A local to national-scale inverse modeling system to assess the potential of spaceborne CO<sub>2</sub> measurements for the monitoring of anthropogenic emissions

Abstract

Abstract. This work presents a flux inversion system for assessing the potential of new satellite imagery measurements of atmospheric CO2 to monitor anthropogenic emissions at scales ranging from local intense point sources to regional and national scales. While the modeling framework keeps the complexity of previous studies focused on individual and large cities, this system encompasses a wide range of sources to extend the scope of the analysis. This atmospheric inversion system uses a zoomed configuration of the regional transport model CHIMERE which covers most of Western Europe with a 2-km resolution grid over Northern France, Western Germany and Benelux. For each day of March and May 2016, over the 6 hours before a given satellite overpass, the inversion controls separately the hourly budgets of anthropogenic emissions in this area from ~300 cities, power plants and regions. The inversion also controls hourly regional budgets of the natural fluxes. This enables the analysis of results at the local to regional scales for a wide range of sources in terms of emission budgets and spatial extent while accounting for the uncertainties associated with natural fluxes and the overlapping of plumes from different sources. The potential of satellite data to monitor CO2 fluxes is quantified by posterior uncertainties or uncertainty reductions (URs) from prior inventory-based statistical knowledge. A first analysis focuses on the hourly to 6-hour budgets of the emissions of the Paris urban area, and on the sensitivity of the results to different characteristics of the images of vertically integrated CO2 (XCO2) corresponding to the spaceborne instrument: the pixel spatial resolution, the precision of the XCO2 retrievals per pixel, and the swath width. This sensitivity analysis provides a correspondence between these parameters and thresholds on the targeted precisions on emission estimates. However, the results indicate a large sensitivity to the wind speed and to the prior flux uncertainties. The analysis is then extended to the large ensemble of point sources, cities and regions in the study domain, with a focus on the inversion system ability to monitor separately neighbor sources whose atmospheric signatures overlap and are also mixed with those produced by natural fluxes. Results highlight the strong dependence of uncertainty reductions to the emission budgets, to the wind speed and whether the focus is on point or area sources. With the system hypothesis that the atmospheric transport is perfectly known, the results indicate that the atmospheric signal overlap is not a critical issue. For the emissions within the 6-hours before a satellite overpass, UR of more than 50 % can only be achieved for power plants and cities whose annual emissions are more than ~2 MtC yr−1. For more regional budgets encompassing more diffuse emissions, this threshold increases up to ~10 MtC yr−1. The results suggest therefore an imbalance of the monitoring capabilities towards high and dense sources.

Complementing XCO2 imagery with ground-based CO2 and 14CO2 measurements to monitor CO2 emissions from fossil fuels on a regional to local scale

“…This section presents the high dimensional inversion framework designed in this study for the co-assimilation of CO 2 and 14 CO 2 data. It has strong similarities with the system developed by Santaren et al (2021), which assimilates CO 2 data only, and it borrows from Wang (2016) to assimilate 14 CO 2 data. The system relies on the following.…”

“…The restriction to 1 d is connected to results of Santaren et al (2021), which show the lack of sensitivity of observations made during a given day to the fluxes during other days over the modelling domain, and to the large computation cost associated with the preparation of a full day of analytical inversion. With such an inversion window, wider than the one chosen in Broquet et al (2018) or Santaren et al (2021), the system tracks the signal from the FF emissions up to 12 h before the satellite overpass (see Sect. 2.6.1) and 10 h before the in situ data assimilation window (see Sect.…”

“…After a few hours, the air masses having been transported over typically ∼ 30-100 km, and the signal from individual FF CO 2 sources (industrial plants, cities, regions) is much diffused and hardly detectable in XCO 2 images. Consequently this 1 d timescale is large enough to represent the full extent of the CO 2 FF plumes that can be exploited in images from CO2M-like instruments to compute the corresponding emissions (Broquet et al, 2018;Santaren et al, 2021). The ability to track large-scale budgets of FF emissions over longer time periods relies on complementary observations of FF emission tracers.…”

“…Previous analyses of the potential of high-resolution satellite imagery of XCO 2 (such as ESA, 2015; Wang et al, 2020;Santaren et al, 2021) have focused on its use as a standalone observation system and on the potential complementarity of images of co-emitted species co-registered with an instrument on board the same satellite or from another mission (Reuter et al, 2019;Kuhlmann et al, 2019Kuhlmann et al, , 2020. However, the distinction between FF and natural CO 2 signals and thus the separation between the FF and natural components in the flux estimates remain difficult, even when using highresolution images and satellite data on co-emitted species (Kuhlmann et al, 2020;Santaren et al, 2021;Sadiq et al, 2021). The separation between the emissions from biofuel (BF) and FF combustion is another challenge because BF emissions can be located in the same hotspots as FF ones (Ciais et al, 2020).…”

“…the relative difference between the posterior uncertainty and the prior uncertainty in the control parameters. The analysis of this uncertainty reduction is made over 1 d at the local scale (urban areas, industrial plants) to the scale of administrative regions in Europe, following the rationale and the general inverse modelling framework of Santaren et al (2021). It focuses on a large part of western Europe, using a regional atmospheric transport model with a 2 km horizontal zoom over northern France, western Germany, Belgium, Luxembourg, and a large part of the Netherlands.…”