Global and regional reconstruction of Holocene vegetation, fire and land-use
Global and regional reconstructions of Holocene vegetation, fi re and land-use are essential for understanding the interactions between climate, terrestrial ecosystems, and human activities. Paleoecological approaches based on sub-fossil pollen, macrofossils and charcoal preserved in sedimentary basins are being increasingly integrated with dynamic modeling efforts, a combination that aids both understanding and global synthesis, but poses challenges of eff ective upscaling of data, downscaling
... f data, downscaling of models, and appropriate data-model comparisons. For vegetation cover and fi re incidence this amalgamation is relatively advanced; but integration of historical and archeological information into global ecological models is at an early stage of development. Equally, extending global ecosystem models to consider the biogeochemical consequences of land cover change is very new. Approaches to vegetation reconstruction have now developed beyond traditional site-based studies into regional mapping based on statistically sophisticated combinations of data. Maps of Holocene vegetation types derived from quantitatively classifi ed pollen data have been compiled for Europe, having developed from single taxon isopoll maps. Modeling of paleovegetation, with the aim of understanding the dominant drivers of vegetation change, began at the global scale and the research challenge has been to upscale site-based pollen data to match the spatial coverage of model output. The BIOME 6000 project developed a successful approach to paleovegetation data-model comparison. Global-scale vegetation biomes were modeled and the site-based pollen data were upscaled to validate the model by biomisation. This was a collective eff ort amongst pollen analysts, where pollen types were re-classifi ed into plant functional types and then biomes. Data coverage was strongly biased towards regions with long sedimentary records, but this upscaling exercise was of suffi cient taxonomic resolution to validate BIOME model output. Biomisation of pollen data is a crude but eff ective way of standardizing pollen data among sites, but its rigid, prescribed categories, do not fully exploit the numerical or taxonomic complexity of pollen data and are rarely sensitive enough to detect human impact. Subse-quent projects have handled pollen data in other ways to generate regional syntheses of higher spatial resolution. Odgaard and Rasmussen (2000) combined pollen data from 10 sites in Denmark and using chord distances between prehistoric and recent samples showed that human activities became the dominant driver of vegetation change by 3 kyr BP in the Late Bronze Age. The POLLAND-CAL project took up the challenge of reconstructing forest clearance, as the traditional use of ratios of tree pollen: non-tree pollen was prone to error, due to the nonlinear relationship between open ground and non-tree pollen. The project modeled pollen production and dispersal based on the classical Prentice-Sugita model (Prentice, 1985) . The exciting possibility of reconstructing human modifi cation of vegetation over large regions is within reach for the fi rst time by modeling changes in forest cover based on the input of pollen loadings from sites of diff erent size within a region (Sugita, 2007).