Mixing layer height and its implications for air pollution over Beijing, China

Abstract. The mixing layer is an important meteorological factor that affects air pollution. In this study, the atmospheric mixing layer height (MLH) was observed in Beijing from July 2009 to December 2012 using a ceilometer. By comparison with radiosonde data, we found that the ceilometer underestimates the MLH under conditions of neutral stratification caused by strong winds, whereas it overestimates the MLH when sand-dust is crossing. Using meteorological, PM2.5, and PM10 observational data, we screened the observed MLH automatically; the ceilometer observations were fairly consistent with the radiosondes, with a correlation coefficient greater than 0.9. Further analysis indicated that the MLH is low in autumn and winter and high in spring and summer in Beijing. There is a significant correlation between the sensible heat flux and MLH, and the diurnal cycle of the MLH in summer is also affected by the circulation of mountainous plain winds. Using visibility as an index to classify the degree of air pollution, we found that the variation in the sensible heat and buoyancy term in turbulent kinetic energy (TKE) is insignificant when visibility decreases from 10 to 5 km, but the reduction of shear term in TKE is near 70 %. When visibility decreases from 5 to 1 km, the variation of the shear term in TKE is insignificant, but the decrease in the sensible heat and buoyancy term in TKE is approximately 60 %. Although the correlation between the daily variation of the MLH and visibility is very poor, the correlation between them is significantly enhanced when the relative humidity increases beyond 80 %. This indicates that humidity-related physicochemical processes is the primary source of atmospheric particles under heavy pollution and that the dissipation of atmospheric particles mainly depends on the MLH. The presented results of the atmospheric mixing layer provide useful empirical information for improving meteorological and atmospheric chemistry models and the forecasting and warning of air pollution.

The paper has been improved considerably: it is much better structured and more precise than before. The "acp-2015-691-author_response-version2.pdf" demonstrates that nothing remains as it had been. The point-to-point response gives further useful insight, so only very few questions remain to be claried.
So, in spite of the improvements, I recommend to address the points listed below to make the paper more convincing. The rst part of the paper dealing with the retrieval of the MLH is certainly easier to optimize; more critical is the second part, when MLH is linked to other parameters. This is due to the fact that there are several parameters (wind eld [local and regional], radiation, aerosol chemistry including hygroscopic growth, and more) interacting with the ML at local and regional scales. Consequently, the discussion must often remain qualitative as it was not possible to measure all parameters at all relevant sites and to investigate what is the reason for what. With respect to the rst part the authors assume that the MLH determined by BL-VIEW is correct under all meteorological conditions (which might not be true I don't suggest this, but it could be). As the complete discussion of the second part is however based on this assumption, the rst part must be very clear.
As a consequence I recommend to explicitly mention in the conclusions that more studies are required to check whether the conclusions of this paper are generally valid, and to include short comments (see below) whenever a statement/conclusion is uncertain or the validity might be limited.
I feel that all gures/tables except Fig. S1 are worthwhile to be included into the main manuscript.
After consideration of these comments, the paper can be published.
Two main points (in the point-to-point response useful information is provided but the issues are not fully solved).

The MLH-error
• Page 9: "In this study, we used the gradient method...". This sounds as if the authors have develop an own retrieval. This is however not the case: they have used BL-VIEW provided by Vaisala. This should be clearly stated here. If there are dierent versions of this software or if there are dierent parameters that can be changed by the user, the selected conguration shall be mentioned here. The reason is, that this selection might inuence the retrieved MLH and thus, that the complete discussion might depend on this selection. So whenever the authors come to a certain (critical) conclusion I recommend to mention that this conclusion is based on certain assumptions (e.g., that the MLH is accurate within 100 m). As an example: in case that the ceilometer fails to retrieve the MLH (too low aerosol backscatter) or if the "wrong" aerosol layer is assumed to be the MLH, then the calculated correlations between MLH and other parameters (second part of the paper) could be erroneous. But I agree that it is beyond the scope of this paper to review MLH-retrievals by ceilometers.
• Page 10: It still remains unclear what the "absolute error (AE) of the measured MLH" is. The reader can understand it either in the sense of "accuracy of BL-VIEW" or as "the dierence between MLH sonde =MLH ref erence and MLH ceilo ". Both errors exist, but the authors only understand it according to the second option. Maybe one sentence can be added to avoid any misunderstanding. By the way: I would avoid the acronym AE, rst, because it is normally used for Angström exponent, and second, because it is not necessary.
• Page 11: "Once the aerosol concentration becomes uniform in the vertical direction, the ceilometer cannot calculate the MLH through sudden changes in the attenuated backscatter proles, resulting in serious underestimations." I don't understand this: If no MLH is retrieved from the ceilometer, why is this an underestimate? Is the MLH set to zero? This would not make sense. Please explain (maybe when BL-VIEW features are discussed in section 2.2.1).
• Page 13 (and many similar cases): "at 739 ± 137 m": it must be explained what the second number means and how it is calculated. Is it the error of the quantity, is it its temporal variability or something else. The expression "consistent" (twice on page 13) is qualitative only; a more precise description would be welcome.

Potential over-interpretation
• Fig. 9a; Page 15: In the point-to-point response it is stated that the gradual decrease of the MLH in summer is really signicant (in a statistical sense). However, a problem remains, when the trend is smaller than the uncertianty of the MLH-retrieval itself. This might lead to an over-interpretation of the data. I am not saying that the observed trend is not real, but one has to be cautious. This point should become clear when the above mentioned amendments has been made, or a brief remark might be added.
• Page 20: The paragraph starting with "In our data, scatters under poor..." is still sort of dicult to understand (what is scatters?), is partly relying on assumptions/speculations and consequently can be omitted or drastically shortened. If the data on the chemical composition are not measured or available, they should not be used in the discussion. Otherwise this might lead to another overinterpretation, in particular as the uncertainties of the MLH are not well characterized in this paper (see previous comments). This make the paper "vulnerable" and rather weakens than strengthens the benet. So I recommend to skip everything that cannot be proven by real facts/measurements. Having this in mind the whole section 3.4.2 should be checked for potential optimizations.
• Page 21: "the critical threshold for the MLH...". I still don't understand what is meant and how it has been derived. I see from the new table, that at relative humidities larger than 80% the correlation between MLH and visibility suddenly becomes very high, but I don't see how the "critical thresholds" are derived. Please explain in more detail and in a conclusive way.
Minor and technical points • Fig. 1: I am not sure if this is really important but I feel that the (city limits) of Beijing should be marked, in particular as it is referenced on page 16.
• Page 5: "in the morning (00:00UTC) and at night (12:00UTC),..." Here, local time should be given (it is however given later in the manuscript). This is also a not-mandatory change.
• Eq. 3: Symbol u * is not explained here • Page 8: "10 km from the station of the Institute of Atmospheric Physics". Suggestion: replace with "10 km from the BJT-site".
• Page 9 (Section 2.2.1): I am not sure if it is worthwhile to mention that the authors are aware of the water absorption issue (see point-to-point response).