Palynological and AVHRR observations of modern vegetational gradients in eastern North America

Abstract

Both fossil pollen records and satellite-based instruments are remote sensors of Earth's vegetation with complementary properties. Satellites supply spatially continuous and highly resolved images for the past several decades, whereas pollen records include local and regional signals of vegetation composition, spanning millennia. Together, pollen and satellite-based observations measure vegetation change across a broad range of temporal scales. Here, we compare pollen percentages of needleleaved and broadleaved plant taxa to AVHRR estimates of percent tree cover, for two regions in eastern North America with well-defined physiognomic gradients. The linear fit between the pollen percentages and percent tree cover is strongest for search window half-widths of 25–75 km and unweighted or inverse-distance weightings, consistent with previous taxon-based studies of regional pollen source area and transport. Variance not explained by the linear model arises primarily from differential properties of the AVHRR and pollen sensors, particularly site-specific variability in the pollen data and intertaxonomic differences in pollen representation. These sources of variance can be minimized by regionally smoothing the pollen data and multivariate analogue approaches. A strong fit between observed tree-cover percentages and best-analogue estimates (r2 = 0.70 to 0.78) suggests that analogue-based methods can be applied to infer past tree-cover proportions from fossil pollen records. Linking pollen and AVHRR observations in this manner effectively extrapolates satellite-derived variables beyond the few decades of direct observation, enabling study of longer-term variations in land cover and impacts upon climate and the terrestrial carbon cycle.