During the last deglaciation, forest cover in the Northern Hemisphere extratropics expands roughly 4000 years earlier in transient Earth system model simulations than in pollen-based biome and tree-cover reconstructions, with the largest mismatch in Asia (Siberia and the East Asian monsoon region). Dallmeyer and colleagues, who named this mismatch the "deglacial forest conundrum" in 2022, could exclude simulated-climate bias as the main driver but could not decide between real multi-millennial ecological disequilibrium of the forest biome (limited seed dispersal, slow soil and permafrost development, small and genetically exhausted glacial-refugium populations, competition retarding tree establishment — none represented in Earth system models) and systematic biases of the pollen-based reconstructions themselves (low pollen productivity of key taxa such as larch, long-distance pollen transport, no-modern-analogue assemblages, sparse record density in Siberia), nor quantify their interaction. The open problem is to quantify the relative contributions of these ecological and methodological causes — and of residual regional climate bias, which cannot be fully excluded for Asia — to the lag, and to determine what the resulting temporal disequilibrium implies for the timescales on which pollen-based climate reconstructions are trustworthy. The direct quantitative model–data comparison that was missing in 2022 has since been delivered by the authors' own follow-up work (a Northern Hemisphere REVEALS tree-cover dataset and a 2026 model evaluation), which sharpens all three candidate explanations but leaves the causal attribution explicitly unresolved.

Background

After the Last Glacial Maximum (LGM, about 21,000 years ago), warming climate, rising atmospheric CO2 and retreating ice sheets transformed the Northern Hemisphere (NH) extratropics from largely open glacial landscapes to the Holocene forests of today. Two independent lines of evidence document this transformation. First, transient simulations with Earth system models (ESMs) — coupled atmosphere–ocean–land models whose land surface includes a dynamic global vegetation module; here MPI-ESM1.2 with the JSBACH land model — represent vegetation through a small set of plant functional types (PFTs, plant groups classified by form and function rather than species) whose grid-cell cover fractions respond to climate, typically within decades, because seeds are assumed available everywhere and soil conditions are held fixed. Second, fossil pollen preserved in lake and bog sediments is used to reconstruct past vegetation: the classical "biomisation" approach converts pollen assemblages into biome types (e.g., tundra, taiga, temperate forest), while the REVEALS model (Regional Estimates of Vegetation Abundance from Large Sites) uses taxon-specific relative pollen productivity (RPP) and models of pollen dispersal to correct pollen percentages into quantitative regional plant-cover fractions. Each reconstruction method has known vulnerabilities: some taxa are strongly underrepresented in pollen (larch, Larix, the dominant Siberian conifer, is a poor pollen producer, so its forests are underestimated); pollen is transported over long distances; glacial-age pollen assemblages may lack modern analogues, weakening calibrated pollen–vegetation relationships; and record density is very low across central and eastern Siberia. In 2022, Dallmeyer, Kleinen, Claussen, Weitzel, Cao and Herzschuh compared a transient MPI-ESM1.2 simulation of the last 22,000 years with a global synthesis of pollen-based biome reconstructions and reported the "deglacial forest conundrum": deglacial forest expansion occurs approximately 4000 years earlier in the model than in the reconstructions, with the turnover from open to forested landscapes correspondingly faster, an early-Holocene model forest maximum (~11 ka) versus a mid-Holocene reconstructed maximum (~7 ka), and the largest mismatch concentrated in Asia. Because the simulated climate agrees with pollen-independent climate reconstructions (e.g., chironomid-based — midge-larval — summer-temperature estimates), the authors excluded climate bias as the main driver, leaving a triad of candidate explanations: (i) true ecological disequilibrium — real NH forests tracked deglacial warming on multi-millennial timescales, driven by processes absent from ESMs such as seed-dispersal limitation, slow postglacial soil buildup (centuries at minimum) and permafrost, small refugial population sizes with genetic exhaustion, and competition from resident steppe-tundra vegetation; (ii) methodological bias in the pollen-based reconstructions (biomisation artefacts, larch under-representation, long-distance transport, no-modern-analogue effects, sparse Siberian data); and (iii) residual regional climate bias, which cannot be fully excluded for Asia because high-quality pollen-independent climate reconstructions there are extremely scarce. The paper also flagged two consequences: the evaluation of time-slice paleoclimate simulations with pollen records assumes vegetation–climate equilibrium and may need reassessment, and ESMs may overstate how fast NH forests will track ongoing anthropogenic warming, with implications for projected carbon storage and vegetation–atmosphere feedbacks. What was missing in 2022 was any quantitative decomposition of the lag among these causes, and a direct quantitative comparison of simulated tree PFTs with REVEALS-type tree-cover estimates covering the full deglaciation and the entire NH.

Problem Statement

Determine, with quantitative evidence, the causes of the multi-millennial (approximately 4000-year) lag by which pollen-based biome and tree-cover reconstructions trail the climate-driven forest expansion simulated by Earth system models across the Northern Hemisphere extratropics during the last deglaciation. The reference ~4000-year magnitude is defined in the source study (Dallmeyer et al., 2022) on pollen-based biome reconstructions of Northern Hemisphere forest cover relative to the transient MPI-ESM1.2 simulation with the JSBACH land model, as the temporal mismatch in the deglacial increase of NH forest cover between an early-Holocene simulated forest maximum (~11 ka) and a mid-Holocene reconstructed maximum (~7 ka, ~6.5 ka in Asia); the extended quantitative REVEALS tree-cover reconstruction is a complementary observable whose mismatch against the same simulation must be explained by the same attribution. Specifically, establish the relative contributions and interaction of (i) real ecological disequilibrium of the forest biome, arising from processes not represented in current vegetation models — limited seed dispersal, slow postglacial soil and permafrost development, small and genetically exhausted glacial-refugium populations, and interspecific competition retarding tree establishment; (ii) systematic biases in the pollen-based reconstructions themselves, including low pollen productivity of key taxa such as larch, long-distance pollen transport, no-modern-analogue pollen assemblages, and sparse record density in Siberia; and (iii) residual biases in the simulated climate, which cannot be fully excluded for Asia where pollen-independent climate reconstructions for the glacial and deglacial period are extremely scarce. Establish which of the three causes dominates and by how much, and quantify how these causes interact rather than only their separate main effects; show the attribution to be robust at the Northern Hemisphere scale and in the Asian region where the mismatch is largest; and determine the implications of the resulting temporal disequilibrium — on which timescales pollen-based climate reconstructions (and vegetation–climate equilibrium assumptions used to evaluate paleoclimate simulations with pollen records) remain trustworthy. The three threads are coupled: the interaction among the causes, its dependence on spatial scale, and the implications of the resulting disequilibrium for the reliability of pollen-based reconstructions cannot be resolved independently of one another.

References

  1. Dallmeyer, A., Kleinen, T., Claussen, M., Weitzel, N., Cao, X., & Herzschuh, U. (2022). The deglacial forest conundrum. Nature Communications, 13, 6035. https://doi.org/10.1038/s41467-022-33646-6
  2. Schild, L., Ewald, P., Li, C., Hébert, R., Laepple, T., & Herzschuh, U. (2025). LegacyVegetation: Northern Hemisphere reconstruction of past plant cover and total tree cover from pollen archives of the last 14 kyr. Earth System Science Data, 17, 2007–2033. https://doi.org/10.5194/essd-17-2007-2025
  3. Dallmeyer, A., Weitzel, N., Schild, L., Herzschuh, U., Kleinen, T., & Claussen, M. (2026). Unravelling the tree cover dynamics over the last 20 000 years on the Northern Hemisphere. Climate of the Past, 22, 1125–1157. https://doi.org/10.5194/cp-22-1125-2026
  4. Dallmeyer, A., Poska, A., Marquer, L., Seim, A., & Gaillard, M.-J. (2023). The challenge of comparing pollen-based quantitative vegetation reconstructions with outputs from vegetation models – a European perspective. Climate of the Past, 19, 1531–1557. https://doi.org/10.5194/cp-19-1531-2023
  5. Li, C., Dallmeyer, A., Ni, J., Chevalier, M., Willeit, M., Andreev, A. A., Cao, X., Schild, L., Heim, B., Wieczorek, M., & Herzschuh, U. (2025). Global biome changes over the last 21 000 years inferred from model–data comparisons. Climate of the Past, 21, 1001–1024. https://doi.org/10.5194/cp-21-1001-2025
  6. Zani, D., Lischke, H., & Lehsten, V. (2023). Climate and dispersal limitation drive tree species range shifts in post-glacial Europe: results from dynamic simulations. Frontiers in Ecology and Evolution, 11, 1321104. https://doi.org/10.3389/fevo.2023.1321104
  7. Zani, D., Lischke, H., & Lehsten, V. (2024). The role of dispersal limitation in the forest biome shifts of Europe in the last 18,000 years. Journal of Biogeography, 51, 1438–1457. https://doi.org/10.1111/jbi.14836
  8. Shen, W., Kruse, S., Liu, S., Stoof-Leichsenring, K., Kühn, I., Li, W., Cao, X., Zhang, Z.-R., Zeng, C.-X., Yang, J.-B., Li, D.-Z., & Herzschuh, U. (2025). Post-Glacial Vegetation Trajectories on the Eastern Tibetan Plateau Reflect Millennial-Scale Migration Lags in Complex Mountain Terrain Based on Sedimentary Ancient DNA and Dynamic Dispersal Modeling. Ecology and Evolution, 15, e70862. https://doi.org/10.1002/ece3.70862
  9. Herzschuh, U., Böhmer, T., Jia, W., & Lisovski, S. (2026). Quantitative climate reconstruction from sedimentary ancient DNA: framework, validation and application. Climate of the Past, 22, 1159–1180. https://doi.org/10.5194/cp-22-1159-2026
  10. Dawson, A., Williams, J. W., Gaillard, M.-J., Goring, S. J., Pirzamanbein, B., Lindstrom, J., Anderson, R. S., Brunelle, A., Foster, D., Gajewski, K., Gavin, D. G., Lacourse, T., Minckley, T. A., Oswald, W., Shuman, B., & Whitlock, C. (2025). Holocene land cover change in North America: continental trends, regional drivers, and implications for vegetation–atmosphere feedbacks. Climate of the Past, 21, 2031–2060. https://doi.org/10.5194/cp-21-2031-2025