Assessing progress toward the Paris climate agreement from space


               <jats:p>N/A</jats:p>

Since the Industrial Revolution, growing 'atmospheric stocks' of greenhouse gas (GHG) concentrations (Joos andSpahni 2008, Dlugokencky andTans 2022) have led to global average temperature increases of over 1 • C with carbon dioxide (CO 2 ) the leading contributor, accounting for roughly half the total warming (IPCC 2014). Left unchecked, continued growth in GHG concentrations due to fossil fuel combustion, deforestation, and other human activities is projected to have catastrophic impacts on the Earth's climate and habitability (IPCC 2014). While the impact of human activity on climate may be well understood globally over several decades, the ability to trace more granular activity, e.g. annual fuel usage of individual countries, to changes in GHG concentrations remains a landmark goal needed to support climate change mitigation efforts. This commentary addresses how this goal might be achieved with a focus on transparency, verifiability, and the challenges that lie ahead.
The 2015 Paris Agreement (PA; UNFCCC 2016), reaffirmed by the 2021 Glasgow Pact (UNFCCC 2022a), plans to limit future global temperature increases through coordinated reduction of GHG emissions. These efforts will be evaluated every 5 years during UNFCCC Global Stocktakes (GSTs; PA Article 14) with the first ongoing. GSTs are intended to 'assess collective progress [and] have no individual Party focus' (UNFCCC 2019). Individual Parties are instead expected to assess their own progress through the Enhanced Transparency Framework (ETF; PA Article 13), a series of biennial reports starting in 2024. Our recent study (Weir et al 2021b) showed that NASA's Orbiting Carbon Observatory 2 (OCO-2; Crisp et al 2004, Eldering et al 2017 was able to detect the impact of human activity reductions on atmospheric CO 2 concentrations during the coronavirus disease 2019  pandemic. This achievement marked the first time that monthly, regional changes in fossil fuel emissions were observed in the atmosphere from space. While emissions decreases due to COVID-19 were historically large, the signal in column CO 2 was just 0.14-0.62 ppm, about 0.1% of background values. Nevertheless, year-toyear decreases in CO 2 growth of similar magnitude, if not greater, will be needed to meet even the moderate goal of less than 4.5 • C warming in 2100 (Forster et al 2020, Le Quéré et al 2020. Current spaceborne technology is therefore at or near the threshold for detecting regional impacts of mitigation efforts on GHG levels. Future advances in coverage, resolution, quality, and co-sampled species from planned missions (Crisp et al 2018, Moore et al 2018, Sierk et al 2019, Tsujihara et al 2021, are expected to further improve the capabilities of GHG monitoring systems like our own. For example, the Geostationary Carbon Cycle Observatory (GeoCarb; Moore III et al 2018) mission will observe the Americas twice daily, whereas OCO-2 soundings repeat after 16 days.
The impact of COVID-19 is just one application of our broader effort to maintain and distribute results from a quasi-operational GHG monitoring system. To support scientific and stakeholder needs, we provide 16 day, column-average CO 2 concentrations, anomalies, and uncertainties on a  -up' simulations (yellow, center) compile process representations from inventories, proxy data, and models to produce gridded fluxes, which are transported through the atmosphere to simulate GHG stocks/concentrations, which then force the climate system. In coupled/interacting systems, those climate outputs, typically radiation, soil moisture, and temperature, can be fed back into the underlying process models (top and bottom arrows). Data assimilation (left, blue) ingests data from the growing fleet of GHG-observing satellites to produce the best possible estimates of GHG concentrations and their uncertainties. The 'top-down' approach, another form of data assimilation, goes directly from GHG observations to gridded fluxes through inversion (grey, left). While flux inversion is a valuable tool, it is not considered an essential part of the proposed assessment methodology. Verification with atmospheric data (red, top right) compares simulated and assimilated values from the other branches to independent in situ and remote-sensing data. The ability to verify concentrations enables further analysis of the detection, attribution, and prediction of those signals and to assess different a priori assumptions about causal processes (see example in figure 2 and discussion below).
nominally 50 km global grid through the trilateral NASA/ESA/JAXA Earth observing (EO) dashboard (https://eodashboard.org) and other NASA assets (Weir andOtt 2022a, 2022b). These fields are the result of assimilating OCO-2 data into NASA's God Data assimilation synthesizes the best available science and observations to infer the state of a system, in our case time-varying fields of atmospheric CO 2 concentrations on a spatial grid. Combined with additional, 'bottom-up' simulations and independent verification data, assimilated products enable inferences about plausible driving processes behind observed changes and uncertainties in GHG concentrations, where the term 'process' is applied as broadly as possible, including human activity, natural variability, and any other phenomena affecting GHGs (figure 1). Notably, in figure 1 all information meets at GHG concentrations where simulations of different processes (center, yellow), assimilated fields (left, blue), and verification data (right, red) are comparable in the same space. A further ability to study interactions across the Earth system (top and bottom center arrows) is possible through ongoing efforts to couple and integrate model components, e.g. sharing biospheric models in GEOS and the Goddard Institute for Space Studies' ModelE (Kelley et al 2020).
A mitigation assessment with a strong focus on atmospheric concentrations over surface fluxes has several advantages. First, an assessment based on surface fluxes alone would be challenging because fluxes are unverifiable with atmospheric data except within a few kilometers of a few hundred measurement sites (Raczka et al 2013) or during intensive aircraft campaigns (Hannun et al 2020). Instead, the primary means for 'verifying' surface fluxes at broad scales is indirect (Schuh et (Hall et al 2021). In the case of fossil fuels, the gridded fluxes are outputs of a spatial disaggregation of inventory totals (transition from processes to gridded fluxes in figure 1), a process with its own uncertainties (Oda et al 2019. Attribution is even more challenging, causing fossil fuel flux estimates derived from atmospheric observations to thus far rely on additional isotope (Basu et al 2020) or tracer (Reuter et al 2019, Pickers et al 2022 data. Second, under almost any circumstances, concentration analyses can run at higher resolutions than flux inversions. Among an intercomparison of several analyses, many running at horizontal resolutions of a few 100 km, only OCO-2/GEOS was able to reproduce observed gradients in column CO 2 from Pasadena to Edwards, California, a distance of roughly 100 km (Schuh et al 2021). This gradient, due to emissions trapped in the Los Angeles basin, is one of the clearest urban-to-background gradients in the world, and underscores the importance of high native resolution if a system hopes to capture signals from fossil fuel emission changes. Third, concentration analyses will always be needed as boundary conditions for yet higher-resolution analyses over limited domains. Such studies are likely necessary at urban (1 km) and finer scales for the foreseeable future due to resource constraints of even the most powerful supercomputers. Finally, concentration analyses can easily integrate multi-species and heterogeneous datasets of almost any size, supporting the coordinated implementation of sustainable development goals through synergies with air quality analyses (Miyazaki et al 2020, Keller et al 2021. For example, our system grew out of a meteorological system monitoring ozone (Wargan et al 2017) and the recovery of the Antarctic ozone hole due to the Montreal Protocol, avoiding a duplication of effort.
There are, nevertheless, cases when a posteriori gridded surface flux estimates may be convenient and/or necessary. Surface fluxes are certainly easier to store and manipulate than concentrations that resolve an additional vertical dimension. Furthermore, anomaly analyses of concentrations must account for transport variability in a non-trivial way (see materials and methods of Weir et al 2021a). When surface fluxes are necessary, they do follow from atmospheric concentrations through simple mass balance relationships (Crevoisier et al 2010). Quantifying the uncertainty of these relationships would likely require, at the very least, a sporadic, baseline surface flux estimate. But even if the exact surface fluxes of every Party to the PA were known, we would be far from its ultimate goal. Effective mitigation requires the ability to assess and predict how different socioeconomic pathways/processes impact GHG concentrations and the subsequent responses of the Earth's climate and biosphere.
To be truly useful for mitigation purposes, any assessment of surface fluxes must have some attribution ability, at the very least distinguishing between natural and anthropogenic origins. Again owing to limitations in direct flux verification, this is currently only possible to the extent that different attributions are evident in concentration data. Figure 2 shows an example of our system successfully detecting and assessing the impact of human activity on GHG concentrations: on 15-30 April 2020, while many countries were under COVID-19 lockdowns, we detected notable decreases (negative anomalies) in CO 2 growth relative to a 2017-2019 baseline. The panels depict the following anomalies: (A) observationbased total CO 2 from our concentration analysis, OCO-2/GEOS, which assimilates OCO-2 into GEOS (Weir et  dominates' fossil fuels, defined here as when the absolute value of the difference of (C) and (D) is greater than (B). Even during a historic short-term decline in emissions, the atmospheric signal of activity reductions is small compared to natural variability over much of the Earth: compare (B), (C), and (D). Nevertheless, over the Eastern United States, Europe, India, and China, the observation derived total CO 2 anomalies (A) compare well with the two simulations of total anomalies (C) and (D), with (A) and (C) agreeing well across the globe. These results suggest that (a) OCO-2 detected regional signals driven by short-term changes in human activity, (b) the corresponding CarbonMonitor fossil fuel emissions are consistent with the observed signal, and (c) the system has skill making assessments in general.
While our qualitative analysis suggests OCO-2 and CarbonMonitor produced consistent estimates of a historically large signal, we have thus far stopped short of making quantitative emissions estimates. This step still requires significant advances in our understanding of remote-sensing , natural surface flux variability (Ott et al 2015, Weir et al 2021b, Peiro et al 2022, atmospheric transport (Schuh et al 2019(Schuh et al , 2022, atmospheric chemistry (Wang et al 2020), and several other factors (Andres et al 2014, Oda et al 2019, Schuh et al 2021. In figure 2, for example, the observational analysis (A) suggests a much smaller signal over India than estimated by CarbonMonitor (B). Yet this is an area with considerable biospheric variability, apparent as differences between (C) and (D), and one in which consistently high aerosol loadings often prevent space-based observations and can introduce retrieval biases (more discussion in Weir et al 2021a). Successful quantitative attribution to fossil fuel emissions will likely require integration of several observational constraints including those from isotopes (Basu et al 2020), atmospheric potential oxygen (Pickers et al 2022), and nitrogen dioxide (Reuter et al 2019).
The ability to assess national and subnational emissions in a transparent and verifiable way remains a landmark, unsolved, long-term scientific goal. The greatest difficulty is posed by the fact that national emissions estimates, and thus stock changes, are not directly verifiable. Atmospheric stocks/concentrations of GHGs, conversely, are verifiable across spatial and temporal scales, especially so with the growing satellite constellation (Crisp and Munro et al 2018) and multi-institution collections of in situ measurements (Masarie et al 2014). When used to evaluate different potential driving processes, we have demonstrated how concentration analyses are powerful tools for mitigation assessments. Even with an improving ability to connect atmospheric observations to surface fluxes, global atmospheric concentration analyses with quantified uncertainties and verified against independent data, such as our own (Weir and Ott 2022a), will continue to serve as a verification baseline for new observations and models. Finally, among all stocks (land, ocean, etc), the atmospheric GHG stock controls the temperature changes (Hansen et al 2010, Gelaro et al 2017, AIRS 2019, Lenssen et al 2019 the PA aims to limit, making it perhaps the most essential climate variable.

Data availability statement
No new data were created or analysed in this study.