A forest in Mäntyharju, Finland. Credit: Jani Riekkinen / EyeEm /Getty Images.
Climate change is threatening almost 60 per cent of European forests by making them more vulnerable to windstorms, fires, and insect outbreaks, a study in Nature Communications shows1. The authors combined satellite data, historical records, and machine-learning techniques to reconstruct the way forest vulnerability to environmental stress has changed in Europe over the last 40 years. They found that a 0.5°C increase in average temperature since 2000 has had a huge impact. “It seems to be a tipping point beyond which forests have begun to lose their natural defence capabilities,” says Giovanni Forzieri, lead author of the study, and a researcher at the Joint Research Center of the European Commission in Ispra, Italy.
Forests cope well with long-term modifications in their environment, but not so well with sudden changes such as the current rate of global warming, in which their trees live for several decades, and cannot be quickly replaced by new, better adapted generations. Climate models have recently begun to incorporate simulations of forest disturbances, but understanding of what drives them is still poor. “The most accurate studies on forest vulnerability concern small areas, so the results cannot be easily be extended to wider regions,” says Forzieri. Other studies aggregate all possible forest disturbances on a national basis, making no difference between, for example, fires and windstorms. “This makes it possible to cover larger areas, but at the expense of the ability to study the effect of a single disturbance” he continues.
The team tried to fill these gaps by integrating real-world data with machine-learning techniques. They used satellite images to study how the key characteristics of European forests (extension, density, biomass content, trees’ age) have changed over the last decades. They used existing databases to reconstruct a history of temperature changes, fires, windstorms, and insect swarms across Europe. Then they created a machine-learning algorithm to look for correlations among these data. In essence, the algorithm learned to predict how much biomass a forest is likely to lose after a specific disturbing event, based on its initial conditions. It found that forest vulnerability mostly depends on the density and height of its trees, the total leaf area, and a few other properties. Most importantly, the researchers calculated that vulnerability is increasing, and that up to 33.4 billion tonnes of biomass (58 per cent of Europe’s total forest mass) could be seriously damaged by future natural disturbances. Windthrows, where trees are uprooted by the force of wind, are the main threats, followed by fires and insects. Forests in Finland, northern European Russia and the Alps emerged as the most fragile ecosystems, followed by warm–dry forests in Spain’s inland.
These findings may inform forest management interventions, in particular to prevent forests from reaching a saturation point where they would stop absorbing carbon dioxide from the atmosphere, further destabilizing the climate system. “The use of big-data techniques, machine learning and modeling will be essential to manage the growing amount of data we have in this area,” Forzieri concludes.
“Cascading effects were not considered in this study, and their inclusion will play an important role in determining biomass loss due to natural disturbances” says Emanuele Lingua, an associate professor of ecology at the Università di Padova who was not involved in the study. “The approach would benefit from a more consistent and accurate database on disturbances, that unfortunately is not available. But this research will be relevant for forest management policies at European level.”
