Review of “Shallow and deep convection characteristics in the Greater Houston, Texas area using cell tracking methodology” by Kristofer S Tuftedal et al.

Summary of manuscript:

The authors describe the statistics of convective cells around Houston derived from 4 summer seasons of operational weather radar data. A tracking algorithm is used to capture cell life cycles and study initiation times and locations. Cells are classed as shallow, modest deep and vigorous deep based on their echo-top height and other characteristics, including GOES brightness temperature. Cells of all types are found to be most likely initiated in the morning, with a clear inland propagation associated with a sea breeze. Composite life cycles indicate distinct variations in timing of height of maximum reflectivity, and stages such as maturity and dissipation are identified. Storm initiation is related to near-storm environment based on coincident HRRR model data.

Review:

The manuscript is generally well-presented and contains useful results that are essential to other studies focusing on this domain (TRACER and ESCAPE campaigns) as well as methodology and results useful to the wider scientific community. A substantial number of clarifications are required to improve the understanding of the results, including the uncertainty. The manuscript feels slightly longer than necessary, and a number of figures are identified for potential removal without loss of key results. A section on the near-storm environment, while of great interest, does not seem to be supported by appropriate methodology or justification, and should be considered for removal. Overall, the manuscript should be considered for future publication after major revisions.

Major comments:

1. Incomplete cell life cycles.

There is some mystery about cells whose entire life cycle is affected by the domain. (i) Only in L260 (and again L292, L302) do the authors mention the possibility of cells leaving the domain. Do these cells still contribute to the statistics? How does that impact the life cycle statistics? Or the cell classification (e.g. a cell that was shallow in the domain but may have grown to be deep)? Would it not be prudent to remove all cells that leave the domain? Given the large number of cells, presumably this would leave a sufficient number for most statistical results. As a benefit, it would provide cleaner result and remove at least three artefacts (as suggested in L260 and L292, L302) and give a clearer picture of the dissipating stage. (ii) Only in L270 do the authors mention the radar “cone of silence” and some mitigation. What, exactly, happens to these cells? Are their statistics still incorporated? Is only their CTH corrected, or also maximum dBZ, etc?

2. Height uncertainty.

A second concern is the uncertainty in height, which is a major variable of interest for this study. There will be uncertainty due to the methodology, which is to pick the centre of the beam where either a detectable signal is found (CTH) or where the highest reflectivity is found (Hdbzmax). This brings both uncertainty due to the beam width and due to the separation between scans for the VCP. Both these uncertainties will increase with range from the radar and will be substantial (>1km) towards the edge of the domain. Hence, the paper would benefit from a consideration of uncertainty due to the VCP. In particular, results such as those in Figure 15 could be considered in context of this uncertainty.

3. Near-storm environment and HRRR.

The third main concern regards the near-storm environment and the use of HRRR to analyse this. Firstly, very little information is provided on HRRR in this study (see minor comments as well). Most importantly, the authors should provide evidence that HRRR is an appropriate tool to capture near-storm environment, particularly CAPE, CIN, and storm-relative helicity. Does the HRRR capture the sharp vertical gradients associated with CIN and dry layers? Won’t such gradients be smoothed out during the data assimilation cycles and/or due to the model’s effective resolution? Does HRRR really capture km-scale variability that the authors rely on for their composite soundings? Is HRRR good enough at predicting the convective storm populations that its model soundings can reasonably be used to relate to observed convective storm properties?

Regarding composite soundings, it is not obvious that this is the appropriate tool to study differences in environment. Any specific features of interest (such as CIN or dry layers) will be smoothed out by compositing. If anything, it would be more helpful to provide PDFs of CAPE and CIN across all the soundings (provided the HRRR soundings can be justified). As it stands, the authors may wish to consider removing this section 3.5, as the paper contains sufficient results of interest without it. Alternatively, the authors could consider discussing 1-3 case studies with significantly different environments, and suggest from that discussion a way forward to do further composite analysis.

Minor comments:

52: “lack of dependence on the larger scale meteorology” – There is, of course, a mesoscale phenomenon (the sea breeze) as an underlying forcing mechanism of the convection in this study (L47-49). To what extent are the findings from this study really generalisable to other synoptic situations?

72: Please expand (and possibly define) NSE on first use.

72-73: “Strict thresholding allows for the analysis of the behaviors of each case type distinct from one another.” – Please expand on (a) what variables you will threshold and (b) what is meant by “each case type”. Is case type shallow or deep, or something rather different? If you threshold on lots of variables, give a couple of examples here.

83-87: Are there any references or specific values the authors can provide here for the level of pollution? There is a difference between “pristine” and “far less polluted”. Presumably, other TRACER or ESCAPE studies have reported on aerosol concentrations?

88: “the HRRR data” – Please provide a separate (sub-)section that briefly describes the HRRR data. Particularly, for the 0900-2100 CDT period, what are the initiation times? For example, given the hourly refresh, are you always using data valid at T+3hrs, or are you using data from a fixed initiation time? What is the grid resolution? What output will be used?

90: Does daytime initiation “ensure” that sea breeze propagation was the mechanism or does it “increase the likelihood”?

88-95: Please provide a bit more information on the KHGX. How long does one VCP take? What is the beam width? Consequently, what is the horizontal width at 125km range? It is important for the reader to understand the original resolution of the data rather than the 500m by 500m horizontal grid spacing used for the regridded data.

104: Is this a standard equation for VIL? If so, please provide a reference. In any case, please provide interpretation of the formula and justification. Should the summation go to imax or imax-1 (and what is imax)? Is the denominator relevant if it is set to 1? Presumably, dh is the vertical spacing between individual sweeps of the VCP. As such, the separation will increase with range and the VIL will be quite uncertain at greater distance. Also, the highest sweep may not capture cloud tops close to the radar. Please mention these limitations and uncertainties in the text and possibly even quantify the uncertainty  Perhaps it is best to incorporate this comment with my previous one, and provide an expanded overview of the KHGX data and the subsequent derivation of VIL.

Later on, values for VIL > -20 dB are mentioned. It would be helpful for the reader to have an understanding of typical values (perhaps referring to the literature and/or more widely known variables such as total water path) for shallow, congestus, and deep convection.

106-114: This mostly describes the input to the cell tracking rather than how the algorithm actually works and any parameter choices. “The MCIT algorithm ingests time series of volume scans and tracks local maxima of VIL by identifying the two cells in consecutive radar scans that have common maximum VIL.” This is a rather sparse description of what appears to be quite a sophisticated algorithm as described in Hu et al. (2019a). The authors should describe basic components of any tracking algorithm, such as the minimum size of a cell, the consideration of displacement between consecutive images, and how merging and splitting is handled (whether by the algorithm or by later analysis, L126, noting how many cells are removed).

112: “isolate” – do you mean “identify”?

117: Best to refer to this as ETH (echo top height) rather than CTH. What if the highest gate with detectable signal is also the highest elevation? What is the uncertainty on CTH with range? Knowing the VCP would help the reader understand.

118: Similar to CTH (ETH): What if the lowest gate is also in the lowest elevation? What is the uncertainty on Hcell with range?

120: What is the uncertainty in CRatio with range?

121: What is the uncertainty in area with range?

123: Table 1. It seems that just using CTH (ETH) already separates the three classes. Are the other criteria really necessary? What would happen if a cell with CTH (ETH) between 8 and 12 km had a lifetime minimum BT warmer than 250 K (if possible) or a lifetime maximum VIL less than 0 dB (if possible) or a lifetime max CRatio less than 0.75 (if possible)? Is the cell then excluded from the statistics? If so, what number of cells is excluded?

164: “accelerate with time” – this is not shown as such in Figure 3. Sure, the fastest motions are found later in the life cycle, but what proportion of cells accelerate? You would need to provide the time of maximum motion for each individual cell (and perhaps check that it is significantly greater than the mean or minimum motion) and analyse the spread.

160-174: “from the south to east” and “from southwest to east” and “the sea breeze along the Gulf Coast plays a part in storm initiation and propagation” – how is direction defined? Is this southward or southerly, which are opposite directions. Please be precise. It would be helpful to refer to the geography (e.g. the map in Fig.5) to indicate if this direction is typically away from or towards the coast (or even parallel) and hint at the typical direction of the sea breeze.

170: “without further analysis” – out of interest, does MCIT not keep track of the number of VIL minima per cell? That could provide some indication (e.g. a PDF of number of VIL minima per cell).

182-191: Figure 5 does not seem essential to the paper, with the interesting results shown in Figure 6. As the paper has quite a large number of figures, please consider removing figure 5 and merging the relevant discussion with that for figure 6 (although most relevant points are repeated already).

184: “Houston” – to those unfamiliar with the geography, this is difficult to place (as presumably the centre of the domain coincides with the radar location, which may not be in the centre of Houston. Could the authors consider adding at least a marker if not an outline of the urban area to those figures that show the map? Similarly, “Galveston Bay” is perhaps a bit more obvious given the shape of the coastline, but would still be helpful to indicate (or describe its location on the map).

188-189: “local maximum over the Houston metropolitan area” – to what extent could aerosol loading be differentiated from an urban heat island effect? Should urban heat island not be mentioned here as a possible cause?

212: “below 5 km, suggesting warm precipitation processes” – Here, or earlier, please provide the average height of the freezing level (and perhaps -20 deg C) during the season.

218-219: “The rapid change in GOESBT and HdBZmax shows the quick vertical evolution of these cells resulting in cold precipitation processes.” – Again, that’s not precisely what is shown. A change in a population of cells does not show a quick change in individual cells. What is shown, instead, is that there is a shift in the population dominated by newly initiated cells (as indicated by Figures 4 and 6) to a population dominated by mature and/or long-lived cells. This is still a very interesting result. The “quick vertical evolution” is still a valid conclusion, but it is more evident from the next Figures and should therefore be mentioned in section 3.4.

239-243: It is not clear why the discontinuity is “unnatural” and not just an indicator of a part of the life cycle. Wouldn’t lower Hdbzmax (whether or not due to bright band) simply be an indicator of the storm evolving into an “orphan anvil” that is precipitating? It’s not clear why this is presented as an artefact of the analysis, rather than a useful result about the storm physics.

249-251: “the feature of low dBZavg and high dBZmax is indicative of anvil generation” – How? The use of dBZ-avg for studying anvil generation requires justification. No doubt it is sensible, but to the reader interested in anvil but not so well versed in radar meteorology, this will be very confusing. Is there a reason why the authors did not use the cell area?

263-273: Given that the interest is in the patterns during the life cycle, Figure 11 and 12 are not essential to the paper and could be removed. Figure 13 is of some interest, but may appear as supplementary information as it explains an oddity, but is not essential to the main results. Removing this paragraph also presents a more natural flow into the discussion of Figure 14.

304: “within these cells” – which cells? Shallow, deep, all?

304-318: This is an interesting result but requires some further serious consideration, particularly in light of discussing uncertainty in height estimates (major comment 2). It is a stretch to suggest clear maxima in ascent rates when the mode is obviously (close to) zero throughout.

Figures 3, 5-15, 18: Please change these to a colour scale that is appropriate for sequential increases and appropriate for colourblind people. Certain colours such as the dark purple, dark green, and black stand out because they contrast with the adjust colours, likely emphasising the wrong data. Most software packages should suggest such colour bars, or one can be designed using a tool such as colorbrewer.

Editorial comments:

17: “the maximum radar reflectivity and its height” – “the height and value of the maximum radar reflectivity”

48: “The land-sea breeze circulation […] have been shown” – “…has been shown”

54: “Previous studies […] suggests” – “suggest”

80: “The area used for this study was selected such that it was centered…” – “Our study domain was centered…”

102: “using the (1)” – “using equation (1)”

111 & 116: Please refer to KHGX rather than WSR here.

126: “ration” – “ratio”

156-157: “All three case types…” – more or less repeats L155, suggest merging.

241: ARL – this acronym is hardly used so please just spell out “above radar level” for the remaining cases.

The study describes basic cloud properties based on the MCIT cell tracking algorithm. This is useful and timely research, but it could benefit from the following comments:

1. The authors should acknowledge that the core algorithm of MCIT is based on Rosenfeld (1987), and an associated very similar study of the relationships between the tracked cloud properties based on it (Gagin et al., 1985).
2. Line 24-37, another key reason models usually fail in a real case simulation is that modelers mostly focus on reflectivity comparison between model and radars. Reflectivity can be a very confusing parameter for cloud microphysical analysis since it is proportional to the first moment of particle concentration and 6^th moment to the size of particles within each radar gate volume. Recent efforts to use forward operators (Ryzhkov et al., 2011, Wolfensberger and Berne., 2018, and Kumjian et al., 2019) to simulate dual polarimetric parameters like ZDR, KDP, and Rhohv demonstrate stronger confidence in cloud microphysical analysis.
3. Line 62-65, Hu et al., 2019b indicate the dataset is roughly 3000 cells during a multi-year window of cell tracking within the greater Houston area. This study did analyze general characteristics (Figs 2,3, 6-8) of cloud lifecycles of many cells. So I won’t say this is only “a few convective clouds”.
4. Line 68, add space “aroundthe”.
5. Line 80-82, it is recommended to use radar site centric domain instead of city landmarks to avoid radar beam size inhomogeneity at the same distance from the center of domain. In addition, the authors domain is over 100 km from the radar. Please justify the vertical extent the authors are focusing on and why.
6. Line 92-93, what is the vertical resolution for VIL calculation?
7. Equation 1, Is there any Z limit for VIL calculation? Say if Z is 60 dBZ, is it still used to calculate VIL?
8. Line 117, for CTH detection, what reflectivity threshold, if any, is used here? Or how did the authors determine if the 88D radar data is noise or weather echoes?
9. Line 118, the lowest gate of radar detectible signal is forced by the range from radar as well, so how did the authors make a correction about that?
10. Table 1, is it possible for the authors to provide a time series or movie of one of the tracked cases to demonstrate the labeling of shallow, modest, and vigorous convective cells? Please include the cell boundaries as contours when generating the figure/movie, in the supplementary materials would be sufficient.
11. Figure 4. Are there any restrictions on the lifetime of cells here? Say at least 30 min? Or 5-6 radar volume scans? In addition, please elaborate on the reasoning for peak differences between the three convection types.
12. Figure 8, for the shallow case, why there’s not much difference at different life stages? A similar concern applies to the vigorous type of convection. For example, in panel i, the majority of cell max Z is still very high over 8 km (from 10-over 50 dBZ) here. This raises concerns about the cell tracking quality of this dataset. Cell dissipating should not have Z still over 50 dBZ, that is a hail signature, but in panel l, Z over 50 dBZ is over 10 km and has the highest frequency, although it is less frequent compared with panel i, but still the highest under this normalized lifetime category. It seems that the tracking was terminated due to a splitting event or continuing under another cell identity. This problem was addressed and fixed by Yin et al. (2022). This is an improved version of MCIT that keeps track of all the splits.

References:

Ryzhkov, A., M. Pinsky, A. Pokrovsky, and A. Khain, 2011: Polarimetric Radar Observation Operator for a Cloud Model with Spectral Microphysics. J. Appl. Meteor. Climatol., 50, 873–894, https://doi.org/10.1175/2010JAMC2363.1.

Wolfensberger, D. and Berne, A.: From model to radar variables: a new forward polarimetric radar operator for COSMO, Atmos. Meas. Tech., 11, 3883–3916, https://doi.org/10.5194/amt-11-3883-2018, 2018.

Kumjian, M. R., C. P. Martinkus, O. P. Prat, S. Collis, M. van Lier-Walqui, and H. C. Morrison, 2019: A Moment-Based Polarimetric Radar Forward Operator for Rain Microphysics. J. Appl. Meteor. Climatol., 58, 113–130, https://doi.org/10.1175/JAMC-D-18-0121.1.

Rosenfeld, D., 1987: Objective method for analysis and tracking of convective cells as seen by radar. Journal of Atmospheric and Oceanographic Technology, 4, 422-434.

Gagin, A., D. Rosenfeld and R.E. Lopez, 1985: The relationship between height and precipitation characteristics of summertime convective cells in south Florida. Journal of Atmospheric Sciences, 42, 84-94.

Yin, J., Pan, Z., Rosenfeld, D., Mao, F., Zang, L., Zhu, Y., Hu, J., Chen, J. and Gong, J., 2022: Full‐tracking Algorithm for Convective Thunderstorm System from Initiation to Complete Dissipation. Journal of Geophysical Research: Atmospheres, p.e2022JD037601.

This study presents an extensive investigation into convective cell characteristics in the Greater Houston area, delivering important insights into their behavior, movement, anvil generation, vertical motions, and the influencing environmental factors. The paper is well-organized and distinguishes itself through its methodical tracking of convective cells across their lifecycle and the in-depth analysis of both shallow and deep convection phenomena. However, some aspects might require further refinement.

Specific comments:

1. While the characteristics of convective cells and the impact of local meteorology are discussed, the exploration of the underlying causes and synoptic patterns of these meteorological conditions could be expanded. Such a discussion could provide greater context and depth to the study.
2. The authors acknowledge potential artifacts and uncertainties in their data, primarily related to cells exiting the tracking domain during their anvil generation analysis. It will be beneficial to include a more comprehensive explanation of how this issue might influence the identified characteristics of shallow and deep convection.
3. The authors suggest that aerosols might not have a significant influence on cell initiation. However, it would be useful to understand the types of aerosols considered and the potential role each may play in convective processes. The current analysis and discussion on the role of aerosols could be made more explicit and comprehensive.
4. The study alludes to other potential factors influencing cell initiation but does not explore these in detail. Incorporating these factors into the analysis or outlining them as future areas of study could add more value and richness to the research.
5. In Section 3.1, the authors state that all types of convective cells reach their peak monthly average in August. It would be beneficial if they could offer more insight into the underlying mechanisms driving these monthly variations. Understanding the causative factors may help further understand the influential factors for the convective cells.