Simulation of Total-Column Footprints with FLEXPART 11 for a Regional Inverse Application in the Lund University Modular Inversion Algorithm (LUMIA)

Abstract

Accurate estimations of regional CO2 fluxes are essential for understanding the global carbon cycle and further mitigating climate change. Satellite-based retrievals of column-averaged carbon dioxide (CO2) from the Orbiting Carbon Observatory 2 (OCO-2) offer a promising avenue for estimating carbon fluxes with atmospheric inversion systems, as they provide higher spatial resolution and observational density than traditional in situ networks. However, the large volume of satellite observations presents the challenge of how the data can be processed in order to make simulations computationally feasible.
In this study, total-column CO2 retrievals from OCO-2 were used to simulate total-column footprints with the Lagrangian FLEXible Particle Dispersion Model (FLEXPART) v.11. To reduce computational costs, an adaptive grid cell averaging scheme was applied to the satellite observations prior to the footprint calculations. The precomputed footprints were then used to assimilate OCO-2 observations for January and August 2022 with the Lund University Modular Inversion Algorithm (LUMIA). The resulting posterior fluxes were validated against ground-based observations from the Integrated Carbon Observation System (ICOS), non-ICOS sites contained in the Observation Package (ObsPack) dataset and the Total Column Carbon Observatory Network (TCCON).

The results show that the OCO-2-derived posterior fluxes reduced both mean biases and root mean square error (RMSE) in simulated CO2 concentrations at many validation sites across Europe. The inversion performance was notably better in August than in January, reflecting the stronger observational constraint associated with the higher density and spatial coverage of OCO-2 observations during summer. Sensitivity experiments investigating the temporal and spatial prior error covariance parameters indicated that a temporal correlation length of 30 days and a spatial correlation length of 100 km yielded the most effective configuration for OCO-2 observations, in terms of reducing posterior mean biases at the validation sites.

Popular Abstract

Climate change is largely driven by human activities such as the burning of fossil fuels, which release large amounts of carbon dioxide (CO2) into the atmosphere. This additional CO2 disrupts the Earth’s natural carbon cycle by altering the balance between carbon sources and carbon sinks, such as forests and oceans, which absorb part of the emitted carbon. In particular, the future development of the land carbon sink remains uncertain under a changing climate, highlighting the need for accurate estimates of carbon fluxes.

One method used to estimate carbon fluxes is atmospheric inverse modeling. This approach combines atmospheric transport models with observations of CO2 concentrations to improve prior estimates of carbon fluxes. The observations can originate either from ground-based measurement stations or from satellites. Ground-based stations provide highly accurate measurements at specific locations, whereas satellites measure the average CO2 concentration throughout an entire air column, reaching from the Earth’s surface to the top of the atmosphere. Hence, the resulting observation is a total-column measurement. Compared with the relatively sparse ground-based observation network in Europe, the Orbiting Carbon Observatory 2 (OCO-2) satellite provides much denser spatial coverage, enabling observations over large areas. However, satellite measurements depend on sunlight and are affected by clouds and aerosols, leading to lower data availability during winter months. As the ability to improve prior carbon flux estimates with atmospheric inverse systems is ultimately limited by observational density, using satellite-based observations is desirable.

The aim of this study was to extend the capabilities of the Lund University Modular Inversion Algorithm (LUMIA) to estimate carbon fluxes constrained by satellite-based CO2 observations. Specifically, the current version of the atmospheric transport model (FLEXible PARTicle Dispersion Model - FLEXPART 11) used within LUMIA enables a simulation of the transport with total-column data, such as satellite observations. These transport simulations are also called total-column footprints. The scope of this project was to create the total-column footprints for LUMIA, which are needed to ultimately estimate the carbon fluxes constrained by observations. However, since satellites generate a very large number of observations, efficient processing methods are required to make the simulations computationally feasible. The inversion system ultimately produces estimates of carbon fluxes, which can then be converted into atmospheric CO2 concentrations and compared with independent observations. This allows the evaluation of whether the optimized flux estimates improve the agreement between simulated and observed concentrations.

The results showed that the satellite-derived posterior fluxes reduced concentration biases at many validation sites across Europe. The performance of the inversion system was generally better during August, when observational coverage from OCO-2 was high, than during January, when the number of available satellite observations was substantially lower. Overall, this study demonstrates the potential of integrating satellite observations into regional atmospheric inversion systems, such as LUMIA, and explores the new capabilities of FLEXPART 11.

• author: Rissbacher, Johanna
• supervisor: Marko Scholze; Yohanna Villalobos; Vishnu Thilakan Manaparambil
• organization: Mathematics (Faculty of Sciences)
• course: BERM02 20261
• year: 2026
• type: H2 - Master's Degree (Two Years)
• subject: Science General
• keywords: Atmospheric Inverse Modelling Greenhouse Gases, CO2 FLEXPART 11 Orbiting Carbon Observatory 2 (OCO-2) Lund University Modular Inversion Algorithm (LUMIA)
• publication/series: Master's Theses in Mathematical Sciences
• report number: LUNFBV-3014-2026
• ISSN: 1404-6342
• other publication id: 2026:E66
• language: English
• id: 9246827
• date added to LUP: 2026-07-27 10:46:48
• date last changed: 2026-07-27 10:46:48