Constraining modelled global vegetation dynamics and carbon turnover using multiple satellite observations

Area of Science:

• Earth System Science
• Ecology
• Climate Modeling

Background:

• Uncertainty in dynamic global vegetation models (DGVMs) hinders accurate predictions of land ecosystem responses to climate change.
• DGVMs struggle to simulate vegetation distribution, productivity, biomass allocation, and carbon turnover under changing climate conditions.
• Satellite observations offer a valuable resource for constraining and improving DGVM simulations.

Purpose of the Study:

• To enhance the LPJmL4 DGVM by integrating multiple satellite-derived datasets.
• To improve the simulation of vegetation phenology, productivity, and dynamics using model-data integration.
• To reduce uncertainties in the global climate-carbon cycle feedback.

Main Methods:

• Utilized satellite data including fraction of absorbed photosynthetic active radiation (FAPAR), sun-induced fluorescence (SIF), above-ground biomass (AGB), land cover, and burned area.
• Constrained LPJmL4 DGVM parameters related to phenology, productivity, and vegetation dynamics.
• Employed a machine learning approach to analyze remaining model errors.

Main Results:

• The optimized LPJmL4 DGVM accurately reproduced present-day land carbon cycle estimates and temporal dynamics of FAPAR, SIF, and gross primary production.
• The optimized model showed improved simulation of spatial patterns in biomass, tree cover, and regional forest carbon turnover compared to the prior model.
• Machine learning identified bioclimatic variables as key factors explaining remaining errors in simulated forest carbon turnover.

Conclusions:

• Integrating multiple satellite observations effectively constrains DGVM simulations of vegetation dynamics and carbon cycling.
• Despite improvements, further refinement of model formulations for climate effects on vegetation turnover and mortality is necessary.
• Enhanced DGVMs are critical for reducing uncertainties in climate-carbon cycle feedbacks and predicting future ecosystem responses.