The paper by Zhenquan Wang presents a new tracking algorithm for tropical convective systems and uses the algorithm to answer a few science questions about convective storms. Most of the paper is devoted to the tracking algorithm, in which variable brightness temperature (BT11) thresholds are used to identify cloud systems, segment them into convective cores and anvil clouds, and track the evolution, merging, and splitting of the segments over time. One of the main results is that colder BT11 is associated with a greater frequency of mergers and splits. In the last part of the paper, the algorithm is used to examine cloud lifetime, precipitation, and anvil cloud area. These properties tend to display log-linear relationships when plotted against BT11.

This is an interesting study and reflects an impressive amount of work by the author. I have no doubt that the tracking algorithm developed here is well motivated and well executed, and it seems like it could produce an interesting dataset from which many questions about convective cloud systems could be examined.

However, there are serious issues regarding the clarity of presentation in this paper. I found much of the writing and descriptions of the methodology to be very unclear, and the terminology used for the tracking algorithm was confusing and difficult to grasp. For these reasons, I do not feel equipped to evaluate the appropriateness of the methodology or to understand what the scientific conclusions really mean. So, please excuse me for being unable to provide much constructive feedback here. I would be happy to do so in the future once the presentation has been clarified. Some general comments are below, followed by line comments.

1. Unclear terminology. Cold-core, cold-center, segmentations, HCSs, organizations, organization segments, mergers & splits. Some of these terms are more clearly defined than others, but the precise meanings need to be clarified (especially HCS). Fig 1a was helpful for understanding centers vs cores…perhaps a similar schematic would help for the other terms.

2. Clarifying the methodology. The description of pattern-matching and the tracking algorithm were both quite confusing to me. The goals of each part of the analysis should be clearly laid out at the beginning of each section. It is confusing how segmentations, mergers, and splits are defined. I wish I could point to more specific aspects that I did not understand, but I am finding it difficult to do so at this point.

3. Mergers & Splits statistics. Another thing to clarify is how statistics are computed for mergers and splits (e.g. Fig 7 and 8). How is a PDF of mergers and splits as a function fo BT11 calculated? What if the two merging cores have different BT11? Which of the merging cores do the precip and anvil statistics represent? This was all very unclear.

4. Cloud property results.

• I cannot find a description of how the anvil area is computed. Is it just the entire nonprecipitating area of each individual segment?
• The study region is (20S-20N, 90E-170E), which I find to be interesting from a cloud property perspective. I imagine this choice was largely motivated the availability of different observations, especially the ARM sites. The region includes some of the western Indian Ocean and Bay of Bengal, the entire Maritime Continent region, and some of the west Pacific warm pool. The characteristics of convective systems can differ significantly between the maritime continent, where land -based convection dominates, and the oceanic regions, where larger mesoscale convective systems are typical (see Fig 9 in Yuan & Houze 2010, doi:10.1175/2010JCLI3671.1). Does it make sense to aggregate the precipitation, duration, and anvil area statistics across this entire region? I imagine there would be considerable differences between the Maritime Continent and the oceanic regions, with smaller cloud systems and fewer mergers/splits for land-based convection. The author could consider stratifying the results by region, or at the very least acknowledging what I imagine are very large spreads within each BT11 bin for the cloud property statistics.

5. Grammar and Structure. As a native English speaker, I found this paper quite difficult to understand at times, and this is likely a major reason for the perceived lack of clarity. I simply want to share that thought with the author, so that they can adjust and edit as they see fit. If editing services are available at the author’s institution, they may wish to pursue them. This is simply a suggestion, and I do not consider it necessary for the paper to be published, as long as the necessary components are greatly clarified.

More Minor Comments:

• What is meant by “organization segments”…does this just mean the different structural components of the storm?
• Line 30: “due to the fact that the…”
• Line 43-46: the author cites three papers as evidence that convective organization and precipitation efficiency (PE) are related, but I am not sure these references are correct. Bao & Sherwood (2019), https://doi.org/10.1002/2018MS001503, seems like a more appropriate reference here. Choi et al (2017) found that greater PE (by their definition of PE) was associated with reduced cirrus cloud area, but this is not the same thing as convective organization. Lindzen et al (2001) and Mauritsen & Stevens (2015) hypothesized about the relationship between PE and anvil cloud area, but did not present any evidence of a relationship between organization and PE.

• Line 51: what are the two distinct modes of convection being referred to here?
• Line 81: replace “190 W” with “170 E”
• Line 128: the equation for the speed bias (eq 1) is incorrect. The subscripts are switched around. See eq 4 in Nieman et al (1997)
• Line 128: latter -> later
• Section 2.5: this section was very unclear to me. Please provide some context for what the goal of this pattern matching is and how it fits in to the tracking algorithm
  • Line 142: “normalized BT11”…normalized in what sense?
  • Define “target scene” and “cross scene”
  • Line 145-146: “for the areas larger…” what areas are you talking about?
  • It seems that SSD would be minimized if the BT11 field does not change at all from one time to the next. Are the fields adjusted in space to overlap? Is this what normalization refers to? This was generally quite confusing.
• Fig 1:
  • the font size in panel 1a is too small at the top of the figure (“centers” and “cores”). The green font color for “connecting depth” and “Developing depth” is hardly visible.
  • What does “after moving” mean in the legend? Aren’t you showing two moments in time, with the dotted lines indicating the later moment? Aren’t the solid lines then showing the “before moving” picture?
  • Panels b,c: do the displacements between the solid and dotted lines reflect displacement over time? Or have the later moments been pattern-matched and adjusted for maximum overlap?
• Line 180: If I am understanding correctly, the algorithm detects the full cloud segment by expanding out from the core in 1K BT11 intervals. I imagine there is some ambiguity at times, in which it is not obvious which core a piece of anvil cloud should be assigned to? How is this dealth with?
• Lines 185-190: this paragraph was quite confusing to read, and I had to read it about 5 times to understand the details here. Cold-center BT11, complex BT11, and cold-core BT11 should be more clearly defined somewhere…at the moment they are buried in Fig 1a.
• Line 200: for clarity, specify “The core-core and segmentation-segmentation DORs are relative to the minimum area… The core-segmentation DORs are relative to the core area”
• Lines 198-205 – this paragraph is also confusing to read. Does “temporal associations” mean that the you consider it to be the same storm at different times?
• Fig 2 middle row: the arrows were a bit confusing, maybe it could be equally effective to just put red and white dots on each panel (optional suggestion).
• Line 241: unclear: “thus ends by less disconnected convection complex”. It looks less connected, not less disconnected.
• Line 244: “evolution of the system structures but not the variations of the connections?”…what does this mean?
• Figure 5 is completely lost on me – I do not know what this figure is trying to show, and the caption is not very helpful here. Please explain this figure.
• Figure 6
  • It would be nice to add a panel showing the sample size for each cold-core-peak BT11 bin.
  • it would be nice to see the spreads in duration, precip, and anvil amount for each cold-core-peak BT11 bin. The t-test for the mean is nice, but I imagine these is a very large spread on these quantities, since convective systems vary greatly in size. It would be good to show the spread if there is a simple way to do so.
• Line 269-270: “and the difference of the duration, precip, and anvils between two stages has exponential increases with the core peaking at colder BT11.” I am struggling to see how this is the case in Fig 6. It does not seem like the difference between the orange and blue lines is exponentially greater at lower BT11, although it is hard to tell because of the log scale.
• Lines 278-281: this sentence is not clear, please revise. Will be helpful once HCS is clearly defined. For example, how do mergers and splits create more HCS? My initial thought was that HCS referred to the entire system BT11<260, including many segmentations?
• Equations 7 & 8: what is N exactly? The definition of HCS seems to be very important here. Is it the number of segments? Also, it would be helpful to explain what the point of this sort of analysis is before showing the results.

This manuscript describes a method to track convective systems in the tropics using a so-called ‘variable-BT’ method. The tracked objectives are evaluated against observations by comparing object drifting speed and direction with those observed from three ARM sites. Contributions of convective activities to precipitation and anvil amount are discussed. I think the topic can be a good contribution to the community by bringing a more flexible tracking framework. However, I believe this manuscript needs substantial improvements before it can be considered for publication.

Major:

1. Grammar and Readability:

There are numerous grammar errors that make the manuscript difficult to read. The author should do a thorough proof-reading or seek help from a professional editing service before submitting the revised manuscript. Particular attention should be paid to the abstract, as a readable abstract is more likely to attract readers’ interest in the method developed and can help increase the paper’s impact.

1. Introduction:

The author spent most of the space describing the importance of segmenting convective systems, but the motivation for the work in this manuscript is not well articulated. While there are already quite several tracking algorithms in the community, why is the tracking method developed here a necessary contribution? What are the major differences/advantages of your tracking method over others? Why is it important to have the extra features (if any) from your tracking algorithm? This information should be added to either the introduction or the discussion.

1. Flow and Logic:

The flow and logic of the manuscript need improvement. For example, the paragraph starting from L245 and Figure 5 should be moved up to before Figure 3 or even earlier. The L245 paragraph introduces one of the key novelties of the method developed in this manuscript compared to previous fixed-BT tracking methods, and thus should be introduced and highlighted earlier before demonstrating and evaluating the results in Figure 4 and Figure 3, respectively.

1. Limitations in ARM observations

MMCR is a millimeter wavelength radar, and the signal attenuates quickly when observing deep convective clouds, especially in convective core and stratiform regions. The cloud top heights from MMCR in these regions are thus underestimated if relying on ARSCL data for detection. Cloud fraction profiles are also significantly impacted in the upper part of the convective systems. This will likely contribute to the discrepancies in the comparison between the HCS-drift winds and radiosonde cloud-top winds in Figure 3.

Minor:

Is BT the only parameter used in identifying segments? How did you segment the objects from the BT thresholds? Was it a watershed-type segmentation? The author does not demonstrate well how the ‘variable-BT11’ method works, with details lacking and thus making it hard to evaluate the method’s appropriateness.

The ‘feature-matching displacement’ section 2.5, how is this matrix used in the method?

What is the minimum temporal resolution required to perform the tracking? You mentioned in Section 2.1 the 1-hour resolution BT images from CERES. Are those the images you used for tracking?

Specific:

L80: GEO should be defined.

L83: Only data from the year 2006 was used?

L130: The symbols in the formulas should be explained.

L138-140: This sentence is unclear. Please rephrase it. When you say ‘irregular segments’, how do you determine the irregularity? What about segments with relatively regular shapes like convective core regions?

Figure 2: How was this figure generated? Is it from hypothetical data or satellite observations? How many years of data are used? Can you add the sample number to the figure?

L184: What is the difference between cold-core BT and cold-center BT? The previous paragraph does not seem to describe the terminology well.

Figure 6: Did you explain how you define the development and dissipation stages somewhere? Are the results shown in Figure 6 (and subsequent figures) from above the three ARM sites, or from the tropics in general as specified at the beginning of Section 2.1? Can you add sample numbers to either the figure or the caption?