hide
Free keywords:
-
Abstract:
We present a global megabiome reconstruction for 43 time slices at 500-year intervals throughout the last 21 000 years based on an updated, and thus currently the most extensive, global taxonomically and temporally standardized fossil pollen dataset of 3455 records. The evaluation with modern potential natural vegetation distributions yields an agreement of ∼ 80 %, suggesting a high reliability of the pollen-based megabiome reconstruction.
We compare the reconstruction with an ensemble of six biomized simulations derived from transient Earth system models (ESMs). Overall, the global spatiotemporal patterns of megabiomes estimated by both the simulation ensemble and the reconstructions are generally consistent. Specifically, they reveal a global shift from open glacial non-forest megabiomes to Holocene forest megabiomes since the Last Glacial Maximum (LGM), in line with the general climate warming trend and continental ice-sheet retreat. The shift to a global megabiome distribution generally similar to today's took place during the early Holocene; furthermore, the reconstructions reveal that enhanced anthropogenic disturbances since the late Holocene have not altered broad-scale megabiome patterns.
However, certain data–model deviations are evident in specific regions and periods, which could be attributed to systematic climate biases in ESMs or biases in the pollen-based biomization method. For example, at a global scale over the last 21 000 years, the largest deviations between the reconstructions and the simulation ensemble are observed during the LGM and the early deglaciation. These discrepancies are probably attributed to the ESM systematic summer cold biases that overestimate tundra in periglacial regions and to the challenging identification of steppes and tundra from the Tibetan Plateau pollen records. Moderate deviations during the Holocene mainly occur in non-forest megabiomes in the Mediterranean and northern Africa, with increasing discrepancies over time. These deviations may result from the underestimation of woody plant functional type (PFT) cover in simulations due to systematic biases, such as overly warm summers with dry winters in the Mediterranean, and the overrepresentation of woody taxa in reconstructions, misclassifying deserts as savanna in northern Africa.
Overall, our reconstruction, with its relatively high temporal and spatial resolution, serves as a robust dataset for evaluating ESM-based paleo-megabiome simulations and provides potential clues for improving systematic model biases.
