Skip to content
Resources

How to fly an orchard with a drone for tree counting

Altitude, resolution, overlap and time of day: what makes a flight usable for inventorying every tree instead of just taking a nice picture. With a resolution calculator per drone.

· 8 min read

First decide what you are going to measure

A flight for counting trees is not the same as a flight for looking at the orchard. To count, every canopy has to stand apart from its neighbor and from the ground, the canopy has to be measurable, and the mosaic must have no holes or ghosts. That fixes three things before take-off: resolution, overlap and light.

Resolution is measured in centimeters per pixel (ground sample distance, GSD). At 5 cm per pixel a 4 m canopy is 80 pixels wide: enough to detect and measure it. At 10 cm per pixel it is 40 pixels, and young 1.5 m canopies become 15-pixel blobs that the model confuses with weeds. Our recommendation for inventory is 2 to 5 cm per pixel; for agave, where the adult rosette is 1.5 to 2 m, 2 to 3 cm.

Altitude: the drone decides the resolution, not you

Every camera has a fixed resolution at each altitude. A Mavic 3 Multispectral or a Mavic 3 Enterprise gives about 2.7 cm per pixel at 100 m with the RGB camera; the multispectral sensor on the same drone gives about 5.3 cm at that altitude, because it has four times fewer pixels. A Phantom 4 RTK sits at 2.7 cm at 100 m. The calculator below gives you the altitude for the resolution you want, and the other way round.

Two practical consequences. First: if you fly multispectral and want 5 cm, you have to come down to about 95 m, not 150. Second: flying lower multiplies photos and time. Going from 100 to 70 m improves resolution by 30 % and doubles the number of photos. For counting adult trees, 100 m with RGB is usually the sweet spot; for young orchards or agave, 70 to 80 m.

Resolution (GSD) calculator by drone

Resolution at that altitude
cm/pixel
Photo width on the ground
m
Altitude for the resolution you want
m

Factory values published by each manufacturer; real resolution varies ±5 % with camera calibration and the effective height above the canopies.

Overlap: what makes the mosaic possible

Photogrammetry reconstructs the terrain from points that appear in several photos. Over tree canopies, which move with the wind and all look alike, it needs more photos per point than over bare ground. Our minimum is 75 % front and 70 % side overlap; for closed-canopy or windy orchards, 80 % and 75 %.

With less overlap the software still finishes the mosaic, but with duplicated canopies, curved rows and holes where it could not match photos. That error is invisible in the overall image and very visible in the count: a tree split in two is two trees.

Time of day and light

The long shadows of early morning and late afternoon hide whole canopies on the side away from the sun and stretch silhouettes. The good window is two hours after sunrise to two hours before sunset, and better still around solar noon, when the shadow falls under the canopy.

Uniform overcast is better than sun with scattered clouds: passing clouds change the exposure between photos and the mosaic comes out patchy. For multispectral, the calibration panel is photographed before and after the flight, and with intermittent clouds the calibration stops being valid.

Avoid wind above 8 m/s (about 30 km/h or 18 mph): canopies move between photos and point matching fails. And do not fly right after rain: wet leaves glare and saturate the sensor.

Flight plan: a checklist for the pilot

What we check before every inventory flight:

  • Plot polygon with a 20 m margin outside the boundary, so the edge rows get full overlap
  • Constant altitude above ground, not above the take-off point: on slopes use terrain following or split the polygon by elevation
  • Flight lines perpendicular to the planting rows when possible: it improves matching between passes
  • A speed that does not cause motion blur: at 100 m and 2.7 cm per pixel, up to 12 m/s with a mechanical shutter; less with an electronic shutter
  • RTK GPS correction if you will compare flights against each other; without RTK, three ground control points per plot
  • Calibration panel before and after (multispectral), and enough batteries to finish the whole polygon in the same light window

How many photos and how much they weigh

At 100 m with a Mavic 3 Enterprise and 80/75 overlap you get about 45 to 55 photos per hectare, 8 to 10 MB each: around half a gigabyte per hectare. A 50 ha orchard is about 25 GB of raw photos. With multispectral, five smaller files are added per shot, and the total is around 0.8 GB per hectare.

This matters for the upload: a normal upload link moves 20 to 50 GB overnight. For large campaigns we provide a bucket of your own and it is uploaded in parts.

Frequently asked questions

Is a drone without RTK good enough?

For counting and measuring, yes. RTK matters when you want to compare two flights of the same plot with decimeter precision, or when you will export the inventory to a GIS with other layers. Without RTK, three well-measured ground control points achieve the same.

Can I fly higher to cover more and then zoom in?

No. Resolution is fixed at flight time; zooming afterwards only enlarges the pixels. If the young tree is 12 pixels, it will still be a 12-pixel blob.

Do I need multispectral to count?

No. Good-resolution RGB is enough to detect and measure. Multispectral adds the vigor lenses (NDVI, NDRE) and helps separate crop from weeds in overgrown plots.

Other guides

Already have the flights? We'll process them

Send us the raw photos of one flight of up to 20 hectares and we return the orthomosaic and the inventory of that plot, so you see the result before deciding anything.

By submitting you accept our privacy notice.

Or write to us at [email protected]