Cold-point tropopause temperature monitoring from radio occultation

The cold-point tropopause temperature is the narrowest bottleneck in Earth's atmosphere: it sets how much water vapour reaches the stratosphere, shaping ozone chemistry and radiative forcing. GNSS radio occultation profiles it with roughly 200 m vertical resolution, all-weather, globally.

Sensors

COSMIC-2 / FORMOSAT-7: Six LEO satellites launched 2019, each carrying a TriG receiver. Provides roughly 5,000 globally distributed occultation soundings per day, concentrated in the tropics between 45°S and 45°N. Vertical resolution near the tropopause approximately 100–200 m. Bending-angle profiles retrievable from the surface to ~60 km.
Spire Global GNSS-RO constellation: Commercial CubeSat constellation of more than 100 satellites as of 2024, delivering upward of 10,000 occultations per day globally. Near-real-time latency under three hours supports operational assimilation. Vertical resolution and accuracy comparable to COSMIC-2 in the upper troposphere and lower stratosphere.
Metop GRAS (GNSS Receiver for Atmospheric Sounding): Flown on Metop-A (2006), -B (2012) and -C (2018). Provides a continuous climate-quality record now exceeding 15 years. EUMETSAT processes and archives Level-2 temperature profiles used for trend studies. Polar orbit gives good high-latitude coverage but fewer tropical soundings per day than COSMIC-2.
PlanetiQ GNSS-RO: Commercial constellation designed for high signal-to-noise bending-angle profiles, with receivers tracking GPS, GLONASS, Galileo and BeiDou simultaneously. Increased signal diversity improves sampling density and reduces the longitude gaps that complicate CPT trend isolation.

What the cold-point tropopause actually controls

The cold-point tropopause (CPT) sits at roughly 16–18 km altitude in the tropics, where temperature reaches its local minimum, typically between 185 K and 200 K. Air ascending through deep convection is freeze-dried at this level: water vapour condenses and falls back as ice, so the mixing ratio entering the stratosphere is set almost entirely by the CPT temperature. A one-kelvin change in CPT temperature translates to a roughly 10–15 per cent change in stratospheric water vapour concentration, according to the Clausius-Clapeyron relation at those temperatures.

That matters because stratospheric water vapour is itself a greenhouse gas and a feedstock for the hydroxyl chemistry that governs ozone destruction rates. The 2010–2011 drop in stratospheric water vapour, documented by Dessler and colleagues using satellite observations, was linked to a temporary CPT cooling and measurably slowed surface warming for about a decade. Monitoring CPT temperature is therefore not a niche academic exercise. It is a direct constraint on stratospheric chemistry models and a sensitive, independent check on tropical climate variability.

Why radio occultation resolves what other sensors miss

The CPT is a sharp feature. Temperature can drop by 10–15 K over just 2–3 km of altitude, and the minimum is often confined to a layer less than 500 m thick. Nadir-viewing infrared sounders such as AIRS or IASI have vertical resolution of 1–3 km in the upper troposphere, which is enough to detect the tropopause layer but not enough to resolve the precise minimum temperature without significant smoothing bias.

GNSS radio occultation works differently. A GPS or Galileo signal from a transmitting satellite bends as it passes through the atmosphere on its way to a receiver in LEO. The bending angle at each tangent height is measured with sub-millimetre precision in the signal phase. Because the measurement is a limb geometry, vertical resolution is set by the Fresnel zone width and the signal bandwidth, reaching approximately 100–200 m in the upper troposphere when full-spectrum inversion methods such as full-spectrum inversion (FSI) or canonical transform are applied. The technique is self-calibrating against the known GNSS frequency, so it requires no external radiance calibration and drifts very little over time. That stability is exactly what climate benchmarking demands.

The honest limit: occultation profiles are point measurements, not images. A single sounding covers a horizontal footprint of roughly 200–300 km along the ray path, so sub-mesoscale CPT structure is smeared. Cloud does not block the signal, which is a genuine advantage over infrared methods in the convectively active tropics.

Seasonal cycles, ENSO and the sampling problem

CPT temperature in the deep tropics follows a clear annual cycle, coldest around January when the ascending branch of the Brewer-Dobson circulation is strongest, warmest around July. The peak-to-trough amplitude is roughly 4–6 K in the zonal mean. Superimposed on this is a strong ENSO signal: El Niño conditions warm the tropical tropopause by shifting convective heating and altering the large-scale upwelling rate, while La Niña tends to cool it.

Isolating a long-term trend from these interannual swings requires multi-year records and careful treatment of sampling. Occultation soundings are not uniformly distributed in longitude. COSMIC-2 concentrates between 45°S and 45°N but the six-satellite geometry produces non-random longitude gaps on any given day. Averaging over 30-day windows reduces the sampling noise substantially, but monthly anomaly maps at finer than 10° longitude resolution should be treated with caution unless the daily sounding count in that cell exceeds roughly 20–30. Spire's larger constellation reduces this problem, though it introduces the separate question of inter-constellation bias correction when combining records for trend analysis.

Building a climate benchmark: what the record currently shows

The longest continuous GNSS-RO record with climate-quality processing runs from the CHAMP mission (2001) through GRACE, COSMIC-1 (2006–2020) and the Metop GRAS series. UCAR's COSMIC Data Analysis and Archive Center (CDAAC) and EUMETSAT both maintain independently processed archives. Trend estimates from these records suggest a slight tropical CPT warming over the 2000s, consistent with tropical tropospheric warming, though the magnitude and significance depend heavily on which inter-mission bias corrections are applied.

The key metric for a climate benchmark is that the measurement uncertainty in a monthly zonal-mean CPT temperature should be below about 0.3 K to detect decadal trends of 0.1–0.2 K per decade. Published assessments of COSMIC and Metop GRAS suggest random errors in individual soundings near the tropopause of roughly 0.5–1.0 K, reducing to below 0.2 K in 30-day zonal averages over the tropics when sounding counts are adequate. That is close to the threshold needed, which is why expanding the constellation size matters for trend science, not just for weather.

From raw profiles to an operational CPT dataset

Processing a raw bending-angle profile into a CPT temperature involves several steps: Abel inversion to refractivity, conversion to temperature and pressure assuming dry air (or using background moisture in the lower troposphere), and then identification of the CPT as the minimum temperature in the 14–20 km layer. Each step introduces choices that affect the result. The dry-air assumption introduces a warm bias below about 10 km where moisture is significant, but near the tropopause the atmosphere is cold and dry enough that the bias is negligible, typically below 0.1 K.

Satellize ingests Level-2 bending-angle and refractivity profiles from CDAAC, EUMETSAT and Spire's commercial feed, applies a consistent CPT-detection algorithm, and delivers gridded monthly anomaly fields and trend diagnostics. For clients building stratospheric chemistry or climate model inputs, the deliverable is a versioned NetCDF dataset with per-cell sounding counts and uncertainty estimates included.

Honest limits and what to watch for

Radio occultation is not perfect for CPT work. The horizontal averaging along the ray path means that sharp longitudinal gradients in CPT temperature, such as those near the edges of the warm pool or over the Andes, are smeared. In regions of strong horizontal temperature gradients, the retrieved CPT may not correspond to any single atmospheric column.

Inter-constellation biases are a real concern when splicing records. Differences in receiver hardware, signal tracking loops and processing software can introduce systematic offsets of 0.1–0.3 K in the upper troposphere. These are small in absolute terms but consequential for trend detection. Any dataset claiming a CPT trend should document which missions were combined and what bias corrections were applied. Finally, the CPT is not the only control on stratospheric water vapour entry. Convective ice lofting above the CPT, documented in observations from the StratoClim campaign, can inject water directly into the stratosphere independent of the CPT temperature. A CPT dataset alone does not capture that pathway.

Typical figures

Vertical resolution near CPT	~100–200 m (FSI or canonical transform inversion); ~1–3 km for nadir IR sounders
Horizontal footprint per sounding	200–300 km along ray path; point measurement in cross-track
Daily sounding count (tropics)	~5,000 (COSMIC-2); >10,000 (Spire full constellation)
Temperature accuracy (individual sounding, CPT layer)	~0.5–1.0 K random error; systematic bias <0.1 K in dry upper troposphere
Monthly zonal-mean uncertainty (30-day average, tropics)	<0.2 K where sounding count >200 per 10° latitude band
Latency (near-real-time commercial feeds)	<3 hours (Spire); ~6 hours (CDAAC near-real-time); ~24 hours (EUMETSAT operational)
All-weather capability	Yes; signal unaffected by cloud or precipitation
Archive depth	CHAMP from 2001; COSMIC-1 from 2006; Metop GRAS from 2007; COSMIC-2 from 2019
CPT altitude range monitored	14–20 km search window in standard processing
Delivery formats	NetCDF (CDAAC, EUMETSAT standard); CSV anomaly tables; GeoTIFF gridded fields on request

Analytics Satellize can run

Monthly gridded CPT temperature anomaly field	Abel inversion of bending-angle profiles; CPT detected as temperature minimum in 14–20 km layer; gridded at 5° × 5° with per-cell sounding counts	Versioned NetCDF file with uncertainty layer, updated monthly
ENSO-composited CPT anomaly maps	Epoch compositing keyed to MEI or ONI ENSO index; statistical separation of annual cycle, ENSO signal and residual trend	PDF report with anomaly maps and time-series plots; underlying data in CSV
Stratospheric water vapour entry flux estimate	Clausius-Clapeyron conversion from CPT temperature to saturation mixing ratio; zonal-mean entry value with propagated uncertainty	Monthly scalar time series in CSV; suitable for direct input to chemistry-transport models
CPT trend detection and significance assessment	Linear regression on deseasonalised monthly anomalies; Mann-Kendall test for monotonic trend; bootstrap confidence intervals accounting for ENSO autocorrelation	Trend report with stated confidence level and sensitivity to inter-mission bias assumptions
Inter-constellation bias audit	Collocation of simultaneous COSMIC-2 and Spire soundings within 50 km and 30 minutes; bias statistics in CPT temperature and altitude	Bias characterisation table used to document dataset provenance for climate-record submissions
Longitudinal CPT structure during convective events	Daily sounding aggregation at 10° longitude resolution; overlay with NOAA OLR or GPM precipitation to contextualise CPT depressions near deep convection	Event GIS layer (GeoJSON) and accompanying briefing note

Sources

Who does the work

We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.