Adequacy of contrasting sampling methods for root mass quantification in a slash-and-burn agroecosystem in the eastern periphery of Amazonia

Root research needs to optimize sampling schemes to address different scales and dimensions of variability within a framework of feasible effort and, in some cases, acceptable soil perturbation. Both large and small sample volumes are utilized in root research, and this study is designed to help field researchers in optimizing root sampling. We evaluate the performance of three contrasting sampling strategies—large monoliths (LM, 25 dm−3), small monoliths (SM, 3 × 1 dm−3), and root auger (RA, 5 × 196.3 cm−3). We compare root biomass estimates obtained in a shifting cultivation agroecosystem in eastern Amazonia. We sampled a slash-and-burn field at 1/2 and 1 1/2 years of cultivation, and a paired 3-year-old spontaneous secondary forest regrowth down to 1 m soil depth, and we distinguish roots by diameter class and taxonomic origin (palm vs. non-palm roots). LMs are the only reliable quantification method for coarse roots. Both SM and RA methods are inadequate due to high variability and the frequent failure to detect the presence of coarse roots in the sampled soil. The SM method is acceptable for mid-sized (2–5 mm) roots, and both methods are adequate for fine root (<2 mm) quantification. Labor costs of LM, SM and RA methods are similar since reduced sampling effort in small sample volumes is compensated by increased costs in sample handling. The small sample volumes of SM or RA may be preferable in situations focusing on fine root dynamics, with the necessity to avoid major perturbations, or in repeated sampling schemes. Both SM and RA samplings gave a strong and systematic overestimation of root biomass in all fractions compared to the LM values, with differences being less pronounced for SM. We establish comparability between the three sampling schemes with transformation equations which are remarkably similar between shifting cultivation and secondary forest, between roots of the babassu palm and those of other origin and down the 1-m soil profile. Thus, future field studies in similar environments can utilize SM or RA fine root sampling and subsequently transform data to LM estimates. We recommend the SM rather than the RA method because of the lower data variability and the lower degree of root mass overestimation.