General comment

In this manuscript, the authors use the instantaneous observations of CALIOP to retrieve faint aerosols missed by CALIOP’s official aerosol layer and profile products. Comparison analyses with SAGE III aerosol product demonstrate good agreement from the middle troposphere to the stratosphere. Also, in the 2019 Siberian fire event, retrieved instantaneous aerosol provided more information on faint aerosol propagation trajectory with higher spatial and temporal resolution than CALIPSO Level 3 monthly-averaged aerosols product.

This study is very interesting because most of the previous studies focused on aerosol and cloud layer retrieval, but the authors propose a novel method to retrieve the generally-ignored faint background aerosol based on CALIOP instantaneous observations. The manuscript is well-written and straightforward. The results are satisfactory and effectively described. This method is expected to provide new data for the investigation of aerosol-cloud interaction, which may offer new insights into the aerosol climate effect that otherwise cannot be seen by studies based on integrated or surface aerosol information (i.e., AOD). Therefore, I suggest this manuscript be published after minor revisions.

Major comments

1. The introduction needs to become more refined and better linked to the scientific literature. It should be clearer where the gaps are in the literature and what the contribution of this study is in this respect.
2. The authors use a lidar ratio of 28.75 sr in the troposphere, referring to Kim et al. The lidar ratio is one of the key parameters for aerosol extinction retrieval and varies with aerosol type. Therefore, I suggested the authors more deeply discuss and analyze the lidar ratio, including the difference between land and ocean.
3. The CALIPSO retrieval of instantaneous faint aerosols is very challenging. Is there a useful way to improve the retrieval in the future, such as using a wavelet to denoise the CALIPSO level 1 data before retrieval? Although I do not recommend using denoising algorithms in the study of this paperbecause we prefer to do original research using the most formal methods first, I suggest discussing it for guiding future work.

Specific comments

1. Line 23: “capture” should be “captures”.
2. Line 38: it is not always true to argue that the aerosol particules in the PBL “can usually be detected by CALIPSO”, which is inconsistent with previous findings. Therefore, this statement can be rephrased as “can only be detected by CALIPSO in the upper PBL in the absence of cloud (doi: 1016/j.atmosres.2016.05.010)”
3. Line 43: The citation may be corrected.
4. Line 49: Clouds interact directly with surrounding aerosols, and in particular sub-cloud aerosols have a more significant effect on cloud production. However, these aerosols are not exactly the same as what the authors refer to as faint aerosol. A more rigorous and accurate expression is recommended.
5. Line 50: This study is not motivated by the aerosol proxy used for aerosol-cloud interaction. I think the ignorance of faint aerosol surrounding high-altitude cloud layers is the culprit to complex the quantification of aerosol climate effect. Therefore, “an improper aerosol proxy (such as AOD)” can be changed to “the ignorance of faint aerosols surrounding high-altitude cloud layers” or something like this.
6. Line 68: “in” should be “since”.
7. Line 73: It is recommended use the full name of the product (e.g. VFM) where it first appears in the manuscript.
8. Line 97: “vertical” should be “vertically”.
9. Line 105: How is the SNR calculated here based on this formula?
10. Line 126: The matching of CALIPSO and SAGE considers spatial distances. What about temporal distances?
11. Line 144: The “red dash boxed area” is not marked in the figure 3a.
12. Line 163: Suggest reword “shows well the consistency” to “shows high consistency”.
13. Line 226: The range described here (60-10°N) does not correspond to the range of the red rectangular box in the diagram (40-10°N), and it is suggested to keep it consistent.
14. Line 240: “and compared them” can be changed to “which are compared”
15. English should be further improved by a native English speaker.

                                                                    Review of “CALIPSO retrieval of instantaneous faint aerosols” by Mao et al.

The space-borne lidar CALIOP on board CALIPSO satellite has been providing global measurements of aerosol and cloud backscatter profiles since 2006. Successive improvements in algorithms and calibration procedures have resulted in several versions of the products. Nonetheless there is always scope for further improvements. In particular, it has been known that aerosols with weak backscatter sometimes fall below the CALIOP layer detection threshold and several works have focused on this aspect as has been pointed out by the authors. In this paper the authors attempt to retrieve what they call “instantaneous faint” aerosols using the CALIOP backscatter data in both the stratosphere and troposphere and compare the results with SAGE III on ISS. Firstly it is not clear to me what the authors mean by “faint” aerosols---are these background aerosols or aerosols that just fall below the detection threshold but otherwise retain the intensive optical properties of the nearby detected layers. The methodology is nothing new and Kim et al. (2017) have used the same 20 km horizontal and 300 m vertical averaging for their extinction retrievals. The Kim et al. (2017) paper did a much more comprehensive study of the weak aerosols unlike this manuscript which lacks in some important details and other aspects, as described below. I do not believe this paper presents enough innovative ideas or interesting new results that will be useful to the community. I regret that I am not able to recommend this manuscript for publication in Atmospheric Chemistry and Physics. I have the following comments in no particular order:

1. The authors seem to use a constant altitude bin of 300 m to average the level 1 (L1) backscatter profiles in the entire altitude range from 0-36 km. This is not tenable, because CALIPSO L1 backscatter profiles have varying resolution with altitude, with 30 m from surface to 8.3 km, 60 m between 8.3-20.2 km, 180 m between 20.2-30.1 km and 300 m above 30.1 km. Properly accounting for these differences will require setting the binning at 900m as was done in Kar et al. (2019) for the level 3 stratosphere aerosol product.
2. I am surprised that the authors are talking about the “instantaneous faint” aerosols in the extinction range 10^-4-10^-5 km^-1 and yet do not mention the estimated uncertainties for these extinction profiles at all. When they are claiming to do better than the standard CALIPSO level 2 (L2) and level 3 (L3) products, they should discuss the resulting uncertainties for their extinctions to bolster their claims. Kim et al. (2017) had discussed in detail the uncertainties in their retrievals particularly those coming from the lidar ratios used. The lack of any such discussion in this paper is a major drawback.
3. In their retrieval algorithm they mention excluding the aerosol and cloud layers detected by the CALIOP L2 algorithm and all data below those layers similar to the methodology employed in Kar et al. (2019). However one important filter has been left out which has to do with the thin cirrus clouds. These clouds often fall below the layer detection threshold and can significantly contaminate the “faint” aerosol profiles they are trying to retrieve particularly in the UTLS area. Kar et al. (2019) had used a filter on the volume depolarization ratio to take out these ice clouds. In fact cloud clearance can be an issue with SAGE occultation measurements as well, which affect the data below 20 km (Thomason and Vernier, 2013, https://doi.org/10.5194/acp-13-4605-2013). It is not clear if the authors have used any filter for cloud-clearing of SAGE III data. Further note that the SAGE III-ISS 521 nm extinction product has low bias between 20 km and 25 km at mid latitudes possibly relating to the ozone interference (Knepp et al., 2022, https://doi.org/10.5194/amt-2021-333). These issues should have repercussions for the CALIPSO/SAGE III comparisons the authors have attempted.
4. The authors use a lidar ratio of 50 sr in the stratosphere and 28.75 sr in the troposphere. In Figure 3, they are retrieving “faint” aerosol in between and in continuity with the smoke layers retrieved by CALIOP L2 product, assuming that the “faint” aerosol is also smoke and referring to “the continuous nature of this aerosol layer”. For smoke, CALIOP L2 uses a lidar ratio of 70 sr, so the “faint” aerosol extinctions retrieved using a lidar ratio of 28.75 sr will have a significantly low bias. I am also intrigued by the distinct band of “faint” aerosols above 10 km extending from 15^oS-55^oS—is this background aerosol, smoke? Indeed how confident are the authors that it is aerosol at all? Similarly in the ASR plot Figure 3e, what are all those clumps between 10-30 km? It all just looks like noise to me although the ASR values are quite significant (~1.3) and about the same as within the pink box. Similarly in Figure 7, the authors are showing the plume of “faint” aerosols “connected with the VFM aerosol features”—i.e Siberian smoke extending from 10^oN-60^oN. Once again, they should use the more appropriate lidar ratio for smoke (70 sr) rather than that (50 sr) for stratospheric background aerosols. Since the authors use coincident extinction data from SAGE III-ISS, have they tried to obtain the lidar ratio (in Figure 4b, for instance) using the extinction from SAGE III and backscatter from CALIOP following the method given in Kar et al. (2019)? It will be interesting to see how  those compare with the values they are using.
5. It seems to me that in Figure 4, the layer at ~15 km is detected and retrieved by standard CALIOP L2---wonder how that compares with the profiles the authors retrieve. In any case the “faint’ layers between 15-20 km in this Figure do not look quite faint to me, and is likely the aerosol that just missed the layering threshold of standard L2.
6. As the authors have mentioned, CALIOP has been experiencing low energy laser shots since late 2016 primarily impacting the SAA region and accordingly they have excluded the SAA region. However those low energy shots have been spreading to other latitudes as well and can lead to artifacts in the data including false layer detections at all altitudes, particularly in the dayside. These effects can impact the extinction retrievals the authors are attempting and can be alleviated using the prescription given in the data advisory.
7. Line 173-174—“Further, we can see that the retrieved aerosol extinction is much less than the detection limit (0.01 km^-1) of the CALIPSO Level 2 product” and lines 234-235—"Instantaneous retrieval of faint aerosol at 20 km horizontal resolution provides a chance to deeper understand and quantify the aerosol impact on climate beyond the current CALIPSO Level 3 Stratospheric Aerosol Profile product”. I think the authors are missing the rationale behind the L2 and L3 products. In my understanding CALIOP L2 first detects a ”layer” using a range-dependent threshold and then assigns an aerosol subtype to it, for which a lidar ratio is available. This lidar ratio varies from 23-70 sr. The layer detection scheme for relatively low SNR measurements like CALIOP is quite complicated (Vaughan et al., 2009, doi:10.1175/2009JTECHA1228.1) and was designed to minimize false positive detections which leads to some undesired missed detections. However, overall it has worked very well and without this and (hopefully) proper assignment of lidar ratios for those different types of aerosols, CALIOP products would not be as useful as they have been. The next generation detection scheme for lidars uses 2D algorithms using both 532 nm and 1064 nm backscatter measurements and will lead to much more accurate detections including those of weakly scattering particulates (see Vaillant de Guelis et al., 2021, https://doi.org/10.5194/amt-14-1593-2021). On the other hand, in the L3 stratospheric aerosol product, which is built from the L1 profiles, the retrieved extinctions are of the same order as the authors retrieve here.  The L3 product has low spatial resolution (5^ox20^o in lat/lon necessary to increase the SNR and produces a reliable picture of the known stratospheric features) and is geared towards modelling applications. In other words, the L2 and L3 products have different goals and limitations. L2 products in the stratosphere employ different lidar ratios for different subtypes (ash, sulfates, smoke) unlike in the L3 product where a constant value is used for background as well as the full-aerosol mode. If the authors propose to retrieve the “instantaneous faint” aerosols in between the layers of different subtypes (as in Figures 4 and 7) then they should use the appropriate lidar ratios as mentioned above---this entails using the subtypes defined in CALIOP L2 or they can define their own subtypes.
8. Section 3.3. The authors assume all of the stratospheric perturbation in the northern mid/high latitude is coming from the Siberian wildfires. Much of this may actually be from the Raikoke volcanic eruption (June 2019) instead (Kloss et al., 2021, https://doi.org/10.5194/acp-21-535-2021, Knepp et al., 2022, etc.).
9. Lines 226-229 and lines 250-252. I don’t understand how from one browse image the authors can show “faint” aerosol “propagating” from 60^oN to 10^oN.  By the way the CALIPSO transect shown in Figure 7b passes through the well-known Asian Tropopause Aerosol Layer or ATAL (Vernier et al., 2011, https://doi.org/10.1029/2010GL046614, Fairlie et al., 2020, https://doi.org/10.1029/2019JD031506,  etc.) region. How do they know it’s all smoke from Siberia (or, sulfates from Raikoke) rather than at least partly being contributed by ATAL? In fact the ATAL feature mostly falls below the CALIOP layer detection and is seen in the adequately averaged L1 data as in CALIOP L3 product.
10. Line 84, line 188—The aerosol product discussed in Thomason et al. (2010) paper related to the SAGE III instrument on Meteor 3M spacecraft, not the ISS.

The analysis presented in the manuscript demonstrates the ability to detect “faint” aerosol that is unreported in CALIPSO level 2 retrievals because it lies below feature detection thresholds. The manuscript explains the importance of quantifying this under-represented aerosol based on literature. In order to “detect” the missing aerosol, the authors follow a similar procedure as is used to construct the CALIPSO level 3 stratospheric aerosol product (Kar et al., 2019): cloud and aerosol layers detected by CALIOP are removed from the level 1 attenuated backscatter and then a Fernald retrieval is performed using a fixed lidar ratio. The difference is that the CALIPSO level 3 product reports monthly averages of aerosol extinction, whereas this manuscript analyzes extinction retrieved from individual level 2 granules. This is what is meant by the “instantaneous” descriptor. The manuscript shows three examples where the CALIOP level 2 algorithms did not detect aerosol layers, but the extinction retrieved by the authors does indicate aerosol enhancement. It goes on to compare their retrieved aerosol extinction values to co-located SAGE III measurements, finding decent correlation at night and a high-bias in their CALIPSO retrievals during the daytime. The logic here is that because their retrieved aerosol extinction at night matches well with SAGE III measurements, then the retrieved aerosol extinction is a fair representation of what was missed by CALIOP level 2 feature detection.

General comments

There are some areas where greater details are needed to avoid confusing readers. For example, the manuscript discusses “detectable extinction” by CALIOP multiple times. This is inaccurate because the CALIOP level 2 algorithms do not detect extinction. They detect aerosol layers using attenuated scattering ratio and then perform an extinction retrieval. The minimum “detectable” extinction is really just the minimum extinction occurring within detected aerosol layers. This distinction is important and needs to be made clear. Based on this and my specific comment below about the lack of details regarding CALIPSO vertical resolution, I recommend that the authors provide more information about the details of the CALIPSO level 1 product, and the steps involved in how the level 2 algorithms ultimately retrieve extinction. That would provide important context for the reader.

The impact of lidar ratio selection is important and inadequately discussed in the manuscript. Two values are used for this analysis (50 sr stratospheric, 28.75 sr tropospheric). The manuscript justifies the two selections for generalized values. However, the aerosol type is known in at least two of the specific cases evaluated: smoke. Since smoke lidar ratios are around 70 sr, this leads to a sizable bias. The manuscript should add a discussion of the limitations of the lidar ratios used by the method. 

The conclusions claim to be able to retrieve aerosol extinction down to 0.0001 /km. At that level, however, it is important to consider uncertainty and biases. It is not discussed how large the relative uncertainties are for such small extinction values for the averaging being used. A greater discussion on uncertainties should be added to specify the value the proposed method yields to capturing undetected aerosol. 

Specific comments

Lines 69-70: “The CALIPSO lidar is highly sensitive to cloud/aerosol layers with a lower bound of optical depth…”. What is the lidar highly sensitive to? Presumably layer detection is meant, but the sentence does not say. 

Line 70: “minimum detected extinction of 0.01 to 0.02 /km.” This statement does not accurately represent the order of the level 2 CALIOP algorithms. The level 2 algorithms do not detect extinction. Layers are first detected using attenuated scattering ratios and then extinction is retrieved. Suggest rewording to clarify.

Line 93-94: Rather than just reference the Kar et al., 2019 paper to explain how clouds and aerosols are removed, it is recommended to add a sentence or two summarizing the removal procedure of that paper. Also, Kar et al., 2019 applies additional filters to remove undetected cloud layers beyond just using the VFM to cloud-clear. Does the method for this manuscript do the same?

Line 95: “The TAB is averaged at a vertical resolution of 300 m…” How is it possible to average the TAB to 300 m vertical resolution from 20.2 to 30.1 km when the range bins reported in the level 1 data product are at 180 m vertical resolution for that altitude region? Averaging two bins together in this region would yield 360 m, not 300 m. Furthermore, the TAB is already reported at 300 m vertical resolution from 30.1 – 40 km, so no average is required. Please clarify if the averaging used for this study considers the vertical resolution of the range bins in the level 1 data products.    

Line 147. According to this line, the extinction retrieval yields a value of 0.01 /km. However, the text suggests the layer is smoke, so the lidar ratio being used is too low by 50/70. Therefore, this extinction value should be larger. 

Figure 3. According to pre-processing step (a), clouds and aerosols detected by CALIPSO are removed, along with the data beneath them. However, the purple boundaries in Fig. 3(c) shows that a smoke layer is detected and there is an extinction coefficient reported there. The text even quotes the extinction value on line 148 and panel (d) shows where these layers are detected. I thought that the backscatter was supposed to be removed where layers are reported. Why are they shown in this figure? They are not shown in Figure 7. This should be made clear somewhere which data is used in the retrieval shown in the extinction figure.

Figure 3 caption. “…additional mean filtering (3x3 window) to highlight the faint aerosol area.” The premise of the paper is that averaging to 20 km x 300 m resolution is enough to highlight the faint aerosol. Why is additional averaging needed? Can these features still be discerned without this additional filtering? If not, then should the 3x3 window filtering be included as part of the methodology? 

Line 173: “…indicating a low bias in the CALIPSO retrieval.” Some clarification should be added here because there could be two interpretations of this statement. (1) Because the CALIPSO level 2 layer detection did not capture these extinction values, there is a low bias in what CALIPSO reports. Or, (2) the retrieval of extinction from the CALIPSO products performed in this study has a low bias. Please clarify which condition this statement is addressing.

Line 176: “…we can see that the retrieved aerosol extinction is much less than the detection limit (0.01 km-1) of the CALIPSO Level 2 product”. More precise language is requested here. The CALIPSO level 2 algorithms do not detect extinction, they detect layers and then retrieve extinction. This study addresses the extinction from aerosol layers below the layer detection limit of the level 2 feature finder.

Lines 198 – 200: “Young et al. (2013) noted that the CALIPSO retrievals with SNR≤1 usually contain a positive bias. The SNR during daytime above 20 km is usually less than 1 for TAB at 20 km horizontal scale (Figure 6b), which leads to a significantly positive bias in the retrieval” It is not immediately clear how an SNR < 1 yields a positive bias. SNR speaks toward the (inverse of the) variability with respect to the average value, but not necessarily a bias. I would assume that a bias is more governed by calibration rather than noise. Or is it that the noise is not Gaussian? Please add information as to why a “significantly positive bias” is expected in the retrieval when SNR < 1.

Line 210: It would be helpful to explain why the white areas of missing data occur between ± 15° in the level 3 panels of Figure 7 (because of the tropopause height).

Section 3.3. Possible smoke from Siberian wildfires is not the only explanation for aerosol enhancement in the stratosphere during this time period. The June 2019 Raikoke volcano eruption also emitted a substantial amount of sulfate at northern latitudes. This should be discussed as part of the explanation and interpretation for aerosol enhancement in August 2019. 

Line 225: “These faint aerosols propagate from 60N to near 10N…” The word “propagate” might be inaccurate for this discussion. 

Lines 243 – 244: “The retrievable aerosol extinction greatly extends to 0.0001 km-1…” What is the relative error on these very low extinction values?

Lines 244 – 246: “The comparison is unavailable at low altitudes, but the retrieval should be more reliable (i.e., in the troposphere) because the SNR is higher.” The improvement in SNR is only part of the story. A far more substantial factor that will cause larger errors in the troposphere is the choice of lidar ratio, which can range from 20 – 70 sr. This can cause the biases up to a factor of three when the wrong lidar ratio is used. It is important to include a discussion on how the choice of lidar ratios for this analysis impacts comparisons with SAGE retrievals.

Lines 247 – 248. Same question as before, how does low SNR yield a positive bias? More should be added here to summarize why this is true.

Lines 249 – 252. A couple of points about conclusion item (3). 

First, the stratospheric aerosol enhancement for this time period includes contributions from the Raikoke volcanic plume in addition to (possible) smoke from Siberian wildfires. This should be included with the discussion of sources for this example. There is some discussion in the literature about the contribution of these aerosol types in August 2019. 

Second, this sentence can easily be interpreted as an over-generalization, “our retrieval shows that these faint aerosols even propagate to near 10°N, which is much beyond the detecting range of the CALIPSO L2 products (50° N and 60° N).” I believe this sentence is a summary of the single level 2 granule evaluated in Figure 7 where the level 2 algorithms did not detect a large extent of the stratospheric aerosol enhancement from 10N to 50N, Fig 7(d). For this specific case, the aerosol was not detected by CALIOP level 2. However, the sentence is written as though this is a general result: faint aerosols following the 2019 Siberian fires (and Raikoke eruption) are not detected as far south as 10N by CALIOP level 2 retrievals. This cannot be concluded based on the one granule examined. To make the possibility of misinterpretation more probable, Figure 7(b) shows nothing reported in the CALIOP level 3 stratospheric aerosol product from about 15 N/S. This is merely because the tropopause is above 15.2 km at those latitudes, but a reader could easily read this sentence and look at Figure 7(b) and conclude that CALIOP level 2 missed detecting all of that aerosol during August 2019. It is unlikely that CALIOP level 2 did not detect all of this aerosol, and even if so, it was not proven in the manuscript. I recommend rephrasing conclusion item (3) to be more specific on the evidence for the conclusion being made.