Apply local-scale atmospheric inversions to infer patch-level fluxes

Lead: Theresia Yazbeck

Apply local-scale atmospheric inversions to infer patch-level fluxes

Spatial heterogeneity plays a major role in surface-atmosphere interactions, and consequently in estimating greenhouse gas emissions. Many natural ecosystems, particularly coastal wetlands and northern peatlands, are known for their heterogenous landscape, as they are usually composed of different patches like e.g. open water, moss, tall shrubs, or grass areas. Each of these patches is associated with its individual biogeochemical, ecological, and environmental characteristics. Ground-based measurement techniques like eddy covariance towers allow to continuously measure net fluxes from such ecosystems; however, their capability in differentiating fluxes coming from each landcover or patch is limited. The recent emergence of UAV applications in atmospheric science facilitates measuring fine-scale spatial variability in greenhouse gas concentrations close to the surface (within 2-10 m a.g.l.), thus opening the door for investigating flux signatures from different landcovers within flux tower footprints.

In this study, we attempt to estimate patch-level or landcover-specific fluxes by combining atmospheric transport modeling with inversion techniques at small spatial scales, exploiting both eddy-covariance and UAV-concentrations measurements. We employ EULAG (EUlerian LAGrangian), a well-established Large-Eddy Simulation model, to simulate detailed flow patterns driven by heterogeneous permafrost surfaces. Then, we apply inverse optimization techniques to estimate CO₂ fluxes associated with the specific land cover patches within the mixed landscape, utilizing UAV-based grid surveys of carbon dioxide concentrations. We run EULAG with similar meteorological conditions for the time of the UAV flight and prescribed land cover classification, where each class is associated with its specific unit flux. Our method results with a relative value of the fluxes associated with each land cover type compared to the prescribed flux unit value. When combined with Eddy-covariance measurements and a flux footprint model, we are able to derive flux values for each land cover type. We apply this method on a pilot project in Stordalen Mire in subarctic Sweden.

Figure 1 Evaluating modeled (EULAG) and observed (UAV) CO2 concentrations — *Figure 1 Evaluating modeled (EULAG) and observed (UAV) CO₂ concentrations*

*Figure 1 Evaluating modeled (EULAG) and observed (UAV) CO₂ concentrations*

The inferred fluxes are validated with chamber measurements of carbon dioxide for each land cover. The results show a good agreement between modeled and observed concentrations, with patch-level fluxes aligning well with chamber observations. This novel method demonstrates significant potential for capturing flux heterogeneity across land cover types and advancing the application of inversion techniques in high-resolution atmospheric modeling. Our methodology was applied in the context of an Arctic tundra site; however, it is not restricted to northern latitudes peatlands, and we are planning to extend it to other ecosystems.

Figure 2 (a) Mapping of inferred fluxes associated with each land cover at Stordalen Mire. (b) The gridded line shows the UAV trajectory and the red triangle shows the Eddy-Covariance tower location. The aerial view of the mire is provided through Google Maps 2025.

Figure 2 (a) Mapping of inferred fluxes associated with each land cover at Stordalen Mire. (b) The gridded line shows the UAV trajectory and the red triangle shows the Eddy-Covariance tower location. The aerial view of the mire is provided through Google Maps 2025.

Reference

Yazbeck, T., Schlutow, M., Bolek, A., Triches, N. Y., Wahl, E., Heimann, M., and Göckede, M.: Quantifying landcover-specific fluxes over a heterogeneous landscape through coupling UAV-measured mixing ratios with a large-eddy simulation model and Eddy-covariance measurements, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2025-3791, 2025.