Regrowing forests drive landscape carbon and nutrient cycling over decades, but whether vegetation models can reproduce these long-term patterns of forest succession is uncertain. A team of researchers led by scientists from the U.S. Department of Energy’s Pacific Northwest National Laboratory, simulated carbon cycling and community composition during 100 years of forest regrowth following disturbance. They examined which processes and parameters are most important to accurately model forest succession in the Upper Midwest region of the United States, along with the relative importance of model structure versus parameters. The researchers found that parameter uncertainty is far more important than structural uncertainty, and that simulating both productivity and plant community composition accurately remains a challenge. These results have implications for robustly simulating future climate change effects with Earth system models.
Understanding the influences behind 20th- and 21st-century forest growth and the ability to predict its future evolution, is essential to shaping global policy around climate, biodiversity, and natural resource management. The ability of even state-of-the-art dynamic vegetation models to simulate successional change and forest regrowth is uncertain, and thus the results of this study provide important bounds on our predictive ability for Upper Midwest forests, as well as the ecological implications of these uncertainties.
Vegetation models capture researchers’ understanding of forest function, but whether models can reproduce multidecadal patterns of forest succession is highly uncertain. This research team tested the accuracy and precision with which a vegetation model can simulate carbon cycling and community composition during 100 years of forest regrowth. To do this, they ran ensembles of an ecosystem demography model with different representations of processes important to competition for light. Then, the researchers compared the magnitude of structural and parameter uncertainty. They also assessed which submodel-parameter combinations best reproduced observed carbon fluxes and community composition. On average, the simulations underestimated observed forest production and leaf area after 100 years and predicted complete dominance by a single plant functional type. Parameter uncertainty was large; the two parameters that consistently contributed most to uncertainty were plant-soil water conductance and growth respiration—both empirical coefficients that cannot be observed. The team concluded that parameter uncertainty is more important than structural uncertainty, at least for this model in Upper Midwest forests, and simulating both productivity and plant community composition accurately without physically unrealistic parameters remains a challenge for demographic vegetation models.
Ben Bond-Lamberty, Pacific Northwest National Laboratory, firstname.lastname@example.org
This project was supported by the National Science Foundation. Cyberinfrastructure capabilities were provided by the Pacific Northwest National Laboratory.
Published: September 18, 2020
A. Shiklomanov, B. Bond-Lamberty, J. Atkins, and C. Gough, “Structure and parameter uncertainty in centennial projections of forest community structure and carbon cycling.” Global Change Biology in press (2020). [DOI: 10.1111/gcb.15164]