the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 05 Dec 2023
Analysis of a newly homogenised ozonesonde dataset from Lauder, New Zealand
Abstract. This study presents an updated and homogenised ozone time series covering 34 years (1987–2020) of ozonesonde measurements at Lauder, New Zealand, and derived vertically resolved ozone trends. Over the period of 1987–1999, the ozone trends in the homogenised ozone data are predominantly negative from the surface to ∼30 km, ranging from −2 to −10 % decade−1, maximising at around 12–13 km. These negative trends are statistically significant at 95 % confidence level below 5 km and above 17 km. For the post-2000 period, ozone at Lauder shows negative trends in the stratosphere (but the trends are only statistically significant above 17 km), maximising just below 20 km (∼ −5 % decade−1), despite stratospheric chlorine and bromine from ozone-depleting substances (ODSs) both declining in this period. In the troposphere, the ozone trends change from negative for 1987–1999 to positive in the post-2000 period. The post-2000 ozone trends from the ozonesonde measurements compare well with those from a low-vertical resolution Fourier-transform infrared spectroscopy (FTIR) ozone time series. A multiple-linear regression analysis indicates that anthropogenic forcing plays a significant role in driving the significant negative trend in the stratospheric ozone at Lauder, in which the effect of greenhouse gas (GHG)-driven dynamical and chemical changes is reflected in the significant positive trends in tropopause height and tropospheric temperature, and significant negative trends of stratospheric temperature observed at Lauder. The interannual variation in lower stratospheric ozone is largely explained by the variation in tropopause height at Lauder, which is highly anti-correlated with stratospheric temperature and correlated with tropospheric temperature. Furthermore, the impact of ODSs and GHGs on ozone over Lauder is assessed in a chemistry-climate model using a series of single forcing simulations. The model simulations show that the predominantly negative modelled trend in ozone for the 1987–1999 period is driven not only by ODSs, but also by increases in GHGs with large but opposing impacts from methane (positive) and CO2 (negative), respectively. Over the 2000–2020 period, although the model underestimates the observed negative ozone trend in the lower stratosphere but clearly shows that CO2-driven dynamical changes have had an increasingly important role in driving ozone trends in this region.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(13919 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(13919 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2534', Anonymous Referee #1, 21 Dec 2023
The paper describes an updated / homogenized ozone sonde dataset from the Lauder station in New Zealand. It shows results from a trend analysis of the updated sonde dataset, as well as trend results for the region around New Zealand from simulations by the NIWA-UKCA chemistry climate model. The Southern Hemisphere is a region with quite sparse ground-based observations, so Lauder is a very important station and the presented material is, in principle, well suited for publication in ACP.
Overall the manuscript is generally well written, the Figures are clear, and the use of English is generally good.
Nevertheless, I have a number of questions and suggestions that should be addressed before the paper can be accepted for ACP.In section 2.2 the authors introduce their multiple linear regression model, which includes a single linear trend and EESC as proxies for long-term variations / trends. However, in section 3.2 / Figs. 3 and 4, they show linear trends for two periods (1987 to 1999 and 2000 to 2020). It seems that these trends were calculated with just one linear trend proxy in a simple linear regression, without additional proxies. Is that correct? If so, that needs to be stated very clearly, and the sequence of the (sub)-sections should maybe be reordered. In the current text, this has confused me, and will probably confuse most readers. Alternatively, two linear trends (one over each period), or a trend and change of trend (hockey stick), could be used in the multiple linear regression to be consistent throughout the paper.
The same kind of question applies to the trends for the CCMI simulations. Were these obtained with just a simple linear trend, over the two different periods, or with the full multiple linear regression model?
In line 134, the authors state that "Observations as well as basis functions are smoothed using a 12-month boxcar filter". This is not a usual approach and could affect the derived uncertainties quite significantly, because essentially the number of independent data points is reduced by a factor of 10. How is this accounted for in the uncertainties? How are the uncertainties derived in the first place? How do the authors account / correct for auto-correlation in the residuals (ε (t) in their Eq. 1)? How would the results and uncertainties look without doing this 12-month boxcar? I would assume that much lower values for R2 would be found compared to the quite high values in Table B1. My suggestion is to not use this 12-month boxcar, and go with the standard approach using monthly means without smoothing. In any case, these questions need more analysis and more discussion in the manuscript.
Figures 3 and 4: I suggest to combine both figures and plot the trends before and after homogenization in the same plot, in different colors. As the figure stands now, it is very difficult to see how the homogenization changed the trends (very little after 2000?).
Figure 11: This is a very good and interesting Figure. However, I sorely miss the observed trends here. Please include those in the two panels. How do the vertical profiles of regressed EESC and GHG / overall linear trends from the multiple linear regression of the observed data look like in the two different time periods? How does that compare to the corresponding simulated trends (orange lines, red lines in the Figure)? How do the overall trends compare (to the black lines in the Figure).
Line 27: Here you write that SH stratospheric ozone trends are dominated (controlled largely might be a better expression) by Antarctic ozone depletion (which is large and significant and controlled by ODS). Yet in your regression you find hardly any significant impact of ODS / EESC (e.g. in Fig. 8). This is quite a big and important discrepancy. Yet in the later parts of the manuscript, in the conclusions and in the abstract, this discrepancy is hardly mentioned at all, let alone resolved. You do mention negative stratospheric ozone trends, but those seem to be from simple linear regression, not multiple linear regression. So what's going wrong / different with the multiple linear regression? Is the right approach used? I think this needs to be cleared / understood.
Line 67ff: Please indicate if these differences are resolved now. It seems that the Lauder sondes trends don't change very much by the homogenization (Fig. 3, post 2000 trends are almost the same, pre 2000 trends have become more negative). So I would assume that your paper does not change the Godin et al. results, except for the Lauder FTIR data trends which have changed and now fit with the sonde trends. I find this question important, and I would like to see answers, both already here, and also later in the paper, e.g. in the conclusions.
Line 130: What happens when tropopause height is not detrended? I think this should be tried and discussed. If non-detrended tropopause height picks up a GHG induced climate-change related ozone trend, that might be the correct way to do the trend analysis. One could argue that the mechanism underlying short term changes of tropopause height and ozone changes also acts on the long time-scale, because climate-change statistically favors high tropopause conditions. So the acting processes could be the same. There may not be the need for a different process acting on the long time scales. E.g. in the annual cycle you also have a close correlation (on a longer time scale) between high tropopause height and low ozone in the lower stratosphere. While this is somewhat different from the short time-scale processes due to high and low pressure systems, both time scales give similar correlation of high tropopause with low ozone.
Line 146: Please explain what forcings are included in RefC2. I assume all forcings.
Lines 147, 150: I think this should be "corresponding fixed forcing simulation" not "corresponding single forcing simulation".
Around Lines 186, 197: I think these different trends need to be explained / need a bit more dicussion. If Godin et al. get different trends from the same data, there must be an explanation. What happens if you try Godin et al.s regression? Could the difference come from excluding a few extreme soundings? Generally, differences from previous findings need more explanation. They should not just be mentioned and then ignored. I have also done multiple linear regression (with hockey-stick trends) on HEGIFTOM data from Lauder and find trends very similar to your trends in Fig. 4. This is reassuring. Maybe the Godin at al. Lauder trends were too negative? Anyways, I think this needs a bit more discussion, and maybe should be mentioned in conclusions and abstract as well. Of course it needs to be worded appropriately.
Line 237: As mentioned above, this really begs the question what happens with non detrended tropopause height in the regression.
Line 242: I am not sure about this linear trend term. It seems like a very unspecific overall collector of various things, picking up a confusing mix of ODS-related, GHG-related and other changes. I think there should be better proxies, e.g. hockey stick, two linear trends, ... For me, the insignificant ozone changes picked up by the EESC term are a warning sign. If, according to the regression, ODS changes had no impact on stratospheric ozone at Lauder, I very much wonder if we can trust this regression.
Around line 258: This is a weird argument. We see EESC effects in the troposphere, but we don't see them in the stratosphere, where they should be coming from? Can that be resolved? Might different proxies in the regression help (as suggested above)?
Around line 320 (and in several other places). Are these simple single linear trends? Or are they from the multiple linear regression? This keeps me confused throughout the manuscript. I would much prefer to just have one type of regression, or a much better explanation of what was done, and why there might have been two approaches.
After line 339: As also mentioned above for Fig. 11, I think there needs to be more discussion about how the observed and simulated trends fit together, or do not fit together.
Citation: https://doi.org/10.5194/egusphere-2023-2534-RC1 -
AC1: 'Comment on egusphere-2023-2534', Guang Zeng, 28 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2534/egusphere-2023-2534-AC1-supplement.pdf
-
AC1: 'Comment on egusphere-2023-2534', Guang Zeng, 28 Feb 2024
-
RC2: 'Comment on egusphere-2023-2534', Anonymous Referee #2, 26 Dec 2023
In the manuscript “Analysis of a newly homogenised ozonesonde dataset from Lauder, New Zealand”, Guang Zeng et al. derive long-term MLR trend estimations from the ozone sonde dataset of Lauder. The dataset, recently homogenised in the frame of HEGIFTOM, shows negative pre-1999 trends in better agreement with the previous literature. Post-2000 trends are shown to be significantly negative in the stratosphere and positive in the troposphere, and are in very good agreement with trends derived from a co-located FTIR instrument. The analysis by MLR imputes the negative post-2000 stratospheric trend to anthropogenic forcing led by CO2 related to positive trends in tropopause height and tropospheric temperature and negative trend in stratospheric temperature. CCM simulations from NIWA-UKCA attribute the negative pre-1999 trends not only to the ODS increase but also to a GHG increase with opposing impacts of CH4 and CO2. For the post-2000 period, the CCM analysis assess the role of the dynamical changes driven by CO2 on the negative lower stratospheric trends.
The manuscript fits well within the scope of ACP and is of high scientific quality. It is generally well written despite some very long sentences which do not read well. The results are well presented. The homogenization of the ozone sonde dataset succeeded in improving the pre-1999 trend agreement with other observation techniques and with the literature.
The detailed analysis of the MLR results is a very good contribution to the understanding of the underlying issues of stratospheric and tropospheric pre- and post-2000 trends. The results derived from CCM analysis and derived from MLR on observational dataset enhance the role of CO2-driven dynamical changes in the lower stratospheric trend which represents a significant step towards understanding trends in this atmospheric region.
I list below general and specific comments. General comments are questions and remarks which need clarifications before the paper can be accepted for ACP. Specific comments are minor revisions which may help to improve the readability of the manuscript.
General Comments
Line 64-74:
The same ozone sonde dataset is used in Godin-Beckmann et al. 2022 and in the present study. The authors say the trend values of Godin-Beckmann et al. to be “exceedingly large” (line 69). However, the present study trends are within the Godin-Beckmann et al. uncertainties and respectively, except for the 25km value. The authors should comment on these differences.
Why are the uncertainties of both studies so different? Is the residuals autocorrelation taken into account in the present study? What is the impact of using EESC as an additional explanatory variable on the trend values of the present study?
Line 74:
“which has been updated from the dataset used in Godin-Beckmann et al. (2022)”. Could you please comment already here (instead of Line 194) on the FTIR dataset update as the trends reported in Godin-Beckmann et al. are very different from the present study?
Line 77:
“into the near future”: I cannot not see any mention of post-2022 results in the manuscript. What do you mean?
Line 89:
Table1 indicates only 3 dual flights to evaluate the effects of the sensing solution change. Is the transfer function/correction factor derived from these 3 dual flights or is a general transfer function used?
Line 122:
« Surface humidity are measured » : please replace by: « Surface humidity is measured »
Is this RH?
“Surface humidity are measured by the radiosonde that has a humidity sensor.”: please be more specific.
Line 126:
Why is QBO10 used for the whole altitude range? Would QBO50 have been more appropriate in the troposphere?
Line 130:
"Fig 6." is the first mention of a figure in the manuscript, in that case, this should be Fig 1.
I suggest to remove this figure or to move it to Appendix B.
Line 140:
“coarse resolution » : do you mean spatial or temporal resolution? Please be more specific.
Line 177:
“marked differences”: the pre-1999 trends values are different but similar within their uncertainties. The marked difference lies in the significance of the trends. This should be made clear. Furthermore, the differences would be clearer if both trend profiles (uncorrected and homogenised) were represented on a single figure.
Line 184:
“Figure 3 and 4”: I suggest to merge the information into one single figure with 2 panels (pre-1999 and post-2000) to make comparisons easier.
Line 187:
same comment as for Lines 64-74. Please explain why the trends values are different in both studies. The role of the additional proxies used is non negligeable. For instance, the significative contributions of the EESC or HTtropo proxies (Table B1) are influencing the remaining linear trend value when compared to a MLR not considering these proxies.
Line 195:
“updated version”: See comment for Line 74.
Line 197:
The agreement between ozone sonde and FTIR trends is good however trends of the present study stay negative and significant above 15 km while WMO 2022 shows positive but non significant trends. A reason or a tentative explanation should be given.
Line 200-207:
At that point, I cannot find a reason for these considerations on the seasonal variation.
I suggest to remove this section unless it is used in the explanation of the trends estimation.
Line 213:
As said in comment of Line 130, Figure 6 could be removed. If kept and moved in Appendix B, please replace “de-trened” by “de-trended” and “12-boxcar” by “12-month boxcar”.
Line 210-217 and Figure 7 8,9 and table B1:
Separate and redundant information is spread over 3 figures and 1 table, I find this difficult to handle. I would suggest the following simplification:
Information on Fig 7 is given in Fig 9 except for the R2 value and the regressed timeseries. You could remove Fig 7 and keep only Fig 9 with the observed timeseries in dashed, the regressed timeseries in black and with the R2 information in the respective subplot titles.
Figure 8 and table B1:
The information is redundant and Figure 8 could be replaced by a highlighting of the major contributor(s) in table B1. Table B1 should then be part of the manuscript and not of the Appendix B.
Line 223:
“The downward trend in the stratospheric ozone is clearly explained by the significant negative linear trend that represents all quasi-linear, monotonic drivers of change”
Please add which drivers and that you will discuss these below.
Why can these drivers not be used as proxies but have to be treated apart?
Why is the contribution of EESC negligeable here?
Can you exclude this negative trend to be due (evt partly) to a drift or step in the timeseries which has not been considered or corrected by the homogenisation?
Line 231:
How is this trend estimated?
Meng et al. trend values are estimated “on the natural variability–removed time series” (volcanic eruption, ENSO and QBO). Are you estimating the trends with a similar model as Meng et al. or with your eq.1? Please comment.
Line 236-7:
“This indicates that the negative contribution to the ozone trend in the lower stratosphere (between ∼9 to 15 km) can largely be projected on the significant increase in tropopause height»
Figure 9 shows that de-trended HTtropo is a significant proxy for explaining the ozone variation between 9 and 15 km height. Instead of making considerations about the correlation between de-trend and non de-trended HTtropo and ozone, would it be possible to use non-detrended HTtropo as a proxy? If not, why?
Line 267:
“The regression function we construct here is more suitable to explain the stratospheric ozone changes.” Does it mean that the trend values estimated below 6 km are not reliable?
Line 269-270:
How are the modelled ozone trends estimated? What do you mean by “separately”? If you apply a ILT multi-linear regression on simulated ozone values, please describe the MLR used in that case. If the model directly outputs the trends, please clarify.
Line 272:
Please specify here that/if the attribution of the changes in modelled O3 is done as in Zeng et al 2022 and Morgenstern et al. 2018.
Line 275:
“ are broadly in agreement with the Lauder observations (Figure 4).” I suggest to add the trend values estimated on the ozone sonde dataset in Figure 11.
Line 318:
“with a maximum of −9% decade−1 around 13 km“: Figure 4 shows a maximum of -12%/dec at 13km.
“with significant trends at the 95% confidence above 12 and below 5 km. » Please adjust to values shown on Figure 4.
Specific Comments
Line 5-7:
The sentence doesn’t read well. It’s too long and contains 3 brackets. Please rephrase.
Line 10-13:
Please, make 2 sentences from this one.
Line 19:
“…but clearly shows …” : Please replace by “… , it clearly shows …”
Line 20:
“have had an increasingly important role”: please replace by “have played an increasingly important role”
Line 20:
“in this region”: do you mean in the lower stratosphere or in New Zealand? Please specify.
Line 23:
“and the radiation budget” : please replace by “and in the radiation budget”
Line 39:
“over the period 2000-2020, but such observed trends are”. Please replace by “over the period 2000-2020. Such observed trends are”
Line 42-45:
please make 2 sentences.
Line 48:
“attempts of attribution using the models”: Please replace by “any attempt at attribution using models”
“well-positioned » please replace by “well-suited”
Line 58:
Please remove the ) after NDACC
Line 60:
“Any heterogeneities the data have”: please replace by. “Any heterogeneities in the dataset »
Line 64:
Make a separate sentence with “although we only take the data from January 1987 to December 2020 for analysis here.”
Line 79:
« In the next section », please replace by “In section 2”
Line79-80:
“section”, “Sect.”: please make it uniform in the manuscript.
Line 98:
“For example » : please replace by “For instance”
“is needed for the change of sensing solution because there was a 2-year period when the EnSci ECCs started to be used, but with the 1% solution, rather than the 0.5% solution which has become the recommendation for the EnSci ECCs.” is difficult to read. You could replace that sentence with:
“a transfer function is applied to the data after the change in sensing solution type from 1% to 0.5% KI following the O3S-DQA recommendation”
Btw, is the transfer function applied to the data before or after the change of sensing solution?
Line 100:
“re-process » : please replace by “re-processing”
Line 105-110:
this paragraph does not read well and need to be rephrased.
For instance: “Both homogenised and the uncorrected datasets have been post-processed for trend calculations and the regression analysis (in the case of homogenised data).”
Shoud be: “Both homogenised and uncorrected datasets have been post-processed for trend calculations and for regression analysis.”
What do you mean by “(in the case of homogenised data)”?
Line 118:
Please add 3 “the”: The tropopause height (HTTrop), the surface relative humidity (RHsurf ), the aerosol optical depth (AOD),
Line 123:
Please replace “WMO (1957)” by “(WMO, 1957)”
Line 124:
“using the co-measured temperature data of each ozonesonde flight”: please replace by “from the temperature measured by the radiosonde during each ozonesonde flight”
Line 129:
“normalized to vanishing means and unit standard deviation”: do you mean “standardized”?
Line 152-154:
not clear, I can see redundant information in the same sentence. Please rephrase.
Do you simply mean that the impact of CO2 equals the impact of combined GHGs minus the impact of CH4 and NO2?
Line 160:
I would use “as measured” instead of “uncorrected” (first mention is on Line 93).
Line 162:
“in the vertical to 1 km grid using piecewise linear regression for each profile.”: please replace by: “to a 1 km vertical grid using piecewise linear regression.”
Line 167:
“The effect of changes to the concentration of the KI solution on the conversion efficiency”
Please replace by:
“The effect of the changes of the KI solution concentration on the conversion efficiency”
Line 169:
Please move “The correction procedure and the impact of each correction are described in more detail in Appendix A.” before the reference to Figure A1(3).
Line172:
“outliers where ozone is outside the 3 standard deviation”: please replace by: “outliers defined for ozone being outside the 3 standard deviation”
Line 202:
“Figure (5)”: please replace by “Figure 5"
Line 205:
“Some slight differences between seasons are below 5 km”: please add “visible” after “are”
Line 219-220:
“, with R2 ranging from 0.27 to 0.49 in the troposphere and 0.50 to 0.73 in the stratosphere, implying that the stratospheric ozone variations and trends are better explained by the MLR model than tropospheric features.”, I would say:
“. With R2 values ranging from 0.27 to 0.49 in the troposphere and 0.50 to 0.73 in the stratosphere, the stratospheric ozone variations and trends are better explained by the MLR model than tropospheric features”
Line 228:
Please remove “the” before “cooling”
Line 229:
Please remove “the” before “reanalysis”
Line 234:
“at the 9-12 km layer”: please replace by: “in the 9-12 km layer”
Line 235:
“at the 12-15 km layer”: please replace by: “in the 12-15 km layer”
Line 238:
Replace “at a correlation” by “with a correlation”
Line 245:
Replace the 2nd “together with” by “and”
Line 256:
“This trend transition follows the evolution of EESC, which after a peak in 1997 has been declining since then (Figure 9), and indicates the stratospheric impact on the tropospheric ozone through stratosphere-to-troposphere transport reflecting the effect of stratospheric ozone depletion and recovery.” Too much information for one single sentence. What is “reflecting the effect of stratospheric ozone depletion and recovery “ the EESC evolution or the stratospheric impact or the transport…? Please rephrase.
Line 258:
Please assign a letter or a number to the panels and refer to them as Figure 9 (a) and so on.
Line 260 and 265:
Please replace “for example” by “for instance“
Line 271:
Please remove “single” between “individual” and “forcings”.
Line 312:
Please replace “height-resolved” by “vertically resolved”
Line 320:
“In both these altitude regions the trends are substantially stronger than trends in the uncorrected data which are largely insignificant ». Trends “in” altitude regions are compared to trends “in” uncorrected data. Please rephrase.
Line 326:
Please replace “altitudes” by “pressure levels”
Line 333:
Please add “the” before “tropospheric height”
Line 340:
Please add “on” before “the zonal mean ozone profiles”
Line 343:
“and increases in CO2 which lead”: please replace by “and to increase in CO2 which leads”
Line 356:
Please remove “The Stratospheric Aerosol Optical Depth data was obtained from the NASA Langley Research Center Atmospheric Science Data Center (https://asdc.larc.nasa.gov/)” from the Acknowledgements but add “(https://asdc.larc.nasa.gov/) in Table 2.
Line 291:
Please replace “attained” by “reached”
Line 302:
Please replace “are negative” by “is negative” and “which maximises” by “and maximises”
Figure 2:
please replace “3 box-car” by “3-month boxcar”
Figure 5:
“ozone sonde” or “ozonesonde” : please make it uniform throughout the paper
Figure 10:
Please label the panels with a, b and c. and refer to this in the text.
Figure B2:
Place label the panels with a, b and c. and refer to this in the text.
Lines 407,409,492
Please check Bodecker et al., Boyd et al. and Seidel et al. for doi
Lines 539-545:
WMO 2011,2014,2018 and 2022: citations are not complete
Citation: https://doi.org/10.5194/egusphere-2023-2534-RC2 -
AC1: 'Comment on egusphere-2023-2534', Guang Zeng, 28 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2534/egusphere-2023-2534-AC1-supplement.pdf
-
AC1: 'Comment on egusphere-2023-2534', Guang Zeng, 28 Feb 2024
-
AC1: 'Comment on egusphere-2023-2534', Guang Zeng, 28 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2534/egusphere-2023-2534-AC1-supplement.pdf
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2534', Anonymous Referee #1, 21 Dec 2023
The paper describes an updated / homogenized ozone sonde dataset from the Lauder station in New Zealand. It shows results from a trend analysis of the updated sonde dataset, as well as trend results for the region around New Zealand from simulations by the NIWA-UKCA chemistry climate model. The Southern Hemisphere is a region with quite sparse ground-based observations, so Lauder is a very important station and the presented material is, in principle, well suited for publication in ACP.
Overall the manuscript is generally well written, the Figures are clear, and the use of English is generally good.
Nevertheless, I have a number of questions and suggestions that should be addressed before the paper can be accepted for ACP.In section 2.2 the authors introduce their multiple linear regression model, which includes a single linear trend and EESC as proxies for long-term variations / trends. However, in section 3.2 / Figs. 3 and 4, they show linear trends for two periods (1987 to 1999 and 2000 to 2020). It seems that these trends were calculated with just one linear trend proxy in a simple linear regression, without additional proxies. Is that correct? If so, that needs to be stated very clearly, and the sequence of the (sub)-sections should maybe be reordered. In the current text, this has confused me, and will probably confuse most readers. Alternatively, two linear trends (one over each period), or a trend and change of trend (hockey stick), could be used in the multiple linear regression to be consistent throughout the paper.
The same kind of question applies to the trends for the CCMI simulations. Were these obtained with just a simple linear trend, over the two different periods, or with the full multiple linear regression model?
In line 134, the authors state that "Observations as well as basis functions are smoothed using a 12-month boxcar filter". This is not a usual approach and could affect the derived uncertainties quite significantly, because essentially the number of independent data points is reduced by a factor of 10. How is this accounted for in the uncertainties? How are the uncertainties derived in the first place? How do the authors account / correct for auto-correlation in the residuals (ε (t) in their Eq. 1)? How would the results and uncertainties look without doing this 12-month boxcar? I would assume that much lower values for R2 would be found compared to the quite high values in Table B1. My suggestion is to not use this 12-month boxcar, and go with the standard approach using monthly means without smoothing. In any case, these questions need more analysis and more discussion in the manuscript.
Figures 3 and 4: I suggest to combine both figures and plot the trends before and after homogenization in the same plot, in different colors. As the figure stands now, it is very difficult to see how the homogenization changed the trends (very little after 2000?).
Figure 11: This is a very good and interesting Figure. However, I sorely miss the observed trends here. Please include those in the two panels. How do the vertical profiles of regressed EESC and GHG / overall linear trends from the multiple linear regression of the observed data look like in the two different time periods? How does that compare to the corresponding simulated trends (orange lines, red lines in the Figure)? How do the overall trends compare (to the black lines in the Figure).
Line 27: Here you write that SH stratospheric ozone trends are dominated (controlled largely might be a better expression) by Antarctic ozone depletion (which is large and significant and controlled by ODS). Yet in your regression you find hardly any significant impact of ODS / EESC (e.g. in Fig. 8). This is quite a big and important discrepancy. Yet in the later parts of the manuscript, in the conclusions and in the abstract, this discrepancy is hardly mentioned at all, let alone resolved. You do mention negative stratospheric ozone trends, but those seem to be from simple linear regression, not multiple linear regression. So what's going wrong / different with the multiple linear regression? Is the right approach used? I think this needs to be cleared / understood.
Line 67ff: Please indicate if these differences are resolved now. It seems that the Lauder sondes trends don't change very much by the homogenization (Fig. 3, post 2000 trends are almost the same, pre 2000 trends have become more negative). So I would assume that your paper does not change the Godin et al. results, except for the Lauder FTIR data trends which have changed and now fit with the sonde trends. I find this question important, and I would like to see answers, both already here, and also later in the paper, e.g. in the conclusions.
Line 130: What happens when tropopause height is not detrended? I think this should be tried and discussed. If non-detrended tropopause height picks up a GHG induced climate-change related ozone trend, that might be the correct way to do the trend analysis. One could argue that the mechanism underlying short term changes of tropopause height and ozone changes also acts on the long time-scale, because climate-change statistically favors high tropopause conditions. So the acting processes could be the same. There may not be the need for a different process acting on the long time scales. E.g. in the annual cycle you also have a close correlation (on a longer time scale) between high tropopause height and low ozone in the lower stratosphere. While this is somewhat different from the short time-scale processes due to high and low pressure systems, both time scales give similar correlation of high tropopause with low ozone.
Line 146: Please explain what forcings are included in RefC2. I assume all forcings.
Lines 147, 150: I think this should be "corresponding fixed forcing simulation" not "corresponding single forcing simulation".
Around Lines 186, 197: I think these different trends need to be explained / need a bit more dicussion. If Godin et al. get different trends from the same data, there must be an explanation. What happens if you try Godin et al.s regression? Could the difference come from excluding a few extreme soundings? Generally, differences from previous findings need more explanation. They should not just be mentioned and then ignored. I have also done multiple linear regression (with hockey-stick trends) on HEGIFTOM data from Lauder and find trends very similar to your trends in Fig. 4. This is reassuring. Maybe the Godin at al. Lauder trends were too negative? Anyways, I think this needs a bit more discussion, and maybe should be mentioned in conclusions and abstract as well. Of course it needs to be worded appropriately.
Line 237: As mentioned above, this really begs the question what happens with non detrended tropopause height in the regression.
Line 242: I am not sure about this linear trend term. It seems like a very unspecific overall collector of various things, picking up a confusing mix of ODS-related, GHG-related and other changes. I think there should be better proxies, e.g. hockey stick, two linear trends, ... For me, the insignificant ozone changes picked up by the EESC term are a warning sign. If, according to the regression, ODS changes had no impact on stratospheric ozone at Lauder, I very much wonder if we can trust this regression.
Around line 258: This is a weird argument. We see EESC effects in the troposphere, but we don't see them in the stratosphere, where they should be coming from? Can that be resolved? Might different proxies in the regression help (as suggested above)?
Around line 320 (and in several other places). Are these simple single linear trends? Or are they from the multiple linear regression? This keeps me confused throughout the manuscript. I would much prefer to just have one type of regression, or a much better explanation of what was done, and why there might have been two approaches.
After line 339: As also mentioned above for Fig. 11, I think there needs to be more discussion about how the observed and simulated trends fit together, or do not fit together.
Citation: https://doi.org/10.5194/egusphere-2023-2534-RC1 -
AC1: 'Comment on egusphere-2023-2534', Guang Zeng, 28 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2534/egusphere-2023-2534-AC1-supplement.pdf
-
AC1: 'Comment on egusphere-2023-2534', Guang Zeng, 28 Feb 2024
-
RC2: 'Comment on egusphere-2023-2534', Anonymous Referee #2, 26 Dec 2023
In the manuscript “Analysis of a newly homogenised ozonesonde dataset from Lauder, New Zealand”, Guang Zeng et al. derive long-term MLR trend estimations from the ozone sonde dataset of Lauder. The dataset, recently homogenised in the frame of HEGIFTOM, shows negative pre-1999 trends in better agreement with the previous literature. Post-2000 trends are shown to be significantly negative in the stratosphere and positive in the troposphere, and are in very good agreement with trends derived from a co-located FTIR instrument. The analysis by MLR imputes the negative post-2000 stratospheric trend to anthropogenic forcing led by CO2 related to positive trends in tropopause height and tropospheric temperature and negative trend in stratospheric temperature. CCM simulations from NIWA-UKCA attribute the negative pre-1999 trends not only to the ODS increase but also to a GHG increase with opposing impacts of CH4 and CO2. For the post-2000 period, the CCM analysis assess the role of the dynamical changes driven by CO2 on the negative lower stratospheric trends.
The manuscript fits well within the scope of ACP and is of high scientific quality. It is generally well written despite some very long sentences which do not read well. The results are well presented. The homogenization of the ozone sonde dataset succeeded in improving the pre-1999 trend agreement with other observation techniques and with the literature.
The detailed analysis of the MLR results is a very good contribution to the understanding of the underlying issues of stratospheric and tropospheric pre- and post-2000 trends. The results derived from CCM analysis and derived from MLR on observational dataset enhance the role of CO2-driven dynamical changes in the lower stratospheric trend which represents a significant step towards understanding trends in this atmospheric region.
I list below general and specific comments. General comments are questions and remarks which need clarifications before the paper can be accepted for ACP. Specific comments are minor revisions which may help to improve the readability of the manuscript.
General Comments
Line 64-74:
The same ozone sonde dataset is used in Godin-Beckmann et al. 2022 and in the present study. The authors say the trend values of Godin-Beckmann et al. to be “exceedingly large” (line 69). However, the present study trends are within the Godin-Beckmann et al. uncertainties and respectively, except for the 25km value. The authors should comment on these differences.
Why are the uncertainties of both studies so different? Is the residuals autocorrelation taken into account in the present study? What is the impact of using EESC as an additional explanatory variable on the trend values of the present study?
Line 74:
“which has been updated from the dataset used in Godin-Beckmann et al. (2022)”. Could you please comment already here (instead of Line 194) on the FTIR dataset update as the trends reported in Godin-Beckmann et al. are very different from the present study?
Line 77:
“into the near future”: I cannot not see any mention of post-2022 results in the manuscript. What do you mean?
Line 89:
Table1 indicates only 3 dual flights to evaluate the effects of the sensing solution change. Is the transfer function/correction factor derived from these 3 dual flights or is a general transfer function used?
Line 122:
« Surface humidity are measured » : please replace by: « Surface humidity is measured »
Is this RH?
“Surface humidity are measured by the radiosonde that has a humidity sensor.”: please be more specific.
Line 126:
Why is QBO10 used for the whole altitude range? Would QBO50 have been more appropriate in the troposphere?
Line 130:
"Fig 6." is the first mention of a figure in the manuscript, in that case, this should be Fig 1.
I suggest to remove this figure or to move it to Appendix B.
Line 140:
“coarse resolution » : do you mean spatial or temporal resolution? Please be more specific.
Line 177:
“marked differences”: the pre-1999 trends values are different but similar within their uncertainties. The marked difference lies in the significance of the trends. This should be made clear. Furthermore, the differences would be clearer if both trend profiles (uncorrected and homogenised) were represented on a single figure.
Line 184:
“Figure 3 and 4”: I suggest to merge the information into one single figure with 2 panels (pre-1999 and post-2000) to make comparisons easier.
Line 187:
same comment as for Lines 64-74. Please explain why the trends values are different in both studies. The role of the additional proxies used is non negligeable. For instance, the significative contributions of the EESC or HTtropo proxies (Table B1) are influencing the remaining linear trend value when compared to a MLR not considering these proxies.
Line 195:
“updated version”: See comment for Line 74.
Line 197:
The agreement between ozone sonde and FTIR trends is good however trends of the present study stay negative and significant above 15 km while WMO 2022 shows positive but non significant trends. A reason or a tentative explanation should be given.
Line 200-207:
At that point, I cannot find a reason for these considerations on the seasonal variation.
I suggest to remove this section unless it is used in the explanation of the trends estimation.
Line 213:
As said in comment of Line 130, Figure 6 could be removed. If kept and moved in Appendix B, please replace “de-trened” by “de-trended” and “12-boxcar” by “12-month boxcar”.
Line 210-217 and Figure 7 8,9 and table B1:
Separate and redundant information is spread over 3 figures and 1 table, I find this difficult to handle. I would suggest the following simplification:
Information on Fig 7 is given in Fig 9 except for the R2 value and the regressed timeseries. You could remove Fig 7 and keep only Fig 9 with the observed timeseries in dashed, the regressed timeseries in black and with the R2 information in the respective subplot titles.
Figure 8 and table B1:
The information is redundant and Figure 8 could be replaced by a highlighting of the major contributor(s) in table B1. Table B1 should then be part of the manuscript and not of the Appendix B.
Line 223:
“The downward trend in the stratospheric ozone is clearly explained by the significant negative linear trend that represents all quasi-linear, monotonic drivers of change”
Please add which drivers and that you will discuss these below.
Why can these drivers not be used as proxies but have to be treated apart?
Why is the contribution of EESC negligeable here?
Can you exclude this negative trend to be due (evt partly) to a drift or step in the timeseries which has not been considered or corrected by the homogenisation?
Line 231:
How is this trend estimated?
Meng et al. trend values are estimated “on the natural variability–removed time series” (volcanic eruption, ENSO and QBO). Are you estimating the trends with a similar model as Meng et al. or with your eq.1? Please comment.
Line 236-7:
“This indicates that the negative contribution to the ozone trend in the lower stratosphere (between ∼9 to 15 km) can largely be projected on the significant increase in tropopause height»
Figure 9 shows that de-trended HTtropo is a significant proxy for explaining the ozone variation between 9 and 15 km height. Instead of making considerations about the correlation between de-trend and non de-trended HTtropo and ozone, would it be possible to use non-detrended HTtropo as a proxy? If not, why?
Line 267:
“The regression function we construct here is more suitable to explain the stratospheric ozone changes.” Does it mean that the trend values estimated below 6 km are not reliable?
Line 269-270:
How are the modelled ozone trends estimated? What do you mean by “separately”? If you apply a ILT multi-linear regression on simulated ozone values, please describe the MLR used in that case. If the model directly outputs the trends, please clarify.
Line 272:
Please specify here that/if the attribution of the changes in modelled O3 is done as in Zeng et al 2022 and Morgenstern et al. 2018.
Line 275:
“ are broadly in agreement with the Lauder observations (Figure 4).” I suggest to add the trend values estimated on the ozone sonde dataset in Figure 11.
Line 318:
“with a maximum of −9% decade−1 around 13 km“: Figure 4 shows a maximum of -12%/dec at 13km.
“with significant trends at the 95% confidence above 12 and below 5 km. » Please adjust to values shown on Figure 4.
Specific Comments
Line 5-7:
The sentence doesn’t read well. It’s too long and contains 3 brackets. Please rephrase.
Line 10-13:
Please, make 2 sentences from this one.
Line 19:
“…but clearly shows …” : Please replace by “… , it clearly shows …”
Line 20:
“have had an increasingly important role”: please replace by “have played an increasingly important role”
Line 20:
“in this region”: do you mean in the lower stratosphere or in New Zealand? Please specify.
Line 23:
“and the radiation budget” : please replace by “and in the radiation budget”
Line 39:
“over the period 2000-2020, but such observed trends are”. Please replace by “over the period 2000-2020. Such observed trends are”
Line 42-45:
please make 2 sentences.
Line 48:
“attempts of attribution using the models”: Please replace by “any attempt at attribution using models”
“well-positioned » please replace by “well-suited”
Line 58:
Please remove the ) after NDACC
Line 60:
“Any heterogeneities the data have”: please replace by. “Any heterogeneities in the dataset »
Line 64:
Make a separate sentence with “although we only take the data from January 1987 to December 2020 for analysis here.”
Line 79:
« In the next section », please replace by “In section 2”
Line79-80:
“section”, “Sect.”: please make it uniform in the manuscript.
Line 98:
“For example » : please replace by “For instance”
“is needed for the change of sensing solution because there was a 2-year period when the EnSci ECCs started to be used, but with the 1% solution, rather than the 0.5% solution which has become the recommendation for the EnSci ECCs.” is difficult to read. You could replace that sentence with:
“a transfer function is applied to the data after the change in sensing solution type from 1% to 0.5% KI following the O3S-DQA recommendation”
Btw, is the transfer function applied to the data before or after the change of sensing solution?
Line 100:
“re-process » : please replace by “re-processing”
Line 105-110:
this paragraph does not read well and need to be rephrased.
For instance: “Both homogenised and the uncorrected datasets have been post-processed for trend calculations and the regression analysis (in the case of homogenised data).”
Shoud be: “Both homogenised and uncorrected datasets have been post-processed for trend calculations and for regression analysis.”
What do you mean by “(in the case of homogenised data)”?
Line 118:
Please add 3 “the”: The tropopause height (HTTrop), the surface relative humidity (RHsurf ), the aerosol optical depth (AOD),
Line 123:
Please replace “WMO (1957)” by “(WMO, 1957)”
Line 124:
“using the co-measured temperature data of each ozonesonde flight”: please replace by “from the temperature measured by the radiosonde during each ozonesonde flight”
Line 129:
“normalized to vanishing means and unit standard deviation”: do you mean “standardized”?
Line 152-154:
not clear, I can see redundant information in the same sentence. Please rephrase.
Do you simply mean that the impact of CO2 equals the impact of combined GHGs minus the impact of CH4 and NO2?
Line 160:
I would use “as measured” instead of “uncorrected” (first mention is on Line 93).
Line 162:
“in the vertical to 1 km grid using piecewise linear regression for each profile.”: please replace by: “to a 1 km vertical grid using piecewise linear regression.”
Line 167:
“The effect of changes to the concentration of the KI solution on the conversion efficiency”
Please replace by:
“The effect of the changes of the KI solution concentration on the conversion efficiency”
Line 169:
Please move “The correction procedure and the impact of each correction are described in more detail in Appendix A.” before the reference to Figure A1(3).
Line172:
“outliers where ozone is outside the 3 standard deviation”: please replace by: “outliers defined for ozone being outside the 3 standard deviation”
Line 202:
“Figure (5)”: please replace by “Figure 5"
Line 205:
“Some slight differences between seasons are below 5 km”: please add “visible” after “are”
Line 219-220:
“, with R2 ranging from 0.27 to 0.49 in the troposphere and 0.50 to 0.73 in the stratosphere, implying that the stratospheric ozone variations and trends are better explained by the MLR model than tropospheric features.”, I would say:
“. With R2 values ranging from 0.27 to 0.49 in the troposphere and 0.50 to 0.73 in the stratosphere, the stratospheric ozone variations and trends are better explained by the MLR model than tropospheric features”
Line 228:
Please remove “the” before “cooling”
Line 229:
Please remove “the” before “reanalysis”
Line 234:
“at the 9-12 km layer”: please replace by: “in the 9-12 km layer”
Line 235:
“at the 12-15 km layer”: please replace by: “in the 12-15 km layer”
Line 238:
Replace “at a correlation” by “with a correlation”
Line 245:
Replace the 2nd “together with” by “and”
Line 256:
“This trend transition follows the evolution of EESC, which after a peak in 1997 has been declining since then (Figure 9), and indicates the stratospheric impact on the tropospheric ozone through stratosphere-to-troposphere transport reflecting the effect of stratospheric ozone depletion and recovery.” Too much information for one single sentence. What is “reflecting the effect of stratospheric ozone depletion and recovery “ the EESC evolution or the stratospheric impact or the transport…? Please rephrase.
Line 258:
Please assign a letter or a number to the panels and refer to them as Figure 9 (a) and so on.
Line 260 and 265:
Please replace “for example” by “for instance“
Line 271:
Please remove “single” between “individual” and “forcings”.
Line 312:
Please replace “height-resolved” by “vertically resolved”
Line 320:
“In both these altitude regions the trends are substantially stronger than trends in the uncorrected data which are largely insignificant ». Trends “in” altitude regions are compared to trends “in” uncorrected data. Please rephrase.
Line 326:
Please replace “altitudes” by “pressure levels”
Line 333:
Please add “the” before “tropospheric height”
Line 340:
Please add “on” before “the zonal mean ozone profiles”
Line 343:
“and increases in CO2 which lead”: please replace by “and to increase in CO2 which leads”
Line 356:
Please remove “The Stratospheric Aerosol Optical Depth data was obtained from the NASA Langley Research Center Atmospheric Science Data Center (https://asdc.larc.nasa.gov/)” from the Acknowledgements but add “(https://asdc.larc.nasa.gov/) in Table 2.
Line 291:
Please replace “attained” by “reached”
Line 302:
Please replace “are negative” by “is negative” and “which maximises” by “and maximises”
Figure 2:
please replace “3 box-car” by “3-month boxcar”
Figure 5:
“ozone sonde” or “ozonesonde” : please make it uniform throughout the paper
Figure 10:
Please label the panels with a, b and c. and refer to this in the text.
Figure B2:
Place label the panels with a, b and c. and refer to this in the text.
Lines 407,409,492
Please check Bodecker et al., Boyd et al. and Seidel et al. for doi
Lines 539-545:
WMO 2011,2014,2018 and 2022: citations are not complete
Citation: https://doi.org/10.5194/egusphere-2023-2534-RC2 -
AC1: 'Comment on egusphere-2023-2534', Guang Zeng, 28 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2534/egusphere-2023-2534-AC1-supplement.pdf
-
AC1: 'Comment on egusphere-2023-2534', Guang Zeng, 28 Feb 2024
-
AC1: 'Comment on egusphere-2023-2534', Guang Zeng, 28 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2534/egusphere-2023-2534-AC1-supplement.pdf
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 261 | 99 | 27 | 387 | 17 | 17 |
- HTML: 261
- PDF: 99
- XML: 27
- Total: 387
- BibTeX: 17
- EndNote: 17
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Richard Querel
Hisako Shiona
Deniz Poyraz
Roeland Van Malderen
Alex Geddes
Penny Smale
Dan Smale
John Robinson
Olaf Morgenstern
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(13919 KB) - Metadata XML