Joint CO2 Mole Fraction and Flux Analysis Confirms Missing Processes in CASA Terrestrial Carbon Uptake Over North America

Terrestrial biosphere models (TBMs) play a key role in the detection and attribution of carbon cycle processes at local to global scales and in projections of the coupled carbon-climate system. TBM evaluation commonly involves direct comparison to eddy-covariance flux measurements. We use atmospheric CO2 mole fraction ([CO2]) measured in situ from aircraft and tower, in addition to flux-measurements from summer 2016 to evaluate the Carnegie-Ames-Stanford-Approach (CASA) TBM. WRF-Chem is used to simulate [CO2] using biogenic CO2 fluxes from a CASA parameter-based ensemble and CarbonTracker version 2017 (CT2017) in addition to transport and CO2 boundary condition ensembles. The resulting super ensemble of modeled [CO2] demonstrates that the biosphere introduces the majority of uncertainty to the simulations. Both aircraft and tower [CO2] data show that the CASA ensemble net ecosystem exchange (NEE) of CO2 is biased high (NEE too positive) and identify the maximum light use efficiency E-max a key parameter that drives the spread of the CASA ensemble in summer 2016. These findings are verified with flux-measurements. The direct comparison of the CASA flux ensemble with flux-measurements confirms missing sink processes in CASA. Separating the daytime and nighttime flux, we discover that the underestimated net uptake results from missing sink processes that result in overestimation of respiration. NEE biases are smaller in the CT2017 posterior biogenic fluxes, which assimilate observed [CO2]. Flux tower analyses reveal an unrealistic overestimation of nighttime respiration in CT2017 which we attribute to limited flexibility in the inversion strategy.

Automated summary

Terrestrial biosphere models play a crucial role in studying carbon cycle processes and the carbon-climate system. Evaluation of the CASA TBM using CO2 measurements and flux data reveal uncertainties introduced by the biosphere and highlight issues with missing sink processes and the sensitivity of the model to key parameters such as maximum light use efficiency.