On paper, Sentinel-2 revisits any point on Earth every five days at 10 m resolution, free of charge. In practice, a project in equatorial Indonesia may end up with a handful of usable cloud-free optical observations in a year. The revisit interval is not the constraint; the weather is.
What the Sensors Actually Offer
| Source | Resolution | Revisit | Cloud limited? |
|---|---|---|---|
| Landsat 8/9 | 30 m | ~8 days combined | Yes |
| Sentinel-2 | 10–20 m | ~5 days | Yes |
| Sentinel-1 (radar) | ~10 m | ~6–12 days | No |
| Commercial optical | Sub-metre to 3 m | Tasked | Yes |
The first two are free and well documented, which is why almost every forest monitoring system is built on them. The third is the answer to the cloud problem and the one most projects underuse.
Why Radar Changes the Design
Synthetic aperture radar penetrates cloud. Sentinel-1 therefore delivers the observation cadence the optical sensors promise but cannot keep in the humid tropics.
The trade is interpretability. Radar backscatter responds to structure and moisture, not to the greenness an optical index measures, so a change in signal requires more careful attribution. Rainfall alone shifts backscatter. Radar is excellent at flagging that something changed and weaker at saying what.
The practical architecture most robust systems converge on: radar for continuous change detection, optical for confirmation and classification when a clear scene arrives, ground visits for the subset that matters. Each layer covers the previous one's blind spot.
Frequency Is Set by What You Need to Detect
A project detecting large-scale clearing can work with annual optical composites. A project needing to catch encroachment early, while intervention is still possible, needs something closer to continuous.
These are different systems with different costs, and the choice should follow from the project's actual management need rather than from what the methodology minimally requires. Methodology compliance sets a floor; operational usefulness usually sets a higher bar.
Where Ground Data Remains Irreducible
Remote sensing tells you where and when the canopy changed. It does not tell you the biomass that was there, and it does not calibrate itself.
Every change-detection system needs ground reference data for two purposes: training the classifier, and independently assessing its accuracy. These must be separate samples. A classifier assessed on its own training data reports an accuracy that means nothing.
The same applies to biomass. Satellite-derived biomass estimates are models fitted to ground plots; without local plots the model is being extrapolated outside its calibration range, and the uncertainty that introduces is frequently larger than the signal being measured.
The Honest Summary
Free satellite data has made continuous monitoring affordable and has not removed the need for fieldwork. What it has changed is the allocation: fieldwork is now targeted where the imagery says something happened, rather than distributed blindly across a concession. That is a large efficiency gain, and it is a different thing from replacement.
Move beyond estimates
Verifiers test sampling design, uncertainty and whether a number traces back to the field. TREEO dMRV captures that evidence in real time, in one auditable chain.



