The forest
remembers.

A forest looks settled from a distance. Its history is less settled. After the last ice age, trees expanded from refuges as climates warmed; species arrived at different speeds and occupied changing combinations of soils, temperatures and disturbance. Later, farming, grazing, fuel demand and construction altered those landscapes again. The forest we recognize today is the outcome of those changes.

The human part of that history is much older than industrial forestry. Farming spread through southeastern and Mediterranean Europe from around 6000 BCE and then north of the Alps. Clearing opened land for crops and settlement, while the woodland that remained supplied construction material, tools and fuel. The arrival of metalworking added demand for charcoal. These were changes in the relationship between a productive landscape and the people drawing on it, not one simultaneous clearance of an entire continent.

By the High Middle Ages, population growth and economic activity intensified pressure on woodland in large parts of Europe. Glassmaking, mining and smelting added to household demand. Yet forest history does not follow a single downward line. Woodland could return when settlement and cultivation contracted; institutions and management practices changed what could be removed, where and by whom. Local histories of use matter because they help explain why the landscapes entering the satellite record are already different.

Sediments retain a record that a modern inventory cannot recover. The REVEALS model translates pollen into estimates of vegetation cover, correcting for differences in pollen production and dispersal. Its European reconstruction draws on 1,128 pollen records. Here, six selected windows show reconstructed woody cover in one-degree cells. Pale gaps mean missing evidence, not a treeless landscape.

Some cells draw on a large lake or several sites; others rely on one small lake or bog. Changing evidence coverage matters as much as changing color. The displayed windows begin around six thousand years before 1950, within the larger reconstruction’s 11,700-year coverage. They do not form a continuous annual census. The recent window is nominally 1850–1950 CE in the supplied labels, while the paper describes its upper boundary as the coring year; neither makes it a map of 2024.

Woody cover combines the reconstruction’s evergreen and summer-green components, including its constituent woody taxa. It is not a national forest-land definition. Component standard errors are retained separately because their covariance is unavailable. A precise-looking uncertainty interval for the sum would conceal an assumption that the source does not support. The historical map is useful precisely when its changing coverage and different truth status remain visible.

In 1713, Hans Carl von Carlowitz published Sylvicultura oeconomica. Its title connects growing trees with households, building, brewing, mining and smelting. The problem was already economic: how can a society use a productive resource without consuming its future supply? Today, the investment horizon has an additional complication. A stand may mature under a climate materially different from the one in which it was established.

Eurostat counted 3.6 million people across forestry and wood-based industries in the EU in 2022. That broad total includes paper, wood products and furniture; forestry and logging itself accounted for about 540,000 employed and self-employed people. EU roundwood production reached 460 million cubic meters in 2023. Disturbance can travel from forest owners to sawmills, transport, construction and local labor markets.

Salvage logging can temporarily increase the supply of wood after trees have been damaged, even as the productive base becomes less dependable. A rise in removals can reflect planned harvesting, commercial demand or the response to a disaster. A timber statistic alone cannot distinguish those mechanisms.

The timing matters for value. An owner may receive money from salvage now but lose the option to sell mature timber later, while facing re-establishment costs before the next harvest. Processors may see plentiful raw material during the salvage phase and less dependable supply afterward. That mechanism is economic interpretation, not a loss estimate derived from these pixels. Measuring it would require prices, wood quality, costs, stand age and the intended management path.

A policy that counts hectares therefore needs a companion account of condition. The same land area can contain mature stands, young regeneration, recently harvested patches and damaged trees awaiting removal. The satellite record adds a history of canopy disturbance, while inventories and field observations are needed to establish whether recovery is restoring the functions that matter. Neither source can perform the other’s entire job.

The 2024 map aggregates every mapped forest pixel into a ten-kilometer equal-area grid. Each cell’s color represents disturbed forest divided by mapped forest within that cell. It does not divide by national land area, nor does a larger country automatically receive a darker color. The national rate is the result of adding those numerators and denominators, not averaging the colored cell percentages.

The official-area check uses Eurostat forest closing stocks for 2023. An administrative inventory and a satellite forest mask need not have identical definitions or boundaries. France’s comparator is metropolitan France, including Corsica. Spain’s EFDA raster excludes the Canary Islands, while the official national total includes them: this is a qualified geographic scale check. Spain’s official value is imputed and France’s is estimated and provisional. A fixed comparator changes levels, but cannot change a country’s year rankings or post/pre ratio.

The horizontal coordinate is the share of mapped forest disturbed at least once from 1985 to 2024. The vertical coordinate is the share disturbed more than once, using EFDA’s distinct-event count. Both use the same mapped-forest denominator and forty-year window. Repeated area is a subset of ever-disturbed area, so a positive relationship is partly mathematical; it cannot establish a mechanism.

A repeated footprint is evidence of recurrence within the observed window. It is insufficient evidence that a forest failed to recover. The interval between events, their severity, their agents and the regeneration between them matter. Two harvests decades apart do not by themselves demonstrate chronic ecological stress.

Looking back to 1985 gives later years more chances to find a previous disturbance. The companion data therefore include a fixed ten-year lookback from 1995 onward. It counts current-year detections with an earlier detection in the preceding decade. This avoids future leakage and equalizes the lookback, while leaving recovery unmeasured.

The main estimates retain observed signed values, including those outside the interval stated in the metadata. Range diagnostics and a restricted-range sensitivity are supplied. All retained values must pass the physical dNBR range check. Negative values remain in the calculation.

The source supplies each pixel’s latest disturbance year. A greatest-disturbance-year layer is absent in v3, so the companion calculation derives the year with the largest valid dNBR among detected events, selecting the earliest year for ties. This is a spectral definition of greatest, with limited ecological meaning.

The principal denominator is the fixed EFDA forest mask. The sensitivity uses Eurostat’s 2023 closing stock of forest hectares, excluding other wooded land and never substituting total national area. Definitions, geographic scope and inventory methods differ, so the result is a scale comparison rather than a harmonized forest-at-risk estimate.

Replacing one positive constant forest denominator with another changes the level of every annual rate in a country by the same factor. It cannot change that country’s year ranking or its ratio of post-period to pre-period means. Across-country levels and area-weighted aggregates can change, which is why both denominators are retained rather than declaring one interchangeable with the other.

Spain’s national official total includes the Canary Islands, while the EFDA extent does not. France is metropolitan, including Corsica. The Spanish official value is flagged as imputed; the French one is estimated and provisional. Those limitations accompany the sensitivity, instead of being hidden by additional decimal places.