How to photograph a shelterbelt for porosity analysis

If the analysis produces a porosity figure you don’t trust, the failure is almost never in the algorithm. It is almost always in the capture. An algorithm can filter out bad frames, but it cannot conjure useful information from a photograph that wasn’t there in the first place — a sky framed against a hillside, a canopy lost in shadow, a lens distortion that bends every branch at the edges.

This guide is the capture side of the method. It covers camera position and framing, light, lens choice, drone settings, and the specific failure modes that occur often enough to have earned named diagnoses. The guide pairs with the measurement method, which covers what happens to the photographs once you have them.

Standing position

Perpendicular to the belt. Not oblique, not diagonal. The line from camera to belt should cross the belt’s running direction at a right angle. Two or three belt-heights back, measured horizontally. For a 6-metre belt that is 12–18 metres. Closer than 2 H and the canopy fills too much of the frame, lens distortion foreshortens the top, and you cannot get sky above the canopy for the analysis to orient against. Further than 3 H and atmospheric haze, intervening crops, and framing inconsistency all start to bite.

The camera should be held level. Tilt introduces trapezoidal distortion — the belt appears wider at the bottom of the frame than the top, or vice versa. The analysis measures vertical slices through the belt, and a trapezoidal projection biases the per-slice porosities systematically. A small bubble level on the camera or a phone’s built-in level indicator is worth the five seconds to check.

Aim the camera at the mid-height of the belt, not the canopy or the base. Roughly 40–60% of the frame height should be occupied by the belt itself, with a narrow strip of sky above (5–15%) and as little ground below as you can manage (15–30%). The ground will be masked off in analysis, but less masking is better than more masking.

Light

Evenly overcast is ideal. A uniformly bright sky is the perfect background against which vegetation silhouettes — no colour cast, no directional shadows, no hot spots on the canopy, no deep sub-canopy shadows that look like gaps but aren’t. The measurement at 10:30 on a grey Tuesday in March is cleaner than the measurement at noon on a clear June day.

If you must shoot in direct sun, keep the sun behind you. Front-lit belts (sun behind the camera) show the canopy structure clearly, and the analysis handles the shadowed north-facing side of the canopy correctly. Back-lit belts (sun behind the belt) can work for well-lit dense canopies but are treacherous for leaking belts, because the sun through gaps creates bright spots that the algorithm reads as extra sky. Side-lit belts, where sunlight skims the face of the canopy from one side, are the worst case: the half facing the sun is bright, the other half is shadowed, and the two halves produce different porosity readings for the same belt.

Avoid the first hour after sunrise and the last hour before sunset. Low-angle light creates long shadows through the belt that look like structural gaps, so the porosity reads high.

Lens and focal length

A 35–50mm equivalent focal length on a full-frame sensor is the sweet spot — the flat, central, near-distortion-free region of most lenses. A phone main camera (typically 24–28mm equivalent) is usable but requires standing further back to stay in the middle of the frame where distortion is lowest. Ultra-wide phone modes (0.5x, around 13mm equivalent) are not usable: the edge distortion inflates porosity readings at the sides of the frame by 5–10 percentage points.

Zoom above 85mm equivalent is usable in theory but forces you to stand so far back that haze and foreground obstruction become dominant problems. It also introduces compression that can make a belt appear denser than it is — a visual effect rather than a real porosity change, but the algorithm cannot distinguish.

Turn off any in-camera HDR, Night mode, Portrait mode, or Scene Optimiser. These apply non-uniform tone mapping and selective sharpening that distort canopy edges. Plain camera app, highest-quality JPEG, default exposure.

Drone capture

Drones earn their keep on belts taller than about 5 metres, where standing far enough back on foot is impossible, or on long belts where consistent framing across 150+ metres needs the drone’s altitude control to maintain a constant shooting plane.

Fly at roughly half the belt height. For a 10-metre belt, 5 metres altitude. For a 15-metre belt, 7–8 metres. This positions the drone at mid-canopy, where the analysis is cleanest: the camera looks straight-on at the belt rather than down at the canopy top or up at the trunk zone. Horizontal distance stays at 2–3 belt-heights.

Fly parallel to the belt, capturing one frame per 10–15 metres of belt length. For a 200-metre belt that is 15–20 frames, which gives plenty of overlap for the batch average to be representative. Maintain constant altitude and constant horizontal distance between frames — a DJI or similar with consistent GPS altitude hold is fine for this.

UK operations: flying a sub-250g drone (DJI Mini series, for example) in the CAA Open category A1 or A3 covers most farm shelterbelt work without operational authorisation, provided you stay 50 metres from uninvolved people and have registered as an operator. Heavier drones need an A2 Certificate of Competency and may need landowner permission for take-off and landing. Check the current rules before you rely on them — CAA requirements evolve.

Foreground and obstruction

Every pixel in the bottom third of the frame that is not part of the belt is a potential problem. Crops (even stubble), fences, gates, water troughs, parked vehicles, livestock — all of these look dark to the algorithm and will artificially drag the porosity number down. The ground-cutoff mask in analysis takes care of most of this, but the lower on the frame you can position it, the more of the actual belt you get to measure.

If the belt is immediately bordered by a tall crop (maize, oilseed rape at flowering, mature barley), shoot from a vantage that puts the crop below the camera rather than between camera and belt. Sometimes that means working from a farm track, a field margin, or a raised verge.

Sky behind the canopy, not another belt

The single most common capture failure in UK farm shelterbelt photography is a second tree line behind the subject belt. In hedgerow-rich landscapes — Devon, Somerset, Herefordshire, many parts of the South East — it is common for the next field over to also have a boundary hedge, and that hedge appears in the photograph as dark pixels where the sky should be. The analysis counts those dark pixels as belt material and undercounts porosity, sometimes by 15 points or more.

The fix is to walk around the belt until you find an angle where the background is open sky, ideally for the whole length of the belt. If you can only find a clear-sky angle for part of the belt, shoot only that part, mark the rest as not-yet-measured, and come back another day when you can approach from the other side.

Five capture failures the algorithm cannot rescue

1. Motion blur. Handheld on a windy day with shutter speed below 1/250s. Silhouettes soften, porosity reads high. Use a higher shutter, bump the ISO, or steady the camera against a gate post.
2. Muddy mid-tone exposure. Underexposed overcast where the canopy and sky are both middle grey. The sky/not-sky distinction the algorithm relies on becomes ambiguous.
3. Background tree line. Already covered — the fix is an angle change, not a post-processing step.
4. Side-lit canopy. Low-angle sun from one side produces bright and shadowed halves in the frame that the algorithm reads inconsistently.
5. Tilted horizon. The belt appears trapezoidal, biasing the per-column porosity averages. Use a level indicator.

A reproducible field protocol

A working protocol for a routine farm shelterbelt inspection:

• Conditions: overcast, mid-morning to mid-afternoon, no rain on the lens
• Camera: phone main camera (not ultra-wide) or a 35–50mm lens on a mirrorless
• Stand perpendicular, 2–3 belt-heights back, camera level
• Frame: belt fills 50–60% of frame height, narrow sky strip above, minimal ground below
• One frame every 10–15 metres of belt length, from 10% in from each end
• Tag each frame with GPS location (most phones and drones do this automatically)
• Flag the season (in-leaf, out-of-leaf) and weather in the filename or a notes file
• Import as a batch to the analyzer

Doing this consistently over several years produces a time series that is genuinely defensible. The measurement noise is small enough that year-on-year changes in the 5–10 percentage point range are real, not artefacts of capture.

Frequently asked questions