QUANTITATIVE RECONSTRUCTIONS OF LATE QUATERNARY TREE COVER FROM MODERN POLLEN-AVHRR CALIBRATIONS

Until recently, the quantitative reconstruction of historical and late-Quaternary land cover change has been limited primarily by the lack of readily available quantitative information about present-day vegetation, rather than by the qualities of the fossil data. Forest-survey data can be difficult to obtain, laborious to process, and their quality and consistency varies among political units. Vegetation classification maps are readily available, but the assignment of quantitative properties to these categorical maps is problematic and incorrectly assumes that vegetation properties only change at ecotones. Without quantitative descriptions of the present-day vegetation, it is unrealistic to expect such reconstructions for the past.

The maturation of remote sensing technology and algorithms has vastly improved our ability to describe present-day gradients in vegetation physiognomy. In particular, Advanced Very High Resolution Radiometer (AVHRR) observations have provided synoptic descriptions of global vegetation at a 1 km resolution. Here we 1) compare the distribution of pollen types in surficial sediments in eastern North America to AVHRR-derived estimates of needleleaved and broadleaved tree cover and 2) reconstruct late-Quaternary variations in tree cover. The modern distribution of needleleaved and broadleaved pollen types is broadly similar to the AVHRR data. The correlation between the AVHRR and pollen data varies with spatial grain, and is strongest for search window half-widths of 25-75 km. A major source of uncertainty is the intertaxonomic differences in pollen representation. To minimize this source of uncertainty, the reconstruction of past tree cover relies upon standard modern-analog techniques, with one important twist: if no good modern analog can be found for a fossil pollen sample, the pollen types are reclassified into plant functional categories and the analog analysis rerun. This allows the assignment of vegetation properties to ‘no-analog’ fossil pollen samples. Tree-cover reconstructions for the late Quaternary are generally consistent with previous interpretations of the fossil pollen data, but nevertheless are qualitatively and quantitatively different from simple mappings of arboreal pollen sums.