Skip to Main content Skip to Navigation

Modélisation inverse des flux de CO2 en Amazonie

Abstract : A better knowledge of the seasonal and inter-annual variations of the Amazon carbon cycle is critical to understand the influence of this terrestrial ecosystem on climate change. Atmospheric inverse modeling is a powerful tool to estimate these variations by extracting the information on the spatio-temporal patterns of surface CO2 fluxes contained in observations of atmospheric CO2. However, the confidence in the Amazon flux estimates obtained from global inversion frameworks is low, given the scarcity of observations in this region.In this context, I started by analyzing in detail the Amazon net ecosystem exchange (NEE) inferred with two global inversions over the period 2002 — 2010. Both inversions assimilated data from the global observation network outside Amazonia, and one of them also assimilated data from four stations in Amazonia that had not been used in previous inversion efforts. I demonstrated that in a global inversion the observations from sites distant from Amazonia, as well as local observations, controlled the NEE inferred through the inversion. The inferred fluxes revealed large-scale structures likely not consistent with the actual NEE in Amazonia. This analysis confirmed the lack of observation sites in Amazonia to provide reliable flux estimates, and exposed the limitations of global frameworks, using low-resolution models to quantify regional fluxes. This limitations justified developing a regional approach.Then I evaluated the benefit of the regional atmospheric model BRAMS, relative to the global forecast system ECMWF, when both models provided the meteorological fields to drive the atmospheric transport model CHIMERE to simulate CO2 transport in tropical South America at high resolution (~35 km). I simulated the CO2 distribution with both transport models―CHIMERE-BRAMS and CHIMERE-ECMWF. I evaluated the model simulations with aircraft measurements in vertical profiles, analyzing the concentrations associated to the individual measurements, but also with horizontal gradients along wind direction between pairs of sites at different altitudes, or vertically integrated. Both transport models simulated the CO2 observations with similar performance, but I found a strong impact of the uncertainty in the transport models. Both individual measurements and horizontal gradients were most sensitive to NEE, but also to biomass burning CO2 emissions (EFIRE) in the dry season. I found that horizontal gradients were more suitable for inversions than individual measurements, since the former were less sensitive fluxes outside South America and further decreased the impact of the transport model uncertainty in altitude.Finally, I developed two analytical regional inversion systems for tropical South America, driven with CHIMERE-BRAMS and CHIMERE-ECMWF, and made inversions with four observation vectors: individual concentration measurements and horizontal gradients at five vertical levels, close to the surface, or horizontal gradients vertically integrated. I found a strong dependency of the inverted regional and sub-regional NEE and EFIRE emissions budgets on both the transport model and the observation vector. Inversions with gradients yielded a better separation of NEE and EFIRE signals. However, the large uncertainties in the inverted fluxes, did not yield high confidence in the estimates. Therefore, even though my study did not improve the knowledge of seasonal and year-to-year variations of the NEE in Amazonia, it demonstrated need of further efforts to improve transport modeling in the region and the inverse modeling strategy, at least through a careful definition of the observation vector that accounts for the characteristics of the available data, and the limitations of the current transport models.
Document type :
Complete list of metadatas

Cited literature [175 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Wednesday, January 31, 2018 - 1:36:07 PM
Last modification on : Friday, August 21, 2020 - 5:02:40 AM
Long-term archiving on: : Friday, May 25, 2018 - 9:33:12 PM


Version validated by the jury (STAR)


  • HAL Id : tel-01697602, version 1


Luis Molina Carpio. Modélisation inverse des flux de CO2 en Amazonie. Océan, Atmosphère. Université Paris Saclay (COmUE), 2017. Français. ⟨NNT : 2017SACLV040⟩. ⟨tel-01697602⟩



Record views


Files downloads