How much cross-sectional area is needed to estimate vessel traits in tropical tree branches?

Wood anatomical traits provide key insights to plant physiology, ecology, and evolution but their quantification is time-consuming. As a result, measurements are often based on small sample areas or few features, potentially reducing accuracy and repeatability of estimates. A better understanding of how sampling intensity influences estimation of anatomical traits would improve the efficiency and reliability of wood anatomical studies.

We annotated all vessels in 164 branch cross-sections from 19 tree species in Puerto Rico and used these to compute the 'true value' of seven vessel traits. We then applied a subsampling approach to quantify how error of estimated trait values depended on the proportional and absolute area of the sample annotated. We used variance partitioning to assess how the contributions of within-sample, within-species, and among-species variation change with area sampled. We also assessed how sampling intensity affected the number of statistically significant comparisons between species.

At low sampling intensity (i.e., ≤10% of the total cross-section area), absolute percent differences of estimates from true values ranged from 6-18%, depending on the trait. Across samples, error declined more strongly as a function of proportional area than absolute area sampled. Nonetheless, among-species and among-individual differences explained the majority of variation in the dataset even at low sampling intensity, while variation among replicate subsamples contributed little to the total variation. Depending on the trait, the proportion of statistically significant differences between species increased by 10-21% from low to high sampling intensity.

Our results suggest that for broad comparisons across species, relatively low sampling intensity can yield robust estimates. However, more sampling may be required for questions demanding greater statistical power, such as within-species or within-individual comparisons. Generally, design of wood anatomy studies should explicitly consider the accuracy and precision required by the research objectives.