Atmosphere Boundary Layer Height (ABLH) Determination under Multiple-Layer Conditions Using Micro-Pulse Lidar

Round 1

Reviewer 1 Report

General comments

                This is an interesting paper and I, personally, will explore the use of published technique with data from my airborne Doppler Wind Lidars.

                The paper is well written with a few grammar errors  and “awkward” word usage.

                The figures are readable.

                The referenced material is an excellent selection.

 

Specific comments and suggestions

33/34: “great concordance” and “great agreement” are not supported by the results. Perhaps “good” or “reasonable” may be better qualitative adjectives.

121/123: While the reference to Welton and Campbell (42) is fine, the resulting NRB needs more “in depth” discussion. The NRB is the basic quantity that is being evaluated in this paper.

143/145: Are the tower instruments sampled every 30 minutes or is there just a 30 minute average value being recorded? If that is the case, how often are the measurements made that go into the average?

173/174: A figure to illustrate the “idealized backscatter profile” would be helpful.

188/201: There is frequent mention of the “top of optically thick clouds”. What about clouds that fully attenuate the lidar signal well below cloud tops?

215: Why 4.3 km?

219: In Figure 1, should there be a 4.3 km value in the flow chart under “no top limiter”?

278: “1.0694” km has too many digits after the decimal for this topic. Perhaps  “1.07”  would be more defensible. Care should be taken throughout the paper not to exaggerate on the last significant digit.

297: “.793” is not “great concordance”.

Section 4.0 is well done with minor grammar edits.

493: What does the last sentence mean? What other lidar is being referenced?

Author Response

Response to Reviewer #1 (remote-sensing 429292)

We thank the reviewer very much for taking time to handle and review our manuscript. Thank you for your recognition of our work. We have carefully considered and addressed all the valuable comments. Please see our detailed reply below in blue.

This is an interesting paper and I, personally, will explore the use of published technique with data from my airborne Doppler Wind Lidars.

The paper is well written with a few grammar errors and “awkward” word usage.

The figures are readable.

The referenced material is an excellent selection.

Specific comments and suggestions

33/34: “great concordance” and “great agreement” are not supported by the results. Perhaps “good” or “reasonable” may be better qualitative adjectives.

Reply: Yes, Thanks. The concordance has been quantified and the description has been written as: “While comparing the lidar-determined ABLH by HM (or CFM) and nearby radiosonde measurements of the ABLH, a reasonable concordance is found with a correlation coefficient of 0.94 (or 0.96) and 0.79 (or 0.74), presenting a mean of the relative absolute differences respect to radiosonde measurements of 10.5% (or 12.3%) and 22.3% (or 17.2%) for cloud-free and cloudy situations. The diurnal variations in the ABLH determined from HM and CFM on 4 selected cases show good agreement with mean correlation coefficient higher than 0.99 and a mean absolute bias of 0.22 km.”

See lines 33-39 of the revised version.

The qualitative results in the text are presented in Table 1, Table 2 and described in section 4.1 and 4.2 of the revised manuscript.

Table 1. Correlation coefficients (R), absolute height differences (mean and standard deviation (std) in km) between ABLH determined by theta gradient and by Haar wavelet covariance transform method (HM) or curve fitting method (CFM) in different situations, as well as mean value of relative absolute differences relative to theta-gradient-determined ABLH in average (rd, in 100%).

Situations	Method	R	mean	std	rd
Cloud-free situation	HM	0.96	0.14	0.11	10.5
	CFM	0.94	0.17	0.13	12.3
Cloudy situation (without top limiter)	HM	0.09	0.83	0.90	66.1
	CFM	0.11	0.67	0.64	53.7
Cloudy situation (with top limiter)	HM	0.74	0.28	0.24	22.3
	CFM	0.79	0.22	0.18	17.2

 

Table 2. Correlation coefficients (R) and absolute height differences (mean and standard deviation (std) in km) between ABLH retrieved by HM and CFM based on lidar data on 4 selected cases.

Date	R	mean	std
2007.07.28	0.998	0.25	0.21
2007.06.09	0.993	0.32	0.37
2007.10.28	0.993	0.13	0.22
2007.06.12	0.997	0.16	0.15

 

121/123: While the reference to Welton and Campbell (42) is fine, the resulting NRB needs more “in depth” discussion. The NRB is the basic quantity that is being evaluated in this paper.

Reply: Yes, we agree. The NRB are detail introduced in the revised manuscript as below:

The Micro-Pulse Lidar system (MPL-4B) installed at SACOL emits a laser pulse at a wavelength of 527 nm. The optical power measured by a lidar is proportional to the signal backscattered by molecules and particles present in the atmosphere [24]. The lidar signal can be expressed as:

                                                   (1)

where  is the backscatter coefficients.  is a system constant for a given lidar.  is the laser output energy and  is the atmosphere transmission. After range correcting and laser energy normalizing, a resulting signal is called normalized relative backscatter  (NRB) which is defined as:

                                               (2)

The MPL at SACOL records signal up to 20+ km at a vertical resolution of 75 m (or 30 m) for each minute [47]. There, the NRB values are interpolated to a 15 m space range using the Lagrange interpolation. The instrument effect including afterpulse, overlap, pulse energy, detector noise has been corrected, detail information about correcting procedures are given by Welton and Campbell [48]. In the study, the lidar data below 4.3 km is used. In addition, because of the existence of the blind zone caused by the incomplete overlap between the lidar laser beam and its receiving optical axis, the application of the backscattering lidar in near-ground is limited [17,33]. The NRB in lowest few hundred meters are removed.

See section 2.1, lines 137-153 of the revised version.

143/145: Are the tower instruments sampled every 30 minutes or is there just a 30 minute average value being recorded? If that is the case, how often are the measurements made that go into the average?

Reply: I’m sorry that the original description is confused. The tower instruments are sampled every 30 minutes (not 30 minute average).

 The description in the revised manuscript has been written as: “The 32 m Micrometeorological tower at SACOL can provide wind speed measurements (014A-L, Met One) at 1, 2, 4, 8, 16 and 32 m with a temporal resolution of 30 min”.

See line 174 of the revised version.

173/174: A figure to illustrate the “idealized backscatter profile” would be helpful.

Reply: Yes, Thanks for your valuable suggestion. A figure illustrating the “idealized backscatter profile” in curve fitting method (Figure 2) as well as a figure presenting the Haar wavelet transform (Figure 1) has been added in the revised manuscript.

Figure 1. (a) A NRB profile (left) and the shape of Haar wavelet, (b) the resulted covariance transform  as .

Figure 2. (a) An idealized backscatter profile, and (b) a real case for curve fitting procedure.

Detail see Figure 1, Figure 2 and related description in line 195, 212 of the revised version.

188/201: There is frequent mention of the “top of optically thick clouds”. What about clouds that fully attenuate the lidar signal well below cloud tops?

Reply: Thanks. For optically thick cloud, the lidar signal may be fully attenuated by the cloud, thus the top of cloud can’t be identified. However, the signal sharply decreases at the called “upper edge” of the cloud similar to at the top of thin cloud. Therefore, for optically thick cloud, it should be called as upper edge.

In the revised manuscript it has been described as “the strongest negative gradient often appears near the upper edge of clouds”.

See lines 227 of the revised version.

215: Why 4.3 km?

Reply: Generally, the daytime depth of the continental ABL is on the order of 1 ~ 2 km, the cloud above the ABLH is considered in the study, so the upper altitude is defined to be approximately 4 km. The lidar signal profile below the 4.3 km is used to retrieve the ABLH in the study.

In the revised manuscript it is described as “In the study, the lidar data below 4.3 km is used”.

See line 150 of the revised version.

219: In Figure 1, should there be a 4.3 km value in the flow chart under “no top limiter”?

Reply: Yes, we agree. For cloud-free situations, the top limiter is 4.3 km and has been changed in the Figure 3 in the revised manuscript.

 

 Figure 3. Major steps for determining the top limiter to eliminate the cloud effect on atmosphere boundary layer height (ABLH) retrieval in the new technique.

See Figure 3 in the revised version.

278: “1.0694” km has too many digits after the decimal for this topic. Perhaps  “1.07”  would be more defensible. Care should be taken throughout the paper not to exaggerate on the last significant digit.

Reply: Yes, Thanks. The issues has been carefully checked and revised throughout the paper. For example, “……the height determined from the theta-gradient is 1.07 km (Figure 5a), which is a little higher than the heights retrieved by lidar-based methods, 0.86 km by HM and 0.93 km by CFM…..”.

See lines 326-327 of the revised version.

297: “.793” is not “great concordance”.

Reply: Yes, we agree. In the revised manuscript, the “great concordance” has been changed to “reasonable concordance” as line 35.

In addition, in cloudy situation, relative to the ABLH determined based on lidar data without top limiter, the ABLH determined with the top limiter is much closer to radiosonde-derived results. The comparison results with a reasonable correlation coefficient can present the improvement of the lidar measurements of the ABLH with the determined top limiter.

Figure 8. Similar to Figure 7, but in cloudy situations. The blue open dots represent the comparison results of ABLH determined by theta gradient (vertical coordinate) and determined by HM (a) or CFM (b) (horizon coordinate) without height limitation, the red stars indicates the comparison results after the top limiter if given for HM and CFM. R represents the correlation coefficients, blue represents no top limiter and red represents with the top limiter.

See Figure 8 and related description as lines 341-349 of the revised manuscript.

Section 4.0 is well done with minor grammar edits.

Reply: Yes, Thanks. Section has been carefully checked and edited. For example, lines 392, 426, 438, 458 …… of the revised version.

493: What does the last sentence mean? What other lidar is being referenced?

Reply: I’m sorry that the original description is confused. The sentence has been rewritten in the revised manuscript as: “……don’t necessarily apply to other observations sites .…… However, the principle for threshold determination is clear and reasonable, it can be referenced for lidar measurements of ABLH in multiple-layer conditions.”

See lines 515-516 of the revised version.

In addition, the last paragraph of the original manuscript is discussion section, which has been merged into results section described as “4.0. Results and Discussion” in the revised manuscript.

Author Response File: Author Response.pdf

Reviewer 2 Report

This article deal with the new technique for the detection of ABLH. They have used old two methods, Harr wavelet transformation and curve fitting method, with the limitations when there are some clouds or residual layer. Even though they used old techniques, this kinds of approach is useful for the real application using MPL. In this article, the important point is determination of limit maximum altitude of ABLH So they must describe more carefully this methods in the section 3-2 “The detail process….” 1) Line 229: relative increase relative gradient(figure2,column2) , or relative gradient relative increase(?) Relative increase(text ) and relative gradient ( in figure 2) are used, which one is corrects? Line 201: the layer is greater than or equal to 55% How can you find this values 55%? Line 206 : threshold value : -2 How can you define this values ? 2) In the figure 1: “define a threshold of -2 NRB gradient”  NRB gradient or relative increase(gradient)? 3) In figure 2(column 2) : relative gradient  Is it right? It should be relative increase? 4) Figure 2( column 3) : the threshold value -2 is important value  I cannot see “-2” : because the x-axis scale is too big (-300~ 300) 5) Line 254 : Covariance transform wavelet transform, wavelet  mother function 6) In section “3-2” I cannot find the method for detecting RL. Please describe more carefully for the determination method of RL.

Author Response

Response to Reviewer #2 (remote-sensing 429292)

We thank the reviewer very much for taking time to handle and review our manuscript. We have carefully considered and addressed all the valuable comments. Please see our detailed reply below in blue.

This article deal with the new technique for the detection of ABLH. They have used old two methods, Harr wavelet transformation and curve fitting method, with the limitations when there are some clouds or residual layer. Even though they used old techniques, this kinds of approach is useful for the real application using MPL. In this article, the important point is determination of limit maximum altitude of ABLH So they must describe more carefully this methods in the section 3-2 “The detail process….”

Reply: Thanks, The section 3.2 (The detail process for improving the ABLH determination) has been carefully complement about the residual identification, the threshold value determination, top limiter determination. Also, a clearer flow chart in Figure 3 has provided in the revised manuscript.

Line 229: relative increase, relative gradient (figure2, column2), or relative gradient, relative increase? Relative increase (text) and relative gradient (in figure 2) are used, which one is correct?

Reply: Thanks, the original description of “relative gradient” in figure2 (column 2) is not suitable, which should be the relative increase in lidar signal (,  is the NRB (normalized relative backscatter)) as described in the text.

In the revised version, the description has been rewritten as “relative increase in NRB” in figure4, column2 (two new figures named figure 1 and figure2 in the revised manuscript are added following the other reviewer’s suggestion). The fourth column is added to more clearly show the gradient of NRB below the cloud base.

Figure 4. Vertical distributions of normalized relative lidar backscatter signal (NRB) (the first column), relative increase in NRB (the second column), gradient of NRB with the determined top limiter (the third column) and the gradient of NRB only below the lowest cloud layer (the last column) on typical situations: (a) cloud-free, observed at 14:15 local standard time (LST), 09 June 2007; (b) one cloud layer is observed at 10:15 LST, 09 June 2007; (c)-(d) more than 1 cloud layers are observed, the lowest-altitude cloud in (c) (at 13:45 LST, 28 October 2007) is decoupled from the ABL while in (d) (at 10:15 LST, 28 October 2007) is within the ABL.

See Figure 4 on page 9 of the revised manuscript

Line 201: the layer is greater than or equal to 55%. How can you find this values 55%?

Reply: The threshold is defined based on previous study and experience. The study by Campbell (1998) indicated that for a cloud boundary to be observed, one or two-bin relative increase in the NRB of at least 55% is required. Testing the Empirical threshold on the MPL data over SACOL we find that cloud layers cloud be identified basically. That is, for a strong signal gradient layer, if the relative increase in NRB at the layer base is greater than or at least equal to 55%, the cloud layers visible to eyes could be identified. Therefore, the Empirical threshold of 55% is utilized in this study.

In the revised manuscript, we supplement the relevant description as: “Campbell [22] noted that for a cloud boundary to be observed, one or two-bin relative increase in the NRB () of at least 55% is required. Testing the Empirical threshold based on many cloudy cases proves that the threshold is applicable over SACOL.”

 See lines 233-235 of the revised version.

Line 206: threshold value: -2 How can you define this values?

Reply: The threshold of “-2” is an empirical value. It is also defined based on numerous experiments on many cloudy days.

 In the revised manuscript, we supplement the relevant description as: “Based on many cloudy cases, it has been found that for MPL signal profile over SACOL, if clouds are visually located above the top of the ABL, a strong signal gradient of at least less than -2 generally exists below the bottom of the cloud. Therefore, in this study, if there is at least one signal gradient that is less than the empirical threshold of -2 below the lowest cloud base, the cloud is classified as outside the ABL. In contrast, the cloud is thought to cap the ABL top. ”

See lines 246-250 of the revised version.

In the figure 1: “define a threshold of -2 NRB gradient”, NRB gradient or relative increase (gradient)?

Reply: Thanks, the Empirical threshold of “-2” is for gradient of NRB. In detail, below the lowest-altitude cloud base, if there is signal gradient that is less than -2, the cloud is classified as outside the ABL. In contrast, the cloud is thought to cap the ABL top.

In the revised version of manuscript, it has been rewritten as: “Whether there is a gradient of NRB below the cloud base” in Figure 3. (The Figure1 in the original manuscript has been revised to show the flow chart more clearly shown as Figure 3 in the revised version)

See Figure 3 on page 8 of the revised version.

In figure 2 (column 2): relative gradient. Is it right? It should be relative increase?

Reply: Yes, We agree. The description has been rewritten as “relative increase in NRB” in Figure 4 (column 2) in the revised version.

See Figure 4, on page 9 of the revised version.

Figure 2 (column 3): the threshold value -2 is important value. I cannot see “-2” : because the x-axis scale is too big (-300~ 300)

Reply: Yes, Thanks. In order to present the threshold value -2, in the revised manuscript, the fourth column in the figure 4 (Figure 2 in the original manuscript) is added which presents the signal profile only below the lowest-altitude cloud base more clearly.

Figure 4. Vertical distributions of normalized relative lidar backscatter signal (NRB) (the first column), relative increase in NRB (the second column), gradient of NRB with the determined top limiter (the third column) and the gradient of NRB only below the lowest cloud layer (the last column) on typical situations: (a) cloud-free, observed at 14:15 local standard time (LST), 09 June 2007; (b) one cloud layer is observed at 10:15 LST, 09 June 2007; (c)-(d) more than 1 cloud layers are observed, the lowest-altitude cloud in (c) (at 13:45 LST, 28 October 2007) is decoupled from the ABL while in (d) (at 10:15 LST, 28 October 2007) is within the ABL.

It can be seen that below the cloud base, some signal gradients less than -2.0 occur, which proves that the lowest-altitude clouds in Figure 4(b)-(c) are above the ABL top and the heights of cloud base are regarded as the top limiters (black solid lines). In contrast, there is no signal gradient less than -2 below the base of the lowest-altitude clouds in Figure 4(d), the cloud caps the ABL top

See Figure 4 on page 9 and related description in lines 261-281 of the revised version.

Line 254 : Covariance transform, wavelet transform, wavelet, mother function

Reply: Yes, I’m sorry that the original description is confused. In the revised version the description of the Haar wavelet transform method (HM) has been written as: “The wavelet covariance transform defined by Gamage and Hagelberg [54] is a means of finding step changes in a signal. The Haar, step function provides a multi-scale local gradient analysis of the lidar signal to isolate the spikes due to aerosol concentrations. The Haar wavelet  and covariance transform  are defined with equations 3 and 4 as follows [53]:……The similarity between the lidar signal () and the Haar wavelet () is described by covariance transform (). The ABLH determined by HM corresponds to the  where the similarity () reaches its maximum……”

In addition, In order to more clearly presents the HM, an example of NRB profile (), the shape of Haar wavelet () and the covariance transform () is shown in Figure 1 in the revised version. When the covariance transform reaches its maximum at dilation of 600, the altitude (about 2000 m) is the resulted ABLH.

Figure 1. (a) A NRB profile (left) and the shape of Haar wavelet, (b) the resulted covariance transform  as.

Detail see Figure 1 on page 5, and lines 183-202 of the revised version.

In section “3-2” I cannot find the method for detecting RL. Please describe more carefully for the determination method of RL. 

Reply: Yes, Thanks. The description of the RL detecting has been added in the revised manuscript as: “As an elevated high concentration aerosol layer, RL often occurs above the surface aerosol layer in the morning and the aerosol concentration strongly decreases near its top. That is, when the cloud layer or RL exists, a strong positive gradient with a strong negative gradient located above can be simultaneously observed. Based on this characteristic, the cloud layers or RL can be found by seeking pairs of strong positive and negative gradients in the NRB profiles. For cloud…….Different to cloud, the magnitude of the lidar signal backscattered by RL is close to the by the near-surface aerosol layer. If the average value of the signal within the strong gradient layer is less than three to five times of the average value of the signal below the bottom of the strong gradient layer, the strong gradient layer is identified as RL.”

See lines 228-232, 237-240 of the revised version.

 

Author Response File: Author Response.pdf

Reviewer 3 Report

see attached file.

Comments for author File: Comments.pdf

Author Response

Response to Reviewer #3 (remote-sensing 426292)

We thank the reviewer very much for taking time to handle and review our manuscript. We have carefully considered and addressed all the valuable comments. Please see our detailed reply below in blue.

The manuscript you submit contains a very interesting study on the atmosphere boundary layer height (ABLH) measurement. It provides a set of signal diagnostic tools for accurately evaluating ABLH from an aerosol backscatter lidar. It surely can improve the actual methodologies used by different research groups because it treats the bias induced by the presence of cloud in and above the planetary boundary layer. I am not a meteorologist, but I am more confident with the basics of laser interaction with the atmosphere. Fc£ this reason, I apologize for my comments. They might not be suitable for the science you intend to publish.

At my opinion, this work deserves to be published after considering several minor changes listed below that I proposed to improve your manuscript.

Proposed minor changes

Line 2, title: It should be improved to better highlight the work that has been done. Suggestion: remove “Exploring a new technique for improving”. This, because the signal treatment techniques are already know but their combination as you proposed brings a real improvement especially in case of cloudy weather.

Reply: Thanks, We agree. The title has been changed to “Atmosphere boundary layer height (ABLH) determination under multiple-layer conditions using Micro-Pulse lidar”.

Line 20: Standard meteorological measurements (eg balloon soundings) are able to determine ABLH. Please improve the sentence.

Reply: Yes, Thanks. The description has been written as: “Accurate estimation of the atmospheric boundary layer height (ABLH) is critically important and it mainly relies on the detection of the vertical profiles of atmosphere variables (temperature, humidity and horizontal wind speed) or aerosols. Aerosol Lidar is a powerful remote sensing instrument frequently used to retrieve ABLH through detecting vertical distribution of aerosol concentration.”

See lines 18-21 of the revised version.

Line 20 -21: what kind of lidar? aerosol lidar? or trace gases lidar (03)? water vapor Lidar? please state.

Reply: Yes, Thanks. The lidar used in the study is aerosol lidar and in the revised manuscript it has been written as “Aerosol Lidar is a powerful remote sensing instrument frequently used to retrieve ABLH through detecting vertical distribution of aerosol concentration.”

See line 20 of the revised version.

Line 24: cloud base, how is it evaluated by considering multiple scattering. What is the request ABLH accuracy?

Reply. Thanks. However, we didn’t take the multiple scattering into consideration in this study. One of advantages of the Micro-Pulse Lidar (MPL) is that the instrument employs a narrow receiver field-of-view (FOV) (Rubio et al., 2001). For small FOV, the effect of higher-order of scattering is small. In addition, some researches have indicated that “multiple scattering (MS) effects are zero at the base of the cloud (where the laser enters the cloud) and increase with the penetration depth. As we approach the top of the cloud, MS effects decrease since the geometrical constraints become the dominant factor.” (Mitrescu, 2005). Therefore, in our study the MPL data is used, only the location of the cloud boundary (base, upper edge) are identified based on strong signal gradient, the MS effect is not taken into consideration.

Line 28 abstract: Cloud top, how is it determined from ground based lidar? In this case what about cloud with strong optical depth (multiple scattering)? You should there specify in term of cloud optical depth the applicability of aerosol lidar to evaluate ABLH. This aspect should be presented in the text and a short indication in the abstract on the limitation should be presented.

Reply: Thanks for your valuable suggestion. However, same to reply to comment (4), because the MPL is used which has a narrow FOV, the MS effect is small. In addition, the MS effect at cloud top is weak (Mitrescu, 2005). The MS effect is not taken into consideration is this study.

Line 34: “great agreement”! please, be quantitative by considering uncertainty.

Reply: Yes, Thanks. The agreement between diurnal variations in the ABLH determined from HM and CFM are quantified through correlation coefficient, mean bias, and standard deviation of the bias.

The description has been written as: “The diurnal variations in the ABLH determined from HM and CFM on 4 selected cases show good agreement with mean correlation coefficient higher than 0.99 and a mean absolute bias of 0.22 km.”

See lines 37-39 of the revised version.

In the text of the revised manuscript, Table 2 is added and the agreement of the ABLH is described as: “the diurnal variation of ABLH can be successfully determined by HM and CFM, and the determined results by two lidar-based methods show good agreement on 4 selected cases with a correlation coefficient higher than 0.99, with a mean bias less than 0.35 km and a low fluctuation of deviation (see Table 2).”

Table 2. Correlation coefficients (R) and height differences (mean and standard deviation (std) in km) between ABLH retrieved by HM and CFM based on lidar data on 4 selected cases.

Date	R	mean	std
2007.07.28	0.998	0.25	0.21
2007.06.09	0.993	0.32	0.37
2007.10.28	0.993	0.13	0.22
2007.06.12	0.997	0.16	0.15

See Table 2 and lines 499-501 of the revised version.

Line 50: this sentence is in contradiction with the next one where meteorological radio soundings are related. Radio-sounding is a standard methodology in meteorology. Moreover, please correct spelling, radiosondes is a French name!

Reply: Yes, we agree. The sentence in the revised manuscript has been rewritten as: “The ABLH can be determined from radiosonde-derived thermodynamic profiles such as temperature, humidity and horizontal wind speed (e.g., [8,9])). However, the detection is infrequently [7]. In recent years, the techniques based on remote sensing are attractive choices for the ABLH determination such as radar (e.g., [10,11]), sodar (e.g., [12]), lidar (e.g., [13,14]), etc”.

The spelling has been checked throughout the text and has changed to “radiosonde”

See lines 54-57 of the revised manuscript.

Line 51: reference to lidar ABLH estimation, see Toledo et al. 2017. Please consider this work. It is a complementary one to your approach.

Reply: Yes, Thanks. The Study by Toledo et al (2017) used six numerical methods to determine the ABLH from lidar measurements under different atmospheric conditions. The results were compared with those obtained from radiosoundings to analyze their reliability for ABL-height retrievals. We have read the valuable research and learned a lot.

The work has been introduced in the introduction as an example of ABLH determination from  lidar measurements in the revised manuscript: “In recent years, the techniques based on remote sensing are attractive choices for the ABLH determination such as radar (e.g., [10,11]), sodar (e.g., [12]), lidar (e.g., [13,14]), etc”. ([14], Toledo et al. 2017)

See lines 57 of the revised manuscript.

Line 56: aerosol as ABLH tracer. What about water vapor? Water vapor Lidar exists still two decades and it can also derive ABLH. Please add some words on this aspect. It will improve your manuscript. However Aerosol backscattering lidar is a more convenient instrument to be implemented in meteorology observatory by considering the price and the dimensions.

Reply: Yes, Thanks for your valuable suggestion. The introduction of the water vapor lidar has been added in the manuscript as: “Lidar measures ABLH usually employing a profile analysis, such as aerosol and water vapor. Aerosols and water vapor have sources near the surface and are confined by the capping inversion at the ABL top. Higher aerosol concentrations are distributed in the ABL than in the free atmosphere, and usually, the ABL is moister than free atmosphere. The strongest negative gradient of aerosol concentration and water vapor occurs near the ABL top. In addition, within the entrainment zone, the entrainment of overlying less-polluted and direr air into the ABL yields increase in the water vapor flux and increase in the variation of aerosol concentrations. Therefore, the vertical distributions of aerosol and water vapor are good tracers for ABLH determination (e.g., [15-19]). The Raman lidar and the DIAL (Differential Absorption Lidar) are able to measure vertical profile of water vapor with good resolution [20]. However, for ABLH determination, the water vapor lidar is less frequently used than the aerosol lidar. Aerosol lidar is more convenient to be implemented in meteorology observation by considering the price and the dimensions.”

See lines 58-69 of the revised version.

 Line 63: please refer the recent work of Toledo et al. 2017 with emphasis of their findings. It will improve your manuscript.

Reply: Yes, Thanks for your valuable suggestion. The work has been introduced in the introduction as: “……For example, Toledo et al [14] analyzed the performance of six numerical usually used methods on ABLH determination from lidar measurements, presented that there are big discrepancies between lidar and radiosounding retrievals when residual layers (RL) are present in the measurements)”.

See lines 79-82 of the revised version.

 Line 69: spelling, mark the end of the sentence with a dot after [28].

Reply: Yes, Thanks. The “;” has been changed to “.” in Line 89 of the revised manuscript.

 Line 97: Improve the sentence. The purpose .. is to propose

Reply: Yes, Thanks. The sentence has been written as: “This study proposes a technique ……. ”

See line 114 of the revised version.

 Line 138: please define all variables used in the equation.

Reply: Yes, we agree. In the revised manuscript, all variables have been defined as: “(, is the potential temperature, is the Latent heat of vaporization, is the specific heat at constant pressure, and  are saturated specific humidity and Temperature) could be calculated based on the pressure, temperature and water vapor profiles.”

See lines 163-167 of the revised version.

 Line 143 and next: please define all variables used in the text even they are text-book like variables.

Reply: Yes, we agree. In the revised manuscript, all variables have been defined as: “the fluxes of momentum (), CO2, latent (H2O) and sensible heat flux () at 3.0 m AGL of 30 min intervals”. Similarly, this issue has been checked throughout the text.

See line 172 of the revised version.

 Line 152: improvement is needed. Explain shortly why Haar wavelet method can be used to evaluate ABLH？HM is based on squared function. Is it adapted to atmosphere Iayering？ Please comment.

Reply: Yes, Thanks. The Haar wavelet transform provides a multi-scale local gradient analysis of the lidar signal to isolate the spikes due to aerosol concentrations. Therefore, it is able to locate the ABLH where sharp transition in the signal occurs.

 In the revised version, the description has been improved as: “The wavelet covariance transform defined by Gamage and Hagelberg [54] is a means of finding step changes in a signal. The Haar, step function provides a multi-scale local gradient analysis of the lidar signal to isolate the spikes due to aerosol concentrations.”

In addition, Figure 1 in the revised manuscript is added to show the method more clearly.

Figure 1. (a) A NRB profile (left) and the shape of Haar wavelet, (b) the resulted covariance transform  as .

Detail see Figure 1, and description in lines 183-185 of the revised version.

 Line 154: the strongest gradient of what?

Reply: Thanks, the “strong gradient” in the original description means the strong gradient of signal corresponding to the sharp transition of aerosol concentrations at the ABL top.

The description is confused and has been written as “The wavelet covariance transform defined by Gamage and Hagelberg [54] is a means of finding step changes in a signal. The Haar, step function provides a multi-scale local gradient analysis of the lidar signal to isolate the spikes due to aerosol concentrations.”

See line lines 183-185 of the revised version.

 Line 156: it is not clear. What is the relation between the mother function, the covariance transform and HM? Please improve the text.

Reply: Yes, Thanks. The introduction of the Haar wavelet transform method has been written as: “The wavelet covariance transform defined by Gamage and Hagelberg [54] is a means of finding step changes in a signal. The Haar, step function provides a multi-scale local gradient analysis of the lidar signal to isolate the spikes due to aerosol concentrations……The similarity between the lidar signal () and the Haar wavelet () is described by covariance transform (). The ABLH determined by HM corresponds to the  where the similarity () reaches its maximum. ”

See lines 183-96 of the revised version.

In addition, Figure 1 in the revised manuscript is added to show the method more clearly.

Figure 1. (a) A NRB profile (left) and the shape of Haar wavelet, (b) the resulted covariance transform  as .

See Figure 1 and related description in lines 195-196 of the revised version.

 Line 160: dilatation? do you mean z extension? please improve or comment.

Reply: Thanks.  is the dilation of the Haar wavelet, which is shown as Figure 1(a) in the revised version. In the revised manuscript, the description has been written as “ is the dilation of the Haar wavelet”.

See Figure 1 and line 189 of the revised version.

 Line 160: “delta h physically represents ...”. Put this explanation earlier in the text.

Reply: Yes, we agree. In the revised manuscript the description has been improved as: “ is the dilation of the Haar wavelet, which physically represents the thickness of the entrainment zone (EZT).”

See line 190 of the revised version.

 Line 243: Figure 1 is a technical one. It seem to be not a necessity for the manuscript comprehension. It adds no science to the manuscript.

Reply: Thanks for your suggestion. However, we think that a flow chart may be helpful for understanding the method when it is described with large number of words in the text (section 3-2). The flow chart in the original manuscript (Figure 1 in the original version) is too complex and lacks clarity, it has been modified and simplified as Figure 3 in the revised manuscript.

 Figure 3. Major steps for determining the top limiter to eliminate the cloud effect on atmosphere boundary layer height (ABLH) retrieval in the new technique.

See Figure 3 in the revised version.

Line 244, Figure 2: Nice figure, it well presents the difficulties to evaluate ABLH from the atmosphere laser light backscattering variation. However, please change the curve line label in the graphs placed in the middle of figure 2. Following the text, the orange dotted curves represent the relative change of the NRB and not a relative gradient of NRBI This mistake makes difficult at first glance the comprehension of the data presented in the figure!

Reply: Yes, we agree. The orange dotted curves represent the relative change of the NRB.

In the revised manuscript, the line label has been changed to “relative increase in NRB” in Figure 4 in the revised manuscript.

Detail see Figure 4 in the revised version.

 Line 269: Why using Theta Gradient notation to depict q-gradient. It is confusing because theta is related to the potential temperature. In figure 3 the orange line seems to show the gradient change of the potential temperature. Please improve!

Reply: Yes, Thanks. In fact, the ABLH is determined by gradient of potential temperature not gradient specific humidity. But when comparing the radiosonde-derived ABLH and lidar measurement of ABLH, the specific humidity profiles are also considered to ensure that there is no significant error in radiosonde-determined ABLH, In Figure 3 (Figure 5 in the revised version) and Figure 4 (Figure 6 in the revised version), the orange line represents the ABLH determined by theta gradient. The specific humidity is only used to reference.

In the revised manuscript, the description has been written as: “In theory, near the ABL top, there will be a sharp increase in potential temperature and meanwhile a sharp decrease in specific humidity [46,57]. The level of the strongest theta gradient is indicative of a transition from a convectively less stable region below to a more stable region above [8]. The theta-gradient is used to determine the ABLH for evaluating the height retrieved from the above lidar-based methods, but meanwhile, the specific humidity profile is taken into consideration to ensure that there is no significant error in radiosonde-determined ABLH.”

 See lines 309-314 of the revised manuscript.

 Line 271: IMPORTANT. Accuracy on ABLH evaluation is never presented! Because experimental data are used, uncertainty on ABLH evaluation exists. Please improve by adding a paragraph on this aspect. It will make the work presented in the manuscript more plausible.

Reply: Thanks for your valuable suggestion. In the revised manuscript, the evaluation of accuracy on lidar measurements of ABLH is done through comparing with the radiosonde-determined ABLH, the statistical variables including correlation coefficient, height difference (mean bias and standard variation of bias), as well as mean value of relative absolute differences relative to theta-gradient-determined ABLH are calculated.

The description about accuracy on ABLH evaluation in the revised manuscript is presented in section 4.1 as: “The L-band radiosonde could provide accurate thermodynamic profiles such as temperature, humidity for the troposphere [58,59]. The ABLH determined from radiosonde-derived profiles are usually used to evaluate the ABLH retrieved from other measurements [14,60,61]. This study also evaluate the accuracy on lidar-measurements of ABLH by comparing with the radiosonde-determined ABLH.…... in cloud-free situations, the ABLH determined by HM (or CFM) have high consistencies with the radisonde-determined ABLH with a correlation coefficient of 0.96 (or 0.94), with a mean bias of 0.14 km (or 0.17 km) and with mean relative absolute differences respect to radiosonde retrievals of 10.5% (or 12.3%) (Table 1)…….in cloudy situations……. Given the top limiter (red star), the height determined by HM (or CFM) show better concordance with those by theta gradient with a good correlation coefficient of 0.74 (or 0.79). The mean bias between HM (or CFM) and theta gradient determined ABLH is 0.28 km (or 0.22 km) with a low std of 0.24 km (or 0.18 km), while the mean relative absolute differences respect to radiosonde retrievals is 22.3% (or 17.2%), below 20% in average (Table 1). Overall, the results of comparison show that relatively reliable ABLH can be determined from the lidar data in the absence of clouds. However, cloud layers, especially those decoupled from the ABL, can seriously interfere the lidar-measurements of ABLH. The top limiter improves the ABLH determination in cloudy situations.”

Detail see Figure 5-8, Table 1, section 4.1 (lines 316-378) of the revised manuscript.

 Line 292: what is the reason of presenting the data correlation with two graphs? Following the text, they depict the same data set. It is confusing.

Reply: Thanks. I’m sorry that the original description is confusing. The correlation coefficient is between the radiosonde-derived ABLH and lidar-determined ABLH, both HM and CFM are used based on lidar data, so there are two graphs. The vertical coordinate of Figure 7 (Figure 5 in the original manuscript) represents the radiosonde-derived ABLH. The horizontal coordinate of Figure 7(a) is the ABLH retrieved from HM and of Figure 7(b) is the ABLH retrieved from CFM.

The sentence has been written as “Shown as Figure 7, in cloud-free situations, the ABLH determined by HM (or CFM) have high consistencies with the radisonde-determined ABLH with a correlation coefficient of 0.96 (or 0.94), with a mean bias of 0.14 km (or 0.17 km) and with mean relative absolute differences respect to radiosonde retrievals of 10.5% (or 12.3%) (Table 1).”

Figure 7. Comparison between radiosonde-determined (vertical coordinate) and lidar measurement (horizontal coordinate) of ABLH by HM (a) or CFM (b) on 41 cases in cloud-free situations. The correlation coefficients are represented by R. The black solid line is the 1:1 line.

See Figure 7 and lines337-340 of the revised manuscript.

Line 300: based on the graph 6, the conclusion on the role of the top limiter seem to be optimistic. The correlation coefficient depicted in figure 6b is higher than one presented in figure 6a but it not showing a strong improvement. It will be perhaps better to plot the correlation with top limiter and without top limiter on two separate graphs. The improvement in the correlation using the top limiter factor will be better highlighted as it can be observed in figure 4c.

Reply: Thanks. I’m sorry that the original description is confusing. The Figure 6(a) (Figure 8(a) in the revised manuscript) and Figure 6(b) (Figure 8(b) in the revised manuscript) doesn’t represent the comparison of the ABLH determined without the top limiter and with the top limiter. The vertical coordinate represents the radiosoned-derived ABLH. The horizontal coordinate of (a) is the ABLH retrieved by HM and of (b) is the ABLH retrieved by CFM.

In order to show the results clearer, the Figure 6 has been redraw as Figure 8 in the revised manuscript. The blue open dots represent the comparisons results without the top limiter, as the red stars represent the comparisons results below the top limiter. R represents the correlation coefficients.

Figure 8. Similar to Figure 7, but in cloudy situations. The blue open dots represent the comparison results of ABLH determined by theta gradient (vertical coordinate) and determined by HM (a) or CFM (b) (horizon coordinate) without height limitation, the red stars indicates the comparison results after the top limiter if given for HM and CFM. R represents the correlation coefficients, blue represents no top limiter and red represents with the top limiter.

See Figure 8 in the revised version.

Line 315, figure 6. Improve the graphic legend. Include the qualification of the limiter "top" ie ”without cloud top limiter” and "with cloud top limiter". This will improve the comprehension.

Reply: Thanks, we agree, the graphic legend has been rewritten as “without top limiter” and “with top limiter” shown as Figure 8 in the revised version.

Figure 8. Similar to Figure 7, but in cloudy situations. The blue open dots represent the comparison results of ABLH determined by theta gradient (vertical coordinate) and determined by HM (a) or CFM (b) (horizon coordinate) without height limitation, the red stars indicates the comparison results after the top limiter if given for HM and CFM. R represents the correlation coefficients, blue represents no top limiter and red represents with the top limiter.

Line 323: averaging times of NRB and ABLH seem to be not consistent (11 and 10 minutes)?

Reply: Thanks. I’m sorry that there was a mistake in writing, the NRB are averaged every 10 min and ABLH are determined in temporal resolution of 10 min.

The description has been changed to “The NRB are averaged every 10 min, correspondingly, the ABLH are calculated at time interval of 10 min.”

See lines 381-382 of the revised version.

 Line 338: How the time-vertical distribution of Theta e (the potential temperature) is evaluated? It is not explained in the text.

Reply: Thanks. In the revised manuscript the calculation of the potential temperature is introduced as: “The Radiometrics Profiling Radiometers (TP/WVP-3000, Radiometrics) installed at SACOL provides temperature and water vapor profiles from the surface to a height of 10 km……(, is the potential temperature, is the Latent heat of vaporization, is the specific heat at constant pressure, and  are saturated specific humidity and Temperature) could be calculated based on the pressure, temperature and water vapor profiles.”

See lines 164-167 of the revised version.

 Line 358: Figure 5 doesn't show any NRB data below 400 meters. Why in Figure 7 NRB plots show data below the overlap range of 400 meters? Same remark in Figure 8 plots.

Reply: Thanks. The L-band radiosonde at Yuzhong launched twice days on 00:00 and 12:00 UTC. But at 00:00 UTC (08:00 LST), ABLH is often a few hundred meters where the lidar signal is inaccurate because of the overlap factor. The radiosonde-provided temperature profiles only at 12:00 UTC (20:00 LST, local standard time) are selected to evaluate the lidar measurements of the ABLH. Therefore, Figure 5 and Figure 6 (Figure 7 and Figure 8 in the revised version) show comparison results only at 12:00 UTC. Generally, at about 20:00 LST, the ABLH is far higher than 400 m as shown in Figure 9 and Figure 10 in the revised version (at about 20:00 LST).

In the manuscript, the description is: “At 00:00 UTC (08:00 LST), ABLH is often a few hundred meters where the lidar signal is inaccurate because of the overlap factor. The radiosonde-derived profiles only at 12:00 UTC (20:00 LST) are selected to evaluate the lidar measurements of the ABLH”.

See lines 321-323 of the revised manuscript.

References

Toledo, D.; Cã³Rdoba-Jabonero, C.; Adame, J.A.; Morena, B.D.L.; Gil-Ojeda, M.; Estimation of the atmospheric boundary layer height during different atmospheric conditions: a comparison on reliability of several methods applied to lidar measurements. International Journal of Remote Sensing 2017, 38 (11), 3203-3218.

Campbell, J.R.;  Hlavka, D.L.;  Spinhirne, J.D.;  Turner, D.D.; Flynn, C.J.; Operational cloud boundary detection and analysis from micropulse lidar data. Proceeding of the Eighth ARM Science Team Meeting. Tucson: US Department of Energy 1998, 119-122.

Mitrescu, C. Lidar model with parameterized multiple scattering for retrieving cloud optical properties. Journal of Quantitative Spectroscopy & Radiative Transfer, 2005, 94.2(2005):201-224.

Rubio M, Reagan J. A. Optical designs for improving performances of aerosol sensing micro-pulse lidars[J]. Proc Spie, 2001, 4484:25-35.

Author Response File: Author Response.pdf