Is the Atkinson discriminant function a reliable method for distinguishing between Betula pendula and B. pubescens (Betulaceae)?

Abstract Betula pendula and B. pubescens are common tree species of Europe that differ in ploidy level. A continuum of morphological variation between the two species makes them hard to differentiate in the field. The Atkinson Discriminant Function (ADF), based on leaf shape, was proposed in 1986 as a metric to distinguish them and has since become a standard approach. Here, we test this method on 944 trees sampled across Britain against species’ discriminations made using twelve microsatellite loci. The ADF misidentified six of 780 B. pubescens trees and 28 of 164 B. pendula trees. This success rate of 96·4% is higher than that found by Atkinson & Codling (1986) based on a smaller sample for which chromosomes had been counted. The success rate can be raised to 97·5% by using an ADF of −2 rather than zero as the boundary line between the species. With this improvement, error rates of over 10% occur for trees with ADF ranging from −11 to +3. We found no evidence for hybrid origin of the majority of trees misidentified by the ADF.