Comment on acp-2020-823:

The manuscript titled “Atmospheric transmission patterns which 1 promote persistent winter haze over Beijing” attempted to illustrate the relationship between North Atlantic oscillation and persistent haze in winter in Beijing, especially on the intraseasonal time-scale. Although the topic is interesting, many major revision must be revised.

Major comments:

1. Title: Possibly, the title is not accurate. From your title, I cannot read any valuable information about the NAO pattern that you focused on. I strongly suggest you to rephrase your title of this manuscript.
2. Line 17: “This study focuses mainly on the role of the NAO+ pattern, because the NAO index shows a closer relationship with winter haze frequency, especially after 1999”. However, I did not find any physical explanations about this arguments that you emphasized in the Abstract and that seemed to be the reasons why you did this research.
3. Line 18–20: The intraseasonal relationship between North Atlantic oscillation and persistent winter haze in Beijing may be a new finding, although the interannual relationship has been revealed by many previous studies. As you argued, the relationship between NAO and haze is stronger after mid-1990s. Thus, can you composted this relationships using the observations of PM2.5 concentrations? As you known, the PM2.5 concentrations were widely observed since 2014.
4. Data and Methods: The details of composite approach should be introduced. For example, how to determine day -10, day 0 and day 4 of a persistent haze events lasting for more than 5 days? and so on……
5. Line 197–207: What is the differences between the definition of NAO and your LW03 NAO? It should be carefully illustrated and marked on Figures. Furthermore, I cannot understand the large differences of the correlation coefficients when you only change the definition of NAO. If this evident differences existed, the LW03 NAO you defined is still the NAO pattern? or something else?
6. Line 207: I did not agreed that the connection between haze and EA/WR pattern is weaker. The correlation coefficients, listed from Line 205 to Line 210, have little difference. In addition, Line 210–219 cannot stand, as you showed decadal changes of connections between NAO/EAWR and haze.
7. Section 3.2: the discussion about intraseasonal relationship needed rewrite to be more compact and clearer.
8. It might be better to move discussion about the impacts of sea surface temperature and Arctic sea ice Section Discussion, because you did not fully explained associated physical mechanisms.
9. Line 478: the relationship between the subtropical Western Pacific SST and haze over North-Central North China Plain also contributed to the variation in haze of Beijing, as well as the Strengthening Relationship between Eurasian Snow Cover and December Haze Days in Central North China after the Mid-1990s.
10. Most of the Figure must be improved. Especially, the country boundaries should be carefully examined.
11. The language must be improved by native speaker.

Review for Atmospheric Chemistry and Physics manuscript acp-2020-823, “Atmospheric transmission patterns which promote persistent winter haze over Beijing,” by Li et al.

This paper analyzes the local and large-scale meteorological conditions that are conducive to haze conditions in Beijing, China. The paper shows that there are physical links between the North Atlantic Oscillation and haze in Beijing, mediated by atmospheric wave patterns that also interact with sea surface temperatures globally and sea ice concentrations in the Arctic.

The paper is written well, the results are clearly presented, and the physical mechanisms proposed are plausible and reasonable. In general, I am supportive of the paper being published. However, I think some revisions are needed in order to justify publication in ACP.

Major comments:

**Novelty**

My main concern with the manuscript as written is the novelty. The manuscript presents an extensive literature review (lines 40-99) that nicely lays out what we know about haze in Beijing and the meteorology that affects it, but I come away from this section wondering: if we know so much already, what is the contribution of this particular manuscript? The authors say they’re interested in the predictability of Beijing haze via the NAO (lines 103-107), but if there are already studies that have illustrated the linkages between the NAO and Beijing haze (lines 59-62), what is new in this study?

My best sense of the novelty of this paper is the notion of predictability. That is, we know that the NAO affects the circulation patterns that are conducive to haze, but does this knowledge give us meaningful predictive power a certain number of days ahead of time? The authors suggest that this may be true at many points, but they do not actually quantitatively test this proposition. Figure 10 is the closest the authors come to this idea, but it’s the last figure in the paper and it feels like it’s under-developed.

To my mind, the best way to test the idea of predictability here would be to examine the conditional probabilities of haze events given NAO conditions. That is, what is the baseline/average probability of haze and what is the probability of haze given that the NAO is positive (above some threshold), etc.? Does the probability of haze substantially increase given NAO+ conditions relative to the baseline/average probability of haze, and is this increase statistically meaningful? This could be done separately for a set of lead times (10 days, then 9 days, then 8 days, etc.).

The point of this analysis would be to ask: If you are a decisionmaker at day X, and you know nothing about the NAO, what is your baseline expectation for the probability of haze on day X+10? Then, if you do know the state of the NAO, how does your expectation for the probability of haze on day X+10 change? The authors could then test the statistical significance of this change with a bootstrap test, a Monte Carlo simulation, or something similar. Figure 10 is the beginning of this type of analysis, but it’s done on interannual timescales, not daily/weekly, and doesn’t have a robust statistical treatment.

**Organization**

My other major comment pertains to the organization of the manuscript. Given the comments I made above, I think there are parts of section 5 that are not necessary or helpful. The paragraphs relating to figures 8 and 9 are somewhat interesting, but do not really add much to the analysis. For example, I am not really persuaded that “Compared with the evolution of the wave train in Fig. 9a, that in Fig. 9b has a longer wavelength, but it is less organized and less persistent. This suggests that the preceding NAO+ pattern could increase the occurrence…” The difference in wave train organization between NAO+ and NAO- conditions is not quantified and difficult to get a handle on. The analysis pertaining to figure 10 is really where things start to get good, and so I think the authors should foreground that type of analysis much more strongly.

The other organization issue that confuses me is the structure of sections 3, 4, and 5. Section 3.1 is focused on interannual timescales. The authors then transition to daily timescales in section 3.2 and then back to interannual timescales in section 4. Then, section 5 discusses the NAO on daily timescales in figure 8 and 9, and then moves back to interannual timescales for figure 10. This seems odd given that the predictability arguments, which seem to be the most novel parts of the manuscript, are primarily relevant on daily-to-weekly timescales.

I recommend the authors think more deeply about which timescale they would like to focus on, given the parts of the manuscript that are the most novel, and describe more explicitly how the sections of the manuscript connect to one another. Some sections may need to be moved around.

I emphasize that the pieces are present in this paper for a nice contribution to the literature. I believe all of my comments can be addressed if the authors invest time in the revision.

Minor comments:

Lines 12 and 27: When you say “significant,” do you mean “statistically significant” or something like “physically meaningful”? I would generally avoid using the word “significant” outside the narrow context of statistical significance, since it can be confusing. Other words you could use here include “pronounced,” “anomalous,” etc.

Line 37: Define PM2.5 before using the acronym (“particulate matter with a diameter of less than 2.5 microns”)

Line 59: A weakened E. Asian trough and Siberian high likely also suppress horizontal diffusion of pollutants, not just vertical diffusion

Line 66: Should this read “increased relative humidity,” rather than decreased?

Lines 140-151: The authors define haze using relative humidity and visibility. Why not define it using certain thresholds of PM2.5 concentrations, since PM2.5 is the primary health hazard associated with haze? (For example: Callahan and Mankin, 2020, Cai et al., 2017.) If it’s because PM2.5 data is not available for a long time period, I think the authors should at least mention this.

Line 157: I would remove the word “associated” or say “large-scale atmospheric circulations associated with haze”. The word “associated” on its own doesn’t mean much.

Line 162: “Significantly different from zero” at what alpha? 0.05, 0.01, etc.?

Lines 273-292: This paragraph/analysis (corresponding to figure 5) does not feel particularly novel to me. Many, many previous studies have described the local meteorology conducive to haze in Beijing. The presence of the anticyclonic circulation pattern over north China is a key factor that shapes many of these local factors (Callahan and Mankin, 2020, Zhong et al., 2019), and you’ve already identified the teleconnection between the NAO and that anticyclonic circulation (figure 3) so I think this additional analysis and figure 5 could be cut.

Line 370: I think this should read “blue line,” not “green line” (there is no green line in figure 8).

Figure 2: I have a hard time judging correlations from time series plots. It would be nice to see scatter plots with each climate index on the x-axis and haze days on the y-axis. The scatter plots could be placed to the right of each timeseries to make a two-column figure. The authors could also point out specific extreme years mentioned in the text (e.g., 2013) in the scatter plots by coloring the points differently or something like that.

References:

Cai, W., Li, K., Liao, H., Wang, H., & Wu, L. (2017) Weather conditions conducive to Beijing severe haze more frequent under climate change. Nature Climate Change, 10.1038/nclimate3249

Callahan, C.W. & Mankin, J.S. (2020) The influence of internal climate variability on projections of synoptically driven Beijing haze. Geophysical Research Letters, 10.1029/2020GL088548

Zhong, W., Yin, Z., & Wang, H. (2019) The relationship between the anticyclonic anomalies in Northeast Asia and severe haze in the Beijing-Tianjin-Hebei region. Atmospheric Chemistry and Physics, 10.5194/acp-19-5941-2019