Sampling roots to capture plant and soil functions

Summary Roots vary in anatomy, morphology and physiology, both spatially (different parts of the same root system) and temporally (plastic changes, root ageing), suggesting that root trait measurements are strongly affected by root sampling categories. In this context, it is urgent to clarify the functional significance of current root sampling categories (e.g. fine roots of the first order, the first three orders, ≤1 mm or ≤2 mm), establish guidelines for choosing between sampling methods and revise root ontology to account for functional differences between traits measured on distinct root categories. Here, we used a worldwide database of fine‐root traits to test the hypothesis that distinct fine‐root trait values – with link to fine‐root functions – were generally affected by different root sampling categories. We observed indeed a clear functional break between first‐order roots and roots of all three other sampling categories, and a smaller but substantial break between roots of the three first orders and the ≤2 mm category, demonstrating globally that different sampling methodologies capture different functional parts of roots. Our synthesis suggests that all current root sampling categories present both advantages and pitfalls and that no single method can appropriately tackle the main current challenge of root functional ecology: i.e. linking fine roots to plant and ecosystem functions in a truly comparable way across all plants. We argue instead that a small set of complementary standardized sampling methods is necessary to capture the linkages between root forms and functions. To assist experimenters selecting adequate sampling we developed a decision table following three logical questions: (i) what plant or ecosystem function must be addressed; (ii) what root categories are involved in this function and (iii) what traits should be measured on these root categories. Challenging, strengthening and expending such common reference framework would be a substantial step towards wider comparability of future functional trait datasets. A lay summary is available for this article.