Representing model uncertainty for global atmospheric CO<sub>2</sub> flux inversions using ECMWF-IFS-46R1

Abstract

Abstract. Atmospheric flux inversions use observations of atmospheric CO2 to provide anthropogenic and biogenic CO2 flux estimates at a range of spatio-temporal scales. Inversions require prior flux, a forward model and observation errors to estimate posterior fluxes and uncertainties. Here, we investigate the forward transport error and the associated biogenic feedback in an Earth system model (ESM) context. These errors can occur from uncertainty in the initial meteorology, the analysis fields used, or the advection schemes and physical parameterisation of the model. We also explore the spatio-temporal variability and flow-dependent error covariances. We then compare the error with the atmospheric response to uncertainty in the prior anthropogenic emissions. Although transport errors are variable, average total-column CO2 (XCO2) transport errors over anthropogenic emission hotspots (0.1–0.8 ppm) are comparable to, and often exceed, prior monthly anthropogenic flux uncertainties projected onto the same space (0.1–1.4 ppm). Average near-surface transport errors at three sites (Paris, Caltech and Tsukuba) range from 1.7 to 7.2 ppm. The global average XCO2 transport error standard deviation plateaus at ∼0.1 ppm after 2–3 d, after which atmospheric mixing significantly dampens the concentration gradients. Error correlations are found to be highly flow dependent, with XCO2 spatio-temporal correlation length scales ranging from 0 to 700 km and 0 to 260 min. Globally, the average model error caused by the biogenic response to atmospheric meteorological uncertainties is small (<0.01 ppm); however, this increases over high flux regions and is seasonally dependent (e.g. the Amazon; January and July: 0.24±0.18 ppm and 0.13±0.07 ppm). In general, flux hotspots are well-correlated with model transport errors. Our model error estimates, combined with the atmospheric response to anthropogenic flux uncertainty, are validated against three Total Carbon Observing Network (TCCON) XCO2 sites. Results indicate that our model and flux uncertainty account for 21 %–65 % of the total uncertainty. The remaining uncertainty originates from additional sources, such as observation, numerical and representation errors, as well as structural errors in the biogenic model. An underrepresentation of transport and flux uncertainties could also contribute to the remaining uncertainty. Our quantification of CO2 transport error can be used to help derive accurate posterior fluxes and error reductions in future inversion systems. The model uncertainty diagnosed here can be used with varying degrees of complexity and with different modelling techniques by the inversion community.

The CO2 Human Emissions (CHE) Project: First Steps Towards a European Operational Capacity to Monitor Anthropogenic CO2 Emissions

“…The availability of reliable CO 2 concentrations and their transport will provide lateral boundary conditions for regional-scale and local scale inversions. Moreover, the availability of ensemblebased CO 2 realisations (McNorton et al, 2020) will enable offline modelling and coupled assimilation efforts to refine emissions detection capabilities, adapted to the CO 2 long-lived atmospheric concentration (Bousserez, 2019). In parallel to the CO 2 developments, exploratory studies for the CH 4 (Barré et al, 2020) have shown the capability of the CAMS system for local emission detection.…”

“…We also have numerically assessed how anthropogenic CO 2 emissions are depending on country of origin based on IPCC 2006 Guidelines and its Refinement of 2019 (Choulga et al, 2020), see Figure 6. These uncertainties gridded globally at 36 and 9 km resolutions, provided prior uncertainty information for CO 2 ensemble runs (McNorton et al, 2020), see Figure 7, and The atmospheric uncertainty in CO 2 , shown in Figure 7, is the combined effect of anthropogenic emission uncertainties (largest over emission hotspots in eastern China, and smaller signals over North America, Europe and the Middle East), as well as biogenic emission uncertainties in areas with high net ecosystem exchange, such as the Amazon and Southern Africa.…”

“…The CHE project started in October 2017, bringing together a consortium of 22 European partners and lasting for over 3 years. By the end of the CHE project, the global-scale developments have 1) demonstrated the high-resolution CO 2 modelling capability in global Nature-runs (Agusti-Panareda et al, 2021), 2) integrated updated global CO 2 sectoral emission datasets (Choulga et al, 2020), 3) constructed a new high-resolution (∼10 km) input dataset for fossil CO 2 emissions since the 1960s (Jones et al, 2021), and 4) advanced the use of Ensemble-based uncertainty characterisation preparing the data assimilation step (McNorton et al, 2020). Sizeable advances at European scale include the realisation of high-resolution CO 2 emission inventories (Super et al, 2020b) that served also as exploratory studies of what can be achievable at global scale, where high quality inventories are made available.…”

WOMBAT v1.0: A fully Bayesian global flux-inversion framework

“…introduces errors that could span hundreds of kilometres and several days (Lauvaux et al, 2019;McNorton et al, 2020).…”

“…The complexity of all modelled processes, from fluxes right up to satellite retrieval errors, inevitably leads to model misspecification (e.g., Engelen et al, 2002). The main causes of misspecification are (i) flux-process dimension-reduction error (e.g., Kaminski et al, 2001), which is a consequence of using a spatio-temporal model for the flux field that is low-dimensional and inflexible; (ii) an inaccurate prior flux mean, variance, and covariance (e.g., Philip et al, 2019); (iii) transport-model errors (e.g., Houweling et al, 2010;Basu et al, 2018;Schuh et al, 2019) arising from the underlying assumed physics, meteorology, and discretisation schemes used (e.g., Lauvaux et al, 2019;McNorton et al, 2020); (iv) retrieval biases (e.g., O'Dell et al, 2018) and incorrect associated measurement-error statistics (e.g., Worden et al, 2017); and (v) measurement-error spatio-temporal correlations that are not fully accounted for (e.g., Chevallier, 2007;Ciais et al, 2010). Two other causes of model misspecification worth noting are an incorrectly specified initial global mole-fraction field, and flux components assumed known in the inversion (i.e., assumed degenerate at their prior mean), such as anthropogenic emissions (e.g., Feng et al, 2019).…”