|Review of revised manuscript Spatial regression analysis on 32 years of total column ozone data by J.S. Knibbe et al. |
The authors addressed many of my comments from my first review, and have made improvements to the manuscript. There is a lot of interesting work here, and though I still worry some of the results are over-interpreted, the changes made were useful. The summary section is much improved as well. I find I still have some questions, and the size and complexity of the manuscript is significant. This paper could be separated into two manuscripts, one describing the PHYS vs. STAT model with more details on the seasonal representation, and the second analysis of the EESC and PWLT fits, but I will not pursue this as I did not suggest it in my first review.
In the figures showing the reconstructed time series at three locations, it would be very helpful to include an additional line in the PHYS results showing the term DAY+EP+PV+GEO so that may be compared directly to the “FOURIER” term in the stat model. It is difficult to do this by eye in the figures. Since one of the highlights is the ability to represent the seasonal cycle with these explanatory variables, such a line plot would be useful and a nice complement to the r^2 comparison. This also cannot be garnered from the plots in Figure 5 because these plots are in different units.
Additionally, it would be useful to note in the appropriate figure captions that the units of “coefficient” in each plot are DU per unit change of the explanatory variable.
In Figure 2, is it possible to plot the absolute value of the correlation. Whether it is positive or negative has no bearing, and the sharp discontinuities at the equator make it difficult to read.
Regarding the explanatory variably DAY, do the authors have any reference to model results that would clarify what we expect to see in terms of transport vs. in situ production? I do understand now that the tropical signal should be zero, because the production – while large – does not vary because the length of day does not vary. At higher latitudes, the larger difference between summer and winter length of day is countered by the increasing solar zenith angle. I agree there is production, but I would not expect it to dominate transport. Also I would expect the results to be hemispherically symmetrical. The SH results make more sense to me, but the authors suggest that in the south the DAY variable is partially accounted for in the PV variable, I think implying that the SH signal should be expected to be higher. I admit to being confused, and wonder if a modeling study can help clarify.
I believe there has only been one spring where something close to an ozone hole developed in the northern hemisphere. The Arctic ozone hole is referred to several times, but such episodes are rare for the reasons the authors state (stability of the vortex). There are a couple incidences where I suggest changing the wording.
For example, L230-232: The concept of ozone hole and ozone accumulation at high latitudes seem to be conflated here. The stronger Brewer-Dobson circulation in winter leads to the build-up of ozone due to net transport not isolation of the vortex. I believe they are two separate concepts, but the EP flux represents both the stability of the vortex and the Brewer-Dobson circulation in the analysis. This should be clarified.
Minor/Technical suggested comments:
L32-33: We observe ozone increases with increasing potential vorticity and day length, ozone decreases with increasing geopotential height, and variable ozone effects …
L47: ozone was later identified
L90: and isolates
L92: correlate with stratospheric ozone
L105: remove dimensions
L119: EESC or piecewise linear
L187: accuracy of a few percent
L210: differences in the phase
L211: We considered adding a proxy to represent the QBO at 50 hPa but rejected this
L237: note here that EP data are from NCEP Reanalysis
L243:244: do you mean mixing ratio here, or mixing between the troposphere and stratosphere?
L277: whereas at high latitudes the high correlations complicate interpretation of the regression results.
L313: aligns with
L329-330: we do not define an alternative variable for the Arctic polar region
L332: al.,2011) such that a well defined Arctic ozone hole is rare. [it is not that we can't detect it, it is that it does not typically exist]
L383: resulting EESC time series
L384: trend analyses
L401 remove occurrence of
L409: barely significant
L429: should be taken into account when interpreting these results
L430: the effects of DAY and PV interact because these variables are strongly correlated.
L436: these effects also interact with the effects of DAY and EP
L460: remove EESC ?
L556: the larger amount of ozone depletion within the Antarctic ozone hole in the September to Novemeber period.
Line 567: This results
L610: add period.
L621: transport driven by the Brewer-Dobson circulation
L626: effect of wave dynamics leading to less stability of the polar vortex
L657: to determine which parameterization
L664-668: These sentences seem repetitive. What about 'this may result from ozone not responding linearly to the EESC in such a way that the fit to the 3 year air age EESC is better, even if the true age of air is higher.'
L705: Our results