A carbon-monitoring strategy through near-real–time data and space technology

Summary In this perspective, we proposed an innovative strategy that coupled near-real-time emission data with satellite observations to make a reliable and precise global carbon-monitoring system.


DRAWING THE NEAR-REAL-TIME GLOBAL EMISSION MAP
The current challenge to coupling the "bottom-up" emission inventory with "top-down" satellite observation is that the inventory cannot match the observation in both the spatial and temporal dimensions. 2 The inventory usually lacks high temporal resolution, with only annual or monthly data available, and has a time lag of at least 1 year. Recent progress on the emission estimates based on near-real-time human activities data (smart meters, grid data, transportation mobility, etc.) provides opportunities to advance the emission estimates into near real time. [3][4][5] By using geographic information signals, grid data, or in situ observations (e.g., NOx) as CO 2 emission proxies, near-real-time daily CO 2 emission maps can be developed 6 to serve as prior information to facilitate the detection of potential point-source emitters and validation of satellite observations. 7 The near-real-time global emission maps display high-resolution spatial information on carbon emissions anywhere and anytime when satellites visit (daily scale), thus providing a credible baseline map for satellite observations.

ADVANCING KEY POINT-SOURCE MONITORING BY SATELLITE
Key point emission sources (cities, power plants, etc.) are the main source of anthropogenic carbon emissions (e.g., cities contribute 70% of anthropogenic emissions). With the near-real-time global emission map being constructed, the satellites could now only target huge emitters like cities and plants. This will help keep satellites' advantages on their high resolution and timeliness without humbling the whole emission coverage. For example, the TROPOspheric Monitoring Instrument with km-level spatial resolution is adequate for detecting extreme methane leakage from accidental blowouts and methane ultra-emitters. 8 Such satellites measure radiances in the spectral bands (both visible bands and near-infrared bands) and detect carbon dioxide concentrations based on molecular absorption spectra. In combination with other information (ground observations, meteorological data, etc.), atmospheric inversion models can convert the observed CO 2 concentration values over the key point sources back to the original concentration field and estimate their carbon emissions. For example, current studies that used Gaussian plume model and OCO-2/3 XCO 2 retrievals to estimate corresponding cross-sectional CO 2 fluxes from large emission sources showed broad consistency between satellite-based results and emission inventories. 9 Continuously and accurately tracking emissions on large emission sources could provide a solid basis for large-scale satellite constellation observations. Future developments can be made in several aspects. (1) Satellites with diverse attributes can work together to meet different requirements by combining satellites with multiple spatial and spectral resolutions to track plumes at different scales. For global observation, satellites such as GOSAT with sparse but global coverage, a 3-day revisit period, and a continuous data stream target global observations. As for point-source observation, GHGSat with high sensitivity but spatial coverage limit or multispectral instruments such as Sentinel-2 and WorldView-3 with the high spatial resolution is capable of factory-level monitoring. (2) With more advanced satellite missions and their more sensitive instruments to observe XCO 2 and XCH 4 continuously, like the Geostationary Carbon Cycle Observatory, Carbon Mapper, GHGSat, and the Copernicus Carbon Dioxide Monitoring, specific goals like wide swath, high spatial resolution, frequent revisit, and high accuracy can be anticipated. (3) Artificial intelligence and machinelearning methods can help dig into information that is not directly reflected in satellite imagery. For example, machine-learning methods have been used to automatically enhance the signals of CO 2 measurements by satellites, improving the accuracy of inversion results, 1 or identifies the operation of factories to track the operational status and emissions of these key point sources.

GLOBAL SATELLITE NETWORKING
The satellite constellation is a growing tendency for coming missions that will keep improving the monitoring through the increase in satellite numbers and technologies. Combine all available satellite observations to improve the spatial and temporal accuracy of near-real-time global emission map and assimilation models could help to finally realize global carbon emission monitoring. An integrated observing system is essential to space-based measurements. Following the roadmap of the Committee on Earth Observation Satellites 10 for integrating surface and airborne measurements to support the global stocktake, ll The Innovation 4(1): 100346, January 30, 2023 several carbon monitoring systems have been proposed, such as the CoCO 2 project with the Copernicus Program (https://coco2-project.eu/, European Union) and the NASA Carbon Monitoring System (https://carbon.nasa.gov/, the USA), to accelerate the integration of multiple observation sources and to provide verifiable and transparent data products.
Given the large-scale data generated by satellites and integrated observing systems, standardization and sustainable reusing of observations are guaranteed for more efficient data use. The Copernicus Climate Change Service (https://climate.copernicus.eu/) has provided a landmark example of largescale climate data sharing for a global data-exchange strategy and climate services. A free and open data-sharing platform led by governments will help break down the current data barriers caused by the monopoly of different stakeholders. Cloud-based platforms provide large-scale storage and supercomputing power, such as the Copernicus Climate Data Store toolbox (https://cds. climate.copernicus.eu/) or Google Earth Engine (https://code.earthengine. google.com/). The cloud-based platforms are needed to pool more available observation data and improve the computing power of the earth system modeling and inversion and assimilation system, to increase the spatial resolution from the current 0.1 3 0.1 to the kilometer level, and to shorten the temporal interval from annually to monthly or daily.

CONCLUDING REMARKS
Human activity data-based emission inventory can provide a sector-specific, systematic dataset that is comparable between countries, while satellite observations can support independent, low-cost, spatially distributed, and directly observed datasets, which are especially beneficial for areas that lack bottomup data. Combining the superiorities of the near-real-time dataset and the satellite observations, an innovative technical route is proposed for transitioning from inventory-based carbon monitoring to satellite-driven carbon monitoring and from point-source carbon monitoring to the global scale. A near-realtime carbon map primarily derived from emission inventory with poor resolution in local areas can provide prior information for satellite observations and constrain satellite retrievals. Continuous point-source monitoring by satellites, in turn, can improve accuracy to form a dual-update mechanism with a longterm goal of quantifying anthropogenic emissions by satellite constellation only as an independent method. Based on this mechanism, gathering in-orbit and soon-to-be-launched satellites into a comprehensive observing network allows us to widen the monitoring space from a key point to a unified globe, jointly contributing to the data foundation for implementing international climate treaties and climate policies, and finally pave the way to a more accurate and transparent global stocktake.