|Review of revised manuscript by Peter Tunved and Johan Stroem for Atmos. Chem. Phys., "On the seasonal variation of observed size distributions in Northern Europe and their changes with decreasing anthropogenic emissions in Europe: climatology and trend analysis based on 17 years data from Aspvreten, Sweden",|
This manuscript presents an analysis of 17 years of near-surface aerosol particle size distribution measurements at the Aspreveten regional background monitoring site in Central/Southern Sweden. The topic of the research is of particular interest to understand how the tropoapheric aerosol layer has responded to reductions in emissions that have occurred in the 1980s and 1990s, re: subsequent climate influences from aerosol-radiation-interaction and aerosol-cloud-interaction radiative effects.
The authors have been through the reviewer comments I made in my review of the ACP-D manuscript in July, and have made some improvements to the manuscript (such as adding in the log-log size distribution inset plots to Figure 2) and have revised some parts of the text in response to my review.
I found major revisions were required (but clarified they were relatively straightforward to address), and the reviewers have replied to these comments, organising their response into "6 major replies".
I am generally satisfied with the replies to these, in particular re: point 5 of the 6 replies, I accept that the authors prefer to retain this within the revised manuscript despite the simplified approach.
However, on the 2nd of the replies, I feel the authors have not replied sufficiently to the main point I was making there, responding to a different point in their reply, perhaps because they have not quite followed the constructive suggestion I was making.
In this review of the revised manuscript I am providing a short list of minor revisions that are still required, which almost exclusively relate to the points I was making re: the summary statistics of the log-normal fits to the hourly measurements (shown in Table 1), which give a substantially different perspective on the measurements than the seasonal average size distributions from the daily-mean measurements (shown in Figure 2).
My comment in the original review was intended to request the authors explain to readers of the paper (by adding an extra sentence in the results and conclusions sections) that a seasonal-average approach over daily-mean measurements (as in Figure 2) misses important variations in the size distribution that occur through the day, which the cluster analysis approach applied to derive the size distributions in Figure 3, and the size-resolved fit statistics in Tables 1, much more effectively retain.
This can be seen for example in the fact that spring-time nucleation and Aitken mode peaks are clearly identified in Table 1 (with there being 618 per cc and 581 per cc in the nuclei and Aitken modes at 33nm and 67nm respectively), whereas in Figure 2 there is apparently no evidence of these two peaks, presumably because the growth through the day, as seen for example within the so-called ''banana-plots'' (e.g. as seen for example at Hyytiala in Kulmala et al., 2004), is averaged into a broader single mode in the daily-mean analysis.
An alternative way to explain this, could be that those summary statistics in Table 1 are identifying the earlier morning peaks as nuclei mode, but the later peaks as Aitken mode, with the two modes simply being younger and older parts of the plume within the same growth curve in the banana-plots, the later particles having had more time to grow to sizes which are then identified as Aitken-sized particles rather than at nuclei-mode sizes. When constructing the daily-mean, one only gets the average over these different points within the growth curve -- within these "new particle formation and growth" events.
To my knowledge, this point about the choice of averaging time affecting the interpretation of size distribution measurements has not been made before in the literature. Although it is to-some-extent obvious when one thinks about what information the daily-mean and hourly-mean retains (and what information it doesn't), there is a clear relevance to this paper, in that it illustrates why the cluster analysis approach applied is really a very effective way to enable to retain the hourly variations in size distribution within a long-term analysis/assessment of aerosol particle trends. By contrast, assessing trends based on daily-mean size distribution information will not be as able to identify such intra-daily-timescale variations, such as nucleation events.
In summary, I'm recommending to accept subject to minor revisions, with the list below almost exclusively related to clarifying the reply the authors make in their "major reply 2".
Once the authors have made these revisions, or convince me they do not need to be made, I confirm I am recommending the revised manuscript will then be suitable for publication in the main Atmospheric Chemistry and Physics journal.
Minor revisions required
1) Introduction, line 9, page 2
The authors now cite correctly the radiative forcing chapter of the IPCC 5th assessment report, but have forgotten to add the citation to Myhre et al. (2013) in the list of references.
2) Introduction, page 2, line 15
Typo, please change "later transform into" to "later transforms into"
3) Caption to Figure 1, page 28
The phrase "day-average" needs to clarified -- you mean "daily-mean" I assume (the word "average" is potentially ambiguous re: other central-measures, e.g. daily-median)?
4) Additional inset Figures to the Caption to Figure 2, page 29
The authors have added the log-log plot, but in the small inset diagram, which is a good way to enable readers to cross-check between the two. However, the axes of the inset Figure have only major tick-marks, not including minor tick-marks. For the x-axis in particular this makes it more difficult for the reader, with there only being 1 tick mark apparent (for the 100 nm particle diameter). For the x-axis please add the 10nm major-tick (and label as in the main Figure-insets) adding also the minor tick-marks on each axis.
5) Caption to Figure 2, page 29
The caption to Figure 2 states "...small graphs give size distribution mean diameter (nm) and average total number concentration...". That's a very good idea to add that info to the Figure, but currently that is not shown. The caption text needs to be clear what "size distribution mean diameter" metric is meant -- I assume the geometric-mean diameter (over the full size range) which I expect will be slightly to the right of the Aitken mode peak, reflecting the influence from the sizes of the fewer number of accumulation mode particles.
Please add the extra info to the inset, or (if this obscures the meaning within the small inset Figures) delete that sentence from the caption.
6) Caption to Table 1, page 30
The caption needs to state explicitly that the size-resolved number and size statistics are from fitting to the hourly measurements. That then enables the reader to see the point I am re-iterating in this review re: changes in nucleation requiring to retain intra-daily-timescale variations (which the cluster analysis technique does, but the daily-mean analysis may not or may make
7) Section 3.2, page 9, line 30
Please add a sentence at the start of this section to alert the reader that the cluster analysis will be based on the hourly measurements rather than daily-means, and refer to Kulmala et al. (2004) to note the new particle formation and growth that occurs within the day at Hyytiala and elsewhere. This will help prime the reader to start to think about the different approach here whereby the cluster analysis can be applied to understand aerosol trends retaining this hourly-resolution size distribution information that the measurements are monitoring.
8) Section 5, page 19
Please add a sentence near the start of this section re-iterating this point about the additional value/virtue within the cluster analysis technique, that (even if this is not yet explored in this particular manuscript) the approach potentially enables to identify changes in nucleation within the long-term measurement record that exists at Aspreveten and other sites.
9) Section 6, page 22
The authors should consider adding an extra bullet point to make this more general point (point 7) within the conclusions section. Even though there is no finding specifically linked to this point, the potential methodological progression can also be an important aspect of the study to re-iterate
when concluding. As I alluded to in my earlier review, the potential parallel within the modelling community when storing only monthly-mean tracer information in global models could indicate an additional note re: the importance to consider when multi-decadal model experiments are analysed to seek to understand historical changes in new particle formation.