Deforestation and Degradation Type Mapping
Field notes on separating clear-cut loss from slow canopy thinning, written for analysts working through the SMFM mapping sequence.
Further Detail of SMFM Mapping Tools Below:
Forest loss is not one event. A stand can be opened by a single harvest, thinned over several seasons, or hollowed out by fire at its edges. Each pattern leaves a different signature in the imagery, and each calls for a different label.
This note sets out how the degradation type mapping separates those signatures, and where the classes tend to blur.

Intact canopy in the reference tile. Closed crowns give a high and stable near-infrared response, which is the baseline every later class is measured against.
Method notes
The mapping runs in three passes. Each pass is checked by hand before the next one starts.
- Clear-cut: abrupt drop in canopy cover within one composite period
- Selective removal: patchy gaps with intact crowns between them
- Thinning: gradual reduction in cover across several years
- Edge fire: loss concentrated along a burn perimeter
Thresholds are set per biome and revised when reference plots disagree with the automated label.

A clearing that has opened in the last composite period. The bright ground and sharp boundary are the usual marks of a clear-cut.
Sequence of review
Step one
Stack and screen
Cloud-free composites are built and the baseline canopy index is computed for each tile.
Step two
Label change
Change events are assigned a degradation type using the rules in the method notes.
Step three
Field check
A sample of labels is compared against reference plots and corrected where needed.

Patchwork landscape at the forest edge. Here the boundary between thinning and clearing is hardest to read, so the edge rule is applied with extra care.
Labels are provisional until the field check is complete. Revisions are logged against the tile identifier.
Continue to the resources page for reference datasets, or return to the top of this note to reread the method.

Regrowing woodland after selective removal. Degradation labels stay in place here, because recovery is tracked as a separate class in later passes.