Stratospheric Water Vapor Assessment

The document identifies the transport of water vapor from the troposphere into the stratosphere as a major source of stratospheric and lower mesospheric water vapor. Several pathways facilitate this transport. The primary pathway involves a slow ascent through the cold tropical tropopause layer, often accompanied by significant horizontal movements. The cold-point temperature along these air parcel trajectories plays a crucial role in determining the amount of water vapor that enters the stratosphere. Another pathway is the convective lofting of ice particles, which upon reaching the stratosphere, evaporate and increase the water vapor content. A third pathway involves transport along isentropic surfaces that extend across both the troposphere and stratosphere. Volcanic eruptions can occasionally inject substantial amounts of water vapor directly into the stratosphere. Studies by Holton et al. (1995), Moyer et al. (1996), Fueglistaler et al. (2009), and Sioris et al. (2016) provide further insights into these transport mechanisms. On an annual average, the stratospheric entry mixing ratios resulting from these processes typically range from 3.5 to 4.0 ppmv, according to Kley et al. (2000).

The second major source of stratospheric and lower mesospheric water vapor is the in situ oxidation of methane. This process's contribution to the overall water vapor budget grows with increasing altitude, reaching its peak in the upper stratosphere. The significance of methane oxidation in the water vapor budget is well-established, and its impact is a function of altitude. Research by Le Texier et al. (1988) and others has contributed to the understanding of this process. The volume mixing ratio of water vapor generally increases with altitude in the stratosphere primarily due to methane oxidation, often culminating in a maximum around the stratopause. Above the stratopause, the volume mixing ratio typically decreases as this major source diminishes. The interplay between these two major sources – tropospheric transport and methane oxidation – shapes the overall distribution of water vapor in the stratosphere and lower mesosphere.

While the focus is on sources, the document briefly addresses the sinks of stratospheric water vapor. The primary sink is the reaction of water vapor with O(1D). As altitude increases, photodissociation becomes a more significant sink, eventually dominating in the mesosphere. Dehydration, the removal of water due to the sedimentation of polar stratospheric cloud (PSC) particles in the polar vortices, is another sink process. However, the importance of dehydration is geographically and temporally limited, as highlighted by Kelly et al. (1989) and Fahey et al. (1990). Excluding dehydration, the volume mixing ratio of water vapor generally shows an increase with altitude in the stratosphere because of methane oxidation, usually peaking near the stratopause before decreasing at higher altitudes due to a lack of significant sources.

This section introduces the second SPARC water vapor assessment (WAVAS-II) and describes the selection of satellite data sets used in the analysis. The study focuses on satellite observations of stratospheric and lower mesospheric water vapor obtained between 2000 and 2014. A total of 33 data sets from 15 individual satellite instruments are included. The selection prioritizes observations from the new millennium, building on the previous WAVAS report in 2000 (Kley et al., 2000). Data from instruments like HALOE, POAM III, and SAGE II, while available from before 2000, were excluded. Similarly, the SABER data, though spanning most of the analyzed period, lacked suitable data sets for inclusion. SAGE III data from the ISS, commencing only in 2017, were also excluded. The WAVAS-II data set overview paper by Walker and Stiller (2019) provides a comprehensive description of each individual data set. Figure 1 visually depicts the temporal coverage of the included data sets, indicating periods of coincident observations between different instruments, which is vital for the comparative analysis conducted.

This section details the individual satellite instruments and their respective data contributions to the WAVAS-II analysis. It lists a range of instruments and their operational periods. For instance, LIMS observations ended in May 1979, while SAMS provided data from 1979 to 1981. SAGE II operated for nearly 21 years (October 1984 – August 2005), whereas ATMOS observations were limited to a brief period in 1985. The Odin, TIMED, and Meteor-3M satellites, launched in 2001, carried SMR, SABER, and SAGE III instruments, respectively. While SMR and SABER continue to collect data, SAGE III stratospheric observations concluded in December 2005, mirroring the end of POAM III data collection. The Envisat satellite (launched March 2002) carried GOMOS, among other instruments contributing to water vapor observations. Other key instruments mentioned include the ILAS (with CLAES, HALOE, ISAMS, and MLS), ADEOS-II with ILAS II, SCISAT-1 with ACE-FTS and MAESTRO, and the Aura satellite with MLS and HIRDLS. Finally, SMILES data from the ISS, from 2009 to 2010, and a newer SAGE III instrument aboard the ISS from 2017 onward, are mentioned. The varied operational lifetimes and data coverage across these instruments are explicitly highlighted, impacting the length and reliability of the comparative analysis.

This section discusses the characteristics of the individual data sets and the preprocessing steps undertaken to ensure compatibility and consistency in the analysis. A key consideration is the differing vertical resolutions of the data sets, particularly relevant in the hygropause region and the lower mesosphere. The data sets are categorized into classes based on their vertical resolution around the hygropause to handle these differences systematically. Data sets with limited hygropause coverage are also identified. Further, the different retrieval methods and spaces used by the instruments are acknowledged. For instance, many data sets use volume mixing ratio (VMR), but some, like certain SCIAMACHY sets, employ number density. The generation of convolution data, using averaging kernels and a priori data, is described for different data sets. The sources of these convolution data, averaging kernel data and the additional temperature and pressure information are detailed. Walker and Stiller (2019) provide a detailed description of the vertical resolution estimation and other data set characteristics. Data preprocessing steps involved screening to remove outliers and ensuring consistency in the natural domains and vertical resolutions used across various instruments and data sets before comparison.

The core methodology for analyzing biases in stratospheric and lower mesospheric water vapor data involved profile-to-profile comparisons of more than 30 satellite data sets. This approach is explicitly chosen to reduce sampling errors compared to using binned data sets (zonal or monthly means). The large number of MIPAS data sets (13 out of 33 total) presented a unique challenge. Due to the relative similarity of these MIPAS data sets despite differences in measurement modes and processing, the authors chose to aggregate the comparison results from MIPAS data sets by default. The effects of this aggregation are investigated and shown throughout the analysis. This aggregation approach is contrasted with results that exclude the aggregation of MIPAS data, enabling the researchers to understand the influence of this numerous and similar data on the overall bias assessment. The analysis aims to provide a contemporary overview of uncertainties within the observational database, a key goal of the WAVAS-II program.

The analysis reveals significant biases across the different data sets. The observational database shows the largest biases below 70 hPa, in both absolute and relative terms. In contrast, the smallest biases are consistently observed between 70 and 5 hPa, particularly at low and mid-latitudes. Above this range, biases tend to increase, showing considerable variability with altitude and latitude. Histograms of biases from all altitudes show the largest occurrences for biases up to 0.2 ppmv or 5%. This pattern, however, is not consistent across all altitude ranges, with notable exceptions in the 10-1 hPa range. The 50th percentile (median) of the biases across comparisons are used extensively to characterize the typical bias levels and variability. The use of median, rather than mean, is explicitly justified by its robustness in the presence of outliers.

The study highlights data-set specific bias issues. The analysis addresses the impact of the numerous MIPAS data sets on the overall results. For example, biases in the SCIAMACHY solar OEM data set are typically within ±1 ppmv or ±20% in relative terms. Both positive and negative biases are observed in different altitude ranges. Systematic negative biases are seen in comparisons with specific MIPAS-Bologna data sets (V5H and V5R NOM) in the uppermost altitude range. The impact of the MIPAS data aggregation is shown with summary profiles. The profiles with and without MIPAS aggregation are compared, demonstrating that including the large number of MIPAS datasets can significantly alter the summary bias results, particularly between 30 and 0.6 hPa. The comparison between the median of all comparisons and the median with MIPAS aggregation illustrates this influence. The study emphasizes that the biases vary across different altitude levels, revealing data-set-specific characteristics and inconsistencies that require cautious interpretation.

The drift analysis in this study focuses on the temporal trends in stratospheric and lower mesospheric water vapor measurements. To perform a reliable drift analysis, a minimum overlap period of 36 months between data sets was required. This ensures sufficient data for trend estimation. The analysis utilizes both profile-to-profile comparisons and comparisons of zonal mean time series. The results from these two approaches are compared to assess their consistency and potential differences. The study explicitly notes that the profile-to-profile method reduces sampling errors relative to the zonal mean time series method. The large number of MIPAS data sets again influenced the analysis, leading to the aggregation of results for these data sets in many of the summary presentations. The impact of this aggregation, similar to the bias analysis, is analyzed by showing both aggregated and unaggregated results.

The analysis of drifts reveals a wide range of values that are statistically significant at the 2σ uncertainty level. The smallest drifts are consistently found in the altitude range between approximately 30 and 10 hPa. Histograms from all altitudes indicate that the most frequent drifts lie between 0.05 and 0.3 ppmv decade⁻¹. There's a notable difference between results considering all altitudes versus a focus on specific altitude ranges. The study also shows that the overall drift sizes vary by latitude. The influence of the overlap period on the drift size is also considered, and a relationship is shown between the overlap period and drift size for periods up to 70 months above 100 hPa. However, beyond 70 months, no clear correlation is observed. The use of median values and the aggregation of MIPAS results also influences the final results, and differences are highlighted in the text.

This section delves into data-set specific drift results, focusing on MIPAS-ESA V7R and MLS data sets as examples. Other results are presented in supplementary materials. The analysis highlights the differences between the V5 and V7 calibrations of MIPAS data; with the V7 calibration showing a reduction in significant drifts compared to its predecessor. However, the V7 calibration exhibits mostly negative drifts, potentially indicating overcompensation. The SMR 489 GHz data set is presented as an example, showing mostly positive drifts around 50 hPa and 0.5 hPa, with many being statistically significant. The document notes that the SOFIE data set shows more pronounced biases in the upper stratosphere and lower mesosphere. The analysis compares drift estimates obtained using the profile-to-profile method with those from zonal mean time series comparisons (Khosrawi et al., 2018). The two methods show similar patterns, but differences, while not very frequent (2.6% - 6% of comparisons showing statistically significant differences at the 2σ level), are noted, particularly at 0.1 hPa. This comparison is presented as a matrix, visually showing similarities and discrepancies between the two drift estimation approaches.

The overall assessment emphasizes the usefulness of many of the analyzed data sets for scientific research, either independently or in conjunction with modeling results. However, a crucial caveat is introduced: data sets exhibiting significant biases or drifts should be used with caution. For studies prioritizing the absolute amount of water vapor, data sets with identified bias issues require careful interpretation and consideration of the potential error introduced. Similarly, data sets with significant drifts should be treated cautiously if studying variability beyond the 36-month overlap period used in the drift analysis. This careful consideration of data quality is essential for ensuring reliable conclusions in scientific studies.

By integrating the bias and drift characteristics identified throughout the study, the researchers pinpoint the altitude range between 50 and 5 hPa as the most suitable for scientific analyses of stratospheric and lower mesospheric water vapor. This altitude range exhibits the fewest issues related to biases and drifts in the observational database, making it the most reliable region for drawing conclusions about water vapor concentrations. The identification of this optimal altitude range is a significant finding, guiding future research on stratospheric and lower mesospheric water vapor and improving the confidence in conclusions drawn from such analyses.

The study compares drift estimates obtained through two distinct methods: profile-to-profile comparisons and comparisons of monthly zonal mean time series. The results reveal a high degree of interchangeability between these methods. Although occasional differences exist, statistically significant differences (at the 2σ uncertainty level) are found in only a small percentage of comparisons (2.6% to 6%, depending on the specific altitude and latitude). This consistency across methods strengthens the robustness of the findings, suggesting that either approach yields reliable drift estimates for stratospheric and lower mesospheric water vapor analysis. The study concludes that there's no compelling reason to favor one method over the other for most analyses.