Field Notes: The March Canopy Census, Twelve Drone Flights a Day

The March census is the fixed point in Tidewright's year. Everything else — seeding windows, storm repairs, harvest runs — is scheduled around the fortnight when we fly every southern-hemisphere project at peak summer canopy and ask the water one question: how much forest is there, exactly, now?

This year the campaign covered five project groups, from the Lumeriah coast down to the old Tinderbox pilot, in fourteen flying days. The rough arithmetic of the exercise is the title: twelve flights a day, per crew, on the good days. This is what that looks like from inside the van.

(As with all field notes, our operational figures are Tidewright's own — illustrative — while the measurement science links out to the real literature.)

Why March, Why Noon, Why the Tide Book

Canopy extent is a moving target, so the census has to be a controlled one. Late southern summer is when the canopy is largest and the water is warmest — which also makes it the season when a marine heatwave does its damage, meaning the census doubles as the year's heat-impact assessment.

Within each day, the flying window is brutally narrow. Kelp canopy is a floating surface feature, and the drone sees it well only under a specific deal with the light: sun high enough to kill reflections, tide low enough that surface mats sit near their mapped footprint rather than streaming sideways on a flood, wind under our airframe limits. The whole campaign is built backwards from the tide book and the solar-noon tables, which is why the pilot roster reads like tide charts taped to a fridge.

We fly a standardised grid over each bed at a fixed altitude, with ground-control targets laid on the shoreline rocks before first flight so the mosaics can be pinned to the same coordinate frame year after year. The method is drone photogrammetry stripped of mystique: overlapping nadir images green-band-enhanced, stitched, classified into canopy / mixed / water. Canopy area, gap statistics, edge geometry. Then, because satellites see the same water at a coarser grain, the drone mosaics become training and validation data for the satellite canopy record — the drones correct the satellites, the satellites fill the gaps between drone days.

A Census Day, From the Van

The day actually starts the night before, in the processing queue. By 9 p.m. the day's flights are stitched and classified, and the analysts are comparing coverage against the historical geotiff stack. Discrepancies get flagged for morning re-flight — a cloud shadow that poisoned a grid cell, a swell set that chopped the surface, a boat track through a bed at the wrong moment. The census is self-auditing because the pipeline is fast enough to catch its own mistakes while the crew is still on site.

Dawn is logistics. Batteries conditioned in the van, firmware sanity-checks, NOTAM and airspace confirmations where a site sits near a regional strip, and the weather go/no-go call that somebody has to own. Then the cycles: launch, twenty-five minutes aloft, recovery, cards out, batteries rotating through the chargers like a lunch rush. A two-person flight crew can keep two airframes in near-continuous rotation if the launch site is forgiving. On a good day — and March gives you good days strung together if you move with the high-pressure systems — twelve flights is not macho, it's the schedule working.

Between flights, on the ground, there is a third job: physical ground-truthing. At fixed snorkel and dive points, the crew checks what the classifier thinks it saw. Glossy surface kelp is easy; the questions live in the weeds — submerged fronds just under the surface, canopy shadowed by its own density, rafts of drift weed from somewhere else entirely. Every March census includes a written judgement call log, and next year's classifier is trained with it.

What the Numbers Do Afterwards

The census output goes three places.

First, the Canopy Ledger: the long-run, site-by-site canopy series that underpins our reporting and, where credits are involved, the impact methodology that keeps those claims honest. A credit underpinned by a time series you can audit back to individual flight logs is a very different object from a credit underpinned by a vibe.

Second, management. A bed that contracted between censuses gets a diagnostic: heat exposure, urchin advance, storm loss, harvest pressure. Each has a signature — heat losses are diffuse and depth-graded, urchin encroachment is a geometry problem you can see from the air as a nibbled edge — and each gets a different operational response in the following season's plan.

Third, learning. Because the census frame doesn't move, year-over-year change detection is apples-to-apples, and the dataset compounds. The fifth March is worth more than five times the first one.

The Fortnight's Quietest Hour

Late in the campaign, on the third-to-last flying day, we ran a re-flight over a bed in the northeast that had flagged thin in the pipeline. The re-flight came back thin too. Ground-truthed from the water, it was genuinely thin: a heat-affected patch from the January spike, sitting in the lee of the headland where flushing is weakest. Nobody enjoyed logging it.

But the odd relief of the census is that the bad news arrives with an address. We know the cell, the depth band, the exposure angle, and the historical record it broke. That's the difference between a loss and a lesson, and it's the entire argument for flying twelve times a day in the short March light. Our earlier drone notes from the Mendocino campaign made the same discovery from the other hemisphere: maps aren't the work, but the work is blind without them.

Key Takeaways

• The census is timed to peak summer canopy and flown in narrow tide/light windows; standardised grids and ground control make years comparable.
• Twelve flights a day is a logistics achievement, not an aviation one: batteries, sites, weather calls and a same-night processing queue.
• Drone mosaics calibrate the satellite record; ground-truthing from the water catches what classifiers miss.
• Outputs feed canopy reporting, credit verification, and next season's management plan — each loss gets a diagnosis, not a shrug.
• A fixed, auditable time series compounds in value; the fifth census beats five firsts.

FAQ

Why drones instead of just satellites?

Satellites give coverage and history; drones give resolution and ground truth. Floating canopy sits at the edge of what moderate-resolution satellites can resolve, especially in mixed pixels. We use both: drones to calibrate, satellites to extend.

What happens in bad weather?

The campaign carries buffer days and moves between sites with the pressure systems. A site that can't be flown inside the tide/light window drops to a satellite-plus-diver estimate and is flagged as such in the record — never silently smoothed over.

How do you turn images into "canopy area"?

Overlapping photos are stitched into an orthomosaic pinned to surveyed ground control, then pixels are classified into canopy, mixed, and open water using spectral signatures validated against in-water ground truth. The classifications are audited annually against diver and snorkel surveys.

Does the census disturb wildlife?

Flights run at altitudes and stand-off distances set by wildlife guidelines, and shorebird or marine-mammal presence can scrub a flight. The noise footprint at operating altitude is modest, but the rule is simple: the data can wait a day; the wildlife can't.

Who sees the results?

Aggregated results feed our public impact dashboard, and site-level summaries go to the partners each project belongs to — councils, ranger groups, fisheries co-ops. The forest belongs to the coast that hosts it; so does its report card.