Background: Metabolic scaling theory (MST) is still in debate because observed allometric exponents often deviate from MST predictions, and can change significantly depending on environment, phylogeny, and disturbance. We assembled published scaling exponents from literatures for three allometric relationships linked to biomass allocation: leaf biomass-diameter (L-D), stem biomass-diameter (S-D), and root biomass-diameter (R-D). We used data from natural forests and plantations across China to test the following hypotheses: 1) the allometric relationships of trees support the predictions of MST on a broad scale; 2) the observed deviations from MST predictions are caused by climate, biotic factors, and/or phylogeny; 3) abiotic and biotic factors influence allometric relationships in plantations and natural forests differently, and different allometric relationships (i.e. L-D, S-D, and R-D) are affected differently. We related these scaling exponents to geographic climate gradient, successional stage, stand density, leaf form and phenology, and phylogeny. We used mixed-effect models to examine the major factors affecting tree allometries. Results: In natural forests, S-D and R-D scaling exponents were consistent with MST predictions in primary forests, but were significantly lower in secondary forests. Both S-D and R-D scaling exponents in plantations had a medium value that fell between those of the secondary and primary forests, despite plantations being similar in species characteristics and age to secondary forests. The S-D and R-D exponents were significantly affected by factors that are not yet considered in MST, including winter coldness which explained 2.76%-3.24% of variations, successional stage (7.91%-8.20% of variations), density (a surrogate for competition, 5.86%-8.54% of variations), and especially phylogeny (45.86%-56.64% of variations explained). However, the L-D scaling exponents conformed to MST predictions in primary, secondary, and plantation forests, and was not strongly explained by most factors. Conclusion: MST is only applicable to primary (steady-state) forests, and climate, biotic factors and phylogeny are causes of the observed deviations of allometric relationships from MST predictions. Forest management practices in plantations have a strong influence on tree allometries. L-D allometry is more strongly controlled by biophysical constraints than S-D and R-D allometries, however, the mechanisms behind this difference still need further examinations.