Skip to main content
Dryad

Prioritizing accuracy or efficiency: Comparing general allometric models for perennial bunchgrass species

Carpenter, Scott M.1; Burke, Ingrid C.1; Schlaepfer, Daniel R.2; Lauenroth, William K.3

Published Oct 15, 2024 on Dryad. https://doi.org/10.5061/dryad.08kprr59f

Data files

Oct 15, 2024 version files 170.55 KB

Abstract

Aboveground biomass is important, yet difficult to estimate in dryland ecosystems due to high spatial heterogeneity and the variability of graminoid growth form and density. Allometric relationships are one method of estimating aboveground biomass of forage resources. These models use growth characteristics such as height or diameter to predict biomass. While allometry in forest ecosystems is common, biomass estimation in grasslands and shrublands is primarily based on harvesting or percent cover. Cover estimates vary among researchers and require double sampling at each site to create a relationship between cover and biomass. Multi-species (general) allometric models for high value forage groups, like perennial bunchgrasses, could increase the efficiency of biomass estimation by eliminating the need for destructive sampling. While some general models exist, few studies focus on the application of these models to locations outside the training populations. We tested the applicability of a general bunchgrass model to locations not included in model training against general models developed using biomass samples from our focal sites. We found that our general bunchgrass model trained on data we collected in 2019 made accurate predictions at 76% of sites and that this model outperformed a general bunchgrass model trained on data collected at different sites and by a different research group, which made accurate predictions at 64% of our sites. Despite the loss in accuracy, our study highlights the potential value in further developing general allometric equations for perennial grasses through the development of a grass database. This database may lead to the development of general models with higher confidence in extrapolation beyond the training populations increasing both efficiency and accuracy for land managers.