Great Basin Naturalist


Regression analyses for plant biomass estimation from physical measurements of individual plant dimensions that are nonlinear have generally used some form of the allometric equation. Use of this equation has most often involved logarithmic transformation of the variables (power regression). Transformation, however, introduces systematic bias into the analyses. Power regression was compared with a bias correction technique and with nonlinear regression for the prediction of the total foliage biomass (phytomass). Crown volumes of one sagebrush and one perennial grass species were used for these evaluations. The bias correction factor was uniformly applied to all the predicted values from power regression. Nonlinear regression avoided this bias by not requiring logarithmic transformation. It was also consistently less variable than either power regression or the correction factor method in estimating actual total phytomass by the allometric equation and equivalent or better in accuracy. The correction factor technique consistently gave the poorest predictions of the methods evaluated. Standard linear regression worked as well for the bunchgrass as the best method based on the allometric equation. Predictions were generally better when sample sizes used to derive the regression equations represented the range of plant size and variability in the data for which the phytomass was estimated.