Understanding the variation in wood densities of trees and its implications for carbon assessments

Abstract

Wood density is a key functional trait used to estimate aboveground biomass (AGB) and carbon stocks. A common practice in forest AGB and carbon estimation is to substitute genus averages (across species with known wood densities) in cases where wood densities of particular species are unknown. However, the extent to which genus-level averages are reflective of species wood densities across tree genera is uncertain, and understanding this is critical for estimating the accuracy of carbon stock estimates. Using primary field data from India and secondary data from a published global dataset, we quantified the extent to which wood density varied among individuals within species (intraspecific variation) at the regional scale and among species within genera (interspecific variation) at regional to global scales. We used a published global database with wood density data for 7743 species belonging to 1741 genera. Linear models were used to compare the species values with the genera averages and the individual values with the species averages, respectively. To estimate the error associated with using genus-level averages for carbon stocks estimation, we compared genus values averaged at the global, old world and continental scales with species values from actually measured data. We also ran a simulation using vegetation data from a published database to calculate the estimation errors in a 1 hectare plot level when genera-averaged wood densities are used. Intraspecific variation was significantly lower than interspecific variation. Continental level genera averages led to estimates closer to the species values for the 10 genera for which most data on species was available. This was also evident from a comparison of genera averages at these three spatial scales with species values from our data. Species within certain hypervarying genera showed relatively high levels of variation, irrespective of the spatial scale of the dataset used. The error in estimation of AGB when genera-averaged values were used for species wood densities was 0.35, 0.71 and 2.43% when 0, 10 and 25% of the girth of the trees in the simulated plot were from hypervariable genera. Our findings indicate that species values provide the most accurate estimates for individuals. Genus average wood density values at the continental scale provided more reliable estimates than those at larger spatial scales. The aboveground biomass estimation error when species wood densities were approximated to the genera-average values was 1.4 to 3.7 tonnes per ha when 10% and 25%, respectively, of the girth of trees was from species from hypervariable genera. Our findings indicate that regional or continental scale genera averages provide more reliable estimates than global data and we propose a method to identify hypervariable genera, for which species values rather than genera averages can provide better estimates of carbon stocks.