Applying plant functional types to construct biome maps from eastern North American pollen data: comparisons with model results

Abstract

Global biome models like BIOME1 convert climate-model simulations of past climates into biome distributions and thus facilitate comparison of both climate and biome model results with biomes estimate from paleoecological data. We adapted a biomization method, recently developed for European pollen data, for use with pollen data in eastern North America and then compared its estimated biomes with those derived from applying BIOME1 to the climate simulations from the NCAR CCM1 (National Center for Atmospheric Research Community Climate Model, Version 1) for 6000 years ago (6 ka). We first tested the biomization method by seeing how well the biomes inferred from modern pollen data match observed biomes. We found that modifications to the method were necessary in part to account for physiological differences between North American and European taxa, and in part to cope with our choice of using just 23 major pollen taxa. Our modifications significantly improved the match between observed modern biomes and pollen-derived biomes, as measured by the kappa statistic. We tested our tuning of the biomization method by matching its inferred 6 ka biomes to biomes estimated from pollen data using the modern analog technique. The degree of agreement at 6 ka is close to that for today, showing that (1) the biomization method and modern analog technique, when applied to the same pollen data, produce consistent results, and (2) the modifications made to the biomization method are robust back to 6 ka. We then used the results of the biomization method to test the biome maps simulated by BIOME1, which derives biome distributions from observed climate values for today and from the climatic simulations of the CCM1 for 6 ka. Only a fair agreement is seen, and significant offsets exist in the placement of biomes by BIOME1. For today BIOME1 simulates the boundary between the temperate deciduous and cool mixed forests to be too far south and the steppe-forest boundary to be too far west. These model biases are also evident in the simulations at 6 ka despite the fact that CCM1 simulates warmer than present temperatures in the central United States. To the north, however, BIOME1 correctly simulates the cool mixed forest and taiga boundary at 6 ka as more northwestward than at present.