RO Climate Data Record

Gridded Climate Data Records from CDAAC Merged Radio Occultation Data

-William Randel and Jon Starr
 

The COSMIC team has developed monthly gridded climate data record (CDR) data sets based on merged CDAAC-processed RO data beginning in 2002. This document provides details on the data processing and gridding, and includes example results derived from these data. Links to the gridded data are provided here:
•    Temperature (T) and tropopause statistics
•    Bending Angle (BA)
•    Refractivity (N)

1. Data contributing to the CDAAC CDR
The available data from CDAAC processing comes from many different satellites over time, as shown in Fig. 1. The primary contributors to the data record include CHAMP (for the pre-2006 component), the series of METOP satellites (A,B,C), COSMIC-1, COSMIC-2 and SPIRE. 
 

Figure 1: Monthly Radio Occultation data count from CDAAC-processed data. 

The latitudinal sampling is different among the different satellites/constellations, as shown in Fig. 2. Most of the satellites/constellations provide globally distributed measurements, although COSMIC-2 (with low inclination satellite orbits) samples mainly in the low latitudes. As discussed further below, the COSMIC-2 data are only used over latitudes 30° N-S.

Figure 2: Latitude sampling for several of the primary RO data sources. The counts refer to the number of occultations within each 5-degree latitude bin for the zonal average monthly data. Results are shown for selected individual months for each primary RO data source.

The data used here are based on CDAAC retrievals for all missions, specifically the so-called ‘dry temperature’ retrievals (atmPrf files). These retrievals are described in Reference. Gridded data include Bending Angle (BA), Refractivity (N) and Temperature (T). We use the standard quality controls included in the atmPrf files and only include data with ‘good’ quality flags.

2. Gridding details

a)    Temperature, Bending Angle and Refractivity
The CDAAC CDR for temperature is provided as a monthly mean data set on a 5° x 10° latitude-longitude grid, with a 200 m vertical resolution (sub-sampled from the original, higher vertical resolution atmPrf files). Time series begin in March 2002. Dry temperature (T) is provided over altitudes 10-40 km. For Bending Angle (BA) and Refractivity (N) fields, only zonal average data are archived. BA is archived for 0-80 km and Refractivity (N) fields are provided over 0-60 km, with vertical sampling at 200 m.  Separate data files are provided for T, BA and N, with data links provided above. The data are gridded in a simple manner by combining all available measurements during each month within individual grid boxes. We do not include sampling error bias corrections (derived from meteorological analyses) to these data, such as included in ROM-SAF (Gleisner et al, 2020) or NOAA STAR (Zhou et al, 2025) data sets; results in those papers show that such sampling corrections are generally small (<0.2 K) for monthly, zonal average statistics outside of polar regions. 

b)    Tropopause statistics 
In addition to temperatures, we provide gridded values of a few metrics related to the tropopause, including altitude and temperature of the global lapse-rate tropopause (LR) and potential temperature gradient tropopause (PTGT). Details of these calculations are described in Tinney et al (2023). We also include the cold point (temperature minimum) tropopause altitude and temperature over latitudes 30° N-S. All tropopause fields are derived from gridding the statistics derived from the individual temperature profiles within each grid box.

3. Differences among RO missions, and sub-selecting data

As a preliminary step before combining data from different satellites/constellations, we evaluated systematic differences (biases) among the different RO measurements. Such differences can arise from different satellite orbits, diurnal sampling, receiver characteristics or other causes. As a reference standard we use the series of METOP satellites (A,B,C) which provide a continuous record over 2008-2025. Results here focus on time average zonal mean temperature differences, calculated using available records of overlap between the separate data sets (with each of the data sets gridded separately, as described above, prior to comparison). Comparisons show that biases among the separate METOP (A,B,C) instruments are small (Fig. 3), with time average zonal mean temperature differences < 0.1 K over most of the globe.
 

Figure 3. Time average zonal mean temperature differences between METOP-A and METOP-B (left) and between METOP-B and METOP-C (right), calculated during the period of overlap between the satellites as noted. Differences are less than 0.1 K almost everywhere.  

Time average differences with METOP data for the other satellites/constellations show somewhat larger values, and each satellite/constellation exhibits distinct spatial patterns of biases (Fig. 4). These comparisons show that mean biases are generally small (< +/- 0.2 K), although there are a few regions of larger biases for individual constellations that are omitted from the merged CDR. Specifically, COSMIC-2 data exhibit relatively larger temperature biases compared to METOP (up to +/- 0.5 K) over latitudes 30-45° N and S, and SPIRE has large positive biases compared to METOP over the Antarctic (60-90° S). The specific causes of these larger biases are still a topic of research, but for the present CDR we choose to simply to omit these latitude regions of COSMIC-2 (30-45° N and S) and SPIRE (60-90° S) data from the merged data. We note that relatively little data is involved in either case (see Fig. 2).  

Figure 4. Time average zonal mean temperature differences for different satellites/constellations compared to the merged METOP (A,B,C) data record. Differences are calculated during the period of overlap between the satellite constellations as noted at the top of each panel. The CHAMP vs. METOP overlap is relatively short (7 months, March – September 2008) and hence the patterns are somewhat noisier. Contour interval is 0.1 K. Because of relatively large temperature biases, the merged CDR omits COSMIC-2 data over latitudes 30- 45° N-S and SPIRE data over 60-90° S. 

References

Gleisner, H., Lauritsen, K. B., Nielsen, J. K., & Syndergaard, S. (2020). Evaluation of the 15‐year ROM SAF monthly mean GPS radio occultation climate data record. Atmospheric Measurement Techniques, 13(6), 3081–3098. https://doi.org/10.5194/amt‐13‐3081‐2020

Tinney, E. N., C. R. Homeyer, L. Elizalde, D. F. Hurst, A. M. Thompson, R. M. Stauffer, H. Vömel, and H. B. Selkirk (2022). A Modern Approach to a Stability-Based Definition of the Tropopause. Mon. Wea. Rev., 150, 3151–3174, https://doi.org/10.1175/MWR-D-22-0174.1

Zhou, J., Ho, S.‐P., Zhou, X., Shao, X., Gu, G., Chen, Y., et al. (2025). Construction of temperature climate data records in the upper troposphere and lower stratosphere using multiple RO missions from September 2006 to July 2023 at NESDIS/STAR. Journal of Geophysical Research:Atmospheres, 130, e2024JD041295. https://doi.org/10.1029/2024JD041295

Example results 
We include a few example results derived from the CDAAC temperature CDR. This is a placeholder for now and this section will be fleshed out a little later.
 

Figure 5: Height vs. time section of deseasonalized temperature anomalies over 10° N-S. The black line denotes the cold point tropopause.