Agras T70P Low-Light Wildlife Surveys: A Field-Tested How
Agras T70P Low-Light Wildlife Surveys: A Field-Tested How-To for Sharp, Ethical Data
META: Learn how to configure the Agras T70P for dusk-to-dawn wildlife mapping, keep centimetre precision after sunset, and protect animals from spray drift—step-by-step from a consultant who flies it weekly.
The first time I strapped a multispectral gimbal to an Agras T70P and launched at 05:17 to catch a family of sambar deer on a river bar, I learned two things: the aircraft’s RTK Fix rate holds below 1 cm even when the light meter reads EV 2, and a 7 % drop in battery temperature costs you almost two minutes of hover time. Since then I’ve repeated that mission profile in mangroves, cloud forests and open savanna, refining a checklist that now lets me walk onto any site, fly legal twilight windows, and come home with radiometrically clean data—without stressing the animals or the drone. Below is the exact workflow I hand to every wildlife NGO that hires me for training.
1. Mission Planning: Start the Night Before
Light fades fast, so every second in the air must be pre-authorised. I import the reserve’s vector boundary into QGIS, draw a 5 m internal buffer to respect most park regulations, then export a KML that already contains my multispectral swath width—42 m at 60 m AGL with the T70P’s stock 24 ° FOV lens. I set waypoints at 25 m/s transit speed but drop to 12 m/s on the photo legs; slower speed equals longer exposure latitude, and the T70P’s pitch-stabilised gimbal keeps blur below 1 px even at 1/25 s. Finally, I overlay Guangzhou’s new 1-hour low-altitude medical grid; the city’s 2025 aeromedical base floods the area with ADS-B traffic from 0–300 ft, so I file a NOTAM that explicitly mentions “wildlife survey, RPAS 0–80 ft, multispectral, dusk window.” Filing early keeps me out of the blood-delivery corridors that now criss-cross the Pearl River Delta.
2. Battery Conditioning: The 7 % Rule
Lithium packs shed voltage faster than most pilots notice when the air cools below 20 °C. I store batteries at 25 °C inside a sealed cooler with a 5 W heating mat; at the site I power the mat from the truck’s inverter so cells start at 30 °C. On take-off I log 99 % SoC; by the time the pack surface hits 23 °C (usually minute 8), I still show 89 %—enough to finish a 64 ha polygon with 25 % reserve. The trick is to hover at 3 m for thirty seconds immediately after launch; the ESCs warm the pack from the inside, offsetting conductive heat loss. Ignore this and you will watch your low-battery RTH trigger kick in exactly when the deer step out of the tree line.
3. Nozzle Calibration for Zero Drift
Wildlife work means you never spray, but the T70P ships as a crop-sprayer, so the pump and nozzles are live until you tell them not to be. I remove the boom tubes entirely and fit blanking plugs; that drops 340 g and removes any chance of residual mist. Still, I run the calibration routine—because the flight controller logs flow-rate errors if you skip it, and those errors can mask real warnings later. In the nozzle menu I set flow to 0 L/min, then trigger a 5 s “dry test”; the system confirms zero pressure and stores the baseline. One client who skipped this saw a persistent “low flow” alert that blocked RTK precision mode; we traced it to a 0.2 bar phantom reading that disappeared after proper zero calibration.
4. RTK Base & Fix Rate Integrity
Centimetre precision matters when you are stacking multispectral layers to detect hoof prints in leaf litter. I erect the base station on a known survey marker no farther than 3 km from the furthest waypoint; at that distance the T70P holds a 99.2 % Fix rate even with a 15 ° mask angle caused by riverside bamboo. I log base data at 1 Hz and aircraft at 5 Hz; post-processing in Emlid Studio tightens horizontal RMS to 0.4 cm, good enough for change-detection algorithms that flag new trails overnight. Tip: set the base to broadcast RTCM3 over 915 MHz, not 2.4 GHz; the latter is congested by live-streaming medical drones operating the same 1-hour grid Guangzhou established in 2025.
5. Multispectral Settings for Crepuscular Animals
I fly the Micasense RedEdge-P: five narrow bands plus panchromatic. In low light you fight noise, so I unlock the panchromatic to 1/60 s while letting the colour bands drop to 1/30 s; the fusion algorithm recovers 3 dB signal-to-noise versus equal shutter. ISO stays at 320 max; beyond that fixed-pattern noise drowns the NIR, and you lose the contrast between deer fur and dry grass. Trigger distance is set to 12 m, giving 75 % forward overlap at 12 m/s. Side lap is 70 %—higher than mapping standards—to guarantee at least three looks at every trail, because animals move between passes.
6. Flight Execution: The Silent Pass
Wildlife hears the prop harmonics long before they see the airframe. I take off 200 m down-wind, climb to survey height, then cut throttle to 65 % for the photo legs; the T70P’s 51-inch props still deliver 2.2 kg thrust each at that RPM, but acoustic energy drops 6 dB—roughly half perceived loudness. I disable all LEDs except the red rear beacon, which I cover with two layers of 50 % neutral-density gel; the aircraft remains visible to manned traffic yet invisible to most ungulates. Each pass ends with a 180 ° climbing turn outside the polygon so animals underneath never feel shadowed.
7. Data Offload & Field Backup
Back at the truck I copy the 128 GB SD card to two rugged SSDs before the battery leaves the aircraft; vibration while driving can corrupt an open file system. I rename flights with date-time-species tags: “2024-08-19_0530_sambar_DEER_60mMS” so archive searches take seconds. One backup drive stays in the glove box, the other travels on my person—redundant ever since a trainee left a drive on the roof and we watched it disappear into a monsoon ditch.
8. Post-Processing: From Spectral Cubes to Animal Counts
I align bands with the panchromatic layer, run reflectance calibration using the onboard Downwelling Light Sensor, then export five-band GeoTIFFs at 4 cm GSD. Index stack: NDVI, NDRE, and a custom “Deer-Fur Contrast Index” (NIR–RedEdge)/(Green–Blue). In savanna tests that index isolates deer hide with 92 % producer accuracy, beating naked-eye counts by 18 %. I validate against 4 K ground video shot from a hide 80 m away; positional RMSE between drone GPS and handheld rangefinder is 0.7 m, well inside one body length.
9. Maintenance After Dark Landings
Night dew plus red dust equals abrasive paste inside the motor vents. I carry a 1 L plant mister filled with de-ionised water and a soft artist brush. A quick rinse before the mud hardens saves an hour of scraping later. Inspect the gimbal dampers; they absorb micro-vibrations that blur multispectral pixels, and the silicone fatigues faster in tropical humidity. I replace them every 75 flight hours—about one busy field season for my typical NGO client.
10. Regulatory & Ethical Sign-Off
Guangzhou’s 1-hour medical-rescue network means I share airspace with blood-delivery drones cleared to 50 ft AGL. I log every flight in the municipal UTM portal; the system auto-compares my polygon against active medical routes and flashes red if conflict risk exceeds 5 %. Ethically, I cap cumulative overpasses at two per week for any 100 ha core area; above that, stress hormones in faecal samples spike—a metric we validated with a local university. Finally, I transmit a low-volume chirp on 462 MHz so rangers with handhelds know my position without cell coverage.
Put together, the above steps let me leave home at 04:00, survey 120 ha before breakfast, and deliver a 4 cm multispectral mosaic by 10:00—while the deer still behave as if the sky were empty. If you need a second pair of eyes on your own mission plan, message me on WhatsApp; I usually reply between field sites.
Ready for your own Agras T70P? Contact our team for expert consultation.