Europe's forests face up to double the damage by century's end if emissions keep rising

Science, 2026. DOI: 10.1126/science.adx6329

Every year, wildfires, storms, and bark beetle outbreaks kill trees across European forests. These disturbances have already reached unprecedented levels in recent decades. According to a new modeling study, they are going to get worse, substantially worse, under every climate scenario examined.

What the model predicts

Marc Grunig and colleagues developed a deep learning framework trained on high-resolution Landsat satellite images to simulate forest disturbance patterns across Europe through the end of the 21st century. The model accounts for complex feedbacks between climate change, forest growth, and interacting disturbances, feedbacks that have made accurate large-scale projections difficult in the past.

Under three climate scenarios, the results were consistent in direction, differing only in magnitude. Disturbance-induced tree mortality will intensify across the continent. In the latter half of the century, if greenhouse gas emissions continue unabated, disturbances are projected to reach levels with no historical precedent.

The numbers are stark. Currently, about 180,000 hectares of European forest are disturbed annually. Under continued high emissions, that figure could nearly double to roughly 370,000 hectares per year by 2100.

Wildfire and bark beetles as twin drivers

Wildfire is the dominant driver of disturbance-induced mortality, particularly in dry Mediterranean regions. But the model also projects fire expanding into areas where it has historically been uncommon. Drier, hotter conditions are pushing fire risk northward and into regions that lack the management infrastructure and experience to respond.

Bark beetle outbreaks are the other major force, especially in temperate central European forests. Warmer and drier conditions accelerate beetle life cycles, allowing more generations per year, while simultaneously weakening tree defenses by inducing drought stress. The combination produces larger and more frequent outbreaks.

These two disturbance types also interact. Fires create dead wood that can fuel beetle populations. Beetle-killed forests become fire-prone. The deep learning model's ability to capture these feedback loops is what distinguishes it from simpler projection methods.

Southern and western Europe hit hardest

The geographic distribution of future damage is not uniform. Southern and western Europe face the strongest changes, driven primarily by wildfire expansion. Northern Europe is expected to be less severely impacted overall, though localized hotspots of future forest damage are also projected to emerge there.

The implications extend beyond ecological concern. European forests are a significant carbon sink, absorbing CO2 from the atmosphere. If disturbances increase tree mortality and reduce forest cover, these ecosystems could shift from absorbing carbon to releasing it, creating a feedback loop that accelerates climate change.

What this means for forest management

The study was led by researchers at the Technical University of Munich, with contributions from the Potsdam Institute for Climate Impact Research and numerous other institutions across Europe. Christopher Reyer of PIK, a co-author, pointed to the practical consequence: if forests take up less carbon, or potentially release more than they absorb, other sectors such as transport and agriculture will need to reduce their emissions more rapidly to compensate.

Forest management will need to shift toward building resilience. This means diversifying tree species, adjusting harvesting practices, and preparing for disturbance types that were previously rare in certain regions.

Model limitations

Deep learning models are powerful but not infallible. The framework was trained on historical satellite observations, and its projections assume that the relationships captured in past data will hold in the future. Under unprecedented climate conditions, novel disturbance dynamics could emerge that the model cannot anticipate. The study also does not account for changes in forest management practices, which could mitigate or exacerbate projected trends.