Some of Europe’s last extensive near-natural forests are found in the Carpathian Mountains (Fig. 1). Across large parts of the range, particularly in Romania, forests of beech, spruce and mixed mountain woodland still support wolves, lynx and some of the largest brown bear populations in Europe. Although these landscapes may appear timeless, they have changed repeatedly over the last 12,000 years as climate shifted and, increasingly during the later Holocene, as humans transformed the environment.
Fig. 1: The location of the Carpathian Mountains in Europe
Fortunately for scientists, the Carpathians also preserve remarkable archives of these environmental changes. When the last Ice Age ended, many glacial basins, volcanic craters and mountain depressions became lakes and peatlands. Over thousands of years, sediments accumulated on their bottoms, trapping mineral particles, organic matter and vast numbers of microscopic pollen grains released by the surrounding vegetation (Fig. 2).
Fig. 2. Sediment coring at Lake Sfânta Ana, the youngest crater lake in the Carpathian Mountains, located in the Ciomadul Mountains of Transylvania
Each layer of sediment therefore contains information about the forests that existed at a particular moment in the past. By studying these fossil pollen records, researchers can reconstruct the long-term history of vegetation and climate.
For decades, researchers have used Carpathian pollen records to investigate forest history. But an important question remained: could these same pollen grains be used to reconstruct quantitatively how warm summers and winters were throughout the Holocene?
An idea born in the Carpathians
The origins of the study go back to 2019, when Heikki Seppä and María J. Ramos-Román from the University of Helsinki visited the Carpathians together with palynologist Enikő Magyari (Fig. 3).
Fig. 3. Transport of a sediment test core in the Bucegi Mts, Southern Carpathians in 2019; back: Enikő Magyari (co-author), Zoltán Szabó (former PhD student of Enikő), front: María J. Ramos-Román (former postdoctoral researcher of Heikki) and Heikki Seppä (co-author)
At the time, their discussions focused on the environmental context of prehistoric population movements, particularly the westward expansion of Yamnaya groups during the Early Bronze Age. However, while examining the region’s exceptional pollen records, a different question emerged.
The Carpathians contain numerous high-quality fossil pollen sequences, yet relatively few had been used for quantitative temperature reconstruction. More importantly, no study had produced a coherent reconstruction of both summer and winter temperatures spanning the entire Holocene.
The region seemed ideal for such a study. Along mountain slopes, tree species replace one another in response to temperature changes. Even small climatic shifts can alter the competitive balance between species, eventually reorganising entire forest communities. These ecological changes are clearly recorded in fossil pollen.
However, not all pollen sites are equally suitable. Lowland records are often influenced by summer drought, soil moisture and human activity, while very high-elevation sites may be dominated by spruce for thousands of years, making subtle climatic changes difficult to detect.
The most promising locations appeared to be mid-altitude sites, where many important European tree species occur close to the limits of their ecological tolerances and where forest composition responds sensitively to seasonal temperature changes (Fig. 4).
It was a good idea. But, as often happens in science, it did not immediately become a paper.
Fig. 4: Mid-altitude forest in the Carpathian Mountains. The Carpathian Mountains preserve some of Europe’s largest relatively natural forest landscapes, but their apparently timeless forests have undergone profound changes during the Holocene; photo courtesy of Gusztáv Jakab
Returning to the question
The opportunity finally arrived through the DOMINO-CLIMATE project, led by Enikő Magyari, which investigates interactions between climate, environmental change and prehistoric societies across East-Central and Southeast Europe.
In 2024, the original idea resurfaced during meetings in Helsinki. By this stage another key researcher had joined the project: Jon Camuera, a former doctoral researcher of Heikki Seppä and a specialist in quantitative pollen-based climate reconstruction. Jon contributed expertise in pollen-climate transfer functions as well as the statistical tools needed to distinguish genuine climatic signals from noise in palaeoclimate records. Together with statistician János Korponai, the team began developing a strategy to transform fossil pollen data into quantitative estimates of past climate. Those days working together in Helsinki transformed a rather broad idea into a testable research strategy (Fig. 5).
Fig. 5: Work in progress in Helsinki with most of the team members working on the pollen-based climate reconstruction. From left to right: Niina Kuosmanen, Heikki Seppä, János Korponai, Jon Camuera and Lu Chenao
One key question guided the study: what exactly is the vegetation recording?
In the humid montane environments of the Carpathians, temperature often has a stronger influence on forest composition than water availability. The Holocene pollen records clearly document major ecological reorganisations. During the Early Holocene, pine and birch were gradually replaced by warmth-demanding deciduous trees such as oak, elm, lime and ash. Hazel became exceptionally abundant in many areas. Later, spruce expanded in parts of the mountain range, followed by hornbeam and eventually beech, which became one of the defining species of Carpathian forests.
These shifts were not random. Each species has distinct climatic preferences, dispersal histories and competitive abilities. When species such as hornbeam or beech became dominant, they often reflected changes in environmental conditions that favoured them over previously dominant trees.
The challenge was to translate these ecological patterns into reliable estimates of temperature.
Choosing the right records
Rather than using every available pollen sequence, we adopted a selective approach. We needed records with reliable chronologies, sufficient temporal resolution and locations where vegetation was especially sensitive to temperature change.
After evaluating the available evidence, we selected five sites distributed across the Carpathian arc: Hypkaňa in Slovakia, and Steregoiu, Mohos, Sfânta Ana and Avrig in Romania (Fig. 6).
Equally important was deciding what not to include. Low-elevation sites were generally excluded because vegetation there is strongly influenced by moisture availability and human land use. High-altitude records were also avoided because long-term spruce dominance can reduce the sensitivity of pollen assemblages to temperature variability.
This careful site selection became one of the most important methodological decisions of the study.
Fig. 6: Three of our Carpathian study sites: a) Hypkaňa mire (Slovakia), b) Mohos peat bog (Romania) and c) Sfânta Ana lake (Romania)
Modern pollen as key to ancient climate
The next step was calibration. Pollen-based climate reconstruction relies on comparing fossil pollen assemblages with modern vegetation-climate relationships.
Rather than using a continental-scale dataset, we developed a regional calibration dataset containing nearly 1,000 modern pollen samples from East-Central Europe. This approach better captured the climatic and ecological conditions represented by the fossil records.
The decision proved crucial. It highlighted an important lesson: successful climate reconstruction depends not only on sophisticated statistical methods but also on choosing appropriate ecological analogues.
Revealing Holocene climate patterns
Once the data had been assembled, the reconstruction process became highly iterative. Preliminary results were repeatedly examined, tested and refined.
The five records were combined into regional summer and winter temperature composites, and advanced statistical methods, including the SnSiZer approach, were used to identify which features in the records represented statistically significant climate changes.
The results revealed a fascinating history of seasonal climate change.
Both summer and winter temperatures increased rapidly during the Early Holocene. However, during the last 6,000 years the seasonal trends diverged. Summer temperatures gradually declined, whereas winter temperatures continued to rise until relatively late in the Holocene before cooling during the last two millennia.
These contrasting trajectories suggest a progressive reduction in continentality across East-Central Europe. The expansion of forest species associated with more oceanic climatic conditions, particularly beech and hornbeam, appears to have played an important role in this pattern.
Another major finding was the strong expression of the 8.2 ka climate event. During this abrupt episode, summer temperatures decreased by about 2 °C and winter temperatures by approximately 2.4 °C. Statistical analyses confirmed that this was the most significant short-term climate excursion in the entire record (Fig. 7).
Fig. 7: Our quantitative summer and winter temperatures reconstructed for East-Central Europe based on mid-altitude pollen records from the Carpathians
Science as a conversation
Although the final paper consists of figures, statistics and climate interpretations, the project itself was built through years of conversations and collaboration.
What began as an idea during a visit to the Carpathians in 2019 ultimately evolved into a collaborative international effort and a quantitative reconstruction of 12,000 years of seasonal climate change.
For us, this is perhaps one of the most important stories behind the paper.
Perhaps the most fascinating aspect of the study is that it relies on microscopic pollen grains. Individually, they are tiny. Together, however, they preserve a remarkably detailed record of changing forests, changing climates and environmental transformations across millennia.
The forests covering the Carpathians today are therefore only the latest frame in a much longer story. Their lakes and peatlands preserve the earlier chapters, and sometimes all that is needed to begin reading them is the right question, the right dataset and a group of researchers willing to return to an idea that first emerged years before.