Response to the Comments from Referee #1

Response to the Comments from Referee #1 1 2 We sincerely thank the reviewer for the valuable feedback that we have used to check our 3 newly-developed algorithm and improve the quality of our manuscript as well. The reviewer 4 comments are laid out below in italicized font and specific concerns have been numbered. Our 5 response is given in normal font. 6 7 This paper presents measurements from a lidar, ceilometer and sodar at Cape Grim, from 8 which the boundary-layer height is derived. Although Cape Grim is an interesting station for 9 measurements of pristine Southern Ocean air, it is not clear to me what the purpose of this 10 paper is meant to be. Most of the results show three case studies which are little more than 11 examples – they don’t lead to any useful conclusion beyond this particular dataset. 12 13 At the heart of the difficulties encountered in this paper, and not discussed at all in the 14 introduction, is what is meant by the BLH over the ocean and how this relates to the 15 distribution of aerosols. The BLH is essentially a thermodynamic concept, but the marine 16 boundary layer can be stable, inhibiting the vertical transport of aerosol. To affect a 17 meaningful comparison with WRF, for example, one would need to understand what WRF 18 means by BLH and whether that is relevant to the distribution of aerosols (i.e. are they likely 19 to be mixed throughout the BL or confined to low levels by inversions in the temperature 20 profile?). The issue is compounded here by the very different ‘BLH’ derived from the two 21 remote sensing instruments, strongly suggesting that a simple definition is inappropriate. 22 23 Main comments 24 Regarding the conclusions of the paper: 25

air masses for the spatial separation of the observing sites, (3) the coarse resolution of ERA5 48 and (4) the presence of the lofted layer or cloud layers. Therefore, no single approach can cover 49 all situations over this campaign. Among these causes, (1) (2) and (3) would influence our 50 whole period. More details could be found in the section 3.2.1 of the revised version. 51 52 Figure 4. Comparison of 1-h averaged BLH estimations based on different instruments and 53 or lofted aerosol layers, and presents less diurnal characteristic of MBL (marine boundary layer). 70 Its BLH fluctuated from 0.52 to 0.72 km in the morning and appeared to collapse immediately 71 afterward before growth at 19:00 LST again. The largest difference between 72 miniMPL/ceilometer and ERA5 occurs during the MBL developing period (from 12:00 to 73 20:00 LST) and the mean nocturnal boundary layer are higher than 0.5 km. 74 75 76 Figure 5. Resulting BLH for the whole observing period with 1-h averages from miniMPL 77 (IEDA), ceilometer (gradient method) and ERA5. ERA5-estimated BLHs are shown as magenta 78

stars. 79
We also replotted the BLH comparison, adding ERA5 reanalysis data and WRF model. Here 81 only the plots of three case studies are presented in Figure  can effectively maintain the image information filtering the superimposed noise. The specific 123 operation step is to convolute the input image with two-dimensional Gaussian kernel, and take 124 the convolution sum as the output pixel value. The output image is obtained after Gaussian 125 filtering. 126 characteristics between the target area to be extracted and its background, the image is regarded 128 as a combination of two types of areas with different gray levels (target area and background 129 area). The thresholds is automatically selected to determine a threshold interval. The pixels 130 within this threshold range belong to the target area, while the pixels not within this threshold 131 range belong to the background area. Thus, the corresponding binary image is generated. 132 gray-scale or binary images. Among them, dilation is to add some pixels to the image boundary 135 to make the image extend in the direction of increasing the gray index. In the contrary, erosion 136 is to delete some pixels on the image boundary, so that the image shrinks in the direction of 137 reducing the gray value. These two operations of corrosion and expansion can cause the greatest 138 changes to the edge points. Based on this point, the corroded and expanded image is compared 139 with the original image to obtain the point with the largest local change, that is, the edge point. 140 Connecting these points and hence the whole edge of the image can be obtained. 141  3.The effect of clouds on BLH measurements is barely mentioned, other than to state without 147 proof that the IEDA method is better in this regard. 148 Response: Yes, low cloud layers impede the detection of the BLH. In our initial results, we 149 found that the IEDA and gradient method will mistakenly identify the large gradient of the low 150 cloud layers as the BLH. Specifically, the gradient method typically found the BLH at the 151 beginning of the large negative gradient (base of the cloud layer), while the IEDA calculated the 152 BLH slightly higher than the gradient method. Differences between these two methods were 153 found to reach the maximum of 0.5 km. Therefore, we improved our algorithm firstly by cloud 154

removal. 155
Cloud removal: 156 First, we used the gliding method to identify clouds. According to the characteristics of the 157 cloud backscattered signal (Zuev et al.,1987;Cadet et al., 2005), it can be known that the 158 integral of noise is close to 0, while in the area where the cloud exists, the signal integral value 159 will be a considerable positive one. Therefore, the signal integral in the cloud can be very 160 different from that in other cloudless areas, which can help effectively distinguish the cloud 161 signal from noises. As shown in the following figure, a filtering window W is presented. Wh 162 and WL are the upper and lower edges of the window respectively to integrate the echo signal in 163 the window, i.e 164 (1)

165
(2) Where C(w) is the integral value of window W, and X (z) is the range-corrected signal. C(w) 168 value approaches 0 in a certain interval if noises exist. For clouds, the integral value C(w) will 169 be significantly larger than that of the aerosols. By properly selecting the threshold C(w), the 170 cloud information can be extracted through the window integral value. Response: Yes, the particle linear depolarization ratio (PLDR) only acts to verify the 213 continental sources in the episode 3 (E3). Besides evaluating the IEDA and gradient methods 214 for BLH identification using miniMPL and ceilometer, another object of our study it to analyze 215 the different characteristic of MBL and aerosols suspended in the layer. Episode 3 presents the 216 stable boundary layer characteristic under the continental sources and the relatively higher 217 PLDR could validate it. Besides that, we also include the radon concentration in Figure 6d to 218 prove it. 219 Secondly the Southern Ocean is a particularly interesting region for boundary layer 236 observations. Though various campaigns have been implemented to study the MBL and aerosol 237 properties, the temporal and spatial resolution of conventional meteorological information and 238 the influences of marine/continental sources on the boundary layer evolution are still scarce. 239 In addition, the different characteristic of MBL evolution from BL (boundary layer) over land 240 should be expected, "The BLH is essentially a thermodynamic concept, but the marine 241 boundary layer can be stable, inhibiting the vertical transport of aerosol", as you mentioned 242 before. However, owing to the observing site located at the height of a cliff 94 m above mean 243 sea level, the MBL we observed suffers less coastal impact than pure coastal sites in westerly 244 sector baseline. For example, considering the low BLH (confined to less than 1 km) in our 245 study, the decoupling trend of the MBL (Luo et al., 2016) was not observed during the 246 campaign. Furthermore, the MBL was characterized with a typical diurnal structure under the 247 marine sources, which is similar to the BL evolution over land. The growth of the convective 248 boundary layer in our two events was also highly associated with the changes in surface 249 temperature and wind speed profiles. In contrast, the MBL tends to be well mixed and stable 250 under the strong continental sources. These results indicate that the MBL near the coastal region 251 in Cape Grim may be more associated with relatively small fetches to mesoscale advection. 252 And the continental sources could also significantly influence the MBL evolution. 253 derived BLH for the miniMPL and the ceilometer (2) to investigate the impact of different 255 sources on the boundary layer evolution in the coastal region. This study aims to depict the 256 spatial-temporal structure of BLH variability in winter Cape Grim, thereby complementing the 257 spatial and/or temporal details of vertical BLH and wind information observational studies that 258 is scarce in the Southern Ocean. 259 We have clarified the objective of this measurement in the last paragraph of the "Introduction" 260 part. 2. According to the final sentence of p.5, the shaded period E1 in fig 5 contains a sea  Response: Yes, there is one-hour gap when the wind direction changed from SE to SW. We 318 have re-written the sentence as "Two hours later, the wind direction began to shift southerly 319 the presence of the lofted layers. We have also stated that in the response to the comment 1. 327 The phrase of "weak sea breeze interacts with the uplifted strong offshore wind" has been 328 removed. 329 For the aerosol extinction coefficient retrieval, the processing steps are as follows. After data 330 pre-processing, including range correction, background subtraction, dead-time correction, AN 331 and PC signal merging, and assumption of reference height in an aerosol-free region (usually 332 the upper troposphere, here we chose the range of 7-8 km as the reference height according to 333 the specific circumstance in every process of data analysis), the aerosol backscatter coefficient 334 at 532 nm was inversed by the Klett method (Klett, 1981). According the previous 335 measurements ( Müller et al., 2007;Omar et al., 2009), the constant lidar ratio was assumed as 336 20 sr and 60 sr (at 532 nm) for marine and continental aerosols, respectively to obtain the 337 aerosol extinction properties. We have added a new subsection 2.2.3 "Aerosol Extinction 338 coefficient" to explain the retrieval method.

Response: 372
We have added more details about the characteristics of aerosols in the section 3.2.2. 373 We have corrected the word" period". 374 We have re-plotted the Figure 9, 10 and 11 (now presented as Figure