A non-destructive sampling strategy for basic density, based on removing 12 mm bark-to-bark cores, was developed in E. globulus and E. nitens. Fifty trees of each species, aged 5 to 9 years, were sampled across a range of sites. Core samples were removed on both a north-south and an east-west axis from 6 fixed heights in the base of the tree (0.5 m, 0.7 m, …, 1.5 m). Whole-tree values were calculated from disc samples removed at eight percentage heights (0, 10, 20, …, 70%) and correlations between the cores and whole-tree values were used to determine the optimal sampling height. Core samples were found to be reliable predictors of whole-tree density, explaining between 84% and 89% of the variation between trees. Core sampling of E. globulus and E. nitens to estimate basic density of whole-trees and stands is feasible; cores from trees at all E. globulus sites gave high correlations with whole-tree values. For E. nitens, site differences were apparent, and it is recommended that a small destructive sampling program should be undertaken prior to commencing a major sampling program. Recommended optimal sampling heights are 1.1 m for E. globulus and 0.7 m for E. nitens. Core orientation was not important and density was not related to tree size. Six whole-tree samples or eight core samples are required for estimating the mean density of a stand at a specific site to an accuracy of ±20 kg m−3 with a 95% confidence interval.