Allometric equations based on non-destructive methods were developed to estimate palm stem biomass. Twenty mature palms of different ages were subject to destructive and non-destructive measurements: stem height from the collar to the bottom of the frond at the 33rd position in the crown (frond 33), dry weight, the diameter and thickness of cylindrical slices sectioned at different heights along the stem, and the dry weight of samples of stem tissue taken from the core of each slice. The densities of the stem slices obtained using the destructive method (Dtrue) and the non-destructive method (Dcore) were linearly correlated: Dtrue¿=¿1.062¿×¿Dcore (R2¿=¿0.99) independently of the height, age and genetic origin of the palm. Stem density varied with height, reaching its maximum at around 1.5 m, after which it remained more or less constant to the top of the stem. Stem linear density (SLD) estimated from the product of the density and the section of the stem fell to a threshold value that remained more or less constant above 1.5 m. SLD between observed slices was estimated by interpolation based on a hyperbolic model to compute the true biomass of each palm. An equation derived from the integral of the hyperbolic model was fitted to estimate palm stem biomass as a function of wood density observed by non-destructive sampling at 1.5 m on the stem and of the diameter at the same height. With this equation, palm stem biomass can be estimated with an error of 5%.