Review for the article „Intercomparison of wind speed, temperature, and humidity data between dropsondes and aircraft in situ measurements“

by Namdari et al.

The article presents an intercomparison of the parameters describing atmospheric dynamics (wind speed, wind direction, temperature and relative humidity) obtained by in-situ aircraft measurements and dropsondes at different altitudes above the North Atlantic. The authors use statistical methods to ensure a spatial and temporal collocation within strict limits to make the values comparable.

The topic is of high relevance, as static calibration of sensors in the laboratory is not sufficient to characterize and understand sensor behavior during flight conditions, and in particular wind measurement via five-hole probe and fusion with GNSS and IMU sensors cannot be calibrated in the lab.

Generally, such statistical intercomparison is important to assess data quality. However, in my opinion, statistical methods can only be applied after carefully checking sensors and methods and determining their uncertainty.

There are several aspects that strongly influence the data, and I would expect a thorough discussion before simply comparing the data sets.

Therefore, I suggest rejection of the manuscript in the current form. Taking into account the suggestions will take a lot of time and effort, so this cannot be done within a revised version.

Below are my major concerns, as well as smaller issues, with reference to the corresponding sections and lines.

Major comments:

• The article is about validating airborne in-situ measurements by dropsondes. Therefore, it is not helpful to discuss comparison between remote sensing/lidar and dropsondes, as done in the introduction and the conclusions, with a large literature review. Please either include lidar measurements in the analyses or omit it also in the introduction and conclusion. Also the literature review in the tables should have the focus on the topic of your article, not lidar/dropsonde comparison.
• Please include a thorough analysis of uncertainty of dropsondes and in-situ measurements, taking into account in particular sensor response time, see e.g. Bärfuss et al., 2018, and Bärfuss et al., 2023
• A common approach to calibrate wind speed for 5-hole sondes is to perform calibration manoeuvres, e.g. flying a square and checking if the wind speed and wind direction is the same for all flight directions. Please comment on this and how the data can be improved. Was this applied to your data set?
• If you discuss in situ wind measurements based on drones, please take into account the first articles about this topic, which include a thorough comparison with conventional measurement systems (tower, radiosonde, remote sensing), e.g. Martin et al., 2011
• It is good that you make use of measurements above the ocean, to have a larger spatial homogeneity. However, please characterize the atmospheric conditions, in particular stability. Up to which altitude does the top of the marine boundary layer reach? Maybe you can show vertical profiles of temperature or potential temperature at least for your case studies?
• Dropsonde data: according to l. 117, there were corrections and smooting algorithms applied. Please summarize shortly what was done to the data. This is highly important in this context.
• RH values above 100% were set to 100%. Please explain this. Why was the obviously erroneous data not corrected or excluded? What does this mean for uncertainty?
• According to l. 135, extensive calibrations were applied. Please summarize shortly.
• 138: “wind data derived from aircraft are sensitive to deviations away from straight and level flight conditions”: please quantify, and indicate how you treated this problem in the data analysis. What is “no rapid changes in altitude”? Please quantify.
• 145/146: The RH values have an uncertainty of 5%. Then the calculated RH values have a higher uncertainty? How were they calculated? Why is the uncertainty higher?
• What is the response time of your sensors, in particular humidity sensor of the radiosonde, and how was this compensated? Again, see publications of Bärfuss.
• What about the high variability of all the parameters within clouds? Could you do a statistics on the variability within cloud sections for straight and level flights? Does it really make sense to compare with values of radiosondes several km away and for temporal differences of several min? Generally I would expect a high variability of all parameters within clouds.
• You state that the altitude plays a major role. How do you determine altitude? It says in l. 208 that it is geopotential height. So probably it was calculated based on pressure data. What is the accuracy of the pressure sensor, in particular for the radiosonde? What does this mean in terms of uncertainty for the altitude?
• You state that RH is strongly dependent on temperature. So why don’t you use a different parameter for humidity, like the water vapour mixing ratio, which is independent of temperature?
• The examples in Fig. 4 and 5 indicate highly dynamic conditions. Maybe they are not so suitable as an illustration?

Minor comments:

• Differentiate between wind speed, wind direction and wind vector instead of simply using “wind” or even “winds”
• Do not use acronyms without explanation
• Put references in chronological order when citing them, e.g. l. 41, 42 Lewis and Schwartz, 2004, Nuijens and Stevens, 2012, Neukermans et al., 2018; throughout the manuscript

References:

Bärfuss, K., Pätzold, F., Altstädter, B., Kathe, E., Nowak, S., Bretschneider, L., Bestmann, U., and Lampert, A.: New Setup of the UAS ALADINA for Measuring Boundary Layer Properties, Atmospheric Particles and Solar Radiation, Atmosphere, 9, 28; doi:10.3390/atmos9010028, 21 pp., 2018.

Bärfuss, K. B., Schmithüsen, H., and Lampert, A.: Drone-based meteorological observations up to the tropopause – a concept study, Atmos. Meas. Tech., 16, 3739–3765, https://doi.org/10.5194/amt-16-3739-2023, 2023.

Martin, S., Bange, J., and Beyrich, F.: Meteorological profiling of the lower troposphere using the research UAV "M²AV Carolo", Atmos. Meas. Tech., 4, 705–716, https://doi.org/10.5194/amt-4-705-2011, 2011.

General

The author team presents a method to detect closely separated measurements with a Falcon aircraft and dropsondes (dropped from a King Air) during a large number of flights and drops in the Northwestern Atlantic during 2020 – 2022. The analysed data cover layers below, in, and above clouds, mostly below 3 km height. The flights took place in different regions, seasons, aerosol loads and meteorological conditions. The data mining and sorting algorithms bear new approaches. The analysis of the detected data pairs of humidity, temperature, and horizontal wind components, respectively, is done with elementary statistical metrics like correlation coefficients, standard deviations, biases etc..

The paper ist mostly fluently written, well and clearly structured, and technically joining the standards of AMT.

Despite the considerable effort the authors put into the analysis of a large heterogeneous data set from multiple aircraft flights and dropsoundings, I have some major concerns about a publication in AMT, since the paper lacks a clear objective and an interpretation on what the derived statistical measures mean for instrument accuracies and atmospheric heterogeneity, and if/how these can be separated.

• The paper gives a lengthy overview on previous results from airborne observation system intercomparisons. It is not made clear, for what purpose that is done. The cited results, such as wing-by-wing aircraft measurements, or comparisons between vertical profiles from dropsondes and wind lidar, are basically much different to those presented here. Almost all studies had a specific objective, such a comparison of a „new“ technique to a proven one. All studies also aimed to keep spatial differences very small in the measurement setups. The present paper does not say, what their own study may provide as added value. Just claiming that „it has not been done before“ does neither justify presenting that overview to be included.
• The presentation of results in terms of statistical measures is a bit of a „technical report“ style. The reader has little chance to follow the presentation and keeping an overview on all the given numbers of correlations, biases etc. This is particularly the case in the discussion of certain quantities like humidity or temperature. Both the quantative results as well as their interpretation must become more focussed.
• The authors compare data from a sophisticated aircraft measurement system (TAMMS, laser diode hygrometer) and a widely used dropsonde system (NRD41). The aircraft instrumentation will have undergone the high quality standards of sensor selection, calibration and fusion (such as for wind) to obtain meteorological quantities. There are certain limitations, of course, such as when measurements are made during certain aircraft maneures. Excluding such cases as done, the general in-flight accuracy can be well estimated. Efficient ways to test for remaining problems, such as alignment errors after (re)installation of instruments, are normally detected by changing flight directions in homogeneous wind. Wing-by-wing flights have been also performed for many research aircraft, to detect - mostly small – systematic or random arrors. It is not pointed out, why „occasional“ near-by-passes with dropsonde can be really used to detect errors in aircraft measurements.
• Similarly, a widely used and proven dropsonde like the NRD41 in general will not be subject to large or unknown systematic errors. There are exceptions, such as for humidity in clouds and in the upper troposphere, but lab calibrations, pre-flight checks, sonde intercomparisons ensure that problems are mostly limited to indidual sondes, such as when a electronics component fails or a unfavourable GPS-constellation is given in a certain height range. A lot of information is also available from radiosonde intercomparisons, such as in the WMO UAII 2022 Campaign. Though not fully applicable to dropsondes, basic information on sensors accuracies is relevant for both for radiosondes and dropsondes, being partly quite similar in electronics, sensors and data handling. Again, the question arises what additional value is provided by measurements from an aircraft passing only occasionally a dropsonde.
• Atmospheric inhomogeneity is presumably a major reason for the differences between aircraft and dropsonde data. Ideal collocated measurements were not possible for reducing differences to measurements system differences. Correlation coefficients and scatter in biases, and other statistical metrics must be considered as functions of horizontal and vertical separation d. Only in the limiting case of d -> 0, knowledge about instrument differences can be obtained. The data sample ist big for large separations, however, but getting very small for small separations. This seems to exclude a robust extrapolation to r -> 0.
• A remaining application option would be optimizing combinations of aircraft and dropsonde data to derive two-dimensional cross sections through the atmosphere. Excluding biases between dropsondes and aircraft data would be helpful for that purpose. But the authors do not address this issue.

Minor issues

Due to the general criticism in the overall part, a detailed sentence-by-sentence

The reviewer did not review the English language, since there are enough native speakers in the author group.

The technical presentation (figures, tables, captions, citations, literature) seems to be consistent to AMT standards.