Blog·

Field notes: the defoliation year nobody called a disturbance

Dense green canopy of a mixed hardwood forest, viewed from below through the leaves and branches.

The first time the number dropped, nobody said anything. A two-percent dip in a 400-hectare hardwood block is the kind of thing that gets waved through. Plot remeasurement error, a slightly different flight window, maybe the allometric model doing what allometric models do when the stand is mid-succession and the diameter distribution is still sorting itself out. You write it down, you move on.

The second year it dropped again, by more, and that's when the questions start sounding different. Not "is this noise" but "what happened out there."

What had happened, it turned out, was a forest tent caterpillar outbreak that nobody on the monitoring side had been told about. The silviculture crew knew. The district forester knew. It showed up as a line item in a pest report that circulated to exactly the people who manage pest reports, which is not the same list as the people who sign off on carbon-stock numbers.

By the time the biomass layer and the pest report ended up on the same desk, the defoliation had already run two growing seasons and the canopy had thinned enough that the regression was picking it up as a straightforward loss. It showed up as fewer leaves, lower NDVI, and a lower biomass estimate coming out the other end of the same model that had been running fine for six years. The cause was insects, not a harvest, blowdown, or boundary error.

This is the part that's easy to miss if you only look at the number and not the shape of the number. A plot-based estimate from a bad field season reads as scattered error. A real defoliation event reads as a spatial pattern: a contiguous footprint that tracks the host species and follows the outbreak front the way insect damage moves, nothing like the straight lines a harvest cut leaves or the storm-track scatter a blowdown leaves.

Harvest boundaries follow property lines. Blowdown follows topography and storm track. Defoliation follows the trees the insect eats, and if you know what you're hosting on that stand, the pattern tells you what you're looking at before any ground crew gets out there to confirm it.

The second thing that's easy to miss is persistence. One bad year of canopy vigor can be weather: a spring drought, a late frost that hit the flush. Defoliation that keeps suppressing growth shows up again the following year, and a third year if the insect population hasn't crashed. Estimation noise doesn't behave that way. Noise wanders. A real disturbance signal, even a slow one like chronic defoliation, holds its shape and keeps moving in the same direction year over year until the stand recovers canopy or something kills the outbreak.

None of this makes the reviewer's job simple, because a carbon project doesn't get to just write "it's probably bugs" into a monitoring report and move on. A documented loss event needs a cause, a footprint, and ideally a tie to whatever pest surveillance exists for the region, which is often thin and often a year behind the imagery anyway. But catching the pattern early, in year one or two instead of year three, changes what kind of conversation you're having. It's the difference between flagging a disturbance proactively and explaining, after a verifier has already asked, why the stock numbers moved and nobody mentioned it.

That gap is exactly why an annual pass matters more than people give it credit for. A biomass layer rerun off the same wide-swath imagery every year, built the same way each time, is what lets you tell drift from event in the first place, which is the whole case for running this through a landscape biomass layer instead of waiting on the next field campaign to tell you what already happened.

If your monitoring cycle has a defoliation-shaped gap in it somewhere, worth having a look at how the numbers would have moved if you'd been watching every year instead of every five.