Accuracy of subsampling for height measurements in loblolly pine plots

Abstract

Estimating heights in research and inventory plots is costly. We examined the feasibility of subsampling tree heights as opposed to measuring all trees. Four sampling intensities (75, 50, 25, and 10%) and four sampling strategies (systematic sampling, simple random sampling without replacement, stratified sampling across the diameter distribution, and sampling the first trees in each plot) were investigated. Data from 600 loblolly pine plots in fertilizer trials in the southeastern United States were used. The application of a height-dbh regression to predict the heights of unmeasured trees was also investigated. Sampling the first trees generally resulted in poorer estimates than the other sampling schemes. Systematic and simple random sampling performed similarly. A 50% sampling intensity with either systematic or simple random sampling and a height-dbh regression predicting the heights of unmeasured trees estimated more than 90% of plots to within 2.2% of the observed plot height and more than 94% of plots to within 2.5% of the observed volume, and they were more accurate than the stratified sampling at the same intensity. Systematic sampling is easy to implement, requiring no prior plot knowledge. We conclude that a 50% systematic sampling combined with a height-dbh regression will reduce costs without compromising accuracy. Copyright © 2009 by the Society of American Foresters.