A framework to compare GNSS-RO and reanalysis equatorial wave spectra of Cold-point tropopause temperature

The tropical Cold-point tropopause temperature (CPT) controls the amount of water vapor that enters the stratosphere, as air masses that cross through the equatorial tropopause are subjected to freeze-drying.

GNSS radio-occultation (GNSS-RO) provide temperature profile measurements with high vertical resolution and global coverage, enabling the monitoring of the CPT evolution outside of the few tropical regions covered by radiosondes. Reanalyses are all known to have a modeled CPT that is on average too warm, compared to GNSS-RO measurements.

The reanalysis CPT warm bias maximizes near the Equator, hinting at a possible role of equatorial waves. However, to date the reanalysis CPT biases have only been studied from a zonal-mean and long-term perspective, without looking at the effects of equatorial waves.

Observed equatorial CPT shows peaks in the wavenumber-frequency spectrum coinciding with equatorial wave’s theoretical dispersion curves. This means that equatorial waves that propagate through the equatorial tropopause are modulating CPT variability. The observational study of CPT wave spectrum by Kim and Son (2012) used the COSMIC RO mission, and had a relatively limited space-time resolution: 10°N-10°S meridional average and 3-day running mean, i.e. only showing the symmetric part of the spectrum. Meanwhile, efforts to compare reanalyses’ wave spectra used data at the standard 100 hPa level close to the tropical tropopause, with no observational dataset as reference.

In our study, we showcase a framework to inter-compare CPT wavenumber-frequency spectra from various reanalyses to that of observed CPT from GNSS-RO. We combine multiple GNSS-RO mission data and grid them on 5° x 5° longitude-latitude daily resolution for the years 2007-2018. Model-level CPT from ERA5, ERA-Interim, JRA55 and MERRA-2 reanalyses, are interpolated/averaged onto the same 5° x 5° daily grid from GNSS-RO, enabling a 1-to-1 comparison on the same space-time grid, at the cold-point. Our goals using this dataset are: a comparison between purely observational and reanalyses’ CPT spectra that is as fair as possible, with a better resolution and longer time-period than previous studies, and the separation of the symmetric and anti-symmetric parts of the spectra. This provides valuable information about what types of CPT variability are most troubling to reproduce by the reanalyses.

Observational CPT wavenumber-frequency spectra of power above background from GNSS-RO show well-defined spectral peaks near the MJO domain and the theoretical dispersion curves of Kelvin and equatorial Rossby waves in the symmetric spectrum, as well as mixed Rossby-gravity waves in the anti-symmetric part.

We show the importance of sampling reanalysis data at the observation locations only, as even at synoptic-scales and frequencies of around a week, this can influence spectral power. Reanalyses increasingly struggle at shorter and faster space-and-time-scales, more markedly in the anti-symmetric part of the spectrum.