SAGE III/ISS Data Assimilation System Effective In Reducing Stratospheric Observational Data Gaps

Water vapor in the lower portions of the atmosphere is an integral part of our global weather systems. However, in the stratosphere, water vapor, along with nitrous oxide and hydrogen chloride, can impact ozone layer chemistry. NASA's Microwave Limb Sounder (MLS) aboard Aura satellite has been the backbone of stratospheric retrievals of ozone-depleting substances and ozone concentrations since 2004. Yet Aura is coming to the end of its life, and our ability to monitor stratospheric water vapor, nitrous oxide and hydrogen chloride from space will decrease, relying on the heritage Atmospheric Chemistry Experiment (ACE) Fourier Transform Spectrometer (FTS) instrument, onboard the Canadian SCISAT satellite since 2004. However, this is less ideal, because with the SCISAT orbit, ACE-FTS observes each latitude during specific times of the year, taking over three months to get full coverage (+/- 85°). Figure 1 (Salawitch et al., 2025), produced using the MERRA-2 Stratospheric Composition Reanalysis of Aura Microwave Limb Sounder (M2-SCREAM), shows time series of anomalies (departures from the long-term average) as functions of latitude and time at selected pressure levels in the stratosphere for these four chemical species. Following a relatively uneventful 15-year period between 2005 and 2019, several unusual events occurred within a span of only four years that left distinct imprints on stratospheric composition.

From unusual dynamics in 2019 to the eruption of the submarine volcano, Hunga in 2022, a variety of different processes led to large and long-lasting anomalies in the concentrations of several gases important for stratospheric chemistry, including ozone chemistry. Notably, the Hunga eruption increased the total water vapor mass in the stratosphere by 10% (Figure 1, G); and the smoke from Australian wildfires in 2020 enabled heterogeneous chemical reactions, details of which are not fully understood, that led to a significant repartitioning of stratospheric chlorine (Figure 1, E). The clustering of at least significant extreme events in recent years highlights the fact that, despite significant scientific advances, the stratosphere is not “a solved problem” and underscores the importance of continuing satellite observations of stratospheric composition beyond ozone.

In Knowland et al. (2025), we show how the infrequent but high quality observations from the Stratospheric Aerosol and Gas Experiment (SAGE) III solar occultation instrument on the International Space Station (SAGE III/ISS) since 2017 can be used in a data assimilation system (DAS) much like M2-SCREAM, hereafter referred to as SAGE DAS, to provide a data-constrained stratospheric water vapor product following the end of the MLS instrument.

Figure 1: Deseasonalized and detrended monthly zonal mean anomalies (departures from long-term averages) in four gases important for studies of stratospheric chemistry and dynamics: water vapor (H2O), nitrous oxide (N2O), hydrogen chloride (HCl), and ozone. The anomalies are shown at selected altitudes in units of standard deviation. Three-sigma values are shown in color and four-sigma events are circled and described in the legend. This figure was made using the MERRA-2 Stratospheric Composition Reanalysis of Aura Microwave Limb Sounder (M2-SCREAM: Wargan et al., 2023) produced at the GMAO, and published in Salawitch et al., 2025.

In Figure 2, M2-SCREAM, SAGE DAS and a chemistry-only (no DAS) Control simulation are compared against the independent ACE-FTS measurements. The difference in the data-constrained model products and the Control to ACE-FTS is consistent with the documented differences between the observations found in the literature. The strength of the DAS -- where the information from the infrequent SAGE III/ISS observations (15 to 30 profiles per day compared to MLS with 3500 profiles) can be propagated in both space and time – is best seen in the standard deviation of the differences and the correlation (Figure 2c and 2d). Throughout the stratosphere, where the correlation is high and standard deviation of the difference is small, the data-constrained model products are capturing the variability also observed by ACE-FTS.

Figure 2: Statistical comparison of 2018 to 2021 ACE-FTS stratospheric water vapor (ppmv; black line) to the model products:  M2-SCREAM (blue line), SAGE DAS (cyan line), and Control (red line). In panel (c), for reference, ACE-FTS instrument error and standard deviation of the ACE-FTS observations given as grey and black lines, respectively.  Figure reproduced from Knowland et al. (2025).

In Figure 2, M2-SCREAM, SAGE DAS and a chemistry-only (no DAS) Control are compared against the independent ACE-FTS measurements.  The difference in the data-constrained model products and the Control to ACE-FTS is consistent with the documented differences between the observations found in the literature. The strength of the DAS – where the information from the infrequent SAGE III/ISS observations (15 to 30 profiles per day compared to MLS with 3500 profiles) can be propagated in both space and time – is best seen in the standard deviation of the differences and the correlation (Figure 2c and 2d). Throughout the stratosphere, where the correlation is high and standard deviation of the difference is small, the data-constrained model products are capturing the variability also observed by ACE-FTS.

Another test of the SAGE DAS results is how well the product captured the tropical “tape recorder”. This is the 18-month cycle of tropospheric water vapor entering into the stratosphere through tropical upwelling as part of the larger Brewer-Dobson circulation (Holton et al., 1995, Mote et al., 1996). By calculating the anomalies of water vapor within 15° latitude of the equator (Figure 3), the year-to-year variability is best captured in the data assimilation products (Figure 3a and 3b) whereas the Control (Figure 3c) indicates an increase over time. When comparing M2-SCREAM and SAGE DAS, the tape recorder signal in the SAGE DAS is not as smooth as M2-SCREAM, and this is attributed to the less frequent observations in the tropics. Also of note is how well SAGE DAS captured the Hunga eruption. While the dry anomaly between 30 to 20 hPa in early 2022 does stop before reaching higher altitudes that year, it is about 2 months after the eruption before SAGE DAS captures the Hunga signal. The response is slow due to both infrequent observations in the tropics at that time of year, but it is further hindered by the volcanic aerosols such that the SAGE observations of high stratospheric water vapor have large instrument uncertainty associated with them. By design, the data assimilation system assigns lower weightings to these observations, and until the aerosols settle down out of the stratosphere, the Hunga eruption signal in the SAGE DAS is reduced.

Figure 3: Stratospheric water vapor anomaly profiles in M-2 SCREAM (a), SAGE DAS (b), and Control (c). These profiles show the effectiveness of both products (3a and 3b) in assimilating water vapor, as well as shortcomings in handling the Hunga signal, appearing as the darker blue areas in the upper corner of these two sub-plots. SAGE DAS specifically handles inter-annual variability well, despite having a more delayed Hunga signal than M-2 SCREAM.

While the community awaits follow-on missions to MLS, the assimilation of SAGE III/ISS stratospheric water vapor profiles should be considered in reanalysis products to bridge the gap, with the caveat that for anomalous events and plumes, like the 2022 Hunga eruption where there were insufficient SAGE profiles (Knowland et al., 2025) and where there are routinely few to no observations (e.g., seasonally in the tropics and the polar regions, respectively), the chemistry-only would be driving the model fields. Shortcomings in handling anomalous events is an expected consequence of having limited geographical coverage with sensors like SAGE III/ISS, but the effectiveness of SAGE DAS in constraining stratospheric water vapor on longer time scales such as inter-annual variability is quite proficient. 

References:

Knowland, K. E., Wales, P. A., Wargan, K., Weir, B., Pawson, S., Damadeo, R., & Flittner, D. (2025). Stratospheric water vapor beyond NASA's Aura MLS: Assimilating SAGE III/ISS profiles for a continued climate record. Geophysical Research Letters, 52, e2024GL112610. https://doi.org/10.1029/2024GL112610

Salawitch, R. J., Smith, J. B., Selkirk, H., Wargan, K., Chipperfield, M. P., Hossaini, R., Levelt, P. F., Livesey, N. J., McBride, L. A., Millán, L. F., Moyer, E., Santee, M. L., Schoeberl, M. R., Solomon, S., Stone, K., & Worden, H. M. (2025). The Imminent Data Desert: The Future of Stratospheric Monitoring in a Rapidly Changing World. Bulletin of the American Meteorological Society, 106(3), E540-E563. https://doi.org/10.1175/BAMS-D-23-0281.1

Wargan, K., B. Weir, G. L. Manney, S. E. Cohn, K. E. Knowland, P. A. Wales, and N. J. Livesey, 2023: M2-SCREAM: A stratospheric composition reanalysis of Aura MLS data with MERRA-2 Transport. Earth Space Sci., 10, e2022EA002632, https://doi.org/10.1029/2022EA002632