Review Deep Convection lifecycle characteristics: a database from GoAmazon experiment 

by

Lopes et al

This paper describes a two year (2014-2015) radar and lightning study of the climatology of mesoscale convective systems (MCS)  in the Amazon. It uses the data from a  single radar using the inner 150 km x 150 km domain centered on the radar.   It using feature analysis and tracking techniques to identify the MCSs.  The results are analyzed in terms of months, seasons and IOPS in a wide variety of ways.

I think it is a comprehensive study and should eventually be published.  There are several issues with the paper.

1. The paper uses a lot of jargon.  Jargon is fine as it shortens the text but it must be clear defined and explained.  This is problematic in several places but the biggest problem with the paper is that I don’t understand  the difference between cluster and convective systems on lines 50 to 53.  I re-read the paragraph and I just don’t understand the distinction.  It needs to be clear.  So,  I had troubled appreciating all the points the authors were making throughout the paper.  
2. I think a figure showing a cluster and  a convective system would be helpful.
3. To confound the problem, the authors introduce “multi-cell” storms and terminology in the summary.  What is a cluster, cell, multi-cell, convective system?  Perhaps a schematic would also help.
4. re jargon… line 54…. It is never good when quotes are used.  There is a sense that these are self-defining terms and no further explanation is needed.  But it is needed.  For example, I don’t understand the term “continuity” as it is used latter in the paper which I believe is also the condition for “dissipation or dying system”.  Again, a graphical schematic can help.  Is “spontaneous generation” the same as convective initiation(CI)  (a la Jim Wilson) or something else…if yes, keep it simple and use CI.  If not, explain.
5. “table” is used a lot but I think you just mean “data”…. no need to introduce unnecessary jargon.  I suggest changing. 
6. “raw” vs “filtered”….  I found this to be my second biggest problem.  You filtered the data for a reason….If I understand the paper correctly,  you want to be able to observe the entire life cycle of clusters or convective systems or whatever it is you are studying… I kept asking myself, why are you presenting presenting any “raw” results?  It doesn’t make no sense to me.  I kept reading on and on about statistics but they are uninterpretable from a scientific perspective.  Make the case for including them or delete them entirely.
7. Since filtering is critical, there should be more discussion on it.  It is only presented in Table 2 without much information.  
8. I would suggest that the authors describe the impact of each element of the filter in a step by step fashion by showing what happens to some distribution of some critical parameter (e.g. max reflectivity, storm height, area, number …).  That is, 40 dBZ filter reduces the number of cells by 20%, longer than 12 minutes reduces the number by another 40% etc.  
9. It is not clear whether the minimum sizes (100 and 40) and reflectivity thresholds (20 and 40) are part of the filtering.  I do not see how they are used in the processing or analysis.
10. Once you filter the data for non-boundary features.  You effectively limit the size of feature (cluster or convective system) that you can resolve.  If you apply information theory (see Claude Shannon - to resolve a feature, you have to sample at least twice the resolution… to resolve a physical day, to need to sample twice, like radiosondes… similarly in space, you have an upper limit in the size of convective system that you can fully resolve - something like 75 km in size and smaller till you reach the Cartesian grid spacing)
11. You use 0 dBZ for echotop.  This begs the question as to the sensitivity of the radar at 150km range.  Is the radar sufficiently sensitive enough.  The radar characteristics is not shown - a reference is needed or a table to briefly describe it is needed.
12. You show the calibration results of Schumacher.  Since this is an internal report and not readily available, this deserves a sentence or two to describe.  Bigger question is does it matter?  Did it create discontinuities in the analysis as there ewer 5 dB differences between calibrations.
13. In terms of initial location (line 171)… is there a topographical or other morphological influential feature.  Perhaps a figure showing the study area at the beginning of the paper would help the reader better understand.  Perhaps a heat map is better than tracks to show initiation.  In any case, can you make the figure full page, hard to see even then but better.
14. Isn’t the wind rose (figure 16) backwards?  Traditionally, with wind rose for winds, it is the “from direction” and so should be the other way around.  It could be confusing to the reader.  
15. I congratulate the authors on the grammar as English is not the first language, it is generally readable.  However, there are many places where there are missing prepositions (i.e., the word “the”), many very long sentences, a lot of mis-use of colons and in some cases of semi-colons.  I suggest removing the use of colons altogether - even native speakers do not get this right.  I would suggest writing simple and short sentences and this would get rid of many of semi-colons which are also difficult to use.  The only suggestion I have for the missing propositions is to have someone read it or wait for the next round of reviews.
16. The title puzzles me a little bit.  “a database from…”.  What is a database in this context?  This paper is a two year radar climatology of convection system morphology.  I think this is a science and not a data paper.
17. The information presented is good and valuable.  It would be much more valuable if there was some context about the years 2014-2015.  Were they anomalous? Were they typical?  I think a precipitation time line for say 20-30 years surrounding 2014-2015 would be most informative.
18. Small point. The results section is very dense.  It would be useful to have sub-sub-sections with short titles to separate topic changes.
19. Note: Table 4 and 5 have exactly the same text.  It would be useful (in general) to have more informative text.
20. There are other small issues and a few of them are:
1. line 188….Gutpa2024-pi reference is improper and missing from References
2. line 196… m/min units are not consistent with figure km/min
3. line 11…. “define industrial pole”
4. equation 2… N_o and rho_i are undefined
21. Many sentences needing attention but will leave this to authors to review and revise for now.
1. line 86-89

This manuscript describes a new convective storm dataset constructed from weather radar observations during the GoAmazon field experiment. It will be useful to those involved with the experiment as well as interested in studying tropical convection. The dataset is novel in that it applies a new cell identification and tracking technique to radar observations of convection, albeit there are some concerns about its implementation and interpretation of the results. The manuscript is generally well written but the length and amount of information needs to be significantly reduced in future revisions. 

Main concerns:

1. Data quality/completeness: The authors take care to use quality controlled radar data. However, the use of CAPPIs at 3-km will exclude some growing and decaying parts of the lifecycle. It would have perhaps been better to use composite reflectivity (i.e., maximum reflectivity throughout the column) as input data to better depict the lower altitude echoes during found during the early stage of a convective cell's lifecycle and higher altitude echoes during the latter stages.
2. Data quality: The average lifetime of typical tropical convection is <60-minutes. Since this dataset allows for a 60-minute gap during cell tracking, an insignificant amount of ambiguity about convective system evolution exists, especially for cells that are in close proximity prior to a 60-minute gap. 
3. Data quality: In Table 5, the sample counts for each CS duration do not add up to the total specified on the last row. In Table 6, the total seasonal counts in the last row do not add up to the total sample size of 91609 (raw) and 5976 (filtered) given in Tables 4 and 5.  
4. Presentation: The authors thoroughly present the main characteristics of the dataset, but 24 figures and 7 tables is too much for this type of publication. The breadth of results detracts from the main purpose of this submission. The authors should focus the revision to expanding on the methodology and reducing the results to only present an overview of the resultant dataset, trimming down the number of tables (e.g., 5, 6, and 7) and limiting figures to only a few that address data quality and show temporal trends relevant to the field experiment (e.g., keep Fig 2, 3, 6, 8, 11, 18, 21, 22). 
5. Presentation:  TATHU is a newer tracking software with the details being published only in a non-English language document and github. Future revisions to this manuscript should include more information about the TAHU methods. The github documentation is great. Consider citing in the text or including the workflow diagram in the revision.

Minor comments:

• Line 54, "spontaneous generation" and "continuity": Define these two descriptors like was done for "merged".
• Lines 64-65, "systems...not deep enough": Ambiguous wording. Specify the echo top height or depth was used to filter the convective systems. 
• Table 2: For "Minimum sizes", it is not clear what the two values separate in the dataset. Are these clusters, convective systems, raw or filtered?
• Lines 92-93, "...because one of the filtering criterial...": This suggests the filtering limits the new dataset to only cells with the full lifecylcle within the study domain. If so, this will exclude many cells that could be useful for studying individual stages of the convective lifecycle. Please clarify.
• Lines 101-103, "Another...": Reword this sentence.
• Figure 4, raw vs filtered: Clarify that this is the relative frequency distribution to avoid questioning why the filtered bars are higher than the raw bars.
• Lines 117-118, "Similar to Table 5,...": This needs to be reworded because in Table 5, less than 50% of the raw systems lasted up to 1 hour.
• Figure 7 and 8 captions, "clusters": The caption calls these clusters whereas the text and figure title refer to this data as "convective systems".
• Line 142, "GLD strokes": Note that there may be multiple strokes detected for a single "flash" in the GLD360 dataset (e.g., Murphy and Nag 2015). Also need to acknowledge the performance of this type of network. They largely detect CG lightning and at best half of intracloud lightning (Murphy and Nag 2015). 
• Lines 144-145, "...convective systems with lightning consists of only a few clusters with lightning": It is not clear what is meant by this. Convective systems are a subset of clusters in this dataset and those with lightning comprise an even smaller number. 
• Line 160, "...as cited previously": Include the citation here.
• Figure 13: Why is this graphic needed if the text (lines 169-170) indicate no interesting trends? Similarly for other figures. If there is nothing interesting, then just mention it in the text but do not show that data in a figure.
• Line 191, "variation rates": Remind the reader what the temporal resolution of the radar data.
• Lines 195-197, Fig 19b: Correct the reference to panel b. It is interesting how the global peak occurs at positive values suggesting this dataset largely captures growing convection. 
• Figures 23-24: Instead of these difference CFADs, it would be more interesting to see the CFADs for convective systems with lightning and for IOP1 vs IOP2.

References: 

Murphy, M. J., & Nag, A. (2015). Cloud lightning performance and climatology of the U.S. based on the upgraded U.S. National Lightning Detection Network, Paper presented at Seventh Conference on Meteorological Applications of Lightning Data, American Meteorological Society, Phoenix, AZ.