management decisions, which is not ideal.

Furthermore, we don’t have hard data on what the current trends are for key species in Fiordland.

Blue cod potting

• In the 2014 survey, 1799 blue cod
  were captured (1237 kg; 96% killed)
• Non-neglible losses for a sensitive,
  low-productivity ecosystem
• No potting survey since 2014

• In the broader stock that includes Southland and Fiordland, there have been some worrying trends over the past 20 years. But this assessment is based on data from Foveaux Strait, not the inner fiords.

Blue cod potting

From Fiordland Marine Guardians (2023) Proposed further amendments to fishing regulations in the Fiordland Marine Area

The Fiordland Marine Guardians, in a document released last year, expressed an interest in investigating alternative, non-destructive methods of surveying key species in Fiordland. They specifically mention a method called Baited Underwater Video.

Which is where I come in.

What is a ‘Stereo BRUV’?

• Stereo Baited Remote Underwater Video
• Stainless steel frame with two video cameras pointing at a bait canister, with lights and a depth sensor
• Used around the world as a standard method of monitoring fishes

Introducing the stereo baited remote underwater video

Each unit comprises a stainless steel frame with two video cameras, lights, a depth and temperature sensor, and a canister of fresh bait sticking out the front.

We have six of these units, and we throw them into the sea and leave them for just over an hour. The bait attracts fish and other animals, and we can count the organisms and record the habitat.

This has become a standard tool for monitoring fishes internationally. I’ve used these things for several years across New Zealand from the Kermadecs to Fiordland, as well as in several other countries, the Galapagos Islands, New Caledonia, Fiji.

We also used this method in a global study of sharks called Global FinPrint, which surveyed sharks on coral reefs all over the world.

Why stereo? This one of the reasons why this method is so powerful…

What is a ‘Stereo BRUV’?

• After calibration with a specialist cube and software,
  

After calibrating a stereo BRUV with a special hardware and software,

you can take accurate measurements of the lenghts of fish and other objects.

Much like our having two eyes allow to perceive depths.

What is a ‘Stereo BRUV’?

• After calibration with a specialist cube and software,
  stereo BRUVs allow for accurate measurements of lengths.

The BRUV process

The full process of a BRUV survey involves first calibrating the cameras, which is usually done in a swimming pool.

Then we go to the field and deploy the BRUVs.

Each video must be annotated, which is the process of reviewing the footage and recording data on abundance, length, habitat, etc. This process takes time, sometimes several hours per video, depending on how fishy it is. We call this “annotation” rather than “analysis” so it is not confuses with the statistical analysis of the data.

And finally, there is the data analysis and reporting.

Our work in Fiordland

In 2021 and 2024, the Department of Conservation funded BRUV surveys in Fiordland.

The Ministry for Primary Industries funded the annotation of the 2024 videos.

In 2021 and 2024, Rich Kinsey invited us on board the Southern Winds to undertake BRUV surveys in Fiordland.

MPI has also supported the work by paying for annotation of the 2024 videos.

Today, I’m going to present some of the videos and some preliminary results, mainly in the form of distribution maps.

Fiordland 2021

90 deployments at depths of 10-95 m

Here’s a map of the sites we surveyed in 2021 – 90 deployments across multiple fiords. This was part of a project by my former postgraduate student, Odette Howarth.

Tamatea/Dusky Sound 2024

136 deployments at depths of 15-86 m

Then, in 2024, Rich Kinsey at DOC commissioned us to do a more intensive survey of Tamatea/Dusky Sound, for a total of 136 deployments.

What did we see?

• Blue cod
• Tarakihi

Alright, video time. Here’s a BRUV that’s just been dropped overboard and is heading to the seafloor, here around 26 metres.

This particular deployment is interesting for a few reasons. There is nothing much to see when it hits the floor, but after a few moments…

There are a lot of juvenile blue cod here. And the fact that so many turn up so quickly means that they are a lot very nearby the camera. This might indicate a habitat that is important to juvenile blue cod.

There’s also a couple of tarakihi here.

Abundance of blue cod

Here’s a map of Dusky Sound showing counts of blue cod as the size of the points. Bigger points mean more blue cod were seen at that site.

The video we just saw is this one in the Tuamoana Marine Reserve (fourth from the top of line).

What do these numbers mean? The numbers are what’s known as the “MaxN” count, which is the highest number seen on a single frame of the video.

We don’t just count every fish we see, because a single fish could circle the gear 100 times and get counted 100 times, and we can’t tell that situation from 100 individuals going by. Hence, the conservative approach of using a single frame.

By this method, we can get some very large counts, up to 38 individual blue cod here, with an average of 5.5 per deployment. We also did some surveys for DOC in the outer Hauraki Gulf this year, and got as many as 56 snapper in one frame.

The MaxN count from a single frame is the standard index of relative abundance that can be compared among sites and over time.

We also take measurements of lengths at different intervals, so that we can get a lot of information about the size-structure of the population.

Presence of tarakihi

For other species that are not so numerous, such as tarakihi, I’m just going to show you maps of presence and absence.

Solid points are sites where we saw tarakihi; open circles are where we didn’t.

Tarakihi turned up in a more than half of the sites, throughout the fiord.

What did we see?

• Blue cod
• Tarakihi
• School shark
• Crayfish

What else did we see?

This is the same deployment as the last video I showed you, but much later on. We can see more species have turned up, notably crayfish and school shark. There are five crays and three schoolies here.

We can also see that there are not so many juvenile blue cod, but a lot more large blue cod. This is a key advantage of the method – it seems we get a much better view of the size structure of the population because we see whatever comes past, not just whatever gets stuck in the pot.

Presence of school shark

I was very surprised at how many school shark we saw throughout the survey – they appeared in more than half the deployments, and throughout Dusky Sound. Up to around five individuals. My PhD student, Alex Burton, is working on school sharks, so he was very excited to see this.

We don’t see school sharks anywhere close to this regularly in the outer Hauraki Gulf.

Presence of crayfish

Crayfish were also fairly widespread.

It looks like like they were observed particularly often in the marine reserve, where they were seen in 11 out of 14 sites. They were seen much less often outside the reserve.

Of course, confirming this difference would require a formal statistical analysis, taking into account habitats, depth, etc., and it is certainly possible.

But I was surprised to see so many, actually. I don’t know if BRUVs have ever been so successful at surveying crayfish.

What did we see?

• Blue cod
• Tarakihi
• School shark
• Crayfish
• Hāpuku

Everyone’s favourite species with the goofy-looking underbite, the elusive hāpuku.

Presence of hāpuku

We saw hāpuku at 10 sites, only one or two individuals at each place.

It was very interesting that they were found so often right up in Supper Cove too.

What did we see?

• Blue cod
• Tarakihi
• School shark
• Crayfish
• Hāpuku
• Broadnose sevengill

Another species that turned up often was the broadnose sevengill shark, which we saw in around 32, around a quarter of the deployments.

Like this big female.

Since presenting at the FishNZ meeting, I’m now in collaboration with Alice Rogers and her postgraduate students at Victoria University, working on a project on habitats and prey of sevengill sharks in Fiordland.

I’m generally not clingy with my data. A lot of scientists are territorial, but I just want it to be used.

Presence of sevengill shark

Again, this species was found throughout the fiord.

What did we see?

• Blue cod
• Tarakihi
• School shark
• Crayfish
• Hāpuku
• Broadnose sevengill
• Black coral and gorgonians

Lyndsay Holland of the Department of Conservation’s Conservation Services Programme contacted me after my presentation to FishNZ, because she was interested in where I saw black corals and gorgonians like these, so I sent her some screenshots and GPS coordinates.

Presence of black coral

Here’s where we saw black coral

What did we see?

• Blue cod
• Tarakihi
• School shark
• Crayfish
• Hāpuku
• Broadnose sevengill
• Black coral, gorgonians
• Much more!

We saw many, many other species. Across the survey we saw 37 species of fish alone.

Number of fish species observed

Here’s a plot of how many fish species we saw in each deployment, a diversity metric known as species richness.

We saw up to 12 species of fish in a single site.

So, this really is is a wealth of information that can be mined for all sorts of interesting ecological questions.

In which depths, habitats and regions does each species occur?

Which depths, habitats and regions support the greatest diversity?

We can look at interactions and associations among species – which species are seen together, or not seen together?

• For example, a predator might decline alongside prey species, giving a possible explanation for that decline.

• Better able to understand ecosystem dynamics, which gives us a stronger basis for more effective, ecosystem-based management.

Advantages of BRUVs

• Cost-effective
• Non-destructive, non-lethal
• Globally established methodology
• Standardised observation period (60 min)
• Broad depth range
• Captures a wealth of information on multiple species, size-classes, and habitats
• Permanent video archive

Photo credit: C. Rob Priester

So, what are some of the advantages of this method?

For a start, it’s much cheaper than a potting survey

Importantly, it’s non-destructive. No fish were harmed to bring you these data.

Because it’s cost effective and non-destructive, we can do it more frequently.

It’s globally established, standardised methodology. I was a co-author on a published paper that established the standardised methods.

They can access all sorts of habitats at a broad range of depths. We’ve sent some systems down to 1200 metres (not these ones).

They’ll record anything that goes near a BRUV, all species of all sizes, as well as habitat information.

The fact that we have all this geographical and habitat information means that we can figure out what which areas and habitats are important to blue cod populations, at different parts of their life history.

Finally, all the videos are stored in a permanent archive. We’ve extracted the data that we want, but the future opportunities are endless. These videos will likely be mined for information and projects we haven’t even thought of yet.

The future afid.io

The future is looking bright for these methods.

I’m in talks with a research group in Australia called afid.io about training machine learning algorithms for automatically annotating BRUV videos taken from the NZ context, which would reduce the costs of this method even further!

Disadvantages of BRUVs

• Some upfront expenses (hardware and software)
• Lack of historical baseline data in Fiordland
• Cannot tell the sex of non-sexually dimorphic species (e.g., blue cod)

Like any method, there are disadvantages too.

The gear is quite expensive to buy. All that specialist calibration hardware and software, there are the frames, cameras, torches, and software for annotating the videos – it’s all quite expensive. But the good news is that I have all the gear already.

We don’t have much in the way of historical baselines of stereo BRUV data for Fiordland. We have two surveys using our gear, and there have been a few surveys by other groups as well, though they used different types of systems.

One disadvantage of not capturing the fish, like you do with pots, is that you cannot tell the sex of a blue cod unless you open it up and look at the gonads.

Indeed, potting surveys suggest that heavily exploited blue cod populations are generally skewed in favour of males. This can be important information for large-scale stock assessment models, but in smaller, more enclosed areas like the fiords, this may not be as important as having good, fine scale information on abundance and size structure.

If this information is indeed crucial, there may be other ways of obtaining it with supplementary methods. I’ve heard mentioned the idea of asking recreational fishers to record their catches, and even the sex of fish.

Recommendations

• Ongoing monitoring of Tamatea/Dusky Sound using BRUVs (ideally every 1-3 years)

• Annotation of the 2021 BRUV data (at least for sites repeated in 2024)

• Statistical analysis and reporting for key species

• Comparison of potting data with BRUV data

• Exploration of the potential for BRUVs and complementary methods to provide the information required by management

So where are we at now?

So far, DOC and MPI have funded the fieldwork and annotation of the 2024 data.

We do not have any funding for the statistical analysis and reporting, and we do not have funding lined up for future surveys.

It seems to me that continuing with these surveys, even if it’s just in Tamatea/Dusky Sound, is a good idea, given the sensitivity of the inner Fiordland ecosystems, concerns about population trends, and in light of recent changes to regulations.

Much like planting trees, the best time to start a good monitoring programme is 20 years ago, the second best time to start a good monitoring programme is right now.

So that’s what we’ve done.

I don’t think it would be wise to wait another decade, or even five years, to do another survey.

Establishing trends through time requires regular monitoring. There’s really no way around this.

If we wait five years between surveys, then after 15 years we’ll get the bare minimum number of time points to look at trends: three. Analysis of trends requires us to quantify variability among years, which requires multiple time points.

Luckily, with BRUVs, we can do regular surveys because it’s relatively cheap and non-destructive.

Every year would be ideal, but every 2-3 years would allow us to get a good idea of trends in a reasonable timeframe.

As I mentioned earlier, the data are already being picked up by other research groups and government agencies. I can think of so many projects I’d like to do with the data.

In particular, I’m planning to do a direct comparison of the BRUV data with the historical potting data. We can examine exactly what information is needed by management and how well each method meets those needs as part of a formal cost-benefit analysis. I have an intern student coming from France next year who is interested in this project.

We could even do paired surveys with pots and BRUVs at the same sites in future. This would mean we have overlap of data with the two methods, potentially enabling us to integrate the historical timeline of the potting surveys with current and future surveys with BRUVs.

And like I said, there may be scope to complement the BRUV data with other sources, such as from recreational fishers and citizen science initiatives like Fish Mainland.

Discussion points

• Establish monitoring objectives
• Is the stereo BRUV the right tool?
• If so, how can we ensure the continuation of this work?

Which brings me to the end of my talk.

In this or some future meeting, I’d love to get a better idea of what the objectives of the monitoring should be.

So we can judge whether stereo BRUVs are the right tool for the job. Do they meet our objectives? Will they tell us what we need to know?

I see a lot of potential here for stereo BRUVs. It is a globally established, cost-effective method that can provide good data not only on the abundance, size structure, and trends of key species, but it also a wealth of other information in Fiordland, one of the world’s most interesting and unique marine ecosystems.

If there is broad agreement that this work should continue, how can we ensure that it happens?

We’ve made a great start, but we have yet to secure funding to continue this work, or even report on what we’ve done so far.

• DOC is very supportive but it has a tonne of projects competing for limited funding and resources
• While MPI have been supportive, I get the sense that they prioritise work on stocks that are commercially harvested; and blue cod are no longer commercially harvested in the inner fiords

So although many people see the value of this project, there’s a risk that it’ll slip through the cracks of the major government agencies.

In my view, it would a shame to abandon this project after such a promising start, given the importance of monitoring to the management of these ecosystems for now and for future generations.

Thank you.