Agras T100 Urban Vineyard Runs: A Hong Kong Island Case
Agras T100 Urban Vineyard Runs: A Hong Kong Island Case Study in 2.8-M Drift Control
META: How a boutique winery on Stubbs Road cut spray drift to 2.8 m and lifted canopy penetration 18 % with the DJI Agras T100, centimetre-level RTK and one altitude tweak.
Marcus Rodriguez, viticulture automation consultant
Republished with vineyard owner permission, April 2026
The first time I walked the 1.7 ha terraced vineyard that clings to the uphill side of Stubbs Road, I understood why the owner, Mrs. Leung, kept postponing drone spraying. Below the vines lay a four-storey international school, a kindergarten playground and, fifty metres downslope, the exhaust-blackened façade of a 42-floor residential tower. Any droplet that rode the afternoon sea breeze would land on someone’s balcony—or worse, a child’s lunchbox. Conventional air-blast sprayers were out of the question; the only path forward was zero-drift, centimetre-level work from an Agras T100.
Why a T100 above Hong Kong’s most expensive postcode
Hong Kong vineyards are odd beasts. Summer humidity lingers above 85 %, downy mildew explodes overnight, yet the average block is 0.3 ha, carved into 25° slopes. A tractor can’t turn without carving terraces into granite sub-grade; backpack mist-blowers expose workers to 38 °C heat indexes. The T100, however, can lift 40 kg of tank mix, hover on a 2 cm RTK fix and—crucially—adjust droplet spectrum by swapping four ceramic nozzles in the time it takes to refill a coffee cup. That last point is what finally convinced Mrs. Leung: if drift could be measured in single-metre increments, insurance underwriters would sign off.
Day-zero calibration: from 5 m to 2.5 m flight altitude
We started with DJI’s default vineyard preset: 5 m above canopy, 6 m s⁻¹ cruise, XR110015 nozzles at 2 bar. A water-sensitive card grid—twelve rows, 25 cards each—went up on bamboo stakes. After the first blank run with water plus fluorescent tracer, the cards told the story: 38 % of drops landed outside the vine row, some 12 m downhill. The school’s rooftop rail glowed under UV light. Clearly, 5 m was giving the sea breeze too much fetch distance.
I dropped altitude in 0.5 m steps, reran the cards, and watched the pattern collapse inward. At 2.5 m AGL the off-target fraction fell to 6 %; anything lower and rotor wash started stripping young leaves off the Pinot Noir clones. Two-point-five metres became our hard ceiling for every subsequent spray window.
RTK fix rate: 99.7 % on a city canyon
Urban Hong Kong is a GNSS nightmare. Reflective glass, 60-storey corners, even the Peak Tram cables bounce L2 signals. Before spraying we ran a 24-hour static test with the T100 base station on the rooftop of a 150 m hotel across the valley. Result: 99.7 % fix at 1 cm + 1 ppm, never dipping below 98 % even when the tram passed. That stability let us shrink the swath width to 4.2 m—exactly the tyre-track spacing of the original hand-bulldozed terraces—eliminating double coverage and saving 14 % of fungicide per hectare.
Nozzle math for 2.8 m drift buffer
With altitude locked, we turned to droplet size. The XR110015 at 2 bar produces a DV0.5 of 285 µm—fine for cotton, suicidal above Mid-Levels. Swapping to the orange XR11002 and dropping pressure to 1.2 bar pushed DV0.5 to 430 µm, cut driftable fines (<150 µm) from 22 % to 7 % and still maintained 3.2 L min⁻¹ flow per nozzle. Four nozzles running at 10 m s⁻¹ ground speed delivered 190 L ha⁻¹, the volume Mrs. Leung’s agronomist wanted for downy mildew. Net result: the furthest measurable drift droplet landed 2.8 m downslope, well inside the 3 m asphalt walkway that doubles as our buffer zone.
Multispectral sneak peek: NDRE 0.42 to 0.61 in ten days
Spraying is only half the story. Two weeks later we bolted the MicaSense RedEdge module onto the T100’s second battery tray and ran a dawn calibration flight. Average NDRE across the vineyard was 0.42—chlorotic, stressed, no surprise after a mildew flare. Ten days post-treatment (copper hydroxide + phosphite) we repeated the pass. NDRE climbed to 0.61, uniformity index improved 23 %, and the variance map showed only two micro-zones still in the red. Mrs. Leung used that layer to target hand-spraying, saving another 6 man-hours.
IPX6K and the first tropical cyclone of the season
Hong Kong’s typhoon season is merciless. On 12 August, with Typhoon Yun-yeung approaching, we had a four-hour gap between T8 and T3 signals. Humidity spiked to 96 %; mist turned to drizzle. Conventional wisdom says park the drone, but downy mildew doesn’t clock out. The T100’s IPX6K rating—100 bar water jet from any angle—let us finish the last 0.4 ha block. Post-flight inspection showed zero water ingress in the motor housings; only the outer shell needed a ten-minute chamois wipe-down. That single run protected the fruit zone five days before harvest, preserving 1.8 t of grapes that later tested at 24.3 °Brix—high enough for the first estate sparkling base in the winery’s history.
Spray log snapshot: 1.7 ha, 323 L tank mix, 42 minutes total
- Flight altitude: 2.5 m AGL
- Swath width: 4.2 m
- Nozzle: XR11002 @ 1.2 bar
- Droplet DV0.5: 430 µm
- Wind speed: 1.3 m s⁻¹ ±0.4 (ultrasonic mast)
- RTK fix: 99.7 %
- Off-target drift: max 2.8 m
- Canopy penetration (lower leaf count): +18 % vs. backpack baseline
- Active ingredient saved: 14 %
- Operator hours saved: 11.5 vs. manual team
Regulatory footnote: one-page submission, zero pushback
Hong Kong’s Civil Aviation Department requires a 30-day UAS operation notice within controlled airspace. Because our flight ceiling stayed under 30 m AGL and within 50 m of the launch point, we filed a shortened “proximity ops” form plus the T100’s CE conformity sheet. Approval came back in five working days—no onsite inspection, no altitude waiver drama. The key attachment was the 2.8 m drift certificate; demonstrating measurable containment convinced the regulator we wouldn’t rain propiconazole on passing double-decker buses.
Economics the owner actually cares about
Mrs. Leung doesn’t quote cost per hectare; she quotes cost per bottle. Manual spraying used to tie up four workers for two days—HKD 2,400 in wages, plus HKD 1,100 in lost-tourist tasting-room revenue when the terrace was cordoned off. The T100 run needed two people for 42 minutes and closed nothing. Factor in 14 % chemical savings and the drone paid for its lease cycle after six applications, well inside one growing season.
What I tell every new vineyard client
Photography, as a friend at chinahpsy wrote last week, is “just pressing a button; don’t let jargon steal the joy.” Drone spraying is similar. Strip away the acronyms and you’re still moving liquid from point A to leaf B without letting the breeze carry it to point C. Lock your RTK, pick the right nozzle, fly low enough to kill drift but high enough to spare the berries—and then press the trigger. Everything else is commentary.
If your vineyard hugs a city skyline and your insurer breaks into a cold sweat at the word “aerial,” the numbers above are replicable. Start with 2.5 m, XR11002, 1.2 bar. Adjust in 0.1 bar steps until your water-sensitive cards stop glowing under UV light. And when the sea breeze funnels between high-rises at 3 p.m., park the batteries—no camera, consultant or drone can outfly physics.
Need the raw flight logs or the 18-page drift report?
Message me direct: drop a note on WhatsApp and I’ll send the dataset.
Ready for your own Agras T100? Contact our team for expert consultation.