Sensitivity of Change-Point Detection and Trend Estimates to GNSS IWV Time Series Properties
Round 1
Reviewer 1 Report
Review of the Manuscript (atmosphere-1342388)
“Sensitivity of change-point detection and trend estimates to GNSS time series properties”
submitted to Atmosphere by K-N. Nguyen & Co-authors
General comment
This paper is an interesting study reporting on the analysis of the performances of an analytical method applied to the time series of ground-based Global Navigation Satellite System (GNSS) integrated water vapour (IWV) measurements.
It is acknowledged that long records of observational data are pivotal in monitoring climate change and understanding the underlying processes. Decadal time series are often affected by inhomogeneities due to several reason (e.g. changes in instrumentation, station location, observation and processing methods, station conditions, etc.).
This study investigates the sensitivity of an updated release of a method (GNSSseg) already published, but recently improved in terms of computational time. The IWV trend are estimated too along with various qualitative and quantitative properties of both GNSS and some reference time series.
The subject is other than interesting, even scientifically and technically sound and funded on rather convincing claims, which are satisfactory supported by the experimental data. Results are extensively discussed, mainly in light of the state of art.
The manuscript is very well written and structured. The presented results are of outmost relevance.
When a reviewer has the chance to comes across a work like this, it is a pleasure to read it and the task is very easy.
What more can be said. The paper would be publishable in its current state, I propose only a few very minor changes which the authors may or may not consider.
Details on the suggested minor changes are given in the annotated version of the manuscript
I recommend publication of the manuscript after minor changes.
Comments for author File: Comments.pdf
Author Response
We thank the reviewer for the very positive comments on our manuscript. We applied all the corrections suggested by the reviewer. See the annotated PDF with our comments and answers.
Author Response File: Author Response.pdf
Reviewer 2 Report
- Please correct the word “reanalyses” as “reanalysis” for all article.
- Page 1, line 28, “GNSS measurements provide among the most accurate and continuous…”, the phrase “provide with” is the fixed usage in English.
- Page 2, line 82, “…taking into account the presentce of a periodic bias…”, please correct “presentce” as “presence”.
- Page 3, line 105, “…and spatial resolution..”, there are 2 dots in the paper, please delete one dot.
- Page 6, line 275. “…we consider as a cluster all the change-points closer than 80 days.” Why do you consider as a cluster all the change-points closer than 80 days, not shorter or longer?
- Page 7, line 297. “The spatial resolution of the reanalyses is 0.75°x0.75° for ERA-Interim and 0.25°x0.25°for ERA5. Reduced representativeness errors are thus expected from ERA5 data.” Of course the spatial resolution of ERA5 is better than ERA-Interim. But, we don’t’ know how better is while comparing the GNSS IWV. Can you compare the time series among CODE REPRO2015 dataset, IGS repro1 dataset, ERA5 dataset, and ERA-interim dataset?
- Page 8, line 369. In this paper, you discussed the segmentation results and how they are impacted by four factors. One of the results is that you compared the short period, 17 years, and long period, 25 years. Why do you discuss 17-years period and 25-years period, not shorter or longer?
Author Response
We would like to thank the reviewer for the detailed assessment of this manuscript and constructive comments. We have revised the manuscript accordingly. See the attachment PDF with our comments and answers.
Author Response File: Author Response.pdf
Reviewer 3 Report
This is a good study in general. Detecting the discontinuities in the ZTD/PWV time series is always a critical issue in pushing the GNSS product to the meteorology field. The article also gives a very detailed description of the GNSS data processing.
The major concerns are as follows.
- Regarding the reference data, ERA5 is an hourly product, but in this study, the 6-hourly product is used. I don't understand. Using the hourly data will certainly help as it improves the temporal resolution significantly.
- I would recommend making a table when describing the IGS and CODE processing strategies.
- For the ERAI and ERA5 products, do they have any discontinuities? Please check.
- In section 3.1.1, the authors attribute the relatively nosier signal of IGS compared to CODE to the mapping function differences. However, this can be hardly verified due to the different processing strategies, for instance, CODE uses GPS+GLONASS and IGS uses only GPS, the time resolution of CODE is 2-hour and that of IGS is 5-min, which both can cause large differences in the ZTD fluctuations.
- Section 3.1.1, line 423. Regarding the change points caused by the mapping function change, how many stations are influenced by this? How to explain that it causes discontinuities at some stations but not on others?
- When comparing the PWV trend between Numerical Weather Model (NWM) and GNSS, did you do it station by station? Is it possible that different stations have different trends?
- As for the impact of a priori ZHD, there should be a rule of thumb to quantify it.
- Line 260, "antenna PWV model", should be “PCO/PCV”.
- Line 131, in GNSS we prefer to use "Earth Rotation Parameter" instead of "Earth Orientation Parameter" as GNSS cannot determine UT1-UTC and nutation and IGS only provides polar motion and LOD.
- I would recommend replotting the figures because the texts are too small and can hardly be read.
Author Response
We would like to thank the reviewer for the detailed assessment of this manuscript and constructive comments. We have revised the manuscript accordingly. See the attachment PDF with our comments and answers.
Author Response File: Author Response.pdf
Reviewer 4 Report
Dear Authors
I really appreciated reading your paper. The manuscript reflects the
necessity to detect GNSS time series change points caused by receiver instrumental or SW changes but also due to other effects related to MF model changes or deficiencies in the backgrouns models. All these circumstands affects reasonable trend analysis of the IWV behaviour.
Please find below a few grammar hints (mostly concerning the first pages) and also a few technical questions which should be addressed in a slight revision.
Grammar hints are flagged with a G, technical questions with a T
comments refer to the line numbers of your document
1) One of my major concers relates to the title of your paper. When reading the title most people might look for a GNSS station coordinate time series analyses. As your major efforts are related to study the IWV series this should be also addressed in the title- at least the acronym IWV should show up in the title
2) G , Abstract , l 1,2
'.. segmentation method to detect changes in the GNSS data processing method, the length of time series ...., the auxiliary data used .... , and to the reference time series...'
3) G, l 4
'.. representatives of the first...'
4) G , l 22
'Inhomogenities often take ...'
5) G , l28
'GNSS measurements provide among ??? the most ..'
a word missing or re-phrase
6) G l29
'.. in all weather conditions but up to now have not yet been,,,'
I recommend to check literature for papers of Heinkelmann, Schuh et al.
who have studied the increase of IWV at reference station Wettzell to look for a trend in the decadal time series
7) page 2, l47, T
here you mention 'on-site electromagnetic coupling with the antenna environment'
which effect in detail do you have in mind there? do you refer to Multipath?
if not then you should at least also add Multipath to the relevant disturbing processes
8) G, l82
'presentce'? do you mean 'presence'?
9) G line 86
'which aim is to...'
10)G , l88
'.. can help to mitigate their..'
11) G, l90
'.. metadata when available, and ...'
12) G, l 93
'mimicking'?
13) G ,l120
'.. and discusses future work.'
14) T
below line 144 you introduce the basic formula ZTD=ZHD+ZWD
(not numbered?)
and discuss later on the influence of the MF; I would appreciate if
you enter here instead a more detailed formula with the basic slant delays
(SHD,SWD) which are mapped by individual dry and wet MF to the zenith;
this equation should also introduce potential gradients related to assymetry
15) T. l 147
here you refer to 'deficiency of the hydrostatic mapping function'
but also problems with the wet mapping function might be essential here
16) T , l152
'... the quality of the ZWD to IWV conversion...'
17) T, l227
maybe I am wrong here
but adding up the full number of years begin 1995-end of 2010
gives me always the number of 16 years which is definitely an even number
18) G ,l270
'sThe' ?
19) G, l280
'positions.tation' ?
20) T, chapter 2.4 formula (3)
here you note that seasonal components are captured by fourier Series of fourth order
Of course there are annual effects and maybe also bi annual ones; but is there really a significant contribution for coefficients of order 3 and 4?
if yes explain if these coeffients of erder 3 and 4 are really significant and which effects are captured by them. At least in the conclusions you refer to
'inter-annual variations' but with no details. I wonder that you never mention the effects of ground water changes which affect the station heights and subsequently also tropospheric delay estimates. But even ground water effects are not of order 3 or 4; but there is a phase-shift with respect to the
winter/summer cycle
21) G, l 380
' In comparison to the CODE series , the IGS series is noisier..'
22) G, l492
'It can be deduced from Table 1...'
23) G , l657
here you note
'smaller than from the raw data..'
I would recommend to substitute the word 'from' by either 'for' or 'between'
best regards
Author Response
We would like to thank the reviewer for the detailed assessment of this manuscript and constructive comments. We have revised the manuscript accordingly. See the attachment PDF with our comments and answers.
Author Response File: Author Response.pdf