The growing need for accurate forest biomass and carbon estimates has sparked a recent re-examination of direct volume estimation and taper modeling techniques. Additionally, the need for increased precision of biomass and carbon estimates drives a need to improve the techniques used to evaluate these models. Taper equations have an advantage over direct volume estimation in that they are flexible in estimating component biomass (e.g., residual tree top biomass post-harvest) and upper stem diameters. Using measurements derived from the destructive sampling of 119 ponderosa pine trees in eastern Oregon and eastern Washington, we evaluate three direct volume (Summerfield 1980, Wiley 1978) equations and three taper models (Kozak 1969, Kozak 2004, Czaplewski 1989) in terms of percent bias and percent RMSE of total volume prediction. Wiley (1978), a very simple direct volume equation, performed better than Summerfield (1980). A simplified version of Wiley (1978), as described by Poudel (2018), produced a smaller RMSE and larger bias. Summerfield (1980) is currently used by the USFS-FIA. Kozak (2004), a variable exponent taper equation, preformed the best of the taper models. Czaplewski (1989), a segmented polynomial equation, preformed very similarly to Kozak (2004). The variable exponent taper equation preformed slightly better than the best direct volume equations in terms of RMSE, and slightly worse in terms of bias.
In evaluating these models it is assumed that total volume measured in the field is the true volume, however we know this is not the case. Volume derived from field measurements are still estimates with some bias included. The shorter the segments are used to calculate the volume, the less bias is present. Operational constraints often limit the minimum feasible segment length. If section lengths are not consistent throughout the dataset, there is an unknown and unequal amount of bias inherent within the “true” volumes that are used to evaluate model performance. We evaluated linear interpolation techniques to predict diameter inside bark between measurement locations to increase the accuracy of the true volume estimate and standardize the inherent bias associated with unequal segment lengths. Linear interpolation of fractional error in diameter prediction between measurements worked well (bias < 1.0 cm) in predicting diameter inside bark at 1.3 meters and at the top of the final log.