A Modified Look-Up Table Based Algorithm with a Self-Posed Scheme for Fine-Mode Aerosol Microphysical Properties Inversion by Multi-Wavelength Lidar

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study proposes a method for retrieving aerosol microphysical properties from multi-wavelength lidar, which is based on lookup-table and uses machine learning methods. It takes the method proposed by predecessors as the basic algorithm, and compares it with the modified algorithm in this study. The accuracy, stability and sensitivity to input errors of the modified algorithm were evaluated by numerical simulation experiments. This method was also applied to a multi-wavelength lidar detection campaign, and the retrieval results were compared with those of in-situ measurement. The quality of the manuscript is generally good, but I have found some improvements in the details that need careful consideration before publication.

 

Specific comments

1.     Line 50: When talking about fine and coarse modes aerosol, the authors could cite the references, e.g.

Zhou, X.; Zhou, T.; Fang, S.;Han, B.; He, Q. Investigation of the Vertical Distribution Characteristics and Microphysical Properties of Summer Mineral Dust Masses over the Taklimakan Desert Using an Unmanned Aerial Vehicle. Remote Sens. 2023, 15, 3556.

2.     What does the phrase “self-posed” mean? Does that mean “well-posed”? Please add an explanation to the text, or change the description.

3.     In Section 2.1 you describe the creation of LUT. What is the significance of the assumption that V_t=1μm³/cm³? Can  be any other value? Equations (4) - (6) deal with optical parameters. How do these processes affect the results? Can I just use the original optical parameters (3β+2α)?

4.     Line 275-276: Is there evidence that instability decreases as the number of decision trees increases? Is this conclusion related to the weighted “bagging” strategy described below?

5.     Line 344: What does “The constraint window is a concept corresponding to the LUT.” mean? Is the constraint window similar to LUT? But then you compare the constraint window to the reduced solution space, which is confusing.

6.     The LUT method is highly dependent on the completeness of LUT. In fact, the search space defined by LUT is incomplete, which leads to varying degrees of error in the results. In Section 2.2, you introduced “local interpolation”, how does it increase the incompleteness of LUT?

7.     Line 462-464: Why divide data into two types? This needs to be explained.

8.     In your method evaluation, you used the time efficiency of the algorithm as a key metric. But with the advent of more powerful processors, speeds are always likely to increase significantly. I think the accuracy of the algorithm is more important than the efficiency.

9.     I noticed that your results in the case study contradicted those in the simulation experiments. The performance of 3β+2α in section 3.4 is the best, but the previous conclusion is not so, why?

 

Minor comments

 

1.     Line 15-16: Single scattering albedo is not a microphysical property, but an optical property that can be calculated from microphysical properties.

2.     Line 42: There is a problem with the description of the fine-mode aerosol radius. What you wrote is usually the range of the effective radius or the mean radius, rather than the radius range.

3.     Line 306-307: The former of what? The latter of what?

4.     Figure 13. Is altitude above ground or above sea level? Clarification is required.

Author Response

Thanks for your review and comments. Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Review of "A modified look-up table based algorithm with a self-posed scheme for fine-mode aerosol microphysical properties inversion by multi-wavelength lidar" by Zhou et al., submitted to Remote Sensing

This mauscript presents an algorithm for retrieval of vertical profiles of aerosol microphysical properties from multi-wavelength lidar data. The algorithm is based on look-up table approach, k-nearest neighbour and random forest algorithm and presents an improvement of an existing algorithm, to increase accuracy and improve stability of results. The algorithm is limited to cases of fine mode spherical particles. The study considers different lidar configurations and its performance is analyzed based on simulated data and those obtained from DISCOVER-AQ field campaign. The algorithm is well explained and the topic is of interest to the lidar community. Please find below some minor comments.   

1. lines 71-72: "... whether the inversion algorithm of aerosol microphysical properties can operate normally and yield stable results remains to be determined." should be changed to "it remains to be determined whether the inversion algorithm of aerosol microphysical properties can operate normally and yield stable results."

2. It should be stated more clearly if FAST is the basic algorithm, described by Wang et al. (reference [30] in the list of references).

3. Please discuss how the performance of the algorithm in this manuscript, when applied to data from DISCOVER-AQ field campaign (Figure 14), compares to the performance of FAST algorithm.

Author Response

Thanks for your review and comments. Please see the attachment.

Author Response File: Author Response.pdf