the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Data quality control and calibration for mini-radiosonde system “Storm Tracker” in Taiwan
Abstract. This study introduced and evaluated the calibration schemes of a newly developed upper-air radiosonde instrument, “Storm Tracker” (ST), with data collected in field observations during 2016–2022. The ST is a radiosonde instrument developed and tested in 2016 (Hwang et al., 2020). In a series of field campaigns in the Taiwan area, more than one thousand co-launches of ST and Vaisala RS41-SGP (VS) are conducted. Using the VS measurements as the reference, we developed data correction methods and examined the characteristics of the ST sounding. The corrected ST soundings have 1-K temperature and 7 % relative humidity root mean square difference to the VS soundings. These error differences can be reduced to 0.66-K and 4.61 % below the 700-hPa height. The GPS estimated ST wind error difference is about 0.05 ms-1. The results suggested that the ST can perform similarly to the reference sounding and has reached the level required for environmental sampling and scientific research. The geostrophic adjustment dynamics indicate that the spatial temperature variation in the free atmosphere may not be large. However, the lower atmosphere may have large wind, temperature, and moisture variations. Due to the relatively low cost and accuracy after correction, ST can complement regular upper-air observations for high spatial and temporal resolution.
- Preprint
(1372 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 14 May 2024)
-
RC1: 'Comment on egusphere-2024-661', Anonymous Referee #1, 29 Apr 2024
reply
The manuscript "Data quality control and calibration for mini-radiosonde system Storm Tracker in Taiwan" describes
a twin sounding campaign of the Storm Tracker and the Vaisala RS41 radiosonde, where the latter is used as reference.
The results of the twinsoundings are used to derive a correction for the Storm Tracker temperature and humidity profiles, using a statistical method based on the cumulative distribution function (CDF).
I have considerable concerns with the way the twinsoundings were performed (payload configuration) as well as the applied analysis method, both of which I think affect the results and conclusions of this study.One concern is with the method applied to analyse the data from the coincident twinsoudings. Since
these soundings are performed with two different radiosondes on the same balloon, this allows for a
direct comparison of the profile data. Using a statistical method like the cumulative distribution
function (CDF) seems to me like an unnecessary complication. To my understanding the CDF-based
method as employed by Ciesielski et al., is applied when comparing radiosonde data taken under
similar meteorological conditions albeit not coincidently on the same rig.
The advantage of coincident twinsounding data is that it allows to directly determine the bias +
associated uncertainty between both systems. Furthermore, the physical mechanism that is causing
the bias between both radiosondes is warming by solar radiation that is counteracted by convective
cooling by the air flowing over the sensor (ventilation). The efficiency of the convective cooling
is directly linked to the altitude-dependent ambient air pressure. The CDF method that is applied
in the manuscript does indeed derive corrections for pressure ranges, but the purpose of further
analysing the differences in sense of ambient temperature is not clear to me.For full comment/report see supplement.
-
RC2: 'Comment on egusphere-2024-661', Anonymous Referee #2, 01 May 2024
reply
This paper discusses the results of corrections of temperature and relative humidity (RH) measurements of a mini-radiosonde system “Storm Tracker” in the troposphere in Taiwan. The Storm Tracker (ST) is described in the paper by Hwang et al. (AMT, 2020, https://doi.org/10.5194/amt-13-5395-2020), which uses a low-cost plug-in temperature-humidity sensor (TE-Connectivity HTU21D, https://www.te.com/) which is not developed specifically for upper-air sounding. Using more than 1000 flights of ST together with Vaisala RS41-SGP radiosonde (VS), the authors developed correction tables/functions using VS measurements as the reference. Furthermore, the authors tested different correction methods and compared the results. The best results are obtained with the simplest method called the cumulative distribution function (CDF) method, rather than more sophisticated machine-learning-based methods, at least for temperature.
While making more than 1000 dual flights with VS to deduce correction tables/functions for the ST measurements is a very good effort and the results are worth being reported in AMT, I tend to feel that the current manuscript is “overselling” the ST. This is mainly because the temperature and RH sensors of the ST, as shown in Figures 2 and 3 of Hwang et al. (2020), are never optimal for balloon upper-air sounding, e.g. with possible large contaminations to the temperature measurements depending on solar radiative heating and thus cloud cover status as well as day-versus-night difference. I would also have concern about the production stability of the sensor, i.e. whether the characteristics and the quality are within the reasonable range of uncertainty in different production batches (e.g. in different years). This may mean that in the end, we may always need dual flights with a well characterized and reliable radiosonde like the VS. In addition, the radiowave used for ST is from 432 to 436.5 MHz (Hwang et al., 2020) which is for amateur radiolocation, not for meteorological aids (e.g. International Telecommunication Union (ITU), “Radio Regulations”, Volume 1, 2020, available from https://www.itu.int/pub/R-REG-RR-2020); thus, in some countries ST cannot be used as a meteorological radiosonde officially. This may mean that the ST cannot be “widely” used in the future. Note also that one-hourly sounding campaigns like the one shown in Section 5.2 are not impossible with modern radiosondes like the VS; thus it is not easy for me to imagine possible applications of the ST aiming at new scientific studies that are only possible with the ST.
Therefore, I think that the manuscript might be published in AMT after reconsidering the “overselling” parts (which will be specifically pointed out below).
Specific comments:
Line 16 (and line 369): I am afraid that “geostrophic adjustment dynamics” is never discussed in the manuscript.
Introduction and Figure 1: Please describe the main technological points of the ST in more detail, including the model and characteristics of the equipped temperature-RH sensor. Figure 1(b) should be much greater, and an enlarged photo for the sensor part may be added. Also, Figure 1(c) is not a good one, because the ST sensor boom (or “sensor box”) is not well shown.
Introduction, the review part of various radiosonde issues: The papers cited here tend to be too old. More recent papers for more recent radiosondes need to be cited. These include:
- Vaisala RS41 radiosonde (but with GRUAN data processing): Sommer et al., GRUAN characterisation and data processing of the Vaisala RS41 radiosonde. GRUAN Technical Document 8 (GRUAN-TD-8), v1.0.0 (2023-06-28), https://www.gruan.org/documentation/gruan/td/gruan-td-8.
- Modem M10 radiosonde: Dupont, J., M. Haeffelin, J. Badosa, G. Clain, C. Raux, and D. Vignelles, 2020: Characterization and Corrections of Relative Humidity Measurement from Meteomodem M10 Radiosondes at Midlatitude Stations. J. Atmos. Oceanic Technol., 37, 857–871, https://doi.org/10.1175/JTECH-D-18-0205.1.
- Meisei RS-11G and iMS-100 radiosondes: Kizu et al., Technical characteristics and GRUAN data processing for the Meisei RS-11G and iMS-100 radiosondes, GRUAN-TD-5, v1.0 (2018-02-21), https://www.gruan.org/documentation/gruan/td/gruan-td-5.
These also include very useful information on the modern radiosonde sensor characteristics and on all necessary corrections to the radiosonde measurements.
Figure 2: Are the ST RH measurements really dry biased in comparison with the VS measurements below the 500 hPa level? They look wet biased in this case.
Lines 98-105: Testing mathematically more sophisticated machine-learning-based methods is good and interesting, but if the sensor characteristics are not optimal for upper air sounding (e.g. ~5 deg.C temperature error as shown in Figure 2 is just too large!), I tend to think that we should improve the sensor itself before reconsidering correction methods.
Table 1: Please also add the information on e.g. climate zone, season, etc., i.e. the information listed at Lines 92-93, to the place e.g. right after Location.
Figure 3: I confused with the three horizontal lines within the gray box. I thought that the data processing of the ST and the VS is independent before “L2_ST-VS”. In other words, I thought that the “Paired Entries” are used to establish correction tables/functions with various different methods. Or, in other words, I thought that for the VS, the authors simply use the manufacturer-processed data set, while I have impression from the current figure that the authors make their own (and perhaps common) corrections from L1_VS to L2_VS. The term Level 2 may be confusing if it is used for the ST at this stage, because the ST data will be further corrected by using the VS data as the reference if my understanding is correct.
Section 3.1 (and Figure 4 and its caption): Please add the explanation how to obtain delta T (and its uncertainty) from the CDF.
Line228: three-dimensional?
Lines 241-244: Please also add “(GLM1)”, “(GLM2)”, and “(GLM3)” here, not later.
Lines 255-258: Solar elevation angle may be a better variable?
Figure 7: Results from which method, CDF, GLM*?
Lines 283-285: I have a difficulty on this discussion. I am not an expert of metrology (i.e. science of measurement), but I think that that the “average errors” here are comparable to the uncertainties of the VS measurements does not mean that the uncertainties of the ST measurements (after the corrections) are comparable to the uncertainties of the VS measurements. Can we say roughly how large are the uncertainties of the corrected ST measurements in this case? Is it ~1.414 times?
Figures 10 and 11: The authors should also show delta T and delta RH profiles as well. In particular, delta T profiles are very important because we need tropospheric temperature measurements with uncertainties less than e.g. 0.5 K or even 0.1 K usually, and this degree of differences is hard to see with the current panels.
Line 330: A wavy pattern. Which pattern are the authors referring to? With this figure, I am afraid that we cannot judge that GLM may be better.
Section 5.2: One-hourly sounding campaign is possible with the VS as well.
Line 369: Geostrophic adjustment dynamics has never been discussed, I am afraid.
Lines 373-377: As discussed above, I personally think that the sensor characteristics (including the sensor covering and orientation) need to be improved and more optimized for upper-air sounding, before considering mathematically more sophisticated correction methods.
Lines 378-386: As discussed in the beginning, I tend to think that the authors are overselling the ST here.
Citation: https://doi.org/10.5194/egusphere-2024-661-RC2
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
99 | 21 | 18 | 138 | 8 | 6 |
- HTML: 99
- PDF: 21
- XML: 18
- Total: 138
- BibTeX: 8
- EndNote: 6
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1