Satellite, model and field observation CO2 and CH4 emissions in the Kati Thanda Lake Eyre basin dryland, Australia

Dryland ecosystems, covering 40–50% of global land surface, experience long dry periods interrupted by episodic floods from extreme rainfall events. These highly variable hydrological conditions strongly influence carbon cycling, yet they remain challenging to monitor due to the remoteness and harsh environmental conditions. Satellite remote sensing, process-based models, such as Lund Potsdam Jena (LPJ) offers a basin wide perspective to track carbon dynamics, complementing the field observations.

In this study, we use monthly satellite observations from OCO-2/OCO-3 and TROPOMI between 2019 and 2024 to estimate CO₂ and CH₄ fluxes from the Kati Thanda Lake Eyre Basin (KTLEB), one of the world’s largest endorheic basins and a significant dryland in Australia. The satellite derived CO₂ and CH₄ fluxes were compared with LPJ model simulations, to evaluate spatial, seasonal, and interannual variability, and further compared against field measurements from flooded and non-flooded sites in 2019, 2022, and 2024. In addition, fluxes were compared against multi sensor water and vegetation indices from Landsat, Sentinel, and MODIS to investigate influence of flooding, and vegetation on carbon fluxes.

We find that satellite derived atmospheric CO₂ (XCO₂) concentrations over KTLEB ranged from 394 to 429 ppm (mean 414 ± 0.3 ppm), showing a significant increase from 2019 to 2024 (τ = 0.84). CH₄ (XCH₄) concentrations ranged from 1.776 to 1.812 ppm (mean 1.794 ± 3.6 ppm), also with a significant increase over time (τ = 0.61). Annual CO₂ fluxes exhibited substantial interannual variability, alternating between net uptake and net emission, whereas CH₄ fluxes were predominantly a net sink. Both satellite and LPJ model fluxes showed similar seasonal trends, with higher CO₂ uptake and CH₄ emissions during wet season, although satellite derived CO₂ estimates showed a stronger seasonal swing and greater variability, and CH₄ estimates were generally higher. Spatially, both datasets showed similar patterns, with CO₂ uptake concentrated in upper catchment and CH₄ emissions prominent along the basin’s major rivers.

Comparison with field measurements showed that CH₄ emissions were higher during wet season, consistent with satellite observations and LPJ model, and annual CH₄ estimates were broadly comparable across field, satellite, and model data during 2019, 2022, and 2024. CO₂ fluxes, however, varied more among the approaches. This is likely because model and satellite may have missed the initial rapid increase in CO₂ fluxes after flooding, and field measurements may have missed some CO₂ uptake by vegetation as the floodplain dried out. This underscores the need to improve models by including flood effects and incorporate vegetation carbon fluxes while upscaling field observations to better reconcile carbon estimates across datasets. Correlation analyses further supported this, as CO₂ emissions were significantly correlated with water and vegetation indices, with consistent NDWI, NDVI, and EVI alignment, while CH₄ was predominantly driven by NDWI.

Overall, combining satellite, model, and field measurements provides complementary insights into dryland carbon dynamics, showing that both CO₂ and CH₄ fluxes are driven by flooding, with CO₂ also influenced by post-flood vegetation activity. This highlights the value of integrated data for understanding carbon fluxes in drylands.