Blog·

Allometric equations vs. remote sensing biomass estimates

Rows of straight pine trunks in a dense planted forest, sunlight catching the bark at the back of the stand

Every biomass number in a carbon registry submission traces back to one of two root methods, or some blend of both. Allometric equations convert a measured tree (diameter at breast height, sometimes height) into a biomass figure using a regression built from trees that were cut down, weighed, and dried. Remote sensing biomass estimates do something structurally different: they regress canopy structure and spectral response against that same destructive-harvest data, then apply the resulting model across an entire landscape. Neither one is "more accurate" in the abstract. They answer different questions, and a reviewer who treats them as interchangeable will miss the specific place each one breaks: an equation mismatch in allometry, or a calibration drift in the remote sensing layer.

What allometric equations measure

A species-specific allometric equation is only as good as the destructive sample it came from. Someone felled a set of trees of a known species, in a known climate zone, within a known diameter range, and fit a curve to the relationship between trunk size and dry biomass. That's the strength: the number is tied directly to physical mass, not inferred from a proxy.

It's also the limitation. Apply a Southeast Asian dipterocarp equation to an Amazonian stand and the diameter-to-biomass curve won't hold, because wood density, stem form, and crown architecture differ by species and region. Project developers often fall back on pantropical or generalized equations when a true species-specific one doesn't exist for their plot network, and that substitution is one of the largest uncontrolled error sources in a biomass stock estimate. It rarely shows up as an obvious red flag. It shows up as a stock number that's plausible on paper and off by a margin nobody can bound, because the equation choice itself was never flagged for review.

Allometry also has a coverage problem that has nothing to do with equation quality. A plot network, even a dense one, samples points. It doesn't see the stand between plots, and running a fresh destructive or even non-destructive field campaign every reporting cycle to extend that coverage is exactly the recurring cost most registries are trying to get away from.

Where model-based biomass estimation changes the question

Model-based biomass estimation, the remote sensing side of this comparison, takes the same underlying plot or destructive-harvest data and uses it to calibrate a relationship between biomass and something measurable wall-to-wall: canopy height, texture, multispectral reflectance. The regression is then applied across the full landscape, not just the plots.

This solves the coverage problem allometry can't solve on its own, but it introduces its own failure mode. A remote sensing model reads canopy structure and spectral signal as a proxy for biomass. Whatever species composition, age structure, and disturbance history were present in the calibration plots carries straight into that model, because that's the data the regression was built on. Push the model into a stand with a different mix, say a recovering secondary forest versus the mature stand it was trained on, and the estimate drifts in ways that are hard to see without an independent check. That's why a reviewer should always ask what the calibration set contained, not just what the output map claims.

Using both without pretending they agree

The honest way to reconcile the two is to stop expecting them to converge on an identical number and instead use each for what it's built to do. Allometry, applied to a well-chosen species-specific equation, gives you a defensible plot-level anchor. The remote sensing layer gives you landscape-scale allocation between verification cycles, built on canopy structure and spectral response rather than a fresh field campaign every time a reporting period closes. Biomass Mapping exists for that second piece: an annual, wide-swath biomass layer regressed from multispectral imagery, meant to sit alongside your plot network rather than replace the judgment calls that go into choosing the right allometric equation in the first place.

A reviewer who understands both methods' blind spots asks better questions of a submission than one looking for a single trusted number. If your plot network already anchors the equation side and you're trying to close the landscape-scale gap between campaigns, it's worth seeing what that annual layer covers.