General comments:

Using in situ measurements, satellite observations and the WRF model，the authors investigated the contributions of chemical and physical processes to the evolution of haze extremes. Findings show that chemical process that plays a leading role in PM production varies with the development stage of the haze event. And the haze-PBL interactions accelerate the accumulation of aerosol particles and water vapor at the ground level, amplifying the haze severity. The findings help to demonstrate the potential for achieving co-benefits for air quality and climate via black carbon mitigation. I recommend it to be accepted after minor revision.

1. 3 show the concurrent increases in OOA, PM_2.5 and Ox concentrations during the transition period. Is that also the case for the clean days? Why or why not?
2. Generally, the PBLH increases from morning to afternoon, and then decreases to the midnight. While it seems that the rate of growth in PM during the polluted period (from afternoon to midnight) is comparable to that during the transition period (from morning to afternoon) in Figs. 3a-b. Does it mean that photochemistry produced more PM? I suggest the authors to provide additional information on the diurnal evolution of the PBL for the two selected cases, and further explore the role of boundary layer evolution and chemical processes in the development of pollution.
3. In my opinion, the relative humidity calculated by the Nozaki’s equation (Lines 402-407 and 501-503) may suffer from large uncertainties that originated from 1). errors in PBL height, especially for the clean cases (Line 257), 2). inappropriate selection of Pasquill stability level (Line 255) due to the lack of cloud fraction measurements. 3) the uncertainty of the empirical formula itself. Since the model simulations agree well with the observations (Fig. s2), why not use the modelled results to show the changes in RH from the ground to the free troposphere?

Technical comments:

1. Line 300, the ‘(Li et al. (2015)’ should be ‘(Li et al., 2015)’.
2. I cannot find tables. 1-3 in the manuscript.

This study investigated the severe regional haze episodes in 2013 over the Northern China Plain through both measurements and simulations. The authors showed strongly correlated OOA and Ox concentrations during transition periods and positive feedback to haze formation by aerosol-PBL interactions during polluted periods. In general, this paper is well written and suitable for publication after the minor comments below are addressed.

1. I couldn’t find any tables in the main text.
2. I couldn’t find in the paper how the OOA concentrations were measured or determined. Can the authors add more details of how the OOA data were obtained? Also, can the authors add OOA time series to Figure 1, on top of PM_2.5 concentrations?
3. Line 246: The authors retrieved PBLH from MERRA2, but PBLH can also be estimated by the Nozaki’s equation provided here. I wonder if the PBL estimations from the two methods were consistent.
4. Figure 3: Can the authors add R² between OOA and total PM_2.5? It seems that even OOA only accounted for less than 1/3 of PM_2.5 mass, their concentrations were highly correlated. In this way, would PM_2.5 also show a good correlation with O_x? In addition, it would be helpful to add the same set of graphs for clean days.
5. Figure 4: Can the authors add the PBLH profile for transition days?
6. Line 405: How did these calculated RH compare to the measured RH in Figure 1?
7. Line 406: where did the 1000 m decrease come from? I only find in Line 371 “the simulated maximal height of PBL under the polluted condition is reduced by more than 300 m relative to the clean condition”.

1. Figure 4&5: Both figures show PBLH diurnal profiles, but with very different magnitude. Only looking at clean days, Figure 5 PBLH is only half of that in Figure 4. Why are the simulated results so different from the MERRA2 reanalysis?
2. Figure 5: Can the authors either add legend or rewrite the caption? It should be clearly stated that the top 6 figures and bottom 6 figures are for two episodes.