Why Tidewright's sharpest fish census is a diver swimming 400 metres at dawn on fin power alone. A field note on roving-diver surveys off Tasmania.
These field notes describe Tidewright's own fictional monitoring program on Tasmania's east coast. The survey methods are real roving-diver technique; the site, its roster of mornings and its numbers are ours.
The survey that tells me the most about a recovering kelp forest fits in a slate the size of a paperback. It is called, without affection, the S-series: four hundred metres of reef swum by one diver at one knot, half a metre off the bottom, counting fish. No tape, no quadrat, no camera rig. The instrument is the person and the pair of eyes they were issued, and the main maintenance requirement is that they show up before the fish wake up. On the Lumeriah Coast project we have been swimming the S-series since the second winter of seeding, and this note is my case for why, in an age of sonar mosaics and satellite canopy maps, the most revealing dataset we own is produced at the speed of a slow breaststroke.
The formal name is the roving-diver technique, and it is one of the oldest standardised fish censuses in marine science. Where a belt transect pins the diver to a fixed line, a roving swim lets the diver wander a defined area or route, logging every species encountered, with abundance and size class estimated on the fly. It captures shy fish that flee quadrats, midwater schools that don't touch the bottom, and large carnivores that only ever appear at the edge of visibility — which is to say, it captures precisely the animals that tell you whether a food web is reassembling.
Our version is unglamorous by design. Entry before civil dawn, so the first fish counted are the nocturnal shifters on their way to bed and not the fish that have learned to inspect divers. A fixed swim path between three permanent markers laid in the seeding year. Finning pace — slow enough to disentangle a weedfish from the weed it is disguised as. Two slates: one for benthic and benthic-adjacent species, one for anything in the water column. Forty minutes if the ocean cooperates. Then seven years of identical mornings stacked like sediment layers, and the forest's recovery becomes legible as a roster of names appearing in an order that ecology predicted.
We have the machines, and we use them hard. The Drifter hulls run acoustic fish surveys over every seeded block monthly — split-beam echo-sounders that paint the water column as biomass fields, calibrated against netted reference schools on our farm sites. They are excellent at how much fish. They are nearly silent on which fish. An echo-sounder returns a school as an acoustic texture; it cannot currently tell a thousand blue rockfish equivalents from the Tasmanian equivalents of midwater perch, and it cannot tell you at all whether the kelp forest's character actors are back — the groper in their caves, the weedfish in the weed, the octopus whose den is only visible because of the midden of shells at the door.
Behaviour is the deeper problem. Fish redline their behaviour around engines. A hull with an electric drive is quieter than diesel but still an event, and the cryptic species — the ones most diagnostic of recovery — respond to events by becoming geology. A single diver in a 7-mil suit moving at finning pace, exhaling upward past the canopy, is not an event in the same way. Fish habituate by twenty minutes in; by thirty, the reef resumes its errands and you are allowed to watch. There is no autonomous vessel yet that spends forty minutes being boring in a forest. Being boring at working depth is, I maintain, a uniquely human field skill, and the S-series numbers keep proving it: on comparable mornings our diver counts record three to four times the species richness of the machine pass over the same water.
S-2 runs across what was, in year zero, a textbook barren — crustose rock, wandering urchins, nothing taller than a thumb. The swim now begins in a fringe of young giant kelp, and the first thing reported on the slate this past winter was behaviour, not biomass: fish holding station in cover rather than fleeing across barren. A pair of eastern blue groper — the big female burnished brown, the male in full absurd electric blue, a fish that looks like it was coloured by a committee — working the urchin rubble at nine metres. Banded morwong, the patient striped elders of the southern reef, stacked in the surge channel in numbers we stopped getting surveys-of-record to call out individually in year four. A wobbegong indistinguishable from the sponge farm it was lying on, betrayed only by the fleshy tassels around its mouth. Juvenile wrasse in a confetti burst over the kelp holdfasts. And, at the end of leg three, a school of pike sliding through the canopy gap like a thrown handful of knitting needles — a species we had not logged on S-2 since the monitoring began.
The slate numbers will join the dataset. The school of pike got a margin note, because some things should still be written with feeling: top-down is returning. Piscivorous fish do not waste fuel over a barren. They appear when the forest has enough prey fish structure — enough small lives in enough hiding places — to make hunting worthwhile. That's canopy doing twenty kinds of unpaid work at once.
The temptation every winter is to upgrade the S-series. Better slates, underwater tablets, stereo cameras on the mask, AI species-calls from the video. We resist, and the resistance is the point, for the same reason Jack has written about the bolted-down measuring tape on transect L-7: a time series is only as good as its stubbornness. If the instrument changes, every apparent change afterwards is suspect — are there more groper, or did we get better at looking? The whole argument of long-term reef records is that comparability beats precision. So the S-series must stay exactly what it is: slow, diver-dependent, and blind to improvement.
Instead, we invest in the other half of the instrument: the diver. New surveyors spend a season shadowing before their counts go into the dataset. Calibration swims run twice a year, where two divers swim back-to-back on the same line and the counts are compared like exam marks. It produces a lovely institutional outcome nobody designed: our newest field staff inherit the reef's whole history as apprenticeship. You cannot swim S-2 blind without learning where the rock was bare.
There is a final reason I protect this survey from automation, and it is not in the methodology document. The S-series is the only block of the monitoring calendar where a scientist is alone in the recovering forest, unhurried, unrecorded, watching everything and measuring almost nothing of it by hand. Ideas live there. Three of our published-on-this-site working papers began as margin notes on wet slates — the pike note above will become a food-web question for next winter. The technology stack is growing more beautiful every year. But a forest on its way back is not a dataset. It is a place that ought to be witnessed, weekly, at dawn, at the speed of fins. That, too, is monitoring.
A standardised underwater fish census where a diver swims a defined route or area at slow, steady pace, logging every species encountered with rough abundance and size. Unlike fixed belt transects it captures shy, midwater and wide-roaming species — the ones most diagnostic of a healthy food web.
Not yet, and not fully. Baited and stationary cameras have their own biases — they over-count bold scavengers and miss cryptic species — and fish behave differently around hardware and engine noise. We use cameras and acoustics alongside the swims; the dataset is strongest where the methods disagree and force us to ask why.
The S-series diver isn't alone: a tender boat shadows at the surface with visual contact on the bubble trail, and survey routes sit within short reach of shore stations. Entries happen in the best window of the day's weather, not the worst. Dawn matters because that's when the reef's busiest changeover — night shift to day shift — is visible.
Because structure rebuilds predictably. Cover and food webs establish first, small browsers and plankton-feeders move in, and only when there are enough of them do hunting fish like pike arrive. Seeing the top of the food pyramid return means the layers below it are functioning, not just present.
Start with our explainer on counting fish with sound, then the case for permanence in long-term reef records, and Jack Callum's note on five years of one bolted-down transect.
Tr L-7 is a single 30-metre transect off Tasmania surveyed every month since 2021. Five years of unglamorous re-measurement turned out to be our sharpest instrument.
Jack Callum ·
How do you count fish you can't see? Acoustic surveys use sonar to turn echoes into biomass estimates. Here's how echosounders work — and their limits.
Dr. Mara Ellison ·
Meet the eastern blue groper, the friendly, urchin-crunching giant wrasse of New South Wales reefs — every adult blue male began life as a female.
Dr. Theo Nakamura ·