The art and science of counting a moveable feast of fish
Counting fish sounds easy enough. But what happens when the same fish swims past your camera ten times? Or a wave knocks the camera over? Here’s a little insight into how I’m counting Norfolk Island’s reef fish for my PhD, and what all those numbers might eventually tell us.
How do you count fish when they’re constantly swimming in and out of view? And how do you know whether you’ve just counted the same fish five times?
These are questions I’ve had to get my head around for my PhD research, which involves spending an extraordinary amount of time watching underwater videos of the fish in Norfolk Island’s lagoon. As I’ve discovered, there’s quite a bit more to counting fish than simply adding them up.
First, catch your fish on camera
Part of my research involves placing stationary underwater cameras at selected locations around the reef in Emily and Slaughter Bays. The cameras record the fish going about their daily business while I leave them to it.
There are some obvious advantages to this. For one thing, the fish don’t have a snorkeller hovering nearby. Some are naturally inquisitive, while others disappear into the nearest hiding place as soon as a person approaches. With a stationary camera, they can get on with whatever they were doing before I came along. It also gives me a permanent record that I can watch again, pause, rewind or slow down when a particularly elusive little fish makes an appearance.
But each camera only captures a small slice of reef life. It can’t see what’s behind a coral bommie, under a ledge or just beyond its field of view. And it certainly can’t tell me how many fish are living in the entire lagoon.
So how do I turn all this footage into something scientifically meaningful?
Meet MaxN
Scientists have devised various methods for counting fish from underwater videos. The one I’m using is called MaxN, which simply means the maximum (Max) number (N) of individuals of a particular species visible on the screen at the same time.
Imagine I’m watching a recording featuring blackback butterflyfish. At five minutes, one swims into and out of view. At ten minutes, there are three together and then they disappear. Later, another two appear. How many butterflyfish have I counted?
Three.
Not six, because I have no idea whether those later arrivals are the same fish I saw earlier. They could be swimming in and out of view all morning. So I record the highest number I see together at the one time, rather than adding up every appearance.
I do this for every species I can identify. The maximum number of butterflyfish might occur ten minutes into the recording, while the banded scalyfins reach their maximum at twenty minutes. And infuriatingly, an elusive or rare species might swim into view just as I’ve finished my count!
Eventually, I end up with a list of species and their individual MaxN values.
There’s a catch, of course. There might be ten butterflyfish living around that particular coral bommie, but if I never see more than three together, my MaxN is still three.
It’s not a perfect system, but it avoids the problem of counting the same fish repeatedly. MaxN gives us an index of what’s called relative abundance. In simple terms, it provides a way of comparing fish counts collected using the same method, rather than telling us precisely how many fish live there.
Why scientists get so fussy
Now, you might think I could just pop my cameras anywhere on the reef, leave them recording and come back later to count the fish. If only!
The trouble starts when I want to compare one season with another.
Suppose I place a camera beside a coral bommie in January, then return six months later to the same area but instead point it towards a sandy patch. I’ll probably record different fish, which is hardly surprising, because I’ve changed the habitat I’m looking at.
Even the time of day can affect what I see. Fish have their own routines, with some species more active at certain times than others. Ideally, I’d film at roughly the same time each visit, but the tides have other ideas. I have to organise my swims around low tide, when the lagoon is generally calmer. At other times, the waves and currents can make deploying cameras difficult, if not downright impossible. And because the tide times shift every day, so do my starting times.
Each field session takes about three hours, and fitting everything into a suitable tidal window can be a bit of a juggling act. Not all low tides are equal, either. Sometimes the water level remains relatively high even at low tide, leaving me with a much shorter window of manageable conditions.
Then there’s the small matter of getting each camera back into its original position. I do my best, but if the sea is particularly rough, I might have to tuck a camera into a more sheltered spot to stop it being knocked over. And occasionally I return to discover that the coral I previously used to position it has died or been smashed by waves. So much for putting it back exactly where it was!
A still from a video recorded on 26 July 2025. Note the Acropora coral on the right
A still from a video recorded on 2 November 2026. Note the Acropora (branching) coral on the right has toppled over
Of course, if that particular coral has disappeared, the habitat itself has changed, ergo the fish using it may have changed too. But a different camera position or angle can also affect what appears on screen. Untangling those possibilities is part of the challenge. To help me keep track, I record the time of day, tide times and heights, weather conditions and other details for each deployment. I can’t persuade the reef to behave exactly the same way every time I visit, but at least I can document the conditions.
The length of the recordings is another consideration. My standard counting period is 30 minutes, and it is important that it is the same each time. If I counted for an hour on one occasion and only 15 minutes on another, I’d have twice as long to encounter different fish in the first recording.
All this fussiness is about making the recordings as comparable as possible.
Which brings me to West Slaughter Bay, where things didn’t always go according to plan.
When things go pear-shaped
When I was a brand new newbie doing this, during my first season of camera deployments, the waves in West Slaughter Bay were particularly troublesome. On more than one occasion, a camera fell over and gave me a lovely 30-minute recording of sand and rubble. Not quite what I had in mind.
If it was still pointing up, you might think I could still make use of the footage, particularly if there were fish swimming past. Unfortunately, once the camera has toppled, it’s looking at something different. It might be pointing at sand instead of coral, or its field of view might be half obscured.
Often these mishaps occur while I am far away in another part of the lagoon setting other cameras, but if I do manage to right the camera, I need to let things settle and begin a new counting period. I can’t simply stitch two bits of footage together and pretend nothing happened. And if there isn’t enough usable footage, I may have to discard the recording altogether.
I’ve recently spent a considerable amount of time going back through my videos, checking exactly when cameras moved, whether they settled again and how much usable footage I could rescue. It’s rather galling to spend three hours in the water collecting recordings, then discover that some of them aren’t suitable for analysis. But that’s fieldwork for you.
Poor visibility can also make identifying fish difficult, and a large school of fast-moving fish can turn counting into something of an endurance test.
Why count the same places again and again?
There’s another complication. Just because I record very few fish on one occasion doesn’t necessarily mean there aren’t many around. Perhaps they’re feeding somewhere else that morning, hiding amongst the corals or simply staying out of camera range. Another day, an enormous school might cruise through and make the place look positively crowded.
That’s why I repeat the surveys across different locations and seasons. That’s why I repeat the surveys across different locations and seasons. It helps me distinguish patterns from one-off observations.
I’m also interested in which species are present, not just how many individuals appear. Scientists call the number of different species recorded species richness. For example, one camera might record twenty species in January and fifteen in July.
But again, there’s a catch. Just because a particular species doesn’t appear in a recording doesn’t mean it has disappeared from the reef. It might simply have been somewhere else. Fish, it seems, aren’t particularly concerned about cooperating with scientific research.
Smoke on the water
Of course, collecting the footage is only the beginning. Then comes the really time-consuming part – watching it all, recording the sightings in a spreadsheet all while trying to stay awake!
Every fish has to be identified and counted, and because they’re constantly swimming in and out of view, I spend a great deal of time pausing, rewinding and watching the same section over again. Some species are easy to recognise, while others require a much closer look. And when several fish are darting around together, it’s quite a challenge to keep track of exactly how many are on the screen at once.
Sometimes you have to take a still from the video to count a fast-moving school of blueback sprats (Spratelloides delicatulus)!
Or in this case, elegant wrasse (Anampses elegans)
I’ve discovered that the best way to combat fish-count fatigue is to put on some old familiar tunes and sing along. ABBA, Elton John, Deep Purple and Cat Stevens are regular companions to my fish-counting sessions. (Now I’m showing my age!) It’s probably not how most people imagine marine science research, but it works for me. Although counting a school of fast-moving fish while belting out Smoke on the water does require a surprising amount of coordination for this old girl!
So what can all these numbers tell us?
Once I’ve finished counting, I’ll have a dataset showing which fish species appeared in the recordings, their MaxN values and how those values vary between locations and seasons. Then comes the statistical analysis, which is another adventure altogether, and one I confess to not really looking forward to. Slide rules and log tables were the thing when I was at high school. Just before I took my physics A level we were told we could use a calculator. At the time, a 1977 Sinclair Cambridge Programmable cost £15.95 + VAT, or around $275 today. That was a lot of pocket money!
But I digress!
I’m interested in whether some species are being recorded in greater numbers than others, whether the fish communities differ between sites and whether there are seasonal patterns or changes over time.
There’s also the question of habitat. Some of my camera locations have complex coral structures, with plenty of nooks and crannies for fish to hide in, while others are more open. Does that make a difference to the fish I record? I’d certainly expect it to, but what I’m interested in is whether the differences show up consistently in my counts.
The statistics will help me work out whether any apparent patterns are sufficiently consistent to draw conclusions, or whether they might simply be due to natural variation. And even if the analysis reveals changes, it won’t necessarily tell me what’s causing them. Fish movements, habitat changes, environmental conditions and plenty of other factors could be involved.
So there’s still a lot of work to do before I can say anything meaningful about what I’ve found. In the meantime, I’ll be back at my computer, counting fish and singing along to ABBA. There are worse ways to spend a working day!
References
Hsu, T.-H. T., Gordon, S., Ferrari, R., Hoey, A. S., & Figueira, W. F. (2025). Optimizing remote underwater video sampling to quantify relative abundance, richness, and corallivory rates of reef fish. Coral Reefs, 44, 435–449. https://doi.org/10.1007/s00338-024-02613-6
Schobernd, Z. H., Bacheler, N. M., & Conn, P. B. (2014). Examining the utility of alternative video monitoring metrics for indexing reef fish abundance. Canadian Journal of Fisheries and Aquatic Sciences, 71(3), 464–471. https://doi.org/10.1139/cjfas-2013-0086