Satellites Take the Guesswork Out of Counting Europe’s Timber Growth

Every year, Europe’s forests quietly add millions of cubic metres of wood, and every year governments are asked to put a number on it. Under the United Nations’ System of Environmental-Economic Accounting Ecosystem Accounting (SEEA EA), countries must record the flows of ecosystem services, including wood provisioning, in physical or monetary terms so that the value of standing forests can sit alongside GDP-style national accounts. Until now, the standard way of doing this has leaned heavily on official forestry statistics, which are slow to compile, coarse in resolution and blind to the patchwork reality on the ground. A new study published in One Ecosystem offers a strikingly different route: measure forest growth directly from space, pixel by pixel, and build the accounts from the bottom up.

The research, led by a team working within the European Commission’s PEOPLE-EA project, introduces a method that transforms freely available Earth Observation (EO) data into high-resolution maps of Net Annual Increment (NAI), the standard forestry measure of how much merchantable timber a forest adds each year after subtracting mortality. Rather than taking a national total from statistics and spreading it across a map, as the existing INCA accounting tool does, the new approach calculates growth for every 20-metre pixel of forest and then aggregates upward. The team demonstrated the method across Slovakia, where high-quality field data allowed rigorous validation, and applied it to Greece, Italy and Norway to prove it scales.

The starting point is a careful map of where living trees actually stood at the beginning of the 2021 accounting year. The researchers began with the Copernicus High Resolution Layer Tree Cover Density from 2018, then masked out every area of tree loss recorded between 2018 and the end of 2020 using the Global Forest Change dataset. For the accounting year itself, they combined global fire maps, national harvest records and disturbance maps to separate natural mortality, which reduces the living stock, from human removals, which are counted as harvest rather than loss. Areas below 10 percent tree cover or smaller than half a hectare were excluded to comply with the Food and Agriculture Organisation’s definition of forest.

With the living forest mapped, the heart of the method begins: converting satellite-derived productivity into timber volume. The Copernicus Gross Dry Matter Productivity (GDMP) product, delivered every ten days at roughly 300-metre resolution, tracks how much dry matter vegetation produces based on radiation, temperature, land cover and the fraction of absorbed photosynthetically active radiation. The team summed the 36 ten-day rasters into an annual total, resampled it to 20 metres, and then applied a conversion factor of 0.464 to account for the roughly half of gross productivity that plants immediately burn through respiration, leaving the Dry Matter Productivity (DMP) that becomes actual structural biomass.

Because wood density differs dramatically between species, the researchers split the forest into hardwood, softwood and mixed categories using the Copernicus Dominant Leaf Type layer. Dividing DMP by category-specific wood densities, 0.573 tonnes per cubic metre for Slovak hardwoods and 0.411 for softwoods, converts mass into volume, yielding what the team calls the total green biomass increment, which includes roots, trunks, branches and foliage. A further correction isolates the merchantable fraction, the stem and branch wood that actually reaches the sawmill. The researchers applied an increment correction factor, estimated at 50 percent, because foliage regrown each year can consume up to half of a young tree’s annual biomass increment, a subtlety that standard stock-based conversion factors miss entirely.

One final adjustment acknowledges that not every pixel is wall-to-wall forest. Where tree cover falls between 10 and 50 percent, the calculated increment is scaled down in steps, from 75 percent of full value at moderate cover to just 25 percent at the sparsest forested pixels. The result is a NAI map at 20-metre resolution, masked to the Forest Available for Wood Supply (FAWS) identified by the Joint Research Centre, which accounts for altitude, slope, distance to roads, protected areas and unproductive ground.

The validation against Slovak forestry records was where the method earned its keep. Drawing on 63,367 forest stands of pure European beech or Norwegian spruce, with basal area, height and age measured by standard practice, the team built an empirical reference dataset of parcel-level NAI. The three methods landed remarkably close on averages: the EO approach estimated 4.88 cubic metres per hectare per year, INCA 5.29 and the empirical reference 5.24. But the averages conceal a crucial difference in behaviour. INCA systematically overestimated NAI in beech stands and underestimated it in spruce, while the EO method tracked the reference values far more consistently across both species.

Statistical tests confirmed the pattern. Both methods correlated significantly with the reference data, and INCA’s regression fit was marginally tighter, with an R-squared of 0.258 against 0.244 for the EO approach. The researchers attribute this to the EO method’s willingness to capture a wider range of values, including the high-growth extremes that INCA’s homogenised, statistics-driven smoothing smooths away. Repeated Wilcoxon signed-rank tests on random subsamples of 500 parcels confirmed the differences between the two methods were real, not artefacts of the enormous sample size. At the regional NUTS-2 level, EO-derived totals matched the reference data more closely than INCA’s estimates.

The method is not without weaknesses, and the authors are candid about them. The GDMP product’s native 300-metre resolution must be resampled upward, blurring estimates in heterogeneous landscapes. The Hansen tree-loss layer records loss only as a binary per-pixel flag, so partial disturbances and selective logging can slip through undetected, potentially inflating the living-tree area. The static respiration and increment correction factors ignore variation with species, age and site quality, and Slovakia’s predominantly old-growth forests may be systematically overestimated, since dense canopies read as high productivity from orbit even when the timber below grows slowly. Incorporating forest age, soil, elevation and microclimate data, the team argues, would sharpen the estimates considerably.

Even so, the implications are substantial. Ecosystem service accounting becomes mandatory across the European Union from 2026, and the EO approach offers member states a way to comply without the enormous field campaigns that traditional inventories demand. Because it rests entirely on open, regularly updated satellite data, it can be refreshed yearly, extended to continental scale as finer tree-genus maps such as the ForestPaths European tree genus map come online, and even applied in regions that currently lack any forestry statistics at all. The authors suggest that with the right inputs, the same pipeline could produce rough NAI estimates anywhere trees grow, turning a bureaucratic reporting burden into a continuously updated, globally consistent picture of one of the planet’s most economically vital ecosystem services.

Subject of Research: Using Earth Observation satellite data to compile wood provisioning ecosystem service accounts by estimating forest Net Annual Increment

Article Title: A bottom-up approach to compiling wood provision ecosystem service accounts with EO data

Article References: Evans, A., Buchhorn, M., Černecký, J., Halabuk, A., & Smets, B. (2026). A bottom-up approach to compiling wood provision ecosystem service accounts with EO data. One Ecosystem, 11, Article e184431. https://doi.org/10.3897/oneeco.11.e184431

Image Credits: AI Generated

DOI: 10.3897/oneeco.11.e184431

Keywords: Earth Observation, ecosystem accounting, Net Annual Increment, wood provisioning, forest monitoring, Copernicus, SEEA EA, remote sensing, forest productivity, INCA tool, Slovakia, satellite data