A satellite vegetation report is about to go into a closure plan, an environmental authorisation, or a lender's file with your name near it. Four technical questions separate a finding that survives review from one that only looks authoritative. None of them require you to be a remote-sensing scientist — but the answers tell you whether one was in the room.
1. Compared to what — and is the comparison controlled?
A vegetation index has no universal pass mark. NDVI — (NIR − Red) / (NIR + Red) — responds to green canopy, but its value at any pixel is pushed around by soil brightness, slope, aspect, the rainfall in the weeks before the image, and the growth stage of the plants. Read as an absolute number, "NDVI = 0.45" means almost nothing.
The defensible design borrows from experiment: measure the rehabilitated footprint against an adjacent, undisturbed reference area under the same geology, rainfall and season, on the same image dates. The reference absorbs the confounders. What you report is then a difference between like and like — the same logic as a control group — not a bare figure lifted from a textbook. If a report cites absolute thresholds with no reference or control, ask why.
2. What did the index actually measure — and where does it saturate?
Indices measure a proxy, and every proxy has a blind spot. NDVI tracks chlorophyll and canopy greenness, but it saturates once the canopy closes (leaf area index above roughly 3), so it can't tell a recovering stand from a vigorous one at the top of the range. It says nothing about species composition — a field of alien invasives can outscore indigenous recovery — and nothing about soil stability, which is often the real closure question.
Good practice matches the index to the question: soil-adjusted indices (SAVI) where cover is sparse and background dominates; red-edge indices, which exploit the sharp rise in reflectance near 700 nm, where you need sensitivity at high biomass. A report that leans on a single index for every site is a report that hasn't asked what it's blind to.
3. How do you know the change is real — and not the sensor or the sky?
Between any two dates, a satellite signal moves for reasons that have nothing to do with the ground: haze and water vapour, the sun's angle, thin cloud and shadow, and differences between sensors (Sentinel-2A vs 2B, Landsat vs Sentinel). A "20% NDVI increase" can be an artefact of atmosphere alone.
The controls are specific and checkable. Work from surface-reflectance (Level-2A) products, where atmospheric correction has already been applied, not raw top-of-atmosphere values. Mask cloud and shadow explicitly (Sentinel-2's scene-classification layer does this). Hold sensor, season and processing constant across the time series. And carry an honest statement of uncertainty. A single-date, two-image comparison with no correction and no error discussion is a headline, not a measurement.
4. What can this not tell you?
This is the question that exposes the analyst. A satellite time series is powerful at describing condition and change; it is silent on cause. It can show, defensibly, that one section of a footprint is not tracking its reference. It cannot say from orbit whether the reason is seed mix, compaction, slope, herbivory or a dry season — attribution needs the ground record alongside the imagery. A report that explains cause from pixels alone has walked past the edge of the evidence.
Why this is worth your time
These four questions are, in order, about controls, proxies, corrections and limits — the same four things a peer reviewer checks in a remote-sensing paper. A report that answers them plainly is one you can put in front of a regulator or a lender and defend line by line. One that can't is a picture with a number on it — and when it's challenged, the picture is what you'll be left holding.
This is the standard we build into every rehabilitation monitoring and environmental baseline deliverable. If you'd like a report that answers all four before you have to ask, send us a brief.