Reconstructing Above Ground Forest Biomass Increment and Uncertainty Using Tree-ring Data

Abstract:

In a changing terrestrial climate, it is becoming increasingly important to be able to quantify Earth systems cycles, including thecarbon cycle. Atmospheric concentrations of carbon dioxide continue toincrease as a result of anthropogenic activity, but less is understood about how forest systems will affect the carbon cycle. In practice, it is difficult to measure carbon flux in a forest system. Flux towers, satellite and remote sensing methods, and dynamic vegetation models have been used to quantify current and future forest net primary productivity.

Tree rings provide us with information about forest carbon storage in the past, and have been used to reconstruct above ground biomass increment (aBI). However, uncertainty from measurement error, assumptions about tree architecture including circular stems and diameter-volume relationships, and the fading record - the challenge of quantifying the growth of previously live trees - are often not accounted for. As a first step towards reconstructing aBI and its uncertainty, we develop a tree ring sampling protocol and a Bayesian hierarchical model toestimate aBI while accounting for measurement and architecture uncertainty.

Tree-ring and repeated census plot data have been collected from several sites using a protocol that allows us toquantify growth dependence across trees in a local area. We also use multiple cores per tree to investigate the number of cores needed to reduce uncertainty from the assumption of stem circularity. For short-time-scale reconstructions, we avoid the fading record issue by coring dead trees and co-locating tree-ring data with censuses, thus avoiding having to make assumptions about stand density andmortality. We also statistically investigate the importance of including census data and of coring dead trees to quantify how uncertainty and bias are affected as we go back further in time. Preliminary results show that the model is able to estimate yearly variation in aBI well for many decades when census data are available. Estimates without census data are accurate for approximately 40 years before the fading record introduces biases in second growth forests dominated by 1-2 tree species.