the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Jan 2024
Stability requirements of observation systems to detect long-term stratospheric ozone trends based upon Monte Carlo simulations
Abstract. For new observing systems, particularly satellites, specifications on the stability required for climate variables are provided in order to be useful for certain applications, for instance, deriving long-term trends. The stability is usually stated in units of percent per decade (%/dec) and is often associated with or termed instrument drift. A stability requirement of 3 %/dec or better has been recently stated for tropospheric and stratospheric ozone. However, the way this number is derived is not clear. In this study we use Monte Carlo simulations to investigate how a stability requirement translates into uncertainties of long-term trends depending on the lifetime of individual observing systems, which are merged into timeseries, and the period of available observations. Depending on the need to observe a certain trend over a given period, e.g. typically +1 %/dec for total ozone and +2 %/dec for stratospheric ozone over thirty years, stability for observation systems can be properly specified and justified in order to achieve statistical significance in the observed long-term trend. Assuming a typical mean lifetime of seven years for an individual observing system and a stability of 3 %/dec results in a 2 %/dec trend uncertainty over a period of 30 years, which is barely sufficient for stratospheric ozone but too high for total ozone. Having two or three observing systems simultaneously reduces the uncertainty by 30 % and 42 %, respectively. The method presented here is applicable to any variable of interest for which long-term changes are to be detected.
- Preprint
(1936 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2023-3070', Anonymous Referee #1, 31 Jan 2024
Summary:
The authors perform a simple study to determine the impact of how potential drifts in datasets and biases between datasets can affect derived long-term trends through the use of Monte Carlo simulations. While the scope is fairly narrow, providing a short paper, it is a useful complement to studies that have attempted to assess the impact of the intrinsic quality of datasets on trend studies. These results are useful in the broader context of helping inform high-level requirements on future observing systems. I do not have any major concerns with the paper (mostly some clarifications and grammar changes), but I would like to recommend some additional questions or comparisons for the author that I believe will improve the final messaging. However, I would not consider them required prior to publication and will leave it to the author’s discretion on whether to include them in the final manuscript.
Comments:
Pg02, Ln028: “Similarly, the spread in the recent total ozone trends from the available merged datasets are on the order of ±0.5%/dec, which are indicative of drifts between the merged datasets. The apparent drifts are not only due to …”
I don’t believe it is a given that the differences in trend results come from drifts. Drifts are certainly a possibility, but we have yet to prove it. Perhaps change “indicative of drifts” to “suggest the possibility of drifts” and “apparent drifts” to “potential drifts”.
Pg03, Ln066: “For each of the timeseries a trend was determined and its distribution of the sample …”
Just for the sake of clarity, perhaps instead of saying “a trend was determined” you could say “a simpler linear least squares trend was fit”? At least I am assuming that is how you are calculating the trends used for this study since it is not explicitly stated in the paper. Similarly, I would modify “and its distribution of the sample is shown …” with “and its corresponding uncertainty, represented by the standard deviation of the distribution, is shown …” This would explicitly state what the trend uncertainties you refer to are.
Pg03, Ln068: “A stability requirement of 1.5%/dec and a life time of 7 years for each observing system results in a trend uncertainty of 0.77%/dec after 30 years (close to half the value of the stability requirement).”
In general, it can be easy to mix up the fact that this study often lists 1σ values but ozone trend uncertainties are often reported as 2σ so it might be good to be more explicit at various points in the paper. Given the context of the sentence that follows, this might be a good one to add “(1σ)” after both “%/dec” units. The last sentence of Section 2 is another good place to make this distinction, perhaps even showing the 2σ trend uncertainty value to better illustrate the point. Lastly, it would also be good to clarify this in the conclusion and abstract.
Pg03, Ln072: I am not sure if the word “typical” is best when referring to satellite lifetimes. Naturally many instruments/spacecraft last longer than their stated prime mission lifetime of ~3 years, but there are also those that do not and no mission ever claims beforehand to ensure their mission will last that long. Perhaps “typical” is too strong a phrase and “not uncommon” might be more appropriate.
Pg03, Ln075: “In our case we set a lower limit of five years to be included in the timeseries.”
On one hand I understand why you did this. However, there is an increasing movement toward the idea of repeatedly launching smaller spacecraft that may not have the long lifetime of a larger one but to do so more frequently. It would be interesting to allow this study to go to shorter lifetimes (such as perhaps 3 years).
Pg03, Ln087: “For a given stability requirement and period of the timeseries, the trend uncertainties increases with lifetime. … A continuous drift over a longer time period will cause larger deviations of the timeseries from the truth and increases trend uncertainties.”
I think this is simply because of the statistical distribution of drifts. Obviously if an instrument can be shown to be more stable, it operating for longer is better. However, in the absence of that guarantee, then more shorter missions is seemingly better. It might be worth noting this distinction/caveat.
Pg04, Ln097: “Redundancy in observing systems is therefore more effective in reducing trend uncertainties than improving the stability of an observing system.”
It may be worth reiterating this point in your conclusions. Although it may also be worth considering the comparison of two observing systems where both lack stability (e.g., 5-10%/dec [2σ]) versus having one observing system but with a better stability (e.g., 2-3%/dec [2σ]).
Pg05, Ln101: “If there are no biases (black curves in all panels) then the trend uncertainty decrease faster with time for lower lifetimes …”
It is also worth noting that, in the absence of biases, the trend uncertainty immediately begins to decrease as soon as more data is collected (unlike the beginning years of data with biases).
Pg05, Ln113: “A rule of thumb …”
This would imply this is already some sort of common knowledge. I think it might be better to say something along the lines of, and I’m rephrasing your whole sentence here, “While there are many possible combinations of observing systems, we show that a reasonable assumption of repeated single instruments with a mean lifetime of 7 years and a stability requirement of 3%/dec (2σ) would yield a trend uncertainty of about 2%/dec (2σ) after thirty years of observations.” Also, am I correct about those sigma values on your sentence?
In your conclusions, I think it is useful (as you have done) to reiterate some of your results, but I also think it would be useful to put it in the context of some of the current trend results and what sort of requirements we should be looking for in future measurement systems. If we are looking at trend in the stratosphere of 1-2%/dec, what sort of stability do we need to reliably extract those trends? It is also worth differentiating the upper stratosphere from the lower stratosphere as well as the difference with total column. Lastly, I think, from a messaging perspective, it is worth reiterating that all of these uncertainties stem purely from modeling the impacts of drifts and biases and that the intrinsic data quality of the different data sets (i.e., precision of the measurements) will only add additional uncertainty on top of this.
Editorial Comments:
Pg01, Ln016: “ … ozone depleting substances …” Needs ODS acronym added here for later use
Pg01, Ln017: Change “region” to “regions”
Pg01, Ln018: “Most significant trends are observed …”
Do you mean “Most statistically significant recovery trends”?
Pg01, Ln019: Insert a comma after “extratropics”
Pg01, Ln020: Insert “the” between “at” and “2σ”
Pg02, Ln030: “… but can also dependent on the way how the individual datasets are merged …”
“… but can also depend on how the individual datasets are merged …”
Pg02, Ln034: Insert a comma after “substances”
Pg02, Ln040: Insert a comma after “detection”
Pg02, Ln040: Change “… were stated …” to “… was stated …”
Pg02, Ln041: Change “area” to “areas”
Pg02, Ln043: Insert “an” before “observing system” and “will be only” should be “will only be”
Pg02, Ln046: Change “… the specification are …” to “… how these specifications are …”
Pg02, Ln046: “… they are leaned from the observed …”
Do you mean “learned from” or “informed by”?
Pg02, Ln047: Change “rise” to “raise”
Pg02, Ln049: Insert a comma after “perspective”
Pg02, Ln053: Change “simulation” to “simulations”
Pg02, Ln054: “… merged into timeseries …”
Given the usage of this in the greater context of the sentence, I believe this should say “how [data/observations] are merged into a timeseries”
Pg02, Ln055: Insert a comma after “Section 2”
Pg03, Ln059: “We assume that a simulated timeseries without any drift and bias after subtraction of the start value is zero.”
Move “is zero” from the end to “… bias is zero after …”
Pg03, Ln060: “Each individual observing system with a given lifetime (segment) of which we
compose a long-term merged dataset (timeseries) has varying drifts …”
Insert a comma after “(segment)” and after “(timeseries)”
Pg03, Ln066: Insert a comma after “timeseries”
Pg03, Ln071: Insert a comma before “individual”
Pg03, Ln075: Insert a comma after “we”
Pg03, Ln083: Replace “into” with “in” and insert a comma after “that”
Pg03, Ln086: Insert a comma before “assuming” and after “1.5%/dec”
Pg03, Ln087: Replace “represents” with “represent”
Pg04, Ln093: Insert a comma after “monitoring”
Pg05, Ln101: Insert a comma after “panels)” and replace “decrease” with “decreases”
Pg06, Ln115: Insert a comma after negligible and the word “the” before the word “case”
Pg07, Ln117: Remove “of course” and insert “are” after “nor”
Pg07, Ln118: Replace “normal” with “normally” and “changes […] is” with “changes […] are”
Pg07, Ln120: Insert “it” before “causes”
Pg07, Ln121: Insert a comma after “years” and change “outgas that change” to “outgas, which changes”
Pg09, Ln123: Insert a comma after “data”
Pg09, Ln140: Insert a comma after “decades”
Citation: https://doi.org/10.5194/egusphere-2023-3070-RC1 -
AC1: 'Replies to both reviewers', Mark Weber, 15 Apr 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2023-3070/egusphere-2023-3070-AC1-supplement.pdf
-
AC1: 'Replies to both reviewers', Mark Weber, 15 Apr 2024
-
RC2: 'Comment on egusphere-2023-3070', Anonymous Referee #2, 19 Feb 2024
This paper presents an interesting analysis based on Monte Carlo simulations of the impact of given stability requirement in ozone observing systems on effective trend detection and assessment of ozone recovery. The main parameters used in the simulations are the lifetime of single satellite missions, stability of the missions and their eventual bias. The author concludes that with the current stability requirements given by GCOS-22, ozone recovery is barely detectable 30 years after the peak of ozone depleting substance in the stratosphere. The manuscript is generally well written and well presented. However, there are some issues and recommendations that need to be considered before publication in Atmospheric Measurement Techniques.
- The study and more specifically its conclusion address mainly the GCOS-22 3%/dec stability requirement. However, 2%/dec breakthrough and 1%/dec target requirements are also mentioned for total ozone. More discussion is needed on the impact of such requirements on total ozone trend uncertainties.
- The study is mainly based on satellite missions and their drift over their limited lifetime. However, the global ozone observing system includes also ground-based measurements with much longer lifetime, which provides an additional constraint to the evaluation of satellite instrument stability. How can such constraint from ground-based instruments be taken into account in the study, considering also the findings on the reduction uncertainties from parallel observing systems?
- The study would benefit from a comparison of the parameters considered in the study (drift, bias, lifetime) to actual satellite measurement time series, e.g. of for total ozone. This would enable an assessment of the total ozone measurements system and provide a more concrete assessment of trend detection capabilities as they currently stand.
Specific comments
- P3 l59-60: The sentence starting with "We assume" should be reformulated since a time series cannot be zero.
- The simulation set-up leading to Fig. 4 is not completely clear about the considered bias. Is it equal to 0%? What is the explanation of the author about the bump in the curves around 1.1 % trend uncertainty? Is a drift of 1.5%/dec detectable from current observing system? Could it be corrected? A refined statistical set up could eventually be envisaged based on e.g. Bayesian statistics reducing the probability of strongly drifting times series over the longest observing periods.
- P5 l114-115: the sentence starting with "A rule of thumb..." should be replaced by a table specifying the trend uncertainty from various parameters, e.g. bias, stability and lifetime, summarizing results of Fig. 6.
- Fig 6. In order to better explain the figure, another version of Fig1 or Fig3 could be shown, including the specified bias between segments.
- P7 l120: replace "outgas that change" by "outgas, which changes".
Citation: https://doi.org/10.5194/egusphere-2023-3070-RC2 -
AC1: 'Replies to both reviewers', Mark Weber, 15 Apr 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2023-3070/egusphere-2023-3070-AC1-supplement.pdf
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 185 | 85 | 18 | 288 | 11 | 8 |
- HTML: 185
- PDF: 85
- XML: 18
- Total: 288
- BibTeX: 11
- EndNote: 8
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1