We developed allometric regressions for predicting aboveground biomass (AGB) in the Sonoran Desert. Information on canopy cover and height was collected and used to predict AGB from plant dimensions in twenty 25 m2plots that were also fully harvested. The comparison of these two methods showed that allometric equations without correction for bias led to gross AGB underestimation (four times lower than the true values for uncorrected logarithmic allometric equations). Among the tested correction factors, the ratio estimator highly reduced bias and increased accuracy. Validation of allometric estimates with whole-plot harvesting defined the best equation and the least biased correction factor. However, simple nonlinear power functions also gave accurate and unbiased estimates of AGB. We recommend the use of nonlinear models in lieu of traditional logarithm-transformed models. Correction for bias and field verification should be considered in allometric regressions used to predict AGB. In the absence of validation by direct biomass measurements, allometric predictions derived from linearization of ln-transformed data should be taken with care.