How to size a monitoring polygon around a certified forest management unit

Most auditors inherit a monitoring polygon rather than draw one. It's whatever shapefile came with the management plan, usually the concession boundary itself, sometimes with a round-number buffer tacked on by someone who isn't on this file anymore. That's a reasonable starting point and a poor stopping point.
The FMU boundary tells you where the certificate's scope ends. It does not tell you where the risk to that scope originates, which is almost always outside the line, on the access roads that get pushed in from an adjacent concession, a community forest, or unallocated land before they cross into the certified area.
Start from the legal boundary, not the compartment map
Pull the coordinates from the management plan or the concession agreement, not from the harvest block layout inside it. Compartment boundaries shift between cutting cycles. The legal FMU boundary doesn't, and it's the one a certification body or a buyer will actually ask you to defend if a complaint comes in.
Once you have that outer line, the question stops being "where is the FMU" and becomes "how far outside the FMU do I need eyes." That's a buffer question, and it's mostly a question about roads.
How much buffer is enough
Skid trails and logging roads don't respect the concession line. A road built two kilometers outside the boundary to reach a timber stand often continues straight across it, because road-building follows terrain and existing track, not the cadastral edge. If your monitoring polygon stops at the FMU line, you see the road the day it crosses into certified territory and nothing before that, which is the same as seeing the clearing after it's already underway.
A workable buffer catches an access road while it's still forming outside the boundary, without pulling in unrelated concessions or burying your review queue in track segments nobody on your team is responsible for. For most upland terrain, that means thinking in access corridors rather than a uniform ring: trace the drainages and ridgelines that lead into the FMU from the nearest existing road network, and make sure the polygon covers those approach routes for a few kilometers past the boundary, even where the rest of the buffer is thinner.
Where the FMU shares a boundary with another certified or permitted operation, you can usually pull the buffer in tight, since a new road there is a cross-boundary coordination issue, not a surprise. Where it borders open-access or weakly governed land, push it out further. The risk isn't symmetric around the perimeter, so the polygon shouldn't be either.
Redraw it when the concession does
Concession boundaries get amended, swapped, or re-surveyed more often than audit teams update their monitoring layers for it. A polygon built around last year's boundary will either leave a gap along a renewed section or carry dead weight over land that's left the certificate. Treat the monitoring polygon as something you revisit at each boundary amendment and at minimum once a year, the same way you'd revisit a sampling plan.
If you're monitoring more than one FMU, keep the buffer logic consistent across them rather than custom-fitting each one by eye. A documented rule, something like "approach corridors extended three to five kilometers past the boundary along drainage lines, tighter along shared certified edges," is easier to defend to an assessor than "we drew it where it looked right."
None of this requires guessing at imagery resolution or scan frequency to make the polygon useful. What it requires is getting the shape right before anything is pointed at it. Logging Road Detection runs a monthly scan at sub-meter resolution over whatever polygon you give it and routes back the new road and skid trail segments it finds inside that area, so the boundary work above is what determines whether the first warning you get is a track still outside the FMU or a stand that's already been opened up.
If your current polygon is just the concession shapefile with no buffer applied, it's worth redrawing before your next monitoring cycle starts.