Four tree species were harvested periodically over a 13-year period from plantations in the humid lowlands of Costa Rica: Cedrela odorata, Cordia alliodora, Hyeronima alchorneoides, and Euterpe oleracea. The soil was a well-drained, volcanic alluvium, and high fertility coupled with 4 m of annual rainfall and high temperatures led to rapid growth rates; at age 13 many individual were >30 cm dbh and >30 m tall. Harvested trees were dissected into their component parts: leaves, rachises (for Cedrela and Euterpe), branches, boles, and coarse roots (i.e., >0.5 cm diameter). Roots of small trees were excavated in their entirety; those of large trees were harvested from a cylinder 1.0 m in diameter, immediately beneath the trunk. Large numbers of trees were sampled: 258–379 per species. Size classes sampled ranged from seedlings too small to have a dbh to trees of � 30 cm dbh. Two separate allometric equations (one for trees having only a basal diameter and another for trees having a dbh), with diametersquared times height as the metrics, were developed for each component of each species. Based on breaks in linear trends of ln–ln plots and deviations of predicted from actual values, we developed separate allometric equations, by component, for trees of different sizes. The resulting 40 equations (with one exception, involving very small trees) fit the data well and enable the user to predict biomass, by component, for each of the four species. A single (non-allometric) linear equation, combining all plant parts of all three dicot species, also fit the data well, but it would not provide either the detail or the accuracy provided by the species-specific, plant-part-specific equations. Large sample sizes, a 13-year run of data collection, and the economic and ecological importance of the species studied make this data set uniquely useful for biomass estimations and for understanding the inherent heterogeneity of tree structure in dynamic tropical environments. # 2006 Elsevier B.V. All rights reserved.