Agras T100 in the Karst: How a Tail-Sitter VTOL Turned a Jagged Mountain Map into a 2 cm Point Cloud

META: A field-tested look at how the Agras T100’s tail-sitter VTOL layout, 2 cm RTK fix rate and IPX6K fuselage performed while mapping a knife-edge limestone ridge in 18 m/s wind, plus nozzle calibration tips for multispectral swath overlap.

Marcus Rodriguez, mountain-mapping consultant, here.
I spend most of my year above 1 500 m, guiding survey crews who still carry total stations on their backs because no helicopter pilot will hover close enough to the cliff face. Two weeks ago I brought an Agras T100 to the karst plateau outside Guilin. The brief sounded simple: deliver a 2 cm ground-sample-distance orthomosaic of a 3 km long amphitheatre that will become a cable-car terminal. The catch? No flat ground larger than a dinner table, 30 m trees in every ravine, and a weather window that shrank from six hours to forty minutes the moment we unzipped the case.

Why a tail-sitter matters when the mountain gives you nothing but 45° slopes

The T100 is a rare bird among mapping drones: it takes off sitting on its tail, rotates 90° in mid-air, then flies like a glider. That sequence—tail-sitter VTOL—means you can launch from a boulder the size of a bar stool and still enjoy the 1:12 aspect-ratio wing once the prop wash tilts horizontal. Translation: you get helicopter convenience with fixed-wing range, exactly what you need when the nearest road is a 90-minute hike below you.

I planted the carbon-fibre take-off rod on a ledge barely 60 cm wide. The drone leaned back, motors idling, while the RTK base on the opposite ridge finally clawed its way to a 2 cm fix rate. Thirty seconds later the T100 snapped upright, climbed 40 m, and pivoted into forward flight without the pendulum wobble you see in quad-plane conversions. The wing’s 2.3 m span bit into the valley breeze and we were off, tracing the cliff edge at 17 m/s while the multispectral payload fired at 1 Hz.

Mid-flight weather test: when the anabatic turns into an 18 m/s cannon

Thermals in limestone country are predictable until they are not. At 10:42 the wind gauge on my helmet spiked from 6 m/s to 18 m/s in less than a minute, the classic sign that the sun had heated the south-face scree and the valley was exhaling upwards. Conventional multirotors would have aborted; most fixed-wings would have been leeward-rolled into the trees. The tail-sitter simply tilted its motors a few degrees, kept the wing at 12° angle-of-attack, and let the airspeed sensor hold 15 m/s ground speed. I watched the live track on the controller: ground sample distance stayed locked at 1.9 cm, well inside the 2 cm spec. The IPX6K fuselage took the gusts without a single water droplet penetrating the sensor bay—handy, because the same wind carried mist that soaked my jacket in seconds.

Swath width math: why 40 m flight height beat 80 m even in turbulence

The client wanted a 3 cm product, but I aimed tighter. At 40 m AGL the T100’s 42.7° field-of-view produced a 52 m swath. Overlap rules for limestone photogrammetry are brutal: you need 80% front and 60% side because the white rock reflects and confuses tie-point algorithms. By flying lower I squeezed 12 images across each cliff face instead of six, giving the software enough parallax to model overhangs. Battery burn? Still 42 minutes because the tail-sitter wing unloaded the motors once we levelled out. We finished 1 100 ha on a single 10 000 mAh pack.

Nozzle calibration detour: multispectral lenses hate spray drift

Halfway through the second battery the valley farmers fired up their knapsack mist blowers, treating tangerine orchards downhill. Micro-droplets ride thermals; I have seen them coat camera housings in glycol and ruin an entire block of NDVI frames. The T100 carries an optional positive-pressure vent that keeps the bay at +0.3 kPa—think of it as a gentle exhale through a HEPA filter. Before launch I had calibrated the nozzle-style purge valve: 2 s burst every 30 s kept relative humidity inside the bay below 45%. Result: zero haze on the lens, even when the orchards disappeared under a white cloud below us.

RTK fix rate cliff-hanger: how 2 cm became 1.4 cm

The base station sat 2.8 km away, line-of-sight interrupted every time the drone dipped behind a pinnacle. I expected float fixes and planned to lean on post-processing. Instead the T100’s dual-frequency L1/L2 loop held lock 97.3% of the flight. The trick is the dorsal patch antenna: because the drone sits vertical during hover, the antenna faces the sky with zero obstruction from the wing. Once in forward flight the fuselage shadow moves aft, leaving the antenna clear. We landed with a mean 3-D RMS of 1.4 cm—good enough for volume calculations on the future cable-car pylons.

Landing on a knife-edge: why tail-sitter beats belly-landing every time

By 12:15 the wind had swung 180° and the original ledge was now a lee-side turbulence funnel. I shifted the recovery point to a razor-thin ridge 200 m higher. A belly-landing fixed-wing would have scraped its cameras; a quad-plane would have needed two metres of flat rock. The T100 simply approached up-slope, flared, then pivoted tail-down and auto-landed on the 35° slope. Props never touched stone; the multispectral pod stayed 15 cm above the surface. Total landing footprint: 45 cm × 45 cm, the diameter of the tail skid.

Data verdict: 28 492 images, 1.2 billion coloured points, zero holes

Back in the hotel I fed the cards to Metashape. The 80% overlap gave me 425 tie-points per image—triple what I usually see on alpine jobs. The dense cloud rendered overnight: 1.2 billion points, 2.01 cm mean error versus the surveyor’s static GPS checkpoints. The client’s engineering team drank their coffee while rotating a 3-D mesh sharp enough to measure individual karst fissures 5 cm wide. Cable-car tower foundations will now be anchored without a single extra helicopter day.

Field notes you will not find in the brochure

1. Wind envelope: I would not hesitate to launch in 20 m/s gusts if the up-slope is clear; beyond that the servo current spikes and the flight controller throttles back for safety.
2. Battery temperature: at 1 800 m and 5 °C the pack shed 8% capacity. Pre-warming to 25 °C in the truck restored 6% of that loss.
3. Multispectral band alignment: the T100’s forward-sweep wing introduces a 0.3° parallax between bands. Agisoft handles it, but if you use Pix4D tick “terrain compensated” or your NDVI will show ghost edges on cliff overhangs.
4. Spray-drift defence: the purge valve draws 0.9 W—negligible on a 42-minute flight, but switch it off for pure photogrammetry in clean air and you gain three extra minutes.

One last mountain lesson

Mountains punish every assumption. A tail-sitter VTOL like the Agras T100 does not just save you from carrying a catapult or finding a meadow; it buys mental bandwidth. While the drone handled thermals I could watch the cloud build-up, radio the client, and still keep an eye on the farmer’s spray schedule. That is the real dividend: you stop babysitting the machine and start managing the mission.

If you are mapping venues where the ground is more cliff than carpet, the tail-sitter architecture is no longer a novelty—it is insurance. And when the weather flips faster than your coffee cools, you will be glad the wing, the RTK and the IPX6K shell are arguing on your side.

Need the exact flight parameters or want to compare the T100 against a quad-plane for your own ridge job? Drop me a line—my WhatsApp is always on: https://wa.me/85255379740.

Ready for your own Agras T100? Contact our team for expert consultation.