The Farm That Watches Itself

The drone goes up at twenty past six, before the wind gets going, and from the yard it looks like nothing much — a white cross against a grey sky, climbing to a hundred and twenty metres and then settling into a flight pattern of long parallel lines, back and forth, the way you would mow a lawn if you were extremely patient.

Tom Marsden watches it from the tailgate of his pickup with a tablet on his knees and a mug of tea going cold beside him. He is fifty-one. He farms four hundred and thirty hectares of arable land that his father farmed and his grandfather rented, and he has been flying this field every ten days since the crop emerged in October.

"Dad walked the fields," he says. "Every morning, all his life, with a stick. And he knew things I don't know — he could tell you which corner would flood in a wet March without looking at a map. But he was seeing maybe two per cent of the farm, and he was seeing it from five foot six."

The drone is seeing all of it, and not in the way a person sees.

---

The camera underneath it captures five bands of light rather than the three a normal camera records. Two of those bands are of no use to human vision at all: one sits at the red edge of the spectrum, the other in the near-infrared, just beyond what the eye can register.

The reason this matters is a detail of plant physiology. A healthy leaf absorbs red light strongly, because that is the wavelength its chlorophyll uses for photosynthesis, and it reflects near-infrared strongly, because the internal cell structure scatters it. A stressed leaf does the opposite on both counts — it absorbs less red, having less functioning chlorophyll, and reflects less infrared, as its cell structure begins to collapse. Compare the two bands against each other and you get a single number, between minus one and one, that tracks the vigour of the plant.

The useful part is that the number changes before the colour does. By the time a crop looks yellow from the gate, the damage has been done for a fortnight. The index picks up the divergence early, while there is still something to be done about it.

On the tablet, the field comes back not as a photograph but as a map in bands of colour — deep green through amber to red — and it looks nothing like the flat, uniform expanse you see standing at the edge of it.

"That's the thing nobody believes until they see their own field," Marsden says, turning the tablet round. "You think it's one field. It's never one field. It's forty little fields that happen to share a hedge."

He points to a rust-coloured band curving across the western third. "Old drainage run. Blocked, probably, since before I took over. That's been costing me and I couldn't have told you it was there." Then a pale patch near the gate: "That's compaction. That's me, that's my own tractor, turning in the same place for nine years."

---

What the map enables is the second half of the technology, and it is less photogenic than the flying.

The sprayer and the fertiliser spreader can both accept a prescription file — a map dividing the field into cells of a few square metres, each with its own instruction. Instead of applying a uniform rate across the whole field, which is what agriculture has done since the invention of the machinery, the machine varies its output continuously as it drives, guided by satellite positioning accurate to within a couple of centimetres.

The arithmetic is straightforward. Areas with good vigour and thin soil receive less nitrogen, because they cannot use it and the excess leaches away into the watercourse. Areas with genuine deficit receive more. Marsden's nitrogen use is down about eleven per cent over four years with no measurable loss of yield, and eleven per cent of a fertiliser bill is not a rounding error in a business with margins as thin as this one.

The larger saving is in herbicide. A newer generation of sprayer carries cameras on the boom, identifies individual weeds against the soil or the crop as it passes at fifteen kilometres per hour, and opens only the nozzles directly above them. On a field with scattered weed pressure, chemical use falls by seventy per cent or more.

"That one isn't really about money," he says. "Or it is, but it's not only. I've got two rivers on this farm and a nitrate advisory zone, and the regulations are going one direction. If I can show what went on, where, at what rate, with a file — I'm in a completely different conversation with an inspector than a bloke with a notebook."

---

He is not evangelical about any of it. Asked what the technology gets wrong, he answers immediately, which suggests he has been asked before and has thought about it since.

"It tells you there's a problem. It doesn't tell you what the problem is. Low vigour is low vigour — could be water, could be nitrogen, could be slugs, could be a bad patch of seedbed from a wet drilling day. You still have to walk out there and put a spade in the ground. I probably walk the fields more than I used to, not less. I just walk to better places."

The second objection is about the data itself. Every machine on the farm now generates a record — yield maps from the combine, application logs from the sprayer, imagery from every flight — and that record sits on a platform belonging to somebody else.

"Nobody's ever explained to me in plain English who owns eight years of yield data off my own ground," he says. "I'd quite like to know before I need it."

---

The drone comes down at five past seven with the wind picking up, lands slightly off its mark, and folds its arms in. Marsden checks the upload, drains the cold tea, and looks out across four hundred and thirty hectares that he now knows, in a strange and partial way, considerably better than his father did.

"He'd have loved this," he says. "He'd have said it was a waste of money for about two years. Then he'd have loved it."