the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 13 Dec 2024
A New Technique to Retrieve Aerosol Vertical Profiles Using Micropulse Lidar and Ground-based Aerosol Measurements
Abstract. Accurately characterizing the vertical distribution of aerosols and their cloud-forming properties is crucial for understanding aerosol-cloud interactions and their impact on climate. This study presents a novel technique for retrieving vertical profiles of aerosols, cloud condensation nuclei (CCN), and ice nucleating particles (INPs) by combiningmicropulse lidar, radiosonde, and ground-based aerosol measurements. Herein, the technique is applied to data collected by our team at Texas A&M University during the Tracking Aerosol Convection Interactions ExpeRiment (TRACER) campaign. Aerosol size distribution and CCN counter data are used to estimate the value of the aerosol hygroscopicity parameter, κ. The derived κ, together with Mie scattering theory and the relative humidity profiles from the radiosonde, are then used to estimate how much the aerosols have grown at each altitude. This estimate is applied inversely to the aerosol backscatter coefficient profile to produce a dry aerosol backscatter coefficient profile. The dry aerosol backscatter coefficient profile is used to linearly scale surface measurements of aerosol, CCN, and INP concentrations. Combining lidar and ground-based aerosol measurements reduces the assumptions typically needed in lidar-based aerosol retrievals, resulting in a more accurate representation of vertical distributions of aerosol properties. The method could be readily applied to measurements in future field campaigns.
- Preprint
(4068 KB) - Metadata XML
-
Supplement
(1728 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-3363', Anonymous Referee #1, 03 Apr 2025
The manuscript presents a methodology to convert lidar backscatter profiles into cloud condensation nuclei (CCN) and ice nuclei (IN) number concentration profiles by scaling the lidar-derived profiles to in situ data measured at ground level. It describes the method with emphasis on the lidar analysis, cloud screening, and hygroscopic correction, and demonstrates its application through several case studies. Additionally, it compares measurements from two locations to evaluate the approach's potential. While the manuscript is well written, it lacks essential details regarding the methodology, particularly in describing error sources, and requires revisions from the authors.
Major Comments:
Although correcting for hygroscopic growth is a necessary step in converting backscatter profiles into number concentration profiles, the manuscript devotes excessive attention to this aspect while providing limited discussion on the relationship between aerosol optical and microphysical properties. It also offers technical details regarding the lidar data analysis, but omits essential aspects of the retrieval scheme.
A major drawback, in my opinion, is the large uncertainty in lidar backscatter retrieval at near ranges. While the authors perform an uncertainty analysis, they do not account for uncertainties arising from the overlap function, which can significantly affect the measurements in the near range. This issue is particularly relevant for a micropulse lidar (MPL), which typically has a full overlap height of about 5–6 km (Campbell et al., 2002), and around 2 km for MiniMPL (as I could only find here: https://www-air.larc.nasa.gov/missions/discover-aq/docs/pub/AMS_Berkoff_Finalv2.pdf ).
Campbell, J. R., D. L. Hlavka, E. J. Welton, C. J. Flynn, D. D. Turner, J. D. Spinhirne, V. S. Scott , and I. H. Hwang: Full-Time, Eye-Safe Cloud and Aerosol Lidar Observation at Atmospheric Radiation Measurement Program Sites: Instruments and Data Processing. J. Atmos. Oceanic Technol., 19, 431–442, https://doi.org/10.1175/1520-0426(2002)019<0431:FTESCA>2.0.CO;2, 2002
Another missing aspect is a discussion on the scaling of the backscatter coefficient to CCN/INP values. Despite having time-resolved lidar and in situ measurements, the study scales a single profile to a single CCN/INP value. How large was the scaling factor at different supersaturations, i.e., CCN/β_dry(R_0) and INP/β_dry(R_0)? More importantly, how did this factor evolve with time? Was there a correlation between backscatter values and CCN concentrations over time? Could you identify distinct patterns for different aerosol types? Addressing these questions could provide insight into the feasibility of retrieving CCN/INP profiles using lidar alone, as intended in previous studies referenced in the introduction.
Specific Comments:
Lines 78-79: Another commonly considered aerosol type is marine aerosol.
Lines 124-126: You mention that models and observations often rely on assumptions that introduce biases. Could you provide references to support this statement?
Line 166: You scale the aerosol backscatter to CCN and INP concentrations at ground level. However, how does this scaling evolve over time? What is the scaling factor? This crucial aspect is missing from the discussion.
Equation (2): Please use \exp instead of exp.
Figures 4e, 7, and 8: The profiles exhibit a steady increase in signal and backscatter products from 3 to 8 km. Is this a systematic artifact? Please address this pattern, which is visible in most presented profiles. Could it be associated with the overlap correction or other corrections or filtering?
Line 379: You assume a refractive index value. How does this assumption impact the humidity-effect correction? How much would your results change if the real part varied between 1.35 and 1.55? What about variations in the imaginary part?
Line 374: The step of computing aerosol size growth once the kappa value is determined is critical. I recommend including more details about this calculation beyond citing Petters and Kreidenweis (2007), such as providing the corresponding expression(s) or procedures.
Line 376: Please describe the Mie simulations in more detail. Which Mie package or code was used? Including the corresponding mathematical expressions would also be beneficial.
Line 381: Do you mean that aerosol size distributions are measured under dry ambient conditions? I assume the particles are dried before size distribution measurements, making the process independent of air humidity. Please clarify.
Line 391: The choice between the extinction and backscatter coefficient is important. While a constant lidar ratio holds for a given humidity, varying humidity also alters the lidar ratio, affecting the growth correction factor differently for the extinction and backscatter coefficients. Additionally, since the backscatter coefficient is the robust retrieved property, your Mie calculations should consider the backscattering cross-section instead of the extinction cross-section.
Lines 429-431: The quadratic fitting used to extend the retrieval beyond blind ranges appears to be an extrapolation. Could you provide more details on how this step works?
Figure 7c, 8d, 9d, 10d: I find it difficult to interpret these profiles. While I understand that you scale to different INP values at different temperatures, the approach differs from CCN concentrations, where supersaturation determines the number. Since INP concentration depends on air temperature, there should be a single INP profile based on the dry backscatter value and temperature. However, this relationship has not been discussed in the manuscript. Further analysis or clarification is needed for the INP retrievals.
Lines 533-534: You attribute differences between the two systems to slight misalignment. Could you clarify which misalignment you are referring to? Since MPL systems typically have a transceiver configuration, how would misalignment affect the near-range signal but not the far-range? These differences might instead stem from errors in the overlap function used to correct the signals. The increased signal between 6 and 8 km in the MiniMPL lidar suggests possible system artifacts (Fig. 11). When presenting the lidar analysis, I recommend including signals before and after overlap correction, as well as the overlap function itself. Additionally, how stable is this function over time?
Reference List: The absence of URLs for each reference makes reading and reviewing more difficult. Additionally, some journal names are incomplete.
Citation: https://doi.org/10.5194/egusphere-2024-3363-RC1 -
AC1: 'Reply on RC1', Bo Chen, 03 Jun 2025
We thank the editor and reviewers for their time and constructive comments. Below, we reproduce each comment followed by our response.
RC1: 'Comment on egusphere-2024-3363', Anonymous Referee #1, 03 Apr 2025
The manuscript presents a methodology to convert lidar backscatter profiles into cloud condensation nuclei (CCN) and ice nuclei (IN) number concentration profiles by scaling the lidar-derived profiles to in situ data measured at ground level. It describes the method with emphasis on the lidar analysis, cloud screening, and hygroscopic correction, and demonstrates its application through several case studies. Additionally, it compares measurements from two locations to evaluate the approach's potential. While the manuscript is well written, it lacks essential details regarding the methodology, particularly in describing error sources, and requires revisions from the authors.
Major Comments:
Referee’s comment: Although correcting for hygroscopic growth is a necessary step in converting backscatter profiles into number concentration profiles, the manuscript devotes excessive attention to this aspect while providing limited discussion on the relationship between aerosol optical and microphysical properties. It also offers technical details regarding the lidar data analysis, but omits essential aspects of the retrieval scheme.
Authors’ response: Thank you for this comment. We agree with the Reviewer that correcting for hygroscopic growth is an important aspect of correctly determining aerosols concentration profiles from the lidar backscatter profiles. We also see Reviewer’s point, that we have developed a very detailed method but had not described the method in enough detail in our original submission. Based on this comment, we have now expanded the text providing additional details and explanations as specified in the comments below.
Referee’s comment: A major drawback, in my opinion, is the large uncertainty in lidar backscatter retrieval at near ranges. While the authors perform an uncertainty analysis, they do not account for uncertainties arising from the overlap function, which can significantly affect the measurements in the near range. This issue is particularly relevant for a micropulse lidar (MPL), which typically has a full overlap height of about 5–6 km (Campbell et al., 2002), and around 2 km for MiniMPL (as I could only find here: https://www-air.larc.nasa.gov/missions/discover-aq/docs/pub/AMS_Berkoff_Finalv2.pdf ).
Authors’ response: We agree with the Reviewer that a major limitation of lidar data arises from uncertainties in the overlap region. In the original submission, the overlap corrections or both the micropulse lidar (MPL) and the mini micropulse lidar (MiniMPL) instruments were discussed in some detail on page 9 in the main text and on page 1-2 in the supplement. Below we provide more detail about the overlap correction function used for this study.
Figure R1 TAMU MiniMPL overlap function used for TRACER is shown in solid light blue line and the overlap function provided by the vendor is shown in solid dark blue line. The difference between the overlap function used for TRACER and the overlap function provided by DOE ARM is shown in dashed red line. ARM supplied MPL overlap function shown in orange line.
For any lidar instrument, the so-called “overlap” is the terminology used to describe the geometric mismatch between the outgoing laser beam and the detector’s field of view, which occurs at near range. In this range, the receiver cannot capture the full backscattered signal. The manufacturer, Droplet Measurement Technologies, provides both an overlap function (generated in December 2021) and a method for recalibrating the overlap function. Changes in the overlap function over time may be caused by gradual mechanical shifts in the optical components due to vibrations, and thermal expansion and contraction. For comparison, Figure R1 here shows the overlap function which came with the unit (shown in solid dark blue line), generated in December 2021, and a post-campaign function generated in our laboratory on February 21, 2023 (shown in solid light blue line). Following Welton and Campbell (2002), we recalibrated the MiniMPL overlap function by positioning the lidar horizontally, so the laser beam was parallel to the ground and free of obstructions while collecting calibration data. The MiniMPL overlap function supplied by the vendor and the post-campaign overlap function are similar overall. However, small differences at close range indicate a slight drift in the overlap function over time. The maximum percentage difference between the pre-campaign vendor and post-campaign MiniMPL overlap functions is 10.9% at 0.53 km. Deviations decrease to below 5% by 1 km and approach zero above roughly 3 km. We selected the post campaign overlap function, as it is closer in time to the TRACER campaign. For the MPL unit at the Department of Energy Atmospheric Radiation Measurement (ARM) site, ARM supplied the overlap correction function for the MPL for the TRACER campaign.
Figure R2 comparison of MiniMPL normalized relative backscatter (NRB) profiles calculated using the original vendor-supplied overlap function shown in blue scatter points and the newly derived overlap function shown in red scatter points.
To further justify the choice of the MiniMPL overlap function, we crosschecked NRB profiles from the ARM MPL and MiniMPL, which the MiniMPL NRB during a time when the ROAM-V was parked at the ARM site and the lidars were collocated, as discussed in the original text on page 25-26. Figure R2 compares the normalized NRB from the TAMU miniMPL to the collocated ARM MPL, showing that the post-TRACER overlap function yields stronger agreement (R² = 0.986) than the vendor-provided function (R² = 0.976), further supporting it as the more accurate choice.
Based on the referee’s comment, we now include the raw MiniMPL and MPL signals before performing overlap correction in Figure 11 to demonstrate the effectiveness of the overlap correction as well as additional discussion in the text.
Revised Figure 11. Comparison of retrieved aerosol backscatter coefficient profiles derived from MiniMPL and ARM AMF-1 MPL data. (a) Raw co-polarized lidar signal of TAMU MiniMPL (red solid line) and ARM MPL (blue solid line) (b) Lidar normalized relative backscatter signal of TAMU MiniMPL (red solid line) and ARM MPL (blue solid line) after applying corrections (c) Retrieved lidar aerosol backscatter coefficient of TAMU MiniMPL (orange solid line and area) and ARM MPL (blue solid line and area) (d) Comparison of the lidar aerosol backscatter coefficients retrieved from NRB. The uncertainty interval of the retrieved aerosol backscatter coefficient is shown as error bars.
On page 27 from line 574, the text now reads, “The TAMU ROAM-V was deployed at the AMF1 (LaPorte, Texas) site on 1 September 2022, allowing MiniMPL and MPL to be collocated and compared directly. The ARM MPL deployed at AMF-1 collects data at a vertical resolution of 15 m and a temporal resolution of 10 s (Muradyan, 2021). During the colocation test, the two lidars were separated horizontally by approximately 30 meters and vertically by less than 10 meters. The data of both lidars are time-averaged between 20:00 and 22:00 UTC. Vertical profiles of the lidar raw signal, the NRB, and the aerosol backscatter coefficient, and a comparison of the lidar aerosol backscatter coefficient are shown in Figure 11, a, b, c, and d, respectively. Figures 11a shows that the raw signals from the 2 lidars differ significantly. However, after applying lidar-specific afterpulse, deadtime, background, and range correction for each lidar, their NRB profiles agree closely (Figure 11b). Figures 11c and 11d show that the MPL and MiniMPL NRB and aerosol backscatter coefficient profiles follow similar shapes and magnitudes. The MiniMPL overestimates aerosol backscatter coefficients between 6 km and 8 km compared to the MPL, suggesting that the MiniMPL-derived profiles may be less reliable at higher altitudes. This artifact is consistent with the spurious high-altitude enhancements discussed earlier and is likely caused by signal noise and overlap correction uncertainty in the MiniMPL retrieval. The MiniMPL and MPL profiles exhibit a slight vertical offset below 4 km, which may result from residual errors introduced during the afterpulse, background, or overlap corrections. The differences between the two aerosol backscatter profiles generally remain within the estimated uncertainty bounds, which primarily arise from the assumed lidar ratio and the scattering ratio at the reference height. In summary, MiniMPL and MPL data are remarkably similar despite differences in their lidar designs and specifications. This agreement suggests that the more compact and less expensive MiniMPL can provide comparable data quality to the more established MPL system. In addition, the use of two lidars with comparable outputs enables coordinated deployment and consistent analysis across different sites over the same period.”
Figure R1 has been add to the supplement with explanation of the overlap functions used for TRACER. The text in supplement at line 37 now reads “‘Overlap’ refers to the near-range mismatch between the outgoing laser beam and the detector’s field of view, which prevents full signal capture. Droplet Measurement Technologies provided a default overlap function (Dec 2021) and a method for recalibration (Welton and Campbell, 2002). Following their method, we recalibrated the miniMPL overlap on February 21, 2023, by aligning the instrument horizontally to collect calibration data. As shown in Figure R1, the vendor and post-campaign overlap functions are similar, but small differences at close range indicate gradual drift up to 10.9% at 0.53 km, decreasing to <5% by 1 km. We used the post-campaign overlap due to its closer timing to TRACER. The ARM MPL overlap function was supplied by the ARM program.”
Referee’s Comment: Another missing aspect is a discussion on the scaling of the backscatter coefficient to CCN/INP values. Despite having time-resolved lidar and in situ measurements, the study scales a single profile to a single CCN/INP value. How large was the scaling factor at different supersaturations, i.e., CCN/β_dry(R_0) and INP/β_dry(R_0)? More importantly, how did this factor evolve with time? Was there a correlation between backscatter values and CCN concentrations over time? Could you identify distinct patterns for different aerosol types? Addressing these questions could provide insight into the feasibility of retrieving CCN/INP profiles using lidar alone, as intended in previous studies referenced in the introduction.
Authors’ response: We did not use a scaling factor to convert backscatter values directly into CCN concentrations, nor did we attempt to derive an empirical relationship between CCN and aerosol optical properties. Instead, we used the aerosol dry backscatter profile to linear scale aerosol, CCN, and INP concentrations measured at the surface at the same time the backscatter profile was taken to get CCN and INP concentrations at higher altitudes. For example, Figure 7, which shows the retrieved CCN profiles at 0.2%, 0.6%, and 1.2% supersaturation, and INP profiles at –20 °C and –25 °C. This approach does not rely on fixed scaling factors and can be applied to other field campaigns using similar lidar and aerosol instrumentation without requiring location-specific calibration. Extending our method to situations without surface aerosol measurements is beyond the scope of this work.
We retrieve one aerosol profile for each radiosonde launch using data averaged around the launch time, instead of a full time series. This is because radiosonde data are needed for hygroscopic growth correction, and they are not collected often. Other instruments also need time to collect enough data—CCN counters take 30 minutes to scan supersaturations, and INP samples are collected over longer periods. MiniMPL data also need about an hour of averaging to reduce noise, especially when there is cloud. Because of these limits, our method focuses on getting reliable, meaningful profiles rather than capturing short-term changes. Studying how CCN and INP profiles change over time is still important and could be explored in future campaigns with more advanced lidar systems.
To further clarify our method, we include more detailed explanations in the manuscript.
Page 8, line 167. The text now reads: “Each profile is retrieved from data collected over a 1 to 2 hour period around radiosonde launch time.”
on page 8, line 170. The text now reads: “This method was developed to address the fact that the aerosol size distribution, composition, shape, and hygroscopic growth, all of which influence aerosol backscatter, are not directly measured and must be inferred through additional information or assumptions. By assuming that the surface aerosol properties are representative of the column, the aerosol volume number density becomes approximately linearly related to the dry backscatter coefficient. This allows surface aerosol, CCN, and INP concentrations to be scaled by the lidar-derived dry backscatter profile.”
Page 16, line 324. A new paragraph is added: “The aerosol, CCN, and INP measurements were averaged over the same 1 to 2 hour window as the lidar data used for backscatter profile retrieval. This averaging period reflects the operational constraints of each instrument: the CCN counter requires around 30 minutes to complete a full scan of supersaturations, INP samples were collected over extended durations, and aerosol size distributions benefit from temporal averaging to obtain more representative size distributions”
Page 22, line 453. A new sentence is added: “One profile each for aerosol, CCN, and INP is retrieved for each time-averaging period.”
Reply to specific comments:
- 1. Lines 78-79: Another commonly considered aerosol type is marine aerosol.
Authors’ response: The three aerosol types mentioned in the text refers to the aerosol types mentioned in a study that did not consider the marine aerosol (Lv et al., 2018). The main text at 78-79 is now edited to reduce ambiguity.
Page 3, Line 77, the text now reads: “The first approach involves using multi-wavelength lidar to retrieve aerosol concentrations by classifying them into different aerosol types (urban, biomass burning, and dust) and then using the prescribed hygroscopicity parameter of each aerosol type to estimate the CCN concentration (Lv et al., 2018)”
- 2. Lines 124-126: You mention that models and observations often rely on assumptions that introduce biases. Could you provide references to support this statement?
Authors’ response: We appreciate the reviewer’s comment. Upon review, we recognized that the original statement lacked clear meaning and sufficient support from the literature. In response, we have deleted this statement.
Page 5, Line 123 now reads: “Despite advancements in understanding aerosol–cloud interactions, significant uncertainties remain in accurately characterizing aerosol vertical distributions and their impact on cloud processes, requiring more comprehensive and vertically resolved measurements.”
- 3. Line 166: You scale the aerosol backscatter to CCN and INP concentrations at ground level. However, how does this scaling evolve over time? What is the scaling factor? This crucial aspect is missing from the discussion.
Authors’ response: As discussed previously, we do not use a prescribed scaling factor. Rather, we use the measured CCN and INP measurements collected on the ground to scale the MPL backscatter profile. We estimate the vertical profile of CCN concentration at a given supersaturation by scaling the lidar-derived dry aerosol backscatter profile relative to its surface value and applying it to the surface CCN measurement. Similarly, we estimate the vertical profile of frozen INP concentration by scaling the aerosol profile using the measured surface INP concentration and the INP concentration at a given temperature measured on the ground. For this study, the limiting factor in time resolution is the frequency of ROAM-V radiosonde launches, from which humidity profiles are observed. We retrieve one aerosol profile for each radiosonde launch, representative of the time average of surface aerosol and lidar measurements over the ~2 hour period surrounding the radiosonde launch.
Page 7, Line 164 now reads: “The resulting dry aerosol backscatter coefficient profile is used to linearly scale time-averaged surface aerosol concentration, CCN concentration, and INP concentration measurements to estimate their vertical distributions.”
- 4. Equation (2): Please use \exp instead of exp.
Authors’ response: We have edited this in the main text. The equation now is:
(2)- 5. Figures 4e, 7, and 8: The profiles exhibit a steady increase in signal and backscatter products from 3 to 8 km. Is this a systematic artifact? Please address this pattern, which is visible in most presented profiles. Could it be associated with the overlap correction or other corrections or filtering?
Authors’ response: Yes, it is likely that the steady increase in aerosol backscatter from 3 to 8 km is a systematic artifact. There are two probable causes for this pattern. The first, as suggested by the referee, is that an overestimated overlap correction at near range may have led to underestimation of the lidar signal at lower altitudes. Although the overlap correction was carefully implemented following established procedures, some residual uncertainty may remain, especially near the instrument’s full-overlap height. The second possible cause is a positive bias in the noise floor at higher altitudes, which can artificially elevate the apparent signal and backscatter coefficients in the upper portion of the profile.
Based on the case studies presented in the paper, the magnitude of the artificial increase in aerosol backscatter at higher altitudes appears to be closely related to the noise level in the raw lidar signal. All three cases (Case 1, 2, and 3) use a total averaging period of at least two hours. However, Case 3 (31 August 2022) is entirely cloud-free, allowing for more profiles to be averaged. This results in a smoother signal with lower noise and the smallest observed artifact. In contrast, Case 2 (26 August 2022) contains only a few cloud-free profiles suitable for averaging, which leads to a higher noise level and the most prominent artifact, particularly between 5 and 6 km altitude. These patterns support the second explanation—that the MiniMPL signal noise at higher altitudes may have a small positive bias. As a result, MPL-derived aerosol profiles above 4 km should be interpreted with caution. We acknowledge this issue and will clarify it in the revised manuscript.
Method section 2.5, Page 23, Line 474, the text now reads: “It is important to acknowledge that lidar-derived aerosol profiles may be affected by the artificial increase in aerosol backscatter at higher altitudes, likely caused by overlap correction errors and noise bias. As seen in Figure 7a and 7b, the aerosol backscatter coefficient shows a steady increase with height above 4 km. This apparent increase is likely a systematic artifact related to lidar signal noise at higher altitudes. As a result, the retrieved aerosol profile above 4 km should be interpreted with caution.”
Result section 3.1, Page 24, line 509, the text now reads: “The increase of dry aerosol backscatter profile as well as the aerosol concentration profile between 6 and 8 km is likely a systematic artifact related to the lidar noise at high altitude.”
Result section 3.2, Page 26, line 538, the text now reads: “The increase in the dry aerosol backscatter and aerosol concentration between 5 and 7 km is a systematic artifact likely caused by high lidar signal noise, as shown in Figure 9a above 5 km. The magnitude of this artifact is likely amplified by the high noise level, which is caused by the limited number of cloud-free profiles available for averaging during this period, as compared to the previous case where more cloud-free profiles led to reduced noise and less pronounced artifacts.”
Result section 3.4, Page 28, line 584, the text now reads: “The MiniMPL overestimates aerosol backscatter coefficients between 6 km and 8 km compared to the MPL, suggesting that the MiniMPL-derived profiles may be less reliable at higher altitudes.”
- 6. Line 379: You assume a refractive index value. How does this assumption impact the humidity-effect correction? How much would your results change if the real part varied between 1.35 and 1.55? What about variations in the imaginary part?
Authors’ response: It is common practice to assume the refractive index of ammonium sulfate when the exact aerosol composition is unknown (Ghan and Collins, 2004; Zieger et al., 2013). In our approach, we compute the aerosol’s refractive index at each relative humidity (RH) by taking a volume-weighted average of the dry aerosol (ammonium sulfate) refractive index and that of water, and assuming the imaginary part of the refractive index is considered as 0 in all cases. It appears that this has a small effect on our calculations of the lidar hygroscopic growth correction factor, as illustrated below (Figure R3). However, this assumption represents a limitation of our analysis, as real atmospheric aerosols can have various different refractive indices depending on their chemical composition and mixing state. In reality, aerosol mixtures often include absorbing species and complex internal or external mixing, making the effective refractive index difficult to constrain.
Figure R3 Lidar hygroscopic growth correction factor calculated with constant refractive with humidity (a), and with volume-weighted average of the dry aerosol refractive index and that of water.
Page 19, line 398, the text now reads: “In the absence of detailed aerosol composition data, the refractive index of ammonium sulfate is frequently adopted as a representative value in aerosol optical calculations, as it provides a reasonable approximation for non-absorbing, hygroscopic particles (Ghan and Collins, 2004; Zieger et al., 2013).”
- 7. The step of computing aerosol size growth once the kappa value is determined is critical. I recommend including more details about this calculation beyond citing Petters and Kreidenweis (2007), such as providing the corresponding expression(s) or procedures.
Authors’ response: To determine the wet diameter at each RH value, we solve for the point at which the saturation ratio predicted by the κ-Köhler theory matches the specified environmental saturation ratio. This is done through an iterative root-finding approach, using the dry diameter as the initial guess. Essentially, we solve for aerosol wet diameter given saturation ratio and the aerosol dry diameter.
Page 19, line 380, the text now reads: “Thus, the problem becomes finding the κ corresponding to a given Dp,c as the dry diameter, to match a specific SS as the output. The κ is then numerically determined using an iterative root-finding method to match the measured SSc-Dp,c pairs.”
Page 19, line 391, the text now reads: “To determine the wet diameter at each RH value, we solve for the point at which the saturation ratio predicted by the κ-Köhler theory matches the specified environmental saturation ratio. This is done through an iterative root-finding approach, using the dry diameter as the initial guess.”
- 8. Line 376: Please describe the Mie simulations in more detail. Which Mie package or code was used? Including the corresponding mathematical expressions would also be beneficial.
Authors’ response: The MiePython package is used to calculate the aerosol extinction coefficient at each RH value (Prahl, 2023). Since this is an open-source software and is a standard Mie scattering code, we chose to leave the mathematical expressions out of the text.
The revised text now reads: Page 19, line 396, the text now reads: “Once the aerosol size is known as a function of RH, the MiePython package is then used to calculate the aerosol extinction coefficient at each RH value (Prahl, 2023).”
- 9. Line 381: Do you mean that aerosol size distributions are measured under dry ambient conditions? I assume the particles are dried before size distribution measurements, making the process independent of air humidity. Please clarify.
Authors’ response: Yes. The inlet air is dried below 30% RH which we measure with an RH meter.
Page 19, line 403, the text now reads: “During the field campaign, aerosol size distribution measurements are made after the sample air is dried to below 30% relative humidity.”
- 10. Line 391: The choice between the extinction and backscatter coefficient is important. While a constant lidar ratio holds for a given humidity, varying humidity also alters the lidar ratio, affecting the growth correction factor differently for the extinction and backscatter coefficients. Additionally, since the backscatter coefficient is the robust retrieved property, your Mie calculations should consider the backscattering cross-section instead of the extinction cross-section.
Authors’ response: The reviewer is correct that in lidar measurements, the backscatter coefficient is the more robustly retrieved property. This is why we use the backscatter coefficient as the primary aerosol optical property in our analysis. However, when doing Mie scattering calculation, the extinction cross-section is usually more robust. The extinction cross-section is smoother and less sensitive to small changes in refractive index or size distribution. The reason is that the extinction cross-section integrates over the entire scattering angle, thus it’s less noisy and has less angular dependence. Backscatter cross-section, on the other hand, depends heavily on the exact phase function at 180°, making it more sensitive to size, refractive index, and numerical precision, thus less robust in Mie calculations, as noted in a previous study (Geisinger et al., 2017). In addition, the extinction cross-section calculation is only used to compute the relative change in aerosol backscatter coefficient. Extinction can be used because the relative growth due to humidity is similar for both extinction and backscatter for real atmospheric aerosols, as noted in previous studies (Ghan and Collins, 2004; Ghan et al., 2006). The extinction-derived lidar hygroscopic growth correction factor f(RH) offers a stable and physically reasonable correction for lidar backscatter profiles.
A justification for using extinction calculation is included in the text.
Page 20, line 411, the text now reads, “The extinction coefficient σ, rather than the backscatter coefficient, is used because the extinction coefficient is numerically more stable than the backscatter coefficient in Mie scattering calculations, exhibiting lower sensitivity to uncertainties in particle size distribution and refractive index (Geisinger et al., 2017).”
- 11. Lines 429-431: The quadratic fitting used to extend the retrieval beyond blind ranges appears to be an extrapolation. Could you provide more details on how this step works?
Authors’ response: Yes, the quadratic fitting is an extrapolation. No direct lidar measurements are available in the blind zone (0-100 m for MiniMPL and 0-200m for MPL). To estimate the aerosol backscatter coefficient profile below the lidar’s blind zone, we perform a second-degree polynomial fit to the dry aerosol backscatter profile from up to 300 m AGL down to the edge of blind zone. This fitted curve is then extrapolated into the blind zone. Since the aerosol profile is later linearly scaled by the dry backscatter profile, having a physically reasonable estimate in the blind zone is necessary to ensure that the scaling reflects realistic near-surface conditions. The extrapolated portions of the dry backscatter coefficient profile within the blind zone are shown as dotted lines in Figure 7a. The manuscript has been revised to explain in more detail.
Page 22, line 455, the text now reads: “To estimate the aerosol backscatter coefficient profile within the lidar’s blind zone, we perform a second-degree polynomial fit to the dry aerosol backscatter profile from up to 300 m AGL down to the edge of blind zone. This fitted curve is then extrapolated into the blind zone. Since the aerosol profile is later linearly scaled by the dry backscatter profile, having a physically reasonable estimation of aerosol profile in the blind zone is necessary to ensure that the scaling reflects realistic near-surface conditions. The extrapolated portions of the dry backscatter coefficient profile within the blind zone are shown as dotted lines in Figure 7a.”
- 12. Figure 7c, 8d, 9d, 10d: I find it difficult to interpret these profiles. While I understand that you scale to different INP values at different temperatures, the approach differs from CCN concentrations, where supersaturation determines the number. Since INP concentration depends on air temperature, there should be a single INP profile based on the dry backscatter value and temperature. However, this relationship has not been discussed in the manuscript. Further analysis or clarification is needed for the INP retrievals.
Authors’ response: As described above, CCN concentrations are measured at the surface at different supersaturations and linearly scaled using the dry aerosol backscatter profile to generate vertical profiles of CCN concentration at corresponding supersaturations. Similarly, INP concentrations are measured at the surface at different temperatures and linearly scaled using the dry aerosol backscatter profile to generate vertical profiles of INP concentration. As a result, both CCN and INP profiles follow the shape of the dry aerosol backscatter profile and can be implemented in models where they are transported and interact with clouds.
Additional explanation is now added to Page 22, line 468: “CCN concentration profiles are presented at different supersaturations, and INP concentration profiles are presented at different activation temperatures. Presenting CCN and INP profiles this way is useful for modeling applications, as it allows the model to compute CCN and INP activation dynamically when the particles are transported to conditions supportive of cloud condensation or ice nucleation within the modelled convection (or other atmospheric processes of interest).”
- 13. Lines 533-534: You attribute differences between the two systems to slight misalignment. Could you clarify which misalignment you are referring to? Since MPL systems typically have a transceiver configuration, how would misalignment affect the near-range signal but not the far-range? These differences might instead stem from errors in the overlap function used to correct the signals. The increased signal between 6 and 8 km in the MiniMPL lidar suggests possible system artifacts (Fig. 11). When presenting the lidar analysis, I recommend including signals before and after overlap correction, as well as the overlap function itself. Additionally, how stable is this function over time?
Authors’ response: Thanks for pointing out the increased signal with height between 6 and 8 km in the MiniMPL lidar. This increase in signal is related to the previously discussed artifact. By “misalignment”, we meant that the shape of the profile seems to have a vertical offset. Since ‘misalignment’ could be misinterpreted, it is changed to ‘offset’ in the main text. Secondly, we like the Referee’s suggestion of including signals before and after overlap correction, and we have added them to Figure 11, as discussed in an earlier comment.
Page 28, line 588, the text now reads: “The MiniMPL and MPL profiles exhibit a slight vertical offset below 4 km, which may result from residual errors introduced during the afterpulse, background, or overlap corrections.”
- 14. Reference List: The absence of URLs for each reference makes reading and reviewing more difficult. Additionally, some journal names are incomplete.
Author’s response: Thanks for helping identifying issues in the reference section. The URL for the doi has been added, and the missing journal names have been completed. The updated references for the manuscript is included in the supplement of this reply.
References:
Geisinger, A., Behrendt, A., Wulfmeyer, V., Strohbach, J., Förstner, J., and Potthast, R.: Development and application of a backscatter lidar forward operator for quantitative validation of aerosol dispersion models and future data assimilation, Atmospheric Measurement Techniques, 10, 4705-4726, https://doi.org/10.5194/amt-10-4705-2017, 2017.
Ghan, S. J. and Collins, D. R.: Use of in situ data to test a Raman lidar–based cloud condensation nuclei remote sensing method, Journal of Atmospheric and Oceanic Technology, 21, 387-394, https://doi.org/10.1175/1520-0426(2004)021<0387:UOISDT>2.0.CO;2, 2004.
Ghan, S. J., Rissman, T. A., Elleman, R., Ferrare, R. A., Turner, D., Flynn, C., Wang, J., Ogren, J., Hudson, J., and Jonsson, H. H.: Use of in situ cloud condensation nuclei, extinction, and aerosol size distribution measurements to test a method for retrieving cloud condensation nuclei profiles from surface measurements, Journal of Geophysical Research: Atmospheres, 111, https://doi.org/10.1029/2004JD005752, 2006.
Lv, M., Wang, Z., Li, Z., Luo, T., Ferrare, R., Liu, D., Wu, D., Mao, J., Wan, B., and Zhang, F.: Retrieval of cloud condensation nuclei number concentration profiles from lidar extinction and backscatter data, Journal of Geophysical Research: Atmospheres, 123, 6082-6098, https://doi.org/10.1029/2017JD028102, 2018.
Prahl, S.: miepython: Pure python implementation of Mie scattering, Version v2, 5, https://doi.org/10.5281/zenodo.8218010, 2023.
Welton, E. J. and Campbell, J. R.: Micropulse lidar signals: Uncertainty analysis, Journal of Atmospheric and Oceanic Technology, 19, 2089-2094, https://doi.org/10.1175/1520-0426(2002)019%3C2089:MLSUA%3E2.0.CO;2, 2002.
Zieger, P., Fierz-Schmidhauser, R., Weingartner, E., and Baltensperger, U.: Effects of relative humidity on aerosol light scattering: results from different European sites, Atmospheric Chemistry and Physics, 13, 10609-10631, https://doi.org/10.5194/acp-13-10609-2013, 2013.
-
AC1: 'Reply on RC1', Bo Chen, 03 Jun 2025
-
RC2: 'Comment on egusphere-2024-3363', Anonymous Referee #2, 08 Jul 2025
Summary: This paper uses ground-based in situ observations of CCN and aerosol concentrations in combination with lidar profiles to create vertical profiles of CCN, aerosol, and INP concentrations. The focus of the analysis is correcting the lidar backscatter profiles observed at varying levels of RH to dry backscatter profiles to avoid the disconnect between optical properties and aerosol concentrations at high RH. The paper is well-written and the methods seem thorough. However, the paper is missing more discussion on the other factors that can cause optics to not be linearly correlated with aerosol concentrations, as well as a discussion of the implications of the time averaging done to retrieve aerosol/CCN concentrations from a single backscatter profile.
Major Comments:
- If I’m understanding correctly (Line 194-196), all cloud-free profiles from 2-4 hours of lidar observations are averaged into a singular backscatter profile used to retrieve the aerosol/CCN/INP profiles. More discussion of the implications of this step would strengthen the analysis, because it seems like a long time to average over. On average, how many profiles are included in a 2-4 hour average? Is homogeneity of the aerosol profile over this time scale in this region a good assumption? How does the standard deviation associated with time averaging compare to the backscatter uncertainty already shown on the figures? In lines 569-574, there is discussion of the spatial heterogeneity of aerosol vertical profiles between Galveston and LaPorte, which are only 46km apart. Given the potential for air mass transport over tens of kilometers during a 2-4 hour time frame, the assumption of temporal heterogeneity may lead to some vertical features being washed out. A quantitative assessment of the temporal variability would enhance confidence in the resultant aerosol profile retrievals.
- I would recommend replacing “accurate” with “realistic” in Line 585 and elsewhere in the paper, because “accurate” implies validation against independent observations, which is not presented here. “Realistic" better reflects that the profiles follow expected physical behavior. Additionally, additional discussion of the scenarios in which this method will not work would strengthen the final section. It is alluded to earlier in the paper (Line 409) that the approach only holds for well-mixed aerosol layers, but I think this needs to be reiterated here. As soon as the aerosol chemical composition or size distribution differs from what is measured at the surface, there will be errors introduced as the exact relationship between aerosol/CCN concentration and optical properties changes. It may also be beneficial (if you have a number) to discuss how often such well-mixed aerosol layer cases for which this method holds were observed in the TRACER campaign.
Minor Comments:
- Line 30: Can you clarify why the focus here is on convective processes? Assuming because it’s mostly what is observed in the Houston area, but this makes it sound like aerosols don’t also impact stratocumulus cloud processes.
- Line 81: “multiwavelength” not capitalized
- Line 85: “relationship between”
- Line 105: Clarify what exactly “dry” ambient conditions mean for this case
- Line 145: “first” not capitalized
- Line 145: What defines an enhanced operation day? Is it dependent on atmospheric conditions or just a pre-determined day to deploy additional platforms?
- Fig 2: Can you write out what NRB stands for? Since I don’t think it’s a super common acronym, this could be confusing for someone scrolling through the figures first before reading.
- Line 245-249: I am assuming this is expected in this region due to a primarily polluted urban boundary layer but maybe state it a little more explicitly if this is what you’re implying.
- Fig 6: Can you make panel (a) wider? The numbers on the top really run together and are difficult to read
- Line 379: How does ammonium sulfate compare to/represent the aerosol types typically observed during TRACER? Somewhere when you’re describing the campaign can you discuss the typical/dominant aerosol types observed during the campaign?
- Should there be a section in here somewhere (even if brief) about the radiosonde observations and the uncertainty of their temperature/RH profiles?
- Fig 7: Would a log-scale for the x-axis be helpful? It’s hard to see what’s going on near the y-axis with all the lines really close together.
- Fig 8: Could panels b, c, and d be put below panel a so all are larger and easier to read?
- Lines 454, 469-471: Previously, you mention this approach only holds for well-mixed layers, but here you are using an example with an elevated aerosol layer. I would mention that this does not impact the main point of this section, which is demonstrating the hygroscopic growth correction. Either that, or maybe don’t even address the elevated aerosol layer or only do it at the end? Right now, it feels like you’re picking a bad case right out of the gate immediately after finishing the methodology.
- Fig 9: Is it worth addressing the multiple aerosol layers here? This case also doesn’t appear super well-mixed to me.
- Line 498: Could this be the first case study you show? It might be good to show the most straightforward application first to give a reader confidence in your method before you introduce complications of multiple cloud layers or more pronounced hygroscopic growth.
- Line 518: The placement of this section between your aerosol profile results felt a little confusing – would it make more sense in the methods section maybe?
Citation: https://doi.org/10.5194/egusphere-2024-3363-RC2 - AC2: 'Reply on RC2', Bo Chen, 18 Aug 2025
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 710 | 149 | 29 | 888 | 168 | 20 | 36 |
- HTML: 710
- PDF: 149
- XML: 29
- Total: 888
- Supplement: 168
- BibTeX: 20
- EndNote: 36
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1