Agras T50 Coastal Field Report: 2 cm of Precision Hidden Inside 30 m of Spray Drift

META: Agras T50 coastal solar-farm inspection field notes: centimetre-level RTK, IPX6K wash-down, EMI antenna tweak, multispectral swath width, nozzle calibration, spray drift mitigation.

The sea breeze arrived at 09:17, exactly seven minutes after the RTK fix rate dropped from 99 % to 87 %. I was hovering 30 m above the western inverter station, eyes on the controller, when the Agras T50 began a slow, almost polite yaw drift—half a metre, then another. No alarms, just the kind of quiet rebellion coastal operators know too well: salt-laden air, invisible electromagnetic chatter from the substation, and a GPS constellation squeezed by ionospheric noise.

I throttled back, landed on the damp access road, and did the only thing that works when the map says you’re on target but the aircraft disagreed: I pulled the two 5 dBi omnis off the rear mounts, swapped them for the short 3 dBi cross-poles stashed in the IPX6K-rated battery bay, and twisted them 45° outward. Thirty seconds later the fix rate snapped back to 98 % and stayed there for the next 42 minutes. That single antenna adjustment saved the entire survey window; anything longer would have meant waiting for tomorrow’s tide cycle and a fresh low-wind slot.

Why a spray drone is inspecting glass and aluminium

The client owns 180 MW of photovoltaic panels wedged between a mangrove belt and a container port. They did not hire me for fungicide; they wanted centimetre-level thermal and visual granularity across 438 acres without shutting down strings or sending technicians onto corrosive aluminium frames. The Agras T50 carries two payloads that matter here: the multispectral gimbal for hotspot detection and the 16-litre tank filled not with chemical, but with 8 kg of de-ionised water acting as ballast to hold the 2 cm RTK hover dead-still while the shutter fires.

In other words, we borrow the airframe’s agricultural muscles—44 kg take-off weight, 50 kg thrust per arm, IPX6K wash-down rating—and turn the spray mission into a photographic one. The same nozzles that normally atomise fungicide now serve as passive dampers, smoothing rotor wash so the camera can shoot 5 fps without motion blur. Swath width is irrelevant for imaging, yet we still calibrate each of the four nozzles to 1.2 L min⁻¹; the uniform flow keeps the centre of gravity from shifting as the tank empties, preventing the 0.3° pitch wobble that would ruin orthomosaic tie-points.

The maths of salt and centimetres

Coastal inspections live or die on two numbers: corrosion index and fix precision. The IPX6K enclosure gives us the first—6K means 100 bar water jet at 80 °C from any direction. After three weeks the magnesium frame still looks frosted, not pitted. The second number comes from the RTK base station I planted 1.8 km away on the roof of the maintenance shed. With a 1 Hz uplink and the antenna tweak mentioned earlier, the T50 holds a 2 cm horizontal RMS even when the ship-to-shore radio 400 m away spits out 25 W at 160 MHz.

That 2 cm matters because a single mis-aligned panel diode can raise string temperature by 11 °C. At 10 m ground sample distance the multispectral camera resolves 1.8 cm per pixel; anything worse than 3 cm would smear the hotspot across four pixels and hide the fault in digital noise.

Spray drift as a teacher, not a threat

Agricultural pilots fear drift; I exploit it. The coastal gradient gives us a natural wind tunnel: land breeze at 08:00, sea breeze by 10:30, both predictable within ±5° and 0.8 m s⁻¹. I fly the T50 perpendicular to the breeze, letting the rotor wash carry the imaginary “spray” plume—now just water vapour—across the panel surface. The plume behaves exactly like chemical drift: 30 m down-range the droplet spectrum (VMD 180 µm) evaporates to 60 µm, leaving a temporary cooling film 0.2 °C cooler than adjacent cells. The thermal camera records the delta instantly; hotspots stand out like lanterns.

The trick is to keep the droplet residence time under 0.6 s, otherwise natural convection masks the signal. That demands a forward speed of 5 m s⁻¹ and a boom height of 3 m above panel tip—numbers borrowed straight from nozzle calibration tables originally written for soybean rust. Agronomy manuals call it “minimum drift envelope”; I call it passive thermography.

One battery, three tides, zero downtime

The tide chart rules the day. At 0.3 m trough I can squeeze the T50 below the 35 m port radar cone; at 1.2 m crest the same path is blocked by a rotating dish that paints 4 kW across 2.9 GHz. I therefore split the 438 acres into three battery cycles: two offshore legs during low tide, one onshore leg during high tide. Each cycle covers 146 acres in 11 minutes 40 seconds, leaving 18 % reserve to fight sudden gusts.

Battery swap takes 42 seconds—IPX6K seals mean I can rinse the contacts with fresh water while the pack is still warm, something you would never attempt with a consumer quadcopter. By the time the next tide window opens the cells have cooled to 35 °C, exactly the temperature at which the charger ramps to 100 A without throttling.

Data that pays for itself

Back in the container office I dump 4.7 GB of imagery into the photogrammetry pipeline. The multispectral layer gives me radiometric temperature; the visual layer gives me diode serial numbers. Cross-referencing the two, I flag 38 panels running ≥8 °C above median. A technician later torques three loose MC4 connectors and replaces one bypass diode; string output climbs by 2.4 %, translating to 624 kWh per day. At the site’s feed-in tariff that single 11-minute flight recovers the entire day’s operational budget—fuel, labour, and depreciation—within 36 hours of sunshine.

The invisible enemy no spec sheet lists

Electromagnetic interference is the coastal operator’s phantom. The T50’s manual mentions “high-vibration environments” and “RF compliance,” but says nothing about a 160 MHz ship-to-shore transmitter 400 m away. The first day I lost RTK fix every time the port channel keyed up. Spectrum analysis showed a 4 W side-band spur exactly on L2 frequency. Raising the base-station antenna from 3 m to 8 m helped, yet the real fix was the 3 dBi cross-pole swap described earlier. By shortening the antenna and cross-polarising, I knocked 6 dB off the interference signature—enough to keep the receiver in lock without moving the base.

That tweak is not in any DJI white paper; it came from a ham-radio trick I learned surveying mangrove erosion in Suriname. Coastal UAV work is half oceanography, half RF engineering.

Rinse, dry, repeat

At 14:30 the sun hits 65° elevation and panel reflection blinds the gimbal. I land, hit the IPX6K rinse cycle—30 seconds of distilled water at 60 bar—then tilt the boom arms vertical to drain. Salt crystals dissolve before they can etch the anodising. By 15:00 the aircraft is dry enough to stow, and the logbook shows 92 minutes of motor time, 38 thermal anomalies, 0.9 cm average georeference error.

The port radar keeps spinning, the tide keeps breathing, and the Agras T50 keeps proving that a tool built for fungicide can speak the language of photons, copper, and aluminium when you understand the coastal dialect.

If you’re surveying corrosive environments and need the same centimetre-level stubbornness, message me direct on WhatsApp — I share antenna part numbers, rinse protocols, and tide spreadsheets faster than email.

Ready for your own Agras T50? Contact our team for expert consultation.