TL;DR
This paper evaluates atmospheric transport models' ability to reproduce CO2 vertical profiles observed during airborne campaigns over Europe. It finds that mesoscale models generally perform better than global models, but both struggle with accurately representing vertical mixing. The study also compares two methods for estimating regional CO2 fluxes, finding good agreement between a tracer-based "Radon method" and an "influence function" modeling approach.
What Problem Does This Paper Address?
Accurately predicting future climate change hinges on understanding the sources and sinks of atmospheric CO2, particularly at regional scales (100-1000 km). However, these regional fluxes are poorly understood. Inverse modeling, a common technique to quantify these fluxes, relies heavily on atmospheric CO2 measurements. A major challenge for these models is accurately simulating the CO2 concentration gradients between the atmospheric boundary layer (ABL) and the free troposphere (FT). These gradients are crucial for correctly partitioning calculated fluxes across different regions. Vertical CO2 profiles are ideal for assessing how well models capture these gradients and, by extension, their transport processes. This paper addresses this by comparing observed CO2 vertical profiles from airborne campaigns with simulations from different transport and biospheric flux models.
Methods
The study employs two primary methods:
- Model-Transport Comparison: The researchers compare observed CO2 vertical profiles from the CAATER airborne campaigns with simulations from two transport models:
- LMDZt: A global atmospheric transport model with a regional zoom over Europe.
-
CHIMERE: A mesoscale chemical transport model. They test these models in combination with three different biospheric flux models (CASA, SiB2, and ORCHIDEE), all incorporating diurnal cycles. The comparison focuses on how well the models reproduce the CO2 gradients between the ABL and FT, as well as the overall profile shape.
-
Regional CO2 Flux Estimation: For a specific case study, the authors compare two methods for quantifying regional CO2 fluxes:
- Radon Method: This dual-tracer approach uses simultaneous measurements of CO2 and Radon-222 (a naturally occurring radioactive gas emitted from soils) to infer CO2 fluxes. It assumes a relatively constant Radon flux and uses the correlation between CO2 and Radon to estimate CO2 fluxes, accounting for transport and radioactive decay.
- Influence Function (IF) Method: This method uses backward-in-time transport simulations from the LMDZt model to calculate "influence functions." These functions quantify the contribution of each surface grid point to the CO2 concentration at a specific observation point. By multiplying these IFs with a priori surface flux maps (combining air-sea, fossil fuel, and biospheric fluxes), they estimate the regional CO2 flux influencing the measurement.
Datasets
The study utilizes data from two CAATER (Campaigns of Atmospheric Trace Gas Transport and Chemistry) airborne campaigns conducted over Western Europe:
- CAATER-1: May 23-26, 2001 (CO2 and semi-continuous Radon-222 measurements).
- CAATER-2: October 2-3, 2002 (CO2 measurements).
The models are driven by:
- Meteorological Data: ECMWF analyzed fields.
- Biospheric Fluxes:
- CASA (v. 1997)
- SiB2 (v. 1996)
- ORCHIDEE (v. 2005)
- Fossil Fuel Emissions: Andres et al. (1996), adjusted for campaign years.
- Air-Sea Fluxes: Takahashi et al. (1999, 2002).
No new benchmark datasets were introduced; the study focuses on comparing existing models against observed airborne data.
Key Insights and Ideas
- Vertical Transport is a Major Hurdle: The study strongly suggests that accurately representing vertical transport, particularly the mixing between the ABL and FT, remains a significant challenge for both global and mesoscale atmospheric models.
- Mesoscale Models Show Promise: While not perfect, the mesoscale CHIMERE model generally captured CO2 vertical gradients and profile shapes better than the global LMDZt model, suggesting that finer spatial resolution can improve transport simulation.
- Fluxes Matter, But Transport Dominates: Biospheric fluxes do influence CO2 profiles, sometimes significantly (especially ORCHIDEE), but the study concludes that errors in transport models are the more frequent cause of mismatches with observations.
- Dual-Tracer and IF Methods Agree: The comparison of the Radon method and the Influence Function method for flux estimation showed remarkable agreement, providing confidence in both approaches for regional flux quantification.
- "Footprint" of Influence: The Influence Function method effectively defines the geographical area (the "footprint") that contributes to CO2 concentrations at a specific observation point, which is crucial for understanding regional flux sources.
Key Results
- Profile Comparison:
- For CAATER-1, all model combinations were too "diffusive," failing to reproduce the observed CO2 gradient between the ABL and FT. CHIMERE-ORCHIDEE showed a better shape but was still too variable.
- For CAATER-2, LMDZt's sensitivity to biospheric fluxes became more apparent, but CHIMERE-ORCHIDEE was still more representative of the observed profile structure, though sometimes too variable.
- ORCHIDEE was consistently the best-performing biospheric flux model across different transport models.
- Flux Estimation:
- For a case study on May 25, 2001, both the Radon method and the IF method estimated a regional CO2 sink of -4.2 to -4.4 gC m⁻² day⁻¹.
- The uncertainty for the Radon method was estimated at 35%, and for the IF method at least 33% on average.
- The IF method identified an influence area of approximately 1500 km x 700 km over Northern France, Germany, and Poland for an observation point at 1700 m a.s.l.
- Deconvolution of fluxes for the case study indicated a biospheric sink contributing 73% and fossil fuel emissions 27%.
Limitations and Future Work
- Model Parameterizations: The authors highlight that the parameterizations for vertical transport, particularly entrainment zones and non-mixing regions at the ABL top, need improvement in models like LMDZt.
- Data Sparsity: The study emphasizes the critical need for more intensive and regular airborne campaigns to gather more vertical profiles. This data is essential for cross-validating models and improving transport parameterizations.
- Flux Optimization: The IF method relies on a priori flux maps. Future work should involve using optimized fluxes derived from atmospheric inversions to reduce uncertainties.
- IF Method Uncertainty: The authors acknowledge that the uncertainties of the influence function method need further assessment, suggesting applications at tall towers where simultaneous CO2 and Radon measurements are available.
- Mesoscale Model Variability: While CHIMERE performed better in representing vertical structure, its occasional excessive variability leading to incoherent structures needs further investigation.
Summary updated