A thermal drone detects warm-blooded animals against vegetation day or night, letting conservation teams count populations, locate nests and detect poaching activity from altitudes that cause no disturbance. The limiting factor is almost never the sensor — it is thermal contrast between the animal and the ground it is standing on, and that changes by the hour.
- 130 gMV-2P payload mass
- <40 mKMV-2P thermal NETD
- 60–120 mCommon survey altitude band
- 640×512Thermal array for deer-sized targets
Key takeaways
- Thermal finds warm bodies; it does not identify species. Pair detection with a zoom or starlight channel for confirmation.
- Contrast, not resolution, sets what you can detect. On a warm afternoon the ground can match animal surface temperature and the survey returns nothing.
- Payload mass drives disturbance more than altitude does — a lighter payload allows a smaller, quieter aircraft.
- Canopy is opaque to thermal. Any survey over timbered ground is a minimum count, and should be reported as one.
On this page
Survey missions thermal makes possible
Four mission types account for most conservation flying, and each has a different constraint.
Nocturnal population counts work because animals are active and the terrain has cooled — the largest thermal contrast of the 24-hour cycle. Fawn detection before mowing is time-critical agricultural work with a very short window at dawn. Den and nest monitoring needs repeatable framing more than range. Post-event searches for injured animals after storms or fires trade precision for coverage.
What unites them is that thermal answers “is something warm there”, and a second sensor answers “what is it”. Building a programme around detection alone produces counts nobody can verify.
Contrast conditions, by hour and surface
The table below is the planning shortcut we give survey teams. It describes when the temperature difference between an animal and its background is large enough to work with — which is the actual constraint, not sensor specification.
| Condition | Animal-to-background ΔT | Detectability | Notes |
|---|---|---|---|
| Pre-dawn, clear, dry ground | 12–20 K | Excellent | The best window of the day, in every season |
| Night, clear, after cool day | 8–15 K | Very good | Standard for population counts |
| Night, overcast | 5–10 K | Good | Cloud slows ground cooling |
| Dawn, first hour after sunrise | 6–12 K | Good | The fawn-detection window |
| Midday, sunlit open ground | 0–3 K | Poor | Ground can exceed animal surface temperature |
| Midday, shaded woodland floor | 4–8 K | Moderate | Shade preserves usable contrast |
| After rain, wet coat | Reduced 3–6 K | Reduced | Wet fur reads several kelvin cooler |
| Snow cover | 15–25 K | Excellent | Highest contrast condition available |
The most common survey error is flying at the wrong hour. A midday count over open ground can return near-zero detections while the animals are demonstrably present. Schedule counts pre-dawn and treat daylight flights as documentation, not census.
What altitude actually costs you
Published survey protocols commonly fly 60–120 m, and a 640×512 sensor keeps deer-sized animals detectable across that whole band. The choice within it is a trade between disturbance and pixels on target.
Higher is quieter and covers more ground per pass, but reduces the number of thermal pixels across each animal. Detection survives this well; counting does not, because animals in contact merge into a single warm blob long before they become invisible. If your deliverable is a number rather than a presence map, fly lower and accept the narrower swath.
Disturbance is driven more by aircraft size and acoustic signature than by altitude alone. This is the practical argument for a light payload: a 130 g thermal unit such as MV-2P allows a smaller, quieter airframe, and the smaller airframe is what the animals respond to.
Anti-poaching and protected-area patrol
Thermal detection works both ways: rangers find intruders the same way they find animals. AI human and vehicle classification separates the two automatically along patrol routes, which is what makes unattended or semi-attended patrol viable over large reserves.
Two operational points matter more than sensor choice. First, a passive thermal sensor emits nothing — the patrol does not announce itself, unlike a spotlight. Second, detection range against a person is far longer than against most animals because a standing human presents more vertical projected area; you will typically see the intruder before you see what they are hunting.
Programmes that combine both missions on one sortie are the ones that survive budget review. The technique overlaps closely with border and perimeter surveillance.
Where wildlife surveys go wrong
- Counting under canopy. Thermal does not penetrate foliage. Timbered ground always undercounts — report it as a minimum, never as a census.
- Counting after rain. Wet coats shed several kelvin, and they do so unevenly across a herd.
- Claiming species from thermal alone. Size and gait class are reliable; species is not. Confirm with the visible or low-light channel.
- Clustered animals. Bodies in contact merge into one signature. A second pass from a different angle usually resolves the count.
- Descending on nesting or calving sites. The one manoeuvre most likely to cause the disturbance the survey was designed to avoid.
- No fixed protocol between years. Different altitude, hour or route makes trend comparison meaningless.
Recommended UAVThermal payloads
MV-2P — 640×512 thermal at 130 g with NETD below 40 mK. The lightest route to a quiet survey aircraft, and the sensitivity matters on the marginal daytime and wet-coat cases.
MV-4X — thermal plus starlight and NIR in a 366 g quad-sensor pod. Detection and species confirmation on the same pass, without visible illumination.
OP-80N — black-light full-colour night vision with long zoom, for standoff species confirmation and intruder identification at night.
Related reading
Technology: how uncooled VOx microbolometers work and starlight and low-light imaging.
Field practice: livestock counting with thermal drones, thermal drone deer recovery and real detection-range data from six payloads.
Adjacent missions: wildfire early detection and livestock and ranch monitoring.
FAQ
What altitude avoids disturbing wildlife?
Species-dependent; published survey protocols commonly fly 60–120 m, and a 640×512 sensor keeps deer-sized animals detectable across that whole band. Disturbance is driven more by aircraft size and noise than by altitude alone, which is why a lighter payload on a smaller airframe often disturbs less than a higher-flying large one.
Can thermal distinguish species?
Reliably by size and gait class, not by fine species. Pair thermal detection with a zoom or starlight channel for confirmation. Surveys that claim species identification from thermal alone are making a claim the sensor cannot support.
Why did our daytime survey find almost nothing?
Almost certainly thermal contrast, not equipment. On sunlit open ground at midday the surface can reach or exceed animal surface temperature, leaving 0–3 K of difference. Move the flight to pre-dawn, where 12–20 K is typical.
Can we count animals under tree cover?
No. Long-wave infrared does not penetrate foliage, so any count over timbered ground is a minimum rather than a census. Report it as such — presenting an under-count as a population figure is how survey programmes lose credibility.