Palaeoenvironmental reconstructions using the modern analogue technique: the effects of sample size and decision rules

Abstract

Palaeoenvironmental information is often extracted from pollen records using the modem analogue technique (MAT). It is largely unknown how sensitive the MAT is to subsampling-induced variation in pollen assemblages, or to the decision rules used to reconstruct climate and vegetation. We examine these issues in a Monte Carlo framework in which simulated pollen assemblages at eight count sizes were created from four fossil pollen assemblages. Simulated and fossil assemblages are compared with modern data sets at squared chord distance cut-off values between 0.05 and 0.55 to: (1) quantify count-size-induced variation in pollen assemblages, and (2) determine how count-size-induced variation affects analogue selection and climate reconstruction. The effects of sample size and decision rules on vegetation reconstructions are examined using two analogue selection schemes and two levels of minimum required analogues. The results show that decision rules have as much impact on reconstruction precision and accuracy as do large differences in count size. At counts of 150 grains, the best-performing cut-off values yield annual precipitation estimates within 75 mm of those produced by the reference fossil assemblages, and January and July temperature estimates within 0.75C. Counts as low as 150 grains yield vegetation reconstruction accuracies of?90% using the best-case reconstruction rules; the worst-case reconstruction rules may not achieve this accuracy at 1000 grain counts. Although larger count sizes do lead to greater reconstruction precision and accuracy, the results indicate that analyses of the kind presented here can inform count-size decisions and allow significant re-allocation of analytical effort.