the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 10 Apr 2026
Upper Air Humidity from Automatic Aircraft Surveillance Data
Abstract. Upper air humidity information is under sampled in the current operational meteorological observing network. Radiosondes observations form the backbone, but radiosondes balloons are typically launched only once or twice per day to limit the costs. The number of aircraft humidity observations are low in Europe, because in Europe only a few aircraft are equipped with water vapour sensors.
In this paper a novel technique is presented to derive humidity information from aircraft Automatic Dependent Surveillance Broadcast (ADS-B) data, whenever an aircraft is descending or ascending. The retrieved virtual temperatures observations, averaged over a vertical layer of 500 m, have an accuracy between 0.5 K and 0.75 K when compared to European Centre for Medium Range Forecast (ECMWF). Using additional external temperature information, estimates of the specific humidity can be calculated with an accuracy of 3–4 g kg-1 and in some cases between 2–3 g kg-1 (that is, when more than 20 estimates are available at the same reference height within 20 minutes). Applying the method to measurements from the Falcon F20 French research aircraft SAFIRE shows that even a single aircraft can be used to derive high-quality virtual temperature information (observation error ≈ 0.5 K). Comparison with Aircraft Meteorological Data Relay (AMDAR) and radiosonde humidity showed similar statistics.
Since ADS-B data is received from all ascending or descending aircraft in the vicinity of an airport, a vast amount of upper air virtual temperatures could be made available, when ADS-B information is gathered by ADS-B receivers installed at, or nearby airports.
- Preprint
(632 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
CC1: 'Comment on egusphere-2026-717', Gert-Jan Marseille, 10 Apr 2026
-
AC2: 'Reply on CC1', Siebren de Haan, 16 May 2026
Dear Gert-Jan,
Thank you for your comments. Below you find my replies.
- " I could not find a definition for \delta_T in the text."
The virtual temperature is defined as T(1+ δ q) , with δ =0.608 and to obtain the right hand side of eq requires the calculation of the derivative to temperature/virtual temperature, and assume that both are close.
- 8 g/kg follows from eq line 127:
(2 /(288)2 / 0.608 2 )1/2 = 0.008 kg/kg
" I guess you can provide estimates of the error of derived specific humidity based on the equations in section 3? Or are these the blue lines in Figure 4?"
Yes the error is described in section 3. And no these are not the blue lines in fig. 4
" In figure 6, the lower right panel shows very good agreement for Tv, but much less so for q. Can you add some lines in the text to explain this discrepency, e.g. by refering to (the potential weaknesses of) the equations in section 3."
The weakness of deriving q from Tv lies in the fact that the error in temperature needs to be small to have some skill. This is explained in Section 3.
Kind regards,
Siebren de Haan
Citation: https://doi.org/10.5194/egusphere-2026-717-AC2
-
AC2: 'Reply on CC1', Siebren de Haan, 16 May 2026
-
RC1: 'Comment on egusphere-2026-717', Anonymous Referee #2, 10 Apr 2026
The manuscript presents a novel technique of extracting real-time in-situ humidity observations from any ADS-B equipped aircraft during ascent and descent. In-situ humidity data is indeed very limited, since there are only 9 WVSS-II equipped aircraft currently in Europe and a bit more than a hundred in the USA, which makes this a technique of interest for potentially any met office, since it would provide a new source of global humidity data.
Going through the script, I have some questions regarding the error estimates for humidity and virtual temperature, and only minor issues elsewhere:
line 29: Enhanced Surveillance (EHS)
line 82: with a temperature of T_0 = 288.15 K
line 85: also define g_0
line 126: "using Taylor" is a bit misleading here, since Gaussian error propagation is used citing Taylor as a source, while line 139 says "applying again Taylors approximation". Even though Gaussian error propagation technically is a Taylor approximation, it wold be more clear to say "can be approximated via Gauß, according to Taylor (1997)" in line 126 and "and an estimate of the error can again be obtained via Gauß (Taylor 1997)" in line 139.
line 127: Explain here, how the \sigma^2_{T_\nu} term disappears. It seems like \sigma^2_{T_\nu} \approx \sigma^2_T was used here, which has to be stated in that case. Also the error should scale with 1/\delta^2 then, i.e.: \sigma^2_q \approx \frac{2}{\delta^2 T^2} \sigma^2_T
line 140: "(neglecting the last term)" actually all terms containing \beta_0 seem to be neglected. Please also elaborate on why those terms can be dropped.
Figure 2: "Two solutions are show: in blue denotes the solution" --> "Two solutions are shown: The blue lines denote the solution"
Figure 2: "The bottom two lines expresses the average difference " --> "The bottom two lines express the average difference"
line 213: "The resulting statistics are show" --> "The resulting statistics are shown"
Figure 6: This can be enlarged to a full page, since the individual Figures are quite small.
line 285: "MetoeFrance" --> "MeteoFrance"
Citation: https://doi.org/10.5194/egusphere-2026-717-RC1 -
AC1: 'Reply on RC1', Siebren de Haan, 11 Apr 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-717/egusphere-2026-717-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Siebren de Haan, 11 Apr 2026
-
CC2: 'Comment on egusphere-2026-717', Bruce Ingleby, 31 May 2026
My main comment is that there should be a discussion of errors in pressure altitude.
Currently the text apparently assumes the pressures are correct, whereas I think they are a major source of uncertainty (worse than the 7m or so for GNSS altitude, is there a reference for the 7m?). Static pressure is a difficult variable
to measure because an aircraft travelling at high speed disturbs the pressure field.
I have struggled to find estimates of uncertainty in hPa and the requirements in
terms of pressure altitude are very loose from a meteorological point of view.
Eg Christmann (2025) file:///home/dabi/Downloads/s13272-025-00843-0.pdf
"According to international standards [20], an aircraft
configured to comply with the requirements of an RVSM
minimum aircraft system performance specification is
non-compliant if it is found through height monitoring to
have an altimetry system error of 245 ft or more."
Some aircraft are not RVSM qualified and have even looser requirements.
If the pressure altitude error is constant then it would cancel out in the
technique presented (taking vertical differences). However below about 800 hPa
I see evidence that seems to point to the error changing with height.
Gracey, William 1980 Measurement of Aircraft Speed and Altitude
https://apps.dtic.mil/sti/html/tr/ADA280006/
p47 "The pressure field created by the airflow may change with the configuration of the aircraft and with Mach number and angle of attack."
I presure that this is the reason for the low level results that I see (where
humidity values are usually largest).
I also see different overall height/pressure biases for different aircraft.
This will cause complications for the proposed use of all ascending/descending
aircraft near an airport (if the sample of aircraft at the top of the 500 m layer
considered differs from the sample at the bottom).
I think a caveat should be added to conclusion 1 (line 261/262) that better
understanding of the errors in pressure altitude is needed.Detailed comments (mainly English usage)
17 "ADS-B receivers installed at, or nearby airports" replace "nearby" with "near to"
78 "Pressure altitude and geometric height are two different notions of the vertical coordinate": "options for" better than "notions of"
99 "derivates" - "derivatives"
104 "ascends and descends" - "ascents and descents"
115 "and the same destination or airport of departure" - "near the same airport"
171 "virtual temperature is measured more easily at temperatures above 0oC" - why?
195 "same airport/runway" I suggest deleting "/runway" (also line 217)
Figure 4 caption: "Amsyerdam"
268 "1. Running a trail" - "trial"
Citation: https://doi.org/10.5194/egusphere-2026-717-CC2 -
RC2: 'Comment on egusphere-2026-717', Bruce Ingleby, 24 Jun 2026
Review of "Upper Air Humidity from Automatic Aircraft Surveillance Data"
by Siebren de Haangeneral comments
This is a useful study about the extraction of extra information from ADS-B
aircraft reports.
I suggest changing the title to
"Upper Air Humidity or Virtual Temperature from Automatic Aircraft Surveillance Data"
This seems to be more in line with the perceived likely usage:
"The virtual temperature itself, could be valuable for assimilation in NWP." (191)
"The impact of virtual temperature assimilation on short range weather forecast on air humidity should be investigated" (line 270)
Q Does the author envisage other use of the data apart from NWP?
On line 9 "Using additional external temperature information, estimates of the specific humidity can be calculated with an accuracy of 3-4 g kg-1"
I think that assimilation users would probably prefer to have separate temperature
and virtual temperature profiles rather than a retrieval of specific humidity -
because retrievals usually have more complex error structure.
There is also the possibility that assimilation users could use GNSS height/altitude
data directly rather than virtual temperature (they would need to estimate and
remove height biases, but could also use cruise level information then).
In the method envisaged in this manuscript the humidity or virtual temperature data
would be a vertical average between two levels and so would need a different
"observation operator" for use in the assimilation system rather than just an
interpolation to a point value, so there would be some work needed before use.
There should be discussion of these points.In the Introduction
"Humidity plays a key role in meteorology ...." perhaps mention
Andersson E , Hólm E , Bauer P , Beljaars A , Kelly GA , McNally AP , Simmons AJ , Thépaut J-N , Tompkins AM. 2007. Analysis and forecast impact of the main humidity observing systems. Q. J. R. Meteorol. Soc. 133: 1473-1485.
Detailed commentsline 2 "Radiosondes observations form the backbone, but radiosondes balloons" -
"Radiosonde observations form the backbone, but radiosonde balloons ..."
i.e. radiosonde should be singluar here49 "from the board computer" - 'from the on-board computer"
71 "All height information is truncated to 25 ft (approximately 7.62 m)."
Is the information truncated or rounded (is there a reference for this)?Figure 1 caption "Brussel" - "Brussels"
137 "therefor" - "therefore"
151 "radiosondes observations" - "radiosonde observations"
187-189 "The signature for ascending aircraft is remarkable. The bias changes with height; this could well be related to changing of aircraft flight phase (e.g. change in airspeed, extra flaps, or landing gear for descending, and pitch for ascending)."
Is it possible to investigate this any further e.g. by comparing with airspeed?
Is the same pattern seen when looking at statistics for individual aircraft or
could it (somehow) be a sampling issue?229 "and one time Essen" delete "one time"?
Figure 6. Most of the panels have a y-axis from 0 to 5 - is this height in km?
"5 Summary, Conclusions and Recommendations"
I would like to see a sentence about the (25 ft) truncation of height data and
its effect. Also some discussion about the combination of information from different
aircraft - I can see that it has advantages, but are there any 'cons' eg different
biases for different aircraft, NWP centres would not be able to monitor results
for individual aircraft.Citation: https://doi.org/10.5194/egusphere-2026-717-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 319 | 96 | 26 | 441 | 14 | 17 |
- HTML: 319
- PDF: 96
- XML: 26
- Total: 441
- BibTeX: 14
- EndNote: 17
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Dear author,
Thanks a lot for this very interesting work. I have a question related to the equation on line 127. How do you arrive at the right hand-side of this equation? I could not find a definition for \delta_T in the text. In addition, how do you arrive at an error "around 8 g/kg" in the calculation example below the equation?
In line 189 you state: "The quality of derived specific humidity (i.e. retrieved from temperature and virtual temperature, both with an error of 1 to 1.5 K) is not good enough to estimate the relative humidity with some kind of skill."
I guess you can provide estimates of the error of derived specific humidity based on the equations in section 3? Or are these the blue lines in Figure 4?
In figure 6, the lower right panel shows very good agreement for Tv, but much less so for q. Can you add some lines in the text to explain this discrepency, e.g. by refering to (the potential weaknesses of) the equations in section 3.