How to Use Satellite Imagery to Track Heather Burning and Peatland Recovery

A keeper walks a line of heather along a slope above the valley and puts a torch to it. By teatime there is a black patch the size of a tennis court. Standing next to it, the burn looks permanent. From orbit it is one data point in a record stretching back years — and if you know where to look, you can follow that patch as it greens up again, season after season, without leaving your desk.

Satellites will not tell you everything about a moor. They cannot name the mosses creeping back across a bare peat flat. But they can show you, cheaply and consistently, where fire has been, how hard it burned, and whether the ground is recovering or slipping backwards. What follows is a workflow you can run yourself.

What satellite imagery can and cannot show you

The workhorse is Sentinel-2, part of Europe's Copernicus programme. It gives you free imagery with 10 m pixels in the visible and near-infrared, plus two shortwave infrared bands at 20 m. Landsat adds a coarser 30 m record going back decades. Sentinel-1 provides radar that cuts straight through cloud, and commercial constellations offer 3 m imagery for a price.

Ten metres sounds precise until you do the arithmetic. One pixel is 100 square metres. A narrow muirburn strip, or a small gully repair, may only partly fill a pixel. Species-level detail is out of reach, and so is anything beneath the surface: a satellite measures reflected light, not peat depth or carbon storage.

What it does superbly is consistency. The same sensor, the same calibration, the same view — every few days, for years. That is what makes change detectable.

Start with a baseline, not a burn

Most people come to this after something has happened: a burn they want to quantify, or a restoration scheme they need to report on. By then the most valuable image — the one from before — may be harder to find.

So do this first. Draw your area of interest as a polygon, ideally following estate or management boundaries rather than a rectangle of convenience. Note the dates of any burning, ditch blocking, or bare-peat work. Then pull imagery from two or three years earlier, in the same season, so you have a reference point waiting when you need it.

Keep a simple log: date, what happened, where, and who did it. Half the confusion in remote sensing comes from a spreadsheet that never got updated.

Choosing imagery that survives a cloudy climate

A single clear image on the right day is a luxury on a British moor. Plan around it, not for it.

• Copernicus Browser and EO Browser — free, fast, good for eyeballing a burn or a restoration site before you commit to any analysis.
• Google Earth Engine — free for research and non-commercial use. The right tool for building monthly composites and long time series across many sites.
• USGS EarthExplorer — the Landsat archive, which reaches back into the 1980s. Noisy at 30 m, but useful for context.
• Sentinel-1 radar — roughly weekly passes and no cloud problem. Radar is sensitive to surface roughness and moisture, so it notices wet ground and bare peat where optical imagery is blind.
• Drones and aerial photos — for validation, not for the whole moor.

Use surface reflectance (Level-2A) products where you can, and build median composites over a month rather than relying on one scene. A median composite knocks out most cloud and haze without you having to hand-pick pixels.

Turning pixels into a burn scar

Freshly burned ground reflects very differently from healthy heather. It loses near-infrared reflectance and gains in shortwave infrared. The Normalised Burn Ratio captures exactly that contrast:

NBR = (NIR − SWIR2) / (NIR + SWIR2)

In Sentinel-2 terms that is band 8 against band 12. Healthy vegetation scores high; a fresh burn scores low. Subtract the post-burn image from the pre-burn image and you get dNBR, a map of severity rather than just extent.

1. Pick a pre-fire image from the same season as the post-fire one — late winter against late winter, ideally.
2. Calculate NBR for both dates.
3. Subtract: dNBR = NBR(pre) − NBR(post).
4. Apply thresholds to split unburned, lightly burned and severely burned ground. The widely used severity ladder starts around 0.1, 0.27 and 0.66, but treat those as a starting point and calibrate against your own site.
5. Clean up obvious nonsense: cloud shadow, standing water, dark rock and wet peat all mimic burns.

One caution. Repeated burning on a rotation creates a patchwork of heather at different ages, and old burns can look quite similar to new ones in a single index. Colour, texture and context — slope, proximity to tracks, the shape of the patch — all help. A burn edge is usually straighter and more angular than natural variation.

Tracking recovery over years

Recovery is a time series, not a before-and-after. Build a stack of composites and follow the same pixels through it.

The simplest measure is NDVI, which tracks greenness, but on peatland it flatters you: a flush of grass can raise NDVI without any improvement in peat condition. Pair it with a moisture index and with the SWIR bands, which respond to bare, dry, eroding surfaces. On restoration sites where ditches have been blocked, radar backscatter often shifts before the vegetation does, because the ground is holding water again.

Compare like with like. A January image against a July image will show you a difference in season, not in recovery. And give a site three to five years before drawing conclusions — heather and cottongrass do not rush.

Checking the map against the ground

No index substitutes for walking the site. Fix a few photo points with a phone or a compact camera and shoot them every year from the same spot, same rough framing. If you have a drone, a flight at two to five centimetres per pixel will settle arguments about a boundary in minutes.

Compare a handful of quadrats or plot photos against your classified map. Where they disagree, the ground wins, and your thresholds need adjusting. That feedback loop is what turns a pretty map into something you can defend in a management plan or a funding report.

A routine you can run this season

Keep it modest and repeatable. One well-built time series beats twenty abandoned experiments.

1. Draw your boundary and log the site history.
2. Gather three to five years of pre-change imagery.
3. After the burning season or the restoration work, build a monthly composite.
4. Run NBR and dNBR for extent and severity; run NDVI and a moisture index for recovery.
5. Check the results against photos from the ground.
6. Repeat each year at the same time of year.

Note that burning on moorland is regulated, and the rules differ across the UK nations and have changed in recent years. Check the current requirements with the relevant authority before planning any burn — the imagery will only tell you what happened, not whether it was permitted.

Do the six steps above and you will end up with something rare on a moor: a long, consistent record of what changed and when, built from free data and a couple of afternoons a year.

Photo: 14398 / Pixabay