Abstract: Quantifying uncertainty around forest carbon storage and uptake: a Bayesian state-space model approach
Forest responses to climate change are highly uncertain, but critical for forecasting and managing forest carbon dynamics. As many countries plan to rely heavily on natural climate solutions, such as forest Carbon sequestration, to offset Carbon emissions, resolving this scientific uncertainty is imperative. The US Forest Service Forest Inventory and Analysis (FIA) program currently provides decadal estimates of standing forest Carbon stocks across space, but lacks detail about how annual climate variation affects Carbon uptake. Tree-ring time series data can fill this gap, providing annually resolved growth responses to climate. We used a Bayesian state-space model to fuse information from tree-ring time series and repeat measurements of tree diameters from FIA of the interior west US. Information from >900 ponderosa pine with tree-ring data and repeated diameter measurements allowed us to estimate annual tree-level growth and tree diameter from 1965-2018, validate our models with held-out forest remeasurement data, forecast future tree growth, and parse the ecological effects of interannually varying climate, density-dependent competition, and site quality on tree growth. We used a Bayesian allometric scaling model to scale up estimates of tree size to both above ground tree- and plot-level biomass estimates. Propagating and parsing uncertainty in these forecasts indicated that the primary causes of uncertainty differ between Carbon stocks (diameter and biomass) versus Carbon fluxes (annual increments), and identified paths to model improvement. Fusion of forest inventory data with tree ring growth advances forecasting of forest carbon in two ways â first, it provides empirically-constrained forecasts of how climate change influences tree growth and above ground biomass over time, including the uncertainty surrounding this response, and second, it provides a framework to quantify how tree- and site-level factors (i.e. tree size, tree density, plot conditions) drive spatial variation in forest carbon dynamics.
