the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 12 Aug 2026
Assessment of humidity ground checks for GRUAN-compliant automatic radiosonde operations
Abstract. Automated Radiosonde Launchers (ARLs) are increasingly adopted for upper-air observations as they offer major operational advantages compared to manual launches, but they introduce new challenges for maintaining reference-quality humidity measurements within the GCOS Reference Upper-Air Network (GRUAN). In particular, the applicability of the pre-launch Standard Humidity Chamber (SHC) test, routinely used for manual radiosonde launches, has not been systematically assessed for radiosondes stored inside ARLs prior to launch.
This study investigates how residence time inside ARLs affects radiosonde humidity sensor performance and the adequacy of SHC tests for automated GRUAN operations. A coordinated experiment was conducted at six sites using three radiosonde models (Vaisala RS41, Meisei IMS-100, and MeteoModem M10) and three ARL systems (Vaisala AS15, Meisei ARS, and MeteoModem Robotsonde). Radiosondes were tested in an SHC prior to storage and again after residence times of 7 and 14 days. Relative humidity biases at saturation (RH = 100 %) were analyzed using aggregated statistics, including a paired RH bias difference—i.e., the change in bias between pre-storage and post-storage measurements, to quantify temporal drift. A non-parametric bootstrap resampling analysis was first applied to assess the robustness and confidence intervals of the paired RH bias difference estimates. Statistical significance of the paired RH bias difference distributions was subsequently evaluated using the Wilcoxon signed-rank test. A normalized storage-effect metric, combining the systematic drift and variability of the paired RH bias differences, was compared with the GRUAN humidity uncertainty contribution associated with ground preparation.
Results indicate that radiosonde humidity performance remains stable for residence times up to 14 days under the tested storage conditions, particularly for the RS41–AS15 configuration, where only small negative drifts were observed (approximately ~ −0.1 % RH week-1). In contrast, the IMS-100–ARS configuration exhibited a pronounced increase in drift with storage time (approximately ~ −0.8 % RH week-1), which was amplified under high-humidity and temperature seasonal conditions, while the M10–Robotsonde configuration showed a reversed drift with partial bias reduction during storage (~ +0.12 % RH week-1). These findings demonstrate that radiosonde behavior during storage strongly depends on the combined radiosonde–ARL system.
Based on the experimental and statistical evidence, an operational upper limit of 14 days for radiosonde storage inside ARLs is recommended for GRUAN-compliant operations, with residence times up to 7 days providing optimal reliability. The study confirms the feasibility of implementing SHC tests for ARL operations while highlighting the need for manufacturer- and site-specific procedures to ensure long-term traceability and consistency of reference humidity measurements. The results demonstrate that residence time inside ARLs constitutes a system-dependent source of uncertainty, warranting explicit inclusion as an additional component in the GRUAN Data Product (GDP) humidity uncertainty budget.
- Preprint
(713 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-3215', Anonymous Referee #1, 13 Oct 2026
-
RC2: 'Comment on egusphere-2026-3215', David Edwards, 28 Oct 2026
Review of Romanens et al, " Assessment of humidity ground checks for GRUAN-compliant automatic radiosonde operations".
General comments
The report appears of good quality and appropriate for AMT.
The scope is narrow, in that it is only relating to a limited set of sites, using a limited selection of equipment over a fairly low sample size, but as this specifically relates to the GRUAN methodology and sites available, it cannot be easily avoided and the limitations are discussed.
The recommendation for a separate test for residency time impact for the combined radiosonde & automated launcher type on the overall RH uncertainty appears appropriate given the evidence discussed, as an additional factor in the total GRUAN uncertainty calculations.
The impact of site climate and internal climate control mechanisms is discussed, but even within radiosonde/ARL pairings, it is indicated that this could lead to additional variability, so any new site would also need to be carefully studied with respect to its individual response to climatic conditions, in addition to the combination of radiosonde/ARL.
Thinking of the practical implications of using different models of radiosondes and ASLs in the future given the findings in this paper, while 14 days is the maximum that a radiosonde is likely to be stored in these models of ARL, this is not the case in newer designs of ARLs from these manufacturers. The current design of Meisei ARS can store 40 radiosondes, leading to a potential maximum of 20 days storage, assuming two flights per day. The current Vaisala design, the AS41 Autosonde, which replaces the AS15 has a capacity increased from 24 to 60. In future, if those are brought into use at GRUAN sites through system life-cycling, it would be beneficial to test a storage duration of up to 30 days to understand if there are any further impacts, or if the recommendation of a maximum storage time of 7 or 14 days is still valid. This would be true of the higher capacity Meisei ARS system and any future higher capacity MeteoModem systems.
I have some minor comments on the experimental design in the specific comments and a few minor technical comments.
Specific comments
Table 1: There are only 6 sites studied with a limited latitude range, leading to a limited range of climatic conditions, so further study may be necessary if an ARL system is deployed to a more extreme climatic environment nearer to the equator or poles. It may be worth noting this in section 4.1.Line 157: While a dry bias was noted by TAT for the Meisei system prior to the test period and leading to the inclusion of their extended test, was there an assumption that seasonality was not necessary to study at the other types of ARL? While the sample size of the study (about 30 radiosondes) was carried out over 2 months at most sites, the sample size for the longer Meisei study was not increased proportionally to ensure robust sampling in each season. I also note that a period of 8 months was chosen to study a period of interest, but what was the reason for choosing only this period to study? It may be worth noting this in section 4.1.
Figure 5: Seasonality with 4 seasons as plotted could be misunderstood to indicate equal analysis of all seasons, but the experimental description only indicated that 8 months were studied, not 12. Perhaps consider a fairer presentation based on months rather than seasons, or to make it clearer that some months are not completely represented in the description. Due to the small number of radiosondes studied at the station over the 8-month period (N = 24 for t7 and N = 28 for t14), each season can only represent a very small sample of radiosondes, and this limitation would carry through in any per-month analysis. It may be worth noting this in section 4.1.
Line 522: As noted, the dedicated residence-time uncertainty is not only radiosonde specific, but radiosonde-ARL pairing specific and it would be worth making this clear for readers who are unfamiliar with GRUAN GDPs. As such, if either component is changed, then re-analysis of the associated uncertainties would be necessary, alongside updated guidance for best practice on the process and order of MGC and SHC testing in the launch and loading process, and maximum recommended storage durations.
Technical comments
Line 8: ARS (the Meisei Automated Radiosonde System) is not defined in the List of acronyms.Table 2: Temperatures and RH are shown as relative ranges, but with no context as to the maximum and minimum values experienced at each site. Including the actual temperature ranges in the table would give context as to the conditions experienced by the stored radiosondes. This would be informative, in my opinion, but is not essential.
Line 436: Is a sample size of 24-30 representative enough of the general population of radiosondes (for example, were they chosen from mixed batches and ages at the time of use – does age of sensor have an impact on uncertainty)? The representivity of the radiosondes used is not discussed. It may be worth noting this limitation in section 4.1.
Line 550: CNR-IMAA and AEMET are not defined in the List of acronyms.
Line 559: AS15 is more accurately described as the Vaisala Autosonde® (which is a registered trademark), or the more generic Vaisala Automatic Radiosonde Launcher instead of “Vaisala Autolauncher”.
Citation: https://doi.org/10.5194/egusphere-2026-3215-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 202 | 112 | 32 | 346 | 35 | 29 |
- HTML: 202
- PDF: 112
- XML: 32
- Total: 346
- BibTeX: 35
- EndNote: 29
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of Romanens et al, " Assessment of humidity ground checks for GRUAN-compliant automatic radiosonde operations".
General comments
The submitted manuscript describes a careful study carried out to investigate the effects of storage time in automated radiosonde launchers on radiosonde humidity measurements, under the auspices of GRUAN, the GCOS Reference Upper Air Network. This is an important question for GRUAN because the number of manual launch sites might decrease significantly in the coming years. There are also some small additional sub-components of the study including the effects of repeated ground checks, and of realistic contaminants.
The general subject, making careful long term measurements of the basic properties of the atmosphere, such as humidity profiles, is of course, fundamental to atmospheric science.
The study has been carefully implemented and is well described and reported.
I recommend publication in AMT. I only have a small number of comments which are all quite minor.
Particularly in the introduction, I found the very heavy use of acronyms and abbreviations made the writing not very reader-friendly. Some sentences had up to four different acronyms in quick succession. Some of them didn't seem especially useful to me.
I would also suggest the study might be limited by the fact that the participating stations were all situated in relatively temperate locations in the world. This is seen in table 2 where the climate control has little trouble at most sites keeping the conditions within a mild range all year. This might not be the case in more extreme locations, such as polar, desert or an equatorial island with very high humidity. I would be interested if the authors have any comment to make on that point?
Specific comments
Line 46 Is ensuring that automation doesn't compromise the integrity of climate datasets really a "key role" of GRUAN?
Line 57 There seems to be some inconsistency here because all the sites except for Hong Kong are listed as agency-site name-country?
Lines 80-84 For this paragraph, you should give the actual names of the stations rather than the three letter codes, to be more reader friendly.
Lines 80-84 They are all essentially mid-latitude sites though. The west to east coverage doesn't necessarily imply different conditions at all. It would be much more relevant to give the climate classification of each site.
Lines 80-84 As I mentioned earlier, do the authors think these results would be the same if carried out in a more extreme environment?
Table 1
I suggest dividing table 1 into two separate tables, to improve the readability. At the moment it is too compressed with too many abbreviations. The site names should be listed in full as well as their three letter codes. The column headings shouldn't be abbreviated (eg "ARL", "CC" etc). The abbreviations "CC", 'A/C", "AD" and "H" aren't explained in the caption and the reader has to refer to the text.
Line 98 (Table 1) "GPS" is not the correct term here. You should list the longitude and latitude as distinct columns (with labels "Longitude" and "Latitude".)
Line 116 You list both "weather forecasting" and "numerical weather prediction" so I assume you are referring to human forecasters in the first case? For human use, how accurate do the humidity values need to be?
Lines 154-157 Would the study have benefitted from being carried out over a longer time period at the other sites as well?
Line 180 "Based on the information provided in Table 1" makes it sound as if the authors have received the information from elsewhere and don't completely trust it. (Do you just mean "As shown in Table 1"?)
Lines 191-192 This sentence doesn't quite ready properly – "the fact that … represents an unbalanced contribution "
Line 197 Should "Quality Controlled" be "Quality control"?
Figure 2 (Left) – The blue line seems to be heavily affected by the first point, without which the gradient would be fairly similar to the yellow line.
Figure 2 What is the explanation for the offset? I was surprised by it.
Lines 346-348 This is the first time the reader has heard about this effect, which seems important. It would be better to discuss it earlier on when the SHC is being introduced. It must limit the ability of the SHC to be a reference to within more than 1-2 %.
Technical comments.
Line 6 Delete apostrophe after sensor
Line 137 Delete repeated "Vaisala"