the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 18 Aug 2026
Impact of Super Typhoon Yagi on Earth’s Atmosphere: from Troposphere to Thermosphere/Ionosphere
Abstract. Typhoons can generate gravity waves that transport energy and momentum from the lower atmosphere to the upper atmosphere, but the spectral characteristics and vertical coupling pathways of these disturbances remain incompletely understood. We investigate the atmospheric response to Super Typhoon Yagi (2024) by a Mesosphere-Stratosphere-Troposphere (MST) radar, two BeiDou Navigation Satellite System (BDS) receivers, and ERA5 reanalysis data. Great wind disturbances were observed in the troposphere and lower stratosphere by the MST radar, as Yagi passed approximately 188 km south of it. Gravity-wave activity was strongest between 5–12 km altitude and was dominated by periods of 0.6–5 h, with the maximum absolute momentum flux reaching 1.98 m2 s-2. ERA5 temperature perturbations revealed concentric wave structures centered near the typhoon eye, with amplitudes increasing from approximately 0.7 K at 20 km to more than 2 K at 50 km altitude. In mesosphere and lower thermosphere, enhanced and mutually consistent oscillations in the zonal, meridional, and vertical winds appeared on 7 Sept., with a dominant period near 2.0 h. Nearly simultaneous disturbances with a dominant period of approximately 2.1 h were detected in the total electron content of two BDS receivers. These observations provide evidence of typhoon-related disturbances extending from troposphere to thermosphere/ionosphere.
- Preprint
(12777 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-4680', Ravindra Pratap Singh, 11 Sep 2026
-
RC2: 'Comment on egusphere-2026-4680', Anonymous Referee #2, 12 Sep 2026
Peer Review of “Impact of Super Typhoon Yagi on Earth’s Atmosphere: from Troposphere to Thermosphere/Ionosphere”
Overall recommendation: Major Revision
This manuscript presents an interesting multi-instrument observational study of the atmospheric response to Super Typhoon Yagi (2024), combining three-dimensional wind observations from the CLAIR MST radar, ERA5 temperature fields, and BDS/GNSS TEC measurements.
The observational opportunity is valuable. In particular, the ability of CLAIR to provide three-dimensional wind observations in both the lower atmosphere and MLT region during the passage of a strong typhoon is potentially an important contribution. The identification of approximately 2 h oscillations in all three wind components in the MLT region on 7 September, together with approximately 2.1 h oscillations in two geographically separated TEC measurements, is intriguing. The manuscript also makes a useful attempt to connect observations at different atmospheric heights.
The authors report strong lower-atmospheric disturbances, with the largest gravity-wave activity occurring approximately between 5–12 km and a maximum absolute momentum flux of 1.98 m² s⁻². They then identify wave activity in the 0.6–4.5 h range in the troposphere and lower stratosphere, approximately 2 h activity in the MLT region, and approximately 2.1 h TEC oscillations.
However, the central scientific conclusion is stronger than what the present analysis appears to demonstrate.
In particular, I am not yet convinced that the manuscript has demonstrated vertical propagation of the same gravity-wave disturbance from Yagi into the MLT and subsequently into the ionosphere. The observations demonstrate temporal associations that are consistent with such a scenario, but establishing the propagation pathway and causality requires additional analysis and/or substantially more cautious interpretation.
I believe the paper could become a valuable contribution if the authors strengthen the spectral methodology, quantify significance and uncertainty, and appropriately distinguish between evidence consistent with vertical coupling and demonstration of vertical coupling.
Major Comments
- The central claim of vertical coupling is not sufficiently demonstrated
The manuscript essentially constructs the following chain:
Typhoon → tropospheric GW activity → stratospheric GW activity → MLT ~2 h wave → ionospheric ~2.1 h TEC disturbance.
The individual observations are interesting, but the manuscript currently does not establish that these are necessarily the same propagating wave.
For example, the tropospheric/stratospheric analysis identifies disturbances over approximately 0.6–4.5 h, whereas the MLT disturbance is concentrated near ~2 h and the TEC disturbance near ~2.1 h. The authors interpret the narrowing of the spectrum with altitude as evidence of progressive filtering.
That is physically plausible, but stronger argument would be to show the correlation in other characteristics like:
- horizontal propagation information;
- background wind information (cross-spectral analysis, see Lu et al. (2005));
- consistency of horizontal wavelength/direction;
- or comparison with a gravity-wave propagation model.
At present, the manuscript mainly establishes spectral similarity and temporal coincidence.
I therefore recommend changing statements such as:
“We track the spectrum variations of the typhoon associated GWs from troposphere to thermosphere/ionosphere”
and
“The GWs with a dominant period of 2 h that ultimately reach ionospheric heights”
to more cautious formulations unless additional propagation evidence can be provided.
- The signal-processing procedure may substantially influence the reported wave spectrum
The authors apply several sequential processing operations, including:
- sliding-window denoising;
- running-mean removal;
- Butterworth band-pass filtering;
- interpolation;
- additional running averaging;
- LOESS detrending;
- thresholding;
- wavelet analysis.
For example, for the troposphere/stratosphere, a 25-point running mean is removed before applying the 0.6–5 h band-pass filter.
The radar temporal resolution is 15 min. Therefore:
25 points × 15 min = 375 min = 6.25 h.
Consequently, the preprocessing itself contains a characteristic timescale that is close to—and actually longer than—the upper limit of the subsequent 0.6–5 h analysis band.
This does not automatically invalidate the result, but the authors need to demonstrate that the preprocessing does not artificially enhance, suppress, or reshape the 0.6–5 h spectrum.
I strongly recommend showing:
- the original time series;
- the background component;
- the residual after background removal;
- the result after band-pass filtering;
- the wavelet spectrum.
The authors should also provide the exact Butterworth filter order and implementation details, including whether the filter is applied forward/backward and how the boundaries are handled.
This is particularly important because the paper's main physical conclusion depends heavily on the spectral evolution with altitude.
Minor Comments
- The conclusion contains: “These observations will deep our understanding…” This should be: “These observations will deepen our understanding…”
- In Figure 2 the F10.7 and Kp information is useful, but the authors should explicitly discuss whether the observed ionospheric period could be affected by background solar/geomagnetic variability.
- In Figure 10 the raw TEC, detrended TEC, thresholded TEC, and final residual would be much more informative than showing only the processed product.
Strengths of the Manuscript
- Excellent observational opportunity
- Multi-instrument approach
- Three-dimensional wind observations
- The MLT result is genuinely interesting
Citation: https://doi.org/10.5194/egusphere-2026-4680-RC2 -
RC3: 'Comment on egusphere-2026-4680', Anonymous Referee #3, 08 Oct 2026
Chen, G., Tao, P., Zhang, S., et al., "Impact of Super Typhoon Yagi on Earth's Atmosphere: from Troposphere to Thermosphere/Ionosphere". The authors use the new Chinese Low-latitude Atmosphere and Ionosphere Radar (CLAIR), a 50 MHz MST radar at Qinzhou (22.11°N, 108.66°E), to observe Super Typhoon Yagi (2024). Yagi passed about 188 km south of the radar at 21:00 UT on 6 September. The radar data are combined with ERA5 temperatures, VIIRS DNB cloud imagery and BDS-GEO (C03) vertical TEC from receivers at Nanning and Dongfang.
The observations are worth publishing. However, the central claim of typhoon-driven coupling from the troposphere to the ionosphere is not yet supported. In my view the revised manuscript should at minimum: 1. characterise the MLT and TEC disturbances (wavelengths, phase speeds, directions) and show, by ray tracing or an equivalent analysis, that they can originate from Yagi at the observed times; 2. search for concentric wave structure around Yagi in the mesosphere (VIIRS DNB nightglow) and the ionosphere (two-dimensional GNSS TEC maps), or explicitly acknowledge its absence; 3. demonstrate statistical significance against a quiet-time baseline and control sites away from the storm; 4. show that interpolation, smoothing, height averaging and TEC clipping do not generate or shape the ~2 h signals, and that the radar data are not contaminated by precipitation or interference; 5. temper the title, abstract and conclusions to match the evidence.
If these analyses support the hypothesis, the paper will make a clear contribution. If they do not, the material could still be published as a careful multi-instrument case study with appropriately cautious conclusions. therefore, I recommend major revision
MAJOR COMMENTS
The link rests on a common observed period (~2 h) and on overlapping time windows. No wave in the MLT or the ionosphere is characterised well enough to be traced back to Yagi.
Wave parameters. Please derive vertical wavelength and direction of phase progression from time–height sections of the filtered MLT perturbations. Use hodograph or Stokes-parameter analysis to obtain intrinsic frequency and propagation azimuth, and use the radar's own background winds to correct for Doppler shift. The same parameters should then be checked against the GW dispersion relation in each layer.
Propagation path and travel time. A ray-tracing analysis (e.g., ERA5 below ~1 hPa, radar winds or an empirical model above) is needed to show that waves launched by Yagi's convection can arrive in the radar's MLT volume and at the two TEC pierce points at the observed times. An illustrative estimate shows why this matters. A 2 h wave with a 300 km horizontal wavelength and N ≈ 0.02 rad s⁻¹, in calm air, has a vertical wavelength near 13 km and a vertical group velocity near 6–7 km h⁻¹. Rising from 10 km to 90 km would take ~12 h, during which the packet moves ~1500–2000 km horizontally. The waves seen above Qinzhou on 7 Sep would then have been launched many hours earlier and far from overhead. The near-simultaneous onset of the tropospheric (08 UT, 6 Sep), TEC (22 UT, 6 Sep) and MLT (00 UT, 7 Sep) signals must be reconciled with such travel times.
Ground-relative versus intrinsic period. The ground-relative period is conserved only in a time-invariant background. Tidal winds of tens to more than 100 m s⁻¹ (Fig. 8) violate this, so matching observed periods between layers is not, by itself, evidence of a common wave.
The MLT onset is set by data availability. MLT echoes exist only from about 00:00 to 09:30 UT. The "7 Sep" identification therefore has a resolution of one day, and the event is one of six analysed days.
GWs and TIDs from tropical cyclones are normally attributed to the storm by their concentric (ring-shaped) morphology around it, traced from layer to layer. Azeem et al. (2015), Chou et al. (2017), Xu et al. (2019) and Li et al. (2022), all cited here, follow this approach, as do Nishioka et al. (2013) and Miller et al. (2015). In this manuscript, concentric structure appears only in ERA5 at 20–50 km. In the MLT and the ionosphere the evidence consists of single-point time series, and two pierce points cannot establish a concentric pattern or locate a source.
I encourage the authors to search for this signature with data that already exist:
1. VIIRS DNB nightglow imagery for the nights of 5–8 Sep. The authors already use VIIRS DNB, but only for cloud snapshots. Mesopause-region concentric waves are routinely visible in DNB nightglow (Miller et al., 2015).
2. AIRS 4.3 µm and 15 µm radiances for stratospheric concentric waves at higher resolution than ERA5.
3. Two-dimensional detrended TEC maps from the dense GNSS networks in southern China and neighbouring regions. These include the Meridian Project's 32 stations, CMONOC, Hong Kong SatRef and IGS, using all visible GPS/BDS/Galileo satellites. Such maps would test for concentric TIDs centred on Yagi and give their wavelength, phase speed and direction.
4. Any meteor radar, ionosonde, all-sky airglow imager or incoherent scatter radar data available in Hainan or Guangxi. A meteor radar would also fill the night-time MLT gap.If concentric TIDs centred on Yagi are found, the conclusions would be greatly strengthened. If they are not found, the conclusions should be weakened. Either way, the propagation directions should be checked for consistency. The ERA5 arcs lie mostly east of the eye, while Qinzhou lies north of the track and the Dongfang pierce point lies south of it. Which part of the wave field reaches each observing volume? With the source lying between the two pierce points, a concentric TID would propagate northward at Nanning and southward at Dongfang. The time lag between the two TEC series can test this directly.
Gap filling in the MLT (L255–263). Up to 14.5 h per day is filled by linear interpolation, then band-pass filtered and analysed with a continuous wavelet across six days. Linear segments and block edges inject broadband power into the 0.6–5 h band. Each daily block (~9.5 h) contains fewer than five cycles at 2 h and fewer than two at 5 h, so the cone of influence should apply to each block, not only to the ends of the six-day series. The statement that the band-pass "is sufficient to weaken its influence" needs demonstration, for example synthetic tests with the same gap pattern, or a gap-tolerant method (Lomb–Scargle, per-block wavelet, or harmonic fits).The 3-point running mean (L258). On 15 min data this filter has an amplitude response of about 0.33 at 1 h, about 0.80 at 2 h, zero at 45 min and a sign reversal at 36 min. This alone suppresses 0.6–1 h waves relative to 2 h waves. The narrower MLT spectrum and the "selective role of the propagation environment" (L268–269, L374–376) may therefore be a processing artefact. Identical processing should be applied in all layers, or the filter's transfer function corrected for.
Altitude averaging of perturbations (L213–216, L259). Averaging perturbations over 75–100 km, 3–16 km or 18–25 km cancels waves whose vertical wavelength is shorter than the averaging depth. What survives is long-vertical-wavelength signal or height-coherent artefacts. Please compute spectra at each height and average the power instead. Please also show time–height sections of the filtered perturbations on 7 Sep and on reference days.
TEC thresholding (L281–283). After detrending, values with |dTEC| > 1 TECU are set to zero. Here the background VTEC is 150–180 TECU near the crest of the equatorial ionization anomaly (EIA) with F10.7 ≈ 230–266, so a 1 % TID already exceeds 1.5 TECU. The threshold, taken from studies under different conditions, can remove real signals, and the zeroing introduces step discontinuities that leak power across the analysed band. Please use relative perturbations (%), justify any rejection objectively, describe the "jumps" (L276) and their cause, and show raw and detrended series for the highlighted window.
Comparability of spectra. Colour scales range from 0–0.02 to 0–40 arbitrary units across Figs. 6, 9 and 10, so layers cannot be compared visually. Please use physical units (m² s⁻², TECU² or %²) and a stated normalisation.
Sampling and aliasing. With 15 min sampling the Nyquist period is 30 min. Please state whether each 15 min value is an average or a snapshot. If it is a snapshot, the 10–30 min waves typical of convection can alias into the 0.6–1 h band.
Critical-level filtering depends on phase speed relative to the background wind, not on period alone. High-frequency waves with large vertical group velocities are generally the ones that reach the thermosphere (Vadas, 2007). The claim that the environment removed 0.6–1 h waves and passed 2 h waves needs a phase-speed-based argument, such as blocking diagrams from ERA5 and radar winds, and must first rule out the artefact in comment 4.2.
Convectively generated concentric TIDs often have periods of 10–30 min. The 0.6 h lower cutoff excludes these, although the 1 Hz TEC data could resolve them. Please extend the TEC analysis to shorter periods.
Secondary-wave generation by breaking primary waves (Vadas et al., 2003, which is in the reference list) is a plausible alternative pathway to the thermosphere and should be discussed.
Both pierce points lie near the northern EIA crest, and the highlighted window spans 06–20 LT. Please discuss how fountain-effect variability, pre-reversal enhancement and plasma bubbles were separated from the GW signal.
Minor Comments
References: Vadas et al. (2003) appears in the list but is not cited in the text. "Chinese Meridian Project (2023)" (L117) is cited but not listed.
Use "CLAIR" or "MST radar" consistently, and "Sep." or "September" consistently.
Units throughout: use superscripts (m² s⁻²).
L13–14: The abstract says vertical coupling pathways are poorly understood, but the study does not resolve a pathway. Please align the motivation with what is actually shown.
L17: "Great wind disturbances" → "Strong wind disturbances".
L19: State that the momentum flux is per unit mass, and give the height and time of the maximum.
L70–71: "Other radio and optical observation systems": only VIIRS is used, and only for cloud images.
L82–83: "Yagi has jumped three grades" → "Yagi intensified by three grades".
L85–90: "Level 17" refers to China's extended Beaufort scale. Please give the wind-speed threshold (≥ 56.1 m s⁻¹) for international readers.
L96–103, Fig. 2a: Kp and F10.7 alone are insufficient. Add AE/SME and SYM-H, and comment specifically on 6–7 Sep.
Fig. 2b: the red arrows need a legend or scale. A simple azimuth-versus-time curve may be clearer.
L110–120: Radar parameters are missing: peak power, pulse length and coding, beam width, coherent and incoherent integrations, ST/M mode time-sharing and its effect on effective sampling, and the wind-retrieval method (DBS and its homogeneity assumption, with the ~48 km beam separation at 90 km). Clarify "maximum detectable wind speed". The reference "Chinese Meridian Project, 2023" is not in the reference list.
L127: "are adopted to investigated" → "were used to investigate".
L130–137: If VIIRS DNB is kept only for cloud snapshots, this section can be shortened. See major comment 2 for a more useful application.
L135: "approximately 106 times". If 10⁶ was intended, the superscript has been lost. Please verify against Miller et al. (2012).
L145–151: Give station coordinates, C03 sub-satellite longitude, elevation angles, the mapping function, DCB handling and the method used to derive VTEC. Justify the 350 km shell height under high solar activity, when hmF2 may be higher.
L160: Specify the "sliding-window denoising": window length, statistic and threshold.
L165–166: "stratosphere intrusion" → "stratospheric intrusion".
L168: missing full stop after "September 5".
L178–180: The inertial period at 22.11°N is ~31.8 h. Give the Butterworth order and state whether zero-phase filtering was used.
L183: "Vincent et al. (Vincent, 1983)" should read "Vincent and Reid (1983)".
Eqs. (1)–(2): The minus sign between the variances is hard to read. Define the averaging operator (time window and height range) and the sign convention.
L193–205, Fig. 5: State the step of the 12 h average and any height smoothing. The height where the flux "decreased rapidly" is given as 11, 12 and 13 km in different places (L201–203, L319).
L199: "respectively, Then," → "respectively. Then,".
L211–216: State whether wavelet power was averaged over height or perturbations were averaged before the transform. Define the "25-point running mean" (time, 6.25 h?).
L226: "can't effectively receive anything echoes" → "cannot receive sufficiently strong echoes".
L228–229: Is the 11 × 11 background a mean or a median? Over which levels does the 7-point vertical average run?
L232: Explain how the propagation direction was determined from daily snapshots.
L241–243 vs L255: Fig. 8 uses 21:00–12:00 UT, the spectral analysis uses 00:00–09:30 UT. Please reconcile.
L245–246: "there was no much disturbances emerged" → "few disturbances were evident".
Figs. 6, 9, 10: label colour bars with physical units and mark significance contours.
L257 vs L262: "The few data points" contradicts "up to 14.5 hours of interpolation per day".
Fig. 7: panel letters are illegible. Add grid lines, a clearer typhoon-centre marker and a colour-bar label.
Fig. 8: "Meridion wind velocity" → "Meridional wind velocity".
L285–288: "Remarkable consistency" should be quantified with cross-correlation, coherence and lag between the two TEC series.
L337–340: Quantify the expected conservative growth and compare it with the observed growth (see major comment 6.2).
L348: double full stop after "7 Sep".
L349: Give a quantitative travel-time estimate for ~2 h waves instead of "tens of minutes to several hours".
L353–354: "The disturbance duration is likely longer" is unsupported, since there are no night-time data.
L377: Wang et al. (2023) is a solar-eclipse study and is not an appropriate source for GW saturation. Cite Fritts and Alexander (2003) or similar.
End matter: Please add a code-availability statement for the processing and wavelet scripts.
L393: "will deep our understanding" → "will deepen our understanding".Citation: https://doi.org/10.5194/egusphere-2026-4680-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 293 | 171 | 86 | 550 | 91 | 85 |
- HTML: 293
- PDF: 171
- XML: 86
- Total: 550
- BibTeX: 91
- EndNote: 85
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Please find the attached PDF file for my comments.