What can LiDAR data tell a forest manager, and what can't it?

LiDAR measures the shape of the forest: tree and canopy height, canopy cover and gaps, and the ground surface under the trees, across every hectare. It does not measure diameter, volume, species or grade. Those come from models fitted to field plots, and they are only as good as the plots and the model. LiDAR can also miss trees, and published work finds the small ones under the canopy hardest to detect. The useful question for any LiDAR figure is whether it was measured, modelled or not seen at all.

What the usual answer says

Descriptions of LiDAR in forestry list what it provides: accurate, wall-to-wall height, canopy cover, terrain, stem counts and timber volume, often presented as a replacement for much of the cruise. Each item on that list is available. One forestry consultancy's explanation goes further. It calls LiDAR outputs "estimations", which it says are "very accurate when calibrated with field measurements and data". What the pages don't show is how those estimates compare with the trees that were cut and the logs that were scaled.

Measured, modelled, not seen

A 2024 study did the rare thing of following LiDAR inventories through to the felled trees and the mill. Aaron Sparks and colleagues, publishing in Forest Science, compared field cruising and three LiDAR methods in a 1.1 hectare stand against measurements of every felled tree and the log scale. LiDAR heights were as good as field heights, with lower error: "ALS measured heights had lower root mean square error (RMSE) and bias". Diameter, modelled from height, had an error of about 8.3 to 8.5 centimetres. The LiDAR methods "underdetected trees", and accounted for 78% to 91% of the field reference harvested merchantable volume and 71% to 99% of the merchantable volume scaled at the mill.

The authors add a caution: "the results also illustrate challenges of using mill-scaled volume estimates as validation data". The scale is the best record of what came off the stand, and it is still a record with its own rules.

Missed trees are not random. A 2025 benchmark of tree segmentation notes that early methods were designed for airborne data "where understory trees are often not visible", and that "accurately detecting small understory trees remains a persistent challenge". A LiDAR stem count describes the upper canopy better than the stand, and any figure built on it, such as average tree size, carries that tilt.

A total can be right by accident

The second finding matters more for anyone reporting LiDAR volume to owners. Janne Räty and colleagues built large-area forest maps from LiDAR and models trained on harvester measurements over an 8.7 million hectare study area, published in a forestry journal in 2023. Because "harvesters operate in mature forests", the training data were not a random sample of the forest. The bias in stem count was 39%. The bias in volume was 1%, and the authors explain why: "The latter was due to an overestimation of deciduous and an underestimation of spruce forests that by chance balanced".

A volume total within 1% would pass most checks. By forest type it was wrong in both directions, and a business that sells by species or product would have felt it. The authors' remedy is standard practice in LiDAR inventory: "a probability sample of reference observations may be required to ensure the unbiasedness of estimators". LiDAR volume calibrated on properly sampled plots is a different product from volume fitted to whatever data was to hand.

What LiDAR can see with more effort

Quality is the frontier. Nicolas Cattaneo and colleagues, in Forest Ecosystems in 2024, predicted the diameter of the largest branch per log and stem diameter at several heights from dense drone laser scans of individual trees. They describe it as "an important step towards improved forest inventories". It depends on survey-grade drone scans and individual tree segmentation, a far denser dataset than the airborne LiDAR most estates have. For most operations today, grade and defect remain field and mill observations.

How to use LiDAR figures

Label each LiDAR layer as measured (height, canopy, terrain), modelled (diameter, basal area, volume, stem count) or not seen (species where not modelled, grade, defect, rot). Use measured layers freely. For modelled layers, check them against the scale record of stands already harvested, split by forest type and product. Never check them against a single total, for the reason Räty found. Keep the model's date and the plots it was fitted to beside the layer, because both age.

Quarri's own map features are in design, not delivered. This piece describes LiDAR data in general.

When it doesn't apply

Uses that need only height or canopy, such as finding gaps, planning roads or mapping regrowth, can rely on LiDAR directly. Estates that have calibrated LiDAR models on a proper sample of their own plots may do better than the figures above, though the Sparks diameter model was already local and still missed by about 8 centimetres. And the Sparks study covered a single small stand, so its figures show the kind of gap to look for, not its size on any other ground.

Quarri for forest management is built around how a forest operation runs, from the cruise to the settled account.

Quarri is an AI-native data platform for the timber supply chain. It connects buying, production, sales and inventory for forest management, sawmill, wood products and pulp, paper and packaging operators.