Pollen Representation of Vegetation in Quaternary Sediments: Theory and Method in Patchy Vegetation

1 According to the Prentice-Sugita model, pollen loading (PL) is linearly related to the distance-weighted plant abundance (DWPA) surrounding a sedimentary basin. Since source trees of pollen far away from a basin have much less influence on pollen representation than source trees near a basin, the correlation between PL and D WPA should approach an asymptote as the vegetation sampling area increases. The 'relevant source area' for pollen can be defined as the area beyond which the correlation does not improve. 2 A simulation experiment using patchy vegetation landscapes illustrates this principle, demonstrating that r2 and likelihood function scores do not improve when vegetation sampling increases beyond certain distances. This suggests that very little additional information on the pollen-plant abundance relationship will be gained by a vegetation survey beyond the 'relevant' distance when data are collected from regions of similar vegetation type and spatial pattern. 3 The 'relevant' source area for pollen in lakes in the simulated landscapes is within 50-100 m from the lake edge for forest hollows (radius of hollow R = 2 m), 300400 m for small lakes (R = 50 m), and 600-800 m for medium size lakes (R = 250 m). Although only about 30-45% of total pollen loading comes from within these distances, the model demonstrates that when the background pollen is consistent, this proportion is adequate to reflect local vegetation composition. 4 The simulation results show that pollen data from large lakes (R = 750 m) show little site-to-site variation, especially for a species that grows in small patches (radii of 80 m in the simulated landscape). Linear regression and maximum likelihood methods therefore do not provide accurate estimations of pollen productivity and background pollen loading. Vegetation may appear homogeneous even when the actual pattern of vegetation is heterogeneous and patchy, depending on the size of lake relative to the size of patches. 5 The pollen-plant abundance data should be collected from regions with similar forest composition to determine the 'relevant' source area and the parameters of a linear relationship between PL and D WPA. Otherwise, either the linear regression or maximum likelihood methods provide inaccurate estimates of the relevant source area and parameters even when the simulation data with no sampling and counting errors are used.