Assimilation of tree ring and forest inventory data to forecast future growth responses of Pinus ponderosa

Forest responses to future climate changes are highly uncertain, but critical for forecasting and managing for forest carbon dynamics. To improve ecological forecasts of forest responses, we harness the strengths of two large ecological datasets: tree-ring time series data that provide annually resolved growth responses, and repeated measurements of tree size measurements from spatially extensive forest inventory (FIA) data. We use a Bayesian state-space model for these two datasets, and quantify the effects of precipitation, maximum temperature, tree size, stand density, and site index on tree growth, and their associated uncertainties. This model allows us to forecast future tree growth, and separate uncertainty in our forecasts into different sources of uncertainty: initial conditions uncertainty, driver uncertainty, parameter uncertainty, residual process error. This approach demonstrates the strengths of combining national forest inventories and tree ring data to forecast forest responses, and quantify uncertainties to improve our forecasts of forest ecosystems and the terrestrial carbon cycle.