the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 05 Aug 2024
Cloud phase estimation and macrophysical properties of low-level clouds using in-situ and radar measurements over the Southern Ocean during the SOCRATES campaign
Abstract. The Southern Ocean (SO) provides a unique natural laboratory for studying cloud formation and cloud-aerosol interactions with minimal anthropogenic influence. The Southern Ocean Clouds, Radiation, Aerosol Transport Experimental Study (SOCRATES), was an aircraft-based campaign conducted from January 15 to February 28, 2018, off the coast of Hobart, Tasmania. During SOCRATES, the NSF/NCAR GV research aircraft, equipped with in-situ probes and remote sensors, observed aerosol, cloud, and precipitation properties, and provided detailed vertical structure of clouds over the SO, particularly for the low-level clouds (below 3 km). The HIAPER Cloud Radar (HCR) and in-situ cloud and drizzle probes (CDP and 2DS) measurements were used to provide comprehensive statistical and phase-relevant macrophysical properties for the low-level clouds sampled by the 15 research flights during SOCRATES. A new method based on HCR reflectivity and spectrum width gradient was developed to estimate cloud boundaries (cloud-base and -top heights) and classify cloud types based on their top and base heights. Low-level clouds were found to be the most prevalent, with an almost 90 % occurrence frequency. A new phase determination method was also developed to identify the single-layered low-level clouds as liquid, ice, and mixed phases, with occurrence frequencies of 45.4 %, 32.5 %, and 22.2 %, respectively. Low-level clouds over the SO have significantly higher SLW concentrations, with liquid being most prevalent at higher temperatures, ice phase dominating at lower temperatures, and mixed-phase being least common due to its thermodynamic instability. Regarding their vertical distributions, the liquid phase occurs most frequently in the lower mid-cloud range (from 500 m to 1 km), the mixed phase dominates at cloud bases lower than 1 km but is well distributed along the vertical cloud layer, while the ice phase is prevalent from the middle to upper cloud levels (1–3 km). The higher occurrence of the mixed phase at the cloud base could be attributed to large drizzle-sized drops and/or ice particles.
- Preprint
(5879 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 16 Oct 2024)
-
RC1: 'Comment on amt-2024-124', Anonymous Referee #1, 04 Sep 2024
reply
General Comments
The authors use in-situ cloud probe data and remote sensing observations collected by instruments onboard the NSF NCAR GV aircraft during the SOCRATES campaign to study low level clouds over the Southern Ocean. While the topic is interesting and highly relevant, unfortunately, there are many issues with the methodology used by the authors:
The authors developed a radar-based methodology to derive cloud base. However, they themselves state that it does not agree with the standard method based on lidar observations. The lidar-based method has been proven to work well in many previous studies and since the authors do not independently validate their new method the reader concludes that the radar based method is not accurate.
The authors then develop a new cloud phase classification method based on in-situ and radar observations. The method heavily relies on spectrum width observations but fails to account for shortcomings in the spectrum width observations, which leads to inaccurate conclusions. It also relies on the liquid water path, which is derived along the aircraft track from in-situ measurements and then applied to the whole vertical column of the radar observations. In my opinion, it is not an accurate assumption to claim that LWP is constant throughout a cloud. The authors themselves state that the resulting 2D classification does not look accurate. Nonetheless, they collapse it back into a 1D classification and use both throughout the rest of the study.
The authors compare some of their results with methods from previous studies (even though for some reason they do not compare it with the phase classification that is provided in the radar/lidar dataset itself) and note that their results do not agree very well with previous studies.
I am not sure why the authors chose to develop their own method, when proven methods are readily available, and then decide to use their own method, even though they themselves have found that it does not produce good results.
Because of these issues with the methodology, the interpretation of the results, and the conclusions drawn, are also highly questionable.
Given all these shortcomings (which are listed in more detail in the Specific Comments below), I am afraid I have to recommend to reject the manuscript in its current form.
Specific Comments
- Line 147: Please explain what is meant with “interpolated to fixed radar-heights at a range gate of 19.2 meters”. The range gates are already fixed. And why 19.2 m?
- Section 2.2: I am very skeptical about the method that derives cloud base. How was that method validated? Using data from the GV-HSRL to determine cloud base is the state-of-the-art methodology. If the method proposed by the authors does not agree with the GV-HSRL measurements, that would generally indicate that it is not accurate. If the authors really want to claim that their method is superior, they need to find an independent validation method. Simply plotting the results on the radar observations is not sufficient.
- I would also like to point out that spectrum width in regions with low signal to noise ratios, which are often observed at the cloud edges, are often noisy and not all that reliable. Using their gradient to derive cloud boundaries is therefore not recommended. You can see in Figure 1c that there is a similar gradient at the top of the cloud as at the bottom. The argument that this gradient implies drizzle or precipitation (as stated at line 177) is therefore highly questionable. The fact that the algorithm finds cloud base in the middle of a cloud layer where there is clearly no precipitation at all (Fig. 1 around 26.44) is also suspicious.
- Table 2: What is meant with “may not be single layer”? Does this mean that they could be multi-layer or deep? I don’t think it makes much sense to put both of these cloud types together in one category. The same is true for the HML category. That seems to be a deep cloud but the name of the category seems to indicate that it is a multi-layer cloud.
- Line 217: What is meant with “due to the selected cloud cases”? What cases were selected? How many? And based on which criteria?
- Section 2.4: I’m afraid I do not understand how LWP is calculated. Do the authors assume that the LWC is vertically constant throughout the cloud? That does not seem like valid assumption.
- Figure 3: How do these results compare to similar classifications in the literature? (Mace and Protat 2018, Truong et al. 2020, Romatschke 2023)
- Table 3: The LWP values have very large SDs. Are they actually meaningful?
- Section 4.1: The authors develop a new methodology to derive cloud phase from radar observations. However, they do not validate the method. This leads to the following questions: a) Why was a new method developed when a phase classification is already provided in the dataset? Why do the authors think their methodology is valid if it is not verified? Why do they believe that their method is better than previous methods (that have actually been validated)? Why was it not validated?
- At line 375 the authors say that they compared with previous work. What exactly does that mean? What were the results of these comparisons? Also, the classification heavily relies on LWP and spectrum width. As mentioned earlier, I am not sure how trustworthy LWP is, as it is assumed to be vertically constant (see first paragraph of Section 4.2). Again, spectrum width is noisy in low SNR regions and some, if not many, of the high values may be unreliable. (Romatschke and Vivekanandan decided not to use spectrum width in their algorithm for this reason.)
- In Section 4.2., the authors state that “phase retrieval method but may be not highly depictive of the actual cloud phase”. In that case, what is the value of the 2D derived cloud phase? What is the purpose of this effort? The 2D cloud phase is then again collapsed into a 1D cloud phase product which apparently “returns reasonably accurate findings compared to other phase detection studies over the SO”. These comparisons are later shown (see below) but they are not convincing.
- What is the purpose of developing this new product if other phase classifications already exist? Why start with the 1D cloud probe observations, expand them to low-quality 2D observations, and then collapse them again into 1D phase estimations? Why not use the cloud probe measurements directly and the phase classifications that were derived from them in previous studies?
- Starting line 442, the 1D phase classification is compared with phase classifications from other studies. Unfortunately, the classifications from the current paper do not agree well with the classifications from previous studies. The authors speculate about the causes of these discrepancies but fail to explain why the reader should trust their results over those from previous studies.
- Starting line 474 the authors suggest tuning thresholds from previous studies to match their results but I am not convinced that this is the right thing to do. Again, why do the authors think that their results are more accurate?
- Section 4.3: Given all the problems pointed out earlier (unproven cloud base, unreliable phase classification), I am not sure how trustworthy and valuable this analysis is.
- Section 4.3.1: For this analysis, is the 2D cloud phase or the 1D cloud phase product used?
- Section 4.3.2: Given that the authors stated earlier that the 2D cloud phase product is not particularly accurate, why is it used here?
- Starting line 551: I disagree with the author’s interpretation of the CFADs. In my opinion, the CFADs for the liquid (and mixed) clouds show a strong bright band signature at 0.2 km normalized height. This bright band signature contradicts the author’s claims that the clouds are predominantly liquid. This confirms that the phase classification algorithm does not produce high-quality results. Looking at Figure 7, one gets the impression that the algorithm tends to classify deep clouds with strong reflectivity as mixed, deep clouds with weak reflectivity as liquid, and thin clouds as ice. This is reflected in the CFAD categories of Figure 10. I do not see a connection with cloud phase. It is not surprising that the authors use the word “surprisingly” a lot in the description of the CFADs since the CFADs do not actually show clouds of the respective phases.
- Line 586: The higher width values indicate that a significant portion of the data comes from low SNR cloud edges where width tends to be noisy.
- Paragraphs starting line 599 and 609: Same as previous comment. Please do not over-interpret the high width values in low SNR regions.
- Conclusions: I am not convinced by the conclusions given all the issues pointed out in the previous comments.
Technical Corrections
- I would like to point out that there are many(!) grammatical issues throughout the paper which are too numerous to list. Sentences are often very long and not only the reader, but apparently also the authors get lost. Two of many examples are the sentence starting line 33 and the sentence starting line 80.
- Line 25: Please define “SLW”.
- Line 109 and throughout the manuscript: The instrument is called the “GV-HSRL”.
- Line 128: The SOCRATES acronym has already been defined earlier.
- Line 132: Type RV -> GV.
- Figure 1: Please correct the values on the time axis. (As far as I know, a day has only 24 hours.)
- Line 340: This sentence is a duplicate of the previous one.
- Line 554: When using CFADs, please cite Yuter and Houze, 1994.
- Figure 10: Altitude unit is missing.
- Data availability: The authors only post a general link to the EOL Field Data Archive. It is important to list the individual datasets with their respective DOIs in this section.
Citation: https://doi.org/10.5194/amt-2024-124-RC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 202 | 40 | 58 | 300 | 8 | 11 |
- HTML: 202
- PDF: 40
- XML: 58
- Total: 300
- BibTeX: 8
- EndNote: 11
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1