Climate has become a subject of global concern, especially in recent decades. Climate models are practical tools that can simulate physical processes and predict future change. However, because of the complexity of atmospheric, ocean, and land processes, scientists are faced with significantly large uncertainties in climate models. As world leaders grapple with the urgency of climate action, the role of space-based technology and data has become increasingly critical. Various observed climatic variables (e.g. temperature, precipitation, and sea ice, etc.) are needed for a better understanding of climate change and to improve climate models. Researchers at the European Space Agency(ESA), and the National Aeronautics and Space Administration (NASA), etc. accelerate the use of Earth-observing satellites and the information they provide to address the pressing challenge of climate change.

Satellite remote sensing has provided major advances in understanding the climate system and its changes, by quantifying processes and spatio-temporal states of the atmosphere, land and oceans (Yang et al. 2013). ESA and the European Commission are joining forces to use of Earth-observing satellites and the information they provide for climate change research. ESA has established the Climate Modelling User Group (CMUG) to place a climate system perspective at the centre of its Climate Change Initiative (CCI), and to provide a dedicated forum through which the Earth observation data community and the climate modelling and reanalysis community can work closely together. CMUG will work with the Essential Climate Variable CCI projects to achieve this goal (CMUG). Figure 1 shows the Essential Climate Variables (ECVs) specified by the Global Climate Observing System (GCOS) currently. The Global Climate Observing System (GCOS) has listed 26 out of 50 essential climate variables (ECVs) as significantly dependent on satellite observations (Yang et al. 2013). Figure 2 shows ESA's Technology CubeSats. The European Commission’s Director General for Climate Action, Kurt Vandenberghe, added,

“Space, and in particular Earth observation, offers a unique perspective on how to tackle climate challenges faced by humanity”(ESA 2023).

26 out of 50 Essential Climate Variables (ECVs) significantly depend on satellite observations!

Earth-observing data assimilate to the climate models for improving accuracy

How can the climate models be advanced by Earth-observing data? Figure 3 shows the concept that researchers incorporate space-based satellite data and other Earth-observing sources (remote sensing, in situ, social sensing data) into ultra-high-resolution Earth system models.

With significant increases in processing power, big data assimilation and processing by machine learning algorithms, multi-source observations became a feasible method to improve climate models. Figure 4 shows the schematic of using observations from multiple sensors to advance climate predictions. It illustrates that the key process linking predictive models and observations together, which is so-called data assimilation. This assimilation system is a way of inverting the observation model to adjust the physical state (xi) of the forecast model (f) to be closer to the observations (y). The forecast model uses the state (x) and set of parameters (ωi) with a set of physical equations to generate a prediction. Algorithms such as deep learning are commonly used to establish the relationship of the observed variable and the simulated one.

The example in Figure 5 shows that ESA-CCI sea ice observations are applied to improve summer predictions in the Arctic. Sea ice is a vital component of the Arctic climate system and a local source of climate predictability. Navarro et al. compared two sets of retrospective seasonal predictions respectively initialized with and without assimilating sea ice concentration (SIC) in the Arctic Sea from satellite observations (SIC-ASSIM and NOSIC-ASSIM) (Navarro et al. 2022). A comparison of predicting performance (anomaly correlation coefficient) between SIC-ASSIM and NOSIC-ASSIM during the first (a) May, the second (b) June, and the third to fifth (c) JAS (July-August-September) forecast months. The predictions in JAS months are in long-range time lead, they are presented in one category. The mean 1992-2019 integrated Arctic Sea ice edge error (IIEE) in SIC-ASSIM (light blue) and NOSIC-ASSIM (light red) hindcasts in (d) Labrador and Baffin, (e) Greenland-Iceland-Nordic (GIN) and (f) Barents Seas. The observational reference is NASA NSIDC (National Snow and Ice Data Center). Dots on the maps (a)-(c) and lines (d)-(f) indicate statistically significant differences (95% confidence level) between SIC-ASSIM and NOSIC-ASSIM. Purple boxes in (a), (b), and (c) indicate the regions of Labrador and Baffin, GIN, and Barents Seas, respectively. Symbols on the left-hand side in (d), (e), and (f) indicate the IIEE values in the initial conditions (1 May). The assimilation of SIC improves the representation of the ice edge for the months of May and June in the Labrador-Baffin, GIN, and Barents Seas (figures 1(d)–(f)). Additionally, IIEE in the GIN and Barents Seas shows improvement in October and November in SIC-ASSIM. This research shows an example of how scientists utilized observational data in climate modelling to enhance the accuracy of sea ice predictions.

Besides improving the forecast quality of Arctic Sea ice, SIC assimilation also leads to improvement in the predicting performance of simulating sea surface temperature (SST ) in the central North Atlantic (50-20W, 35-55N) (Navarro et al. 2022). Figure 6 shows the difference predicting performances in SST in

(a) May,
(b) June,
(c) JAS (July-August-September).
(d) Monthly SST skill (anomaly correlation coefficient).
(e)-(h) and
(i)-(l) as (a)-(d) but for T2m (2m temperature) and GPH500 (GeoPotentialHeight), respectively.

The observational reference is the mean of ECMWF Reanalysis v5(ERA5) (Hersbach et al. 2020), the Japanese 55-year Reanalysis (JRA55) (Kobayashi et al. 2015) and National Centers for Environmental Prediction (NCEP) (Saha et al. 2010) reanalysis for T2m and GPH500, and Hadley Centre Sea Ice and Sea Surface Temperature (HadISSTv1.1) (Rayner et al. 2003) for SST.

NASA is committed to this field as well. The Global Modelling and Assimilation Office (GMAO) uses climate models and data assimilation techniques to enhance NASA’s program of Earth Observations. GMAO uses coupled Earth-System models and analyses, along with a broad range of satellite observations, to study and predict phenomena that evolve on seasonal to decadal timescales.(GMAO) Modern-Era Retrospective analysis for Research and Applications, v2 (MERRA-2) is the representative dataset of GMAO. MERRA-2, based-on climate model, assimilating satellite and other space-based observations, provides reliable reference for Earth System Reanalysis. Data assimilation with space-based observation is widely valued and well applied for climate study over the world.

Takeaways

• The use of satellites allows the observation of states and processes of the atmosphere, land and ocean at several spatio-temporal scales. It makes possible that satellite data can be assimilated to climate models to improve simulating the dynamics of Earth system and climate projections.
• Satellite data also contribute significantly to the improvement of meteorological reanalysis products (i.e. MERRA-2) that are widely used for climate change research.
• Advances in computer science, including artificial intelligence (AI), and supercomputing promote the development of big data assimilation (multisource, multiform) in the climate models
• However, no model is perfect, and satellite data can contain errors: uncertainties must be properly characterized so that users have an accurate picture of the model's reliability.

Conclusion

To cope with the impact of climate change, data assimilation with space-based observations in climate models is practical technologies, and provides crucial scientific references for climate study. From my opinion, as the computing power and intelligent algorithms develop, big data assimilation with satellite data and other sources will enhance the accuracy of forecast and reduce uncertainties in climate modelling.