Predicting future productivity of Southeastern U.S. pine ecosystems in a changing climate using data assimilation with diverse data sources

Abstract:

Data-assimilation (DA) is a powerful tool that can provide estimates of parameter uncertainty and model uncertainty. When combined with output from multiple climate models and climate scenarios, DA can also be used to produce estimates of future forest dynamics in a change climate that robustly estimate a hierarchy of the known unknowns. Our project aim is to comprehensively predict future loblolly pine productivity and carbon sequestration, along with a quantified uncertainty, across the Southeastern U.S. Using the computationally tractable Physiological Principles Predicting Growth model (3-PG), we integrated multiple observation types from geographically dispersed data, including ecosystem-scale experiments, to constrain our predictions. Specifically, we used a recently developed productivity dataset based on decades forest industry research to constrain slow responses of biomass accumulation and allocation to climate, flux tower estimates of gross primary productivity to constrain higher-time frequency climate sensitivity, recent through fall-exclusion experiments in four loblolly pine stands to constrain precipitation sensitivity, and the Duke Free-Air Carbon dioxide Enrichment (FACE) to constrain productivity response to elevated CO₂. We describe results from the DA of the 3-PG model by exploring how estimates of future forest productivity depend on each of these data sources. The most important constraint was provided by the geographically widespread annual forest plot measurements but the Duke FACE site was critical for constraining the parameter governing CO₂ enhancement of photosynthesis. We present results for future Southeastern U.S. pine forest productivity that are consistent with decades of forest research in the region and that comprehensively represents uncertainty in future predictions of forest dynamics.