Ozone viewed in space with Earth below. Credit: studio023

Tropospheric warming and stratospheric cooling in the 21st century


Ozone viewed in space with Earth below. Credit: studio023
Ozone viewed in space with Earth below. Credit: studio023

Radio occultation data show that in the first two decades of the 21st century, the Earth’s lowest atmospheric layer – the troposphere – warmed significantly, by up to 0.5 K per decade, while the layer above that – the stratosphere - cooled by about the same amount.

Last Updated

03 May 2023

Published on

30 January 2023

By Hans Gleisner, Axel Von Engeln, Rob Roebeling

Over the last century, the amount of greenhouse gases has accumulated in the Earth’s atmosphere at a gradually accelerating rate. It is expected that the warming caused by greenhouse gases impacts the structure of Earth's atmosphere (Meng et al., 2021). This is precisely what we observe. During the past decades the atmospheric layer closest to the Earth surface, the troposphere, has warmed, while the layer above it, the stratosphere, has cooled (see Figure 1), and related to these changes, the tropopause height has risen globally (IPCC, 2021).

Graphical representation of Earth's atmospheric layers for industrial times.
Figure 1: Graphical representation of Earth's atmospheric layers for industrial times. The increased concentrations of greenhouse gases during industrial times cause 1) the troposphere to warm,­ 2)  the stratosphere to cool, 3) the height of the tropopause to increase.

Quantifying the rate of warming and cooling in the different layers of the free atmosphere is a difficult task, as evidenced by the low confidence the Intergovernmental Panel on Climate Change (IPCC) assigned, in its fifth assessment report, to the temperature changes in the free atmo­sphere (IPCC, 2013). However, in its sixth assessment report IPCC saw an improved confidence assigned to these temperature changes, in particular for the period after 2001 (IPCC, 2021). This is largely due to the addition of a new satellite-based data record, i.e. the Global Navigation Satellite System (GNSS) Radio Occultation (RO) data.

Since around the 1850s, the temperatures in the lowest layers of the troposphere have been measured with ground-based thermo­meters. However, these time series are quite uncertain in the earliest decades, and only provide measurements close to the surface. Apart from sparse balloon-borne measurements gathered since the 1940s, getting a detailed view of upper-atmosphere temperatures had to wait for the start of the satellite era.

Since the end of the 1970s, satellite-borne instruments, flown in Earth's orbit, have been providing regular observations of upper-atmosphere temperatures. Currently, several satellite-based techniques exist to measure these temperatures. Most techniques are based on radiance measurements of the Earth’s atmosphere at infrared or microwave wavelengths, such as those derived from measurements from the IASI and AMSU-A instruments onboard Metop satellites.

At beginning of this millennium, a relatively new technique, called GNSS Radio Occultation, was introduced. This technique derives atmospheric temperature profiles from bending of GNSS signals caused by atmospheric refraction. The unique long-term stability of these profiles makes them excellently suited for studying changes in tropospheric and stratospheric temperatures over long time period. Thus, they help to increase our confidence in the response of the atmosphere to increased greenhouse gas concentrations.

Observing the atmosphere with RO

The principle of the GNSS-RO technique is based on measuring shifts in phase and amplitude of radio waves as they propagate through the atmosphere from a GNSS satellite — flown at an altitude around 20,000km — to a receiver onboard a polar orbiting satellite — flown at altitudes between 300 and 1,300km (Figure 2). From the perspective of the receiver, a GNSS satellite is seen to set or rise at the Earth’s limb. Receiving a GNSS radio signal over a time-span of about two minutes is enough to make an atmospheric scan from near surface to well above the neutral atmosphere (~100km). During such a scan, the amount of refraction — or bending — of the ray can be computed from the phase shifts, and the refractive index of the air as function of height can be deter­mined. In the upper air, where humidity is low, the refractive index is proportional to air density. Hence, from a single occultation event, a near-vertical profile of temperature, pressure, and density from the middle troposphere to the upper stratosphere can be retrieved.

Measurement principle of the GNSS RO technique (satellite orbits not to scale).
Figure 2: Measurement principle of the GNSS-RO technique (satellite orbits not to scale). The refraction, or bending, of a radio wave propagating through the atmosphere is measured as a function of height during an occultation event, and near-vertical atmospheric profiles of density, pressure, and temperature can be retrieved at the ray’s tangent point. From a large number – millions – of such profiles obtained over many years, the evolution of Earth’s troposphere and stratosphere can be monitored (Gleisner et al., 2022)

As noted, different satellite-based techniques can provide profiles of atmospheric temperature and water vapour. These different techniques largely provide complementary information on our atmosphere. The RO technique measures the troposphere and stratosphere with a high vertical resolution and a global coverage. Unlike infrared measurements, the RO measurements are not affected by clouds or land-sea differences. Moreover, the RO measurements have a unique long-term stability — the advantage of measuring time differences rather than radiances — and, thus, require no inter-calibration between satellites or instruments. The latter is an important characteristic for studies of the climate where calibration issues can be a limiting factor. The weaknesses of the RO technique include a limited horizontal resolution of a few hundred kilometers, reduced accuracy in the lower troposphere, in particular in the tropics, and that the quasi-random distribution of the profiles makes provision of uniformly gridded data somewhat complicated.

Preparing RO data for climate studies

Today we have a continuous time-series of more than 20 years of RO measurements, that starts in 2001. These measurements were provided by operational, as well as research, missions. In 2006, the data numbers increased dramatically, first with the launch of the COSMIC Taiwan-United States six-satellite mission, and later, with the launch of EUMETSAT’s Metop satellites that carry the GRAS RO instrument (Figure 3). Currently, these, and other missions operating RO instruments, have  provided around 20 million atmospheric profiles that can be used to construct time series for climate applications.

Daily number of observed atmospheric profiles in version 1 of the ROM SAF CDR and Interim CDR
Figure 3: Daily number of observed atmospheric profiles in version 1 of the ROM SAF CDR and Interim CDR. Data from four RO satellite missions were included in the data records: CHAMP, GRACE, COSMIC, and Metop. The Interim CDR is presently updated regularly with a lag of about one month.

EUMETSAT's Radio Occultation Meteorology Satellite Application Facility (ROM SAF) provides Climate Data Records (CDRs) and Interim CDRs (ICDRs) of RO data from GPS receivers operated onboard the CHAMP, GRACE, COSMIC, and Metop missions. The data are provided as four single-mission data records, as well as one multi-mission data record (see data record details below in the section Data used). In the free atmosphere, above about 5–6km, these data records have an excellent stability in time that surpasses that of, eg current reanalysis data records (Gleisner et al., 2020).

One factor that needs to be addressed for any satellite-based measurement technique is sampling issues related to the orbits of the satellites. For example, the Metop satellites are placed in sun-synchronous orbits and always pass the equator at the same local time, while other satellites (eg CHAMP, GRACE, COSMIC) slowly drift in local time. As a consequence of the evolution of the observing system, where old satellite missions are gradually replaced by new missions, the local-time coverage may change over time with consequences for observations of, eg the diurnal cycle in the atmosphere. This is handled by a correction for the impacts of sampling errors based on an estimation of those errors. Such sampling-error corrected data are provided as a part the ROM SAF gridded monthly-mean data, and is the preferred choice for studies of climate variability and trends (Gleisner et al., 2020).

Variability and trends in 20 years of atmospheric temperatures

An accurate description of climate variability and trends requires time-series that i) are long enough to minimise the effect of variations associated with the weather, the seasons, and any other types of quasi-periodic variability (eg the El Niño-Southern Oscillation (ENSO)), and ii) are long enough to reach the required long-term stability of the measurements. With 20 years of RO data of sufficient accuracy and stability, we are now in a position where RO data contributes to our understanding of the climate of the Earth’s atmosphere (Ringer and Healy, 2008; Gleisner et al., 2022).

Figure 4 shows vertical distributions of monthly-mean temperature anomalies between 8km and 40km for the low-latitude region between 20°S and 20°N over the period 2002 to 2022. The title of the plot is dry temperature as an indication that the temperatures are retrieved under the assumption that the humidity is negligible. This is also why the lowest 8km is masked out. Sampling-error corrected data were selected from the data files, and we used the 10-year period 2007 to 2016 as reference for the anomaly calculations.

Monthly mean temperature anomalies based on RO data from ROM SAF CDR and ICDR ver. 1, covering the period January 2002 to October 2022.
Figure 4:  Monthly mean temperature anomalies based on RO data from ROM SAF CDR and ICDR ver. 1, covering the period January 2002 to October 2022. Long-term trends in the data must be assessed against a background of other types of variability at different timescales.

The figure clearly reveals a lot of structure in the temperature data. The height of the tropopause is seen as a relatively sharp boundary around 18km. Below about 18km, in the troposphere, the temperature anomalies exhibit variations associated with the ENSO and other modes of variability. ENSO is the most influential natural climate pattern on Earth, it refers to three to seven years repeating phases of warming (El Niño) and cooling (La Niña) sea surface temperatures across the tropical Pacific Ocean (NOAA, assessed 20 Jan 2023). Above about 18km, in the stratosphere, the temperature anomalies are dominated by the so called Quasi-Biennial Oscillation (QBO) — these are the regular oscillations in the equatorial stratosphere with an average period of ~28 months. Any long-term trend in the data must be detected against a background of variability at a range of different time scales, which is why the time series need to be sufficiently long.

Figure 5 shows the RO-based temperature trends over all latitude bands for the period 2002-2022. We have restricted the trends to the altitude region 8–30km, where the accuracy of the RO-based temperature data is optimal. The global structure of temperature changes stands out in great detail (Ladstädter et al., 2023). We find a wide-spread cooling in the lower stratosphere, and a transition to tropospheric warming across a relatively sharp vertical gradient in the tropopause region. The warming exhibits a north-south asymmetry that is also apparent in other observational data records, but that is absent in commonly-used climate model simulations (see Figure 7). However, we also find that with longer time series that asymmetry in the observational record is reduced.

Decadal temperature trends over the period 2002 to 2022.
Figure 5:  Decadal temperature trends over the period 2002 to 2022. The trends are computed from the gridded monthly mean RO data available in the ROM SAF CDR and ICDR version 1.

Role of RO data in IPCC assessments

IPCC’s sixth assessment report saw an improved confidence in the vertical structure of temperature changes, enabling a detailed analysis of upper-troposphere and lower-stratosphere trends (IPCC, 2021). This relied heavily on the very first contribution of RO data to be used in an IPCC assessment. The contribution was organised as a collaborative effort amongst several leading processing centres, with the ROM SAF as a key contributor. The trends derived from RO data, as well as infrared measurements and radiosonde data, helped quantify the changes taking place in the vertical structures of temperatures in the Earth’s atmosphere.

One set of findings of the RO contribution is summarised in Figure 6, which was included in Chapter 2 of the IPCC WG1 AR6 report (IPCC, 2021). It tells us that the observed warming rates are faster in the tropical upper troposphere than at, or near, the surface. Furthermore, the trends found in RO data agree reasonably well with the trends found in AIRS data and in radiosonde data, apart from a discrepancy between RO and radiosonde data around 15km. Steiner et al. (2020) suggests that this discrepancy is reduced to near zero if the radiosonde data is restricted to only high-quality instruments.

Global (left panel) and tropics (right panel) upper air temperature trends (2002 to 2019)
Figure 6: Global (left panel) and tropics (right panel) upper air temperature trends (2002 to 2019) in degrees C per decade from three RO-based data records, one infrared-based data record, and two radiosonde-based data records (Credit: IPCC, 2021).

Another set of findings is presented in the Technical Summary of the IPCC WG1 AR6 report. In this summary the RO observed trends are related to two different climate modelled projections (see Figure 7). The similarities between the observations and the climate model projections are pointed out in in Chapter 2 of the IPCC WG1 AR6 report (IPCC, 2021) as follows:

In the tropics, since at least 2001 (when new techniques permit more robust quantification), the upper troposphere has warmed faster than the near-surface (medium confidence). There is medium confidence that most CMIP5 and CMIP6 models overestimate the observed warming in the upper tropical troposphere .

The conclusions from these assessments clearly show the potential of the RO climate data records to contribute to an improved understanding of the atmospheric changes associated with global warming.

Zonal cross-section of projected long-term temperature trends under the emissions scenarios SSP1-2.6 and SSP3-7.0 (the two panels to the right) compared to the radio occultation observed temperature trends (panel to the left).
Figure 7:  Zonal cross-section of projected long-term temperature trends under the emissions scenarios SSP1-2.6 and SSP3-7.0 (the two panels to the right) compared to the RO observed temperature trends (panel to the left). The colour scales are chosen such that similar colours roughly indicate similar changes (Credit: IPCC, 2021).

Future climate observations with RO data

More than 20 years of RO data have already shown the relevance and added value of these data for climate studies. However, 20 years is still a limited period. Continued observations from RO instruments are needed to further increase our confidence in the observed trends. One of the first milestones to reach is in the beginning of the 2030ies. It will then be possible to provide RO-derived products for a 30-year time period, very close to the World Meteorological Organization (WMO) standard reference period 2001 to 2030

Currently, the ongoing retrieval of RO data is guaranteed through several missions until the mid-2040s. For example, through the Radio Occultation sounder (RO) on Metop Second Generation (Metop-SG), with planned operation until 2046, the GNSS-RO on Sentinel-6, with a planned operation till 2032, or the GNSS Radio Occultation Sounder (GNOS) and GNOS-2 on the Chinese Feng-Yun-3 satellites, with a planned operation until 2035.

Early RO instruments relied on the observation of GPS satellites only, limiting the number of observations, these future ones can provide up to four times more data by observing all available GNSS systems (GPS, Galileo, Beidou, Glonass and QZSS). Moreover, several commercial RO providers are already providing data for weather forecasting, these could also contribute to the long-term climate data record. Although it requires full access to low level data, as well as access to relevant instrument documentation — requirements not needed for weather forecasting use. In addition, several upcoming operational and research satellite missions also assess whether an RO instrument can be added to the platform, as most already have the core capability onboard (GNSS-based positioning) and would only require to add occultation antennas.


This study demonstrates how the advent of RO, at the beginning of this millennium, has helped to improve our understanding of the climate of the Earth’s atmosphere, has increased our confidence in recent changes in atmospheric temperatures, and has enabled a more detailed analysis of upper-troposphere and lower-stratosphere trends.

Data used

This use case is based on gridded monthly-mean RO data records available from the ROM SAF. We combined data from the multi-mission data record GRM-28 (included in CDR v1) with data from the Metop data record GRM-29 (included in ICDR v1). The CDRs cover the time period 2002 to 2016 while the ICDR covers the time period from 2017 and onward.

The data are available here:

Satellite-based climate data records based on RO observations of multiple missions (CHAMP, COSMIC, GRACE, and Metop) over the period 2000-2016 by the ROM SAF that are cited as:

→ ROM SAF (2019): ROM SAF Radio Occultation Reprocessed Climate Data Record — Multimission, doi:10.15770/EUM_SAF_GRM_0001 (accessed 20 Jan 2023)

Satellite-based climate data records based on radio occultation observations from the Metop mission over the period 2017-2022 by the ROM SAF that are cited as: → ROM SAF (2019): ROM SAF Radio Occultation Interim Climate Data Record — Metop, doi:10.15770/EUM_SAF_GRM_0006 (accessed 20 Jan 2023)


IPCC, 2013: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 1535 pp.

IPCC, 2021: Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [Masson-Delmotte, V., P. Zhai, A. Pirani, S.L. Connors, C. Péan, S. Berger, N. Caud, Y. Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthews, T.K. Maycock, T. Waterfield, O. Yelekçi, R. Yu, and B. Zhou (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 2391 pp., DOI: https://doi.org/10.1017/9781009157896

Gleisner, H., K.B. Lauritsen, J.K. Nielsen, and S. Syndergaard, 2020: Evaluation of the 15-year ROM SAF monthly mean GPS radio occultation climate data record, Atmos. Meas. Tech., 13, 3081–3098, DOI: https://doi.org/10.5194/amt-13-3081-2020

Gleisner, H., M. Ringer, and S.B. Healy, 2022: Monitoring global climate change using GNSS radio occultation, npj Clim Atmos Sci 5, 6. https://doi.org/10.1038/s41612-022-00229-7

Ladstädter, F., A.K. Steiner, and H. Gleisner, 2023: Resolving the 21st century temperature trends of the upper troposphere–lower stratosphere with satellite observations, Sci Rep 13, 1306, https://doi.org/10.1038/s41598-023-28222-x

Meng, L., J. Liu, D.W. Tarasick, W.J. Randel, A.K. Steiner, H. Wilhelmsen, L. Wang, and L. Haimberger, 2021: Continuous rise of the tropopause in the Northern Hemisphere over 1980-2020, Sci Adv., Vol 7, Issue 45, DOI: https://doi.org/10.1126/sciadv.abi8065

Ringer, M., and S.B. Healy, 2008: Monitoring twenty-first century climate using GPS radio occultation bending angles, Geophys. Res. Lett. 35(5), https://doi.org/10.1029/2007GL032462 EXT

Steiner, A. K., F. Ladstädter, W. J. Randel, A. C. Maycock, Q. Fu, C. Claud, H. Gleisner, L. Haimberger, S.-P. Ho, P. Keckhut, T. Leblanc, C. Mears, L. M. Polvani, B. D. Santer, T. Schmidt, V. Sofieva, R. Wing, and C.-Z. Zou., 2020: Observed Temperature Changes in the Troposphere and Stratosphere from 1979 to 2018, Journal of Climate, Volume 33: Issue 19, DOI: https://doi.org/10.1175/JCLI-D-19-0998.1