the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 07 Nov 2023
Lifecycle of Updrafts and Mass Flux in Isolated Deep Convection over the Amazon Rainforest: Insights from Cell Tracking
Abstract. Long term observations of deep convective cloud (DCC) vertical velocity and mass flux were collected during the GoAmazon2014/5 experiment. Precipitation echoes from a surveillance weather radar near Manaus, Brazil are tracked to identify and evaluate the isolated DCC lifecycle evolution during the dry and wet seasons. A Radar Wind Profiler (RWP) provides precipitation and air motion profiles to estimate the vertical velocity, mass flux, and mass transport rates within overpassing DCC cores as a function of the tracked cell lifecycle stage. The average radar reflectivity factor (Z), DCC area (A), and surface rainfall rate (R) increased with DCC lifetime as convective cells were developing, reached a peak as the cells matured, and decreased thereafter as cells dissipated.
As the convective cells mature, cumulative DCC properties exhibit stronger updraft behaviors with higher upward mass flux and transport rates above the melting layer (compared to initial and later lifecycle stages). In comparison, developing DCCs have the lowest Z associated with weak updrafts, and negative mass flux and transport rates above the melting layer. Over the DCC lifetime, the height of the maximum downward mass flux decreased whereas the height of maximum net mass flux increased. During the dry season, the tracked DCCs had higher Z, propagation speed, and DCC area, and were more isolated spatially compared to the wet season. Dry season DCCs exhibit higher Z, mass flux, and mass transport rate while developing whereas wet season DCCs exhibit higher Z, mass flux, and mass transport rates at later stages.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(2310 KB)
-
Supplement
(429 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(2310 KB) - Metadata XML
-
Supplement
(429 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2410', Anonymous Referee #1, 09 Dec 2023
The authors present novel, insightful and pertinent research into the relation between convective intensity, lifecycle, season and convective mass flux using a combination of RWP and geostationary satellite observations with a cloud-tracking approach. This research is highly relevant as an improved understanding of how convective-scale processes impact convective mass flux, and how these processes change over the convective lifecycle, is vital to understanding their connected to larger scale changes such as anvil area and lifetime. In addition, the study is performed using data from the Amazon region, where deep convective dynamics are important but factors affecting convective-scale processes are less well understood.
In general I find the paper to be insightful and provide useful information, and the results are generally well described. Clear conclusions are reached, and these are linked to prior research. However, I have several concerns regarding how the lifecycle stages of the tracked DCCs are defined. These lifecycle stages are crucial to the further analysis of the paper. In addition, while the results are generally described clearly in the text, the figures are less clear. I would like to see improvements in how they are presented, and the inclusion of additional figures to show how features are detected using the radar data. My recommendation is that the article be reconsidered after major revisions.
Major comments:
Section 2.4: It is very unclear exactly how the lifetime stages are defined. There are several references to the stages being “determined” by Z, cell area or their tendencies, but it seems from L208-211 that the actual definition of the five-lifetime stages is quintiles of the total cell lifetime. Is this correct? The labelling of these bins is then justified by the average Z and area of features within each lifetime quintile. L213 justifies the third quintile as being the mature stage as it has the highest average Z and area, although figure 3c shows that the 4th bin has the largest average area. Have the authors considered whether the different lifecycle stages make up the same proportion of cell lifetime in all cases? In particular, are longer-lived cells expected to have the same proportions as shorter-lived cells?
The second major issue with section 2.4 is a potential sampling bias on the lifecycle bins for cell lifetime due to the similar number of bins as the average number of observations of most features. In particular, the most common observed lifetime of 36 minutes means that there are four radar scans of the tracked cell. In this case, therefore, there will always be one lifecycle bin that is never classified for cells of this length. Based on a quintile normalisation, that will be bin 3 or the “mature” stage. If we assume that longer lived cells are also more intense, then we would expect to have a bias in the Z of “mature” DCC observations because short-lived DCCs are never classified into bin 3. Is this the case, and if so, could the authors look into whether this does affect the analysis? I am not sure how to resolve this issue, other than reducing the number of bins or increasing the minimum length of cells selected for analysis.
Figures: In general I found that the figures were not very clear, particularly from figure 8 onwards. Use of similar colours for the lines in figures 8 onwards makes them difficult to distinguish compared to the line colours used in figures 6 & 7. In addition, multiple different lines are plotted in the same charts without using different line styles (e.g. dashed, dotted) to distinguish them. In many plots, “lifetime bin” is used as the x axis. It would be clearer to a reader looking first at the figures if this was changed to “lifetime stage” with labels “developing”, “early mature”, “mature” etc. In addition, several of the figures show the same properties but for the wet or dry seasons (figures 9 & 10, 14 & 14). It would be easier to compare these cases if both wet & dry season plots were included on the same figure. In addition, I think it would help the reader to understand the features being tracked if an example of the gridded Z field with tracked features was included as an additional figure.
Minor comments:
L35: It may be informative to include a more recent reference showing that km-scale convective resolving models do not fully resolve this issue
L71: Jeyaratnam et al., 2020 missing from references
L77: Minor: The acronym for ‘Manacapuru, Brazil’ is listed here as T3/MAO, but MAO is used throughout the text. It would be clearer simply to list the acronym as MAO here.
L111: How much is the gridded Z affected by low-level convection and congestus at 2km?
L120: For reproducibility it would be good to state which version of tobac was used. Also, if a newer version of tobac is used than that presented in the 2019 paper, it may be appropriate to also cite Sokolowsky & Freeman et al. 2023 (https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1722/, in review)
L124: It would be good to also discuss previous radar cloud tracking, possibly noting that much of the application of cloud tracking with cloud radar data has focused on nowcasting rather than weather (e.g. Wilson et al. 1998 [https://doi.org/10.1175/1520-0477(1998)079%3C2079:NTASR%3E2.0.CO;2], Keenan et al. 2003 [https://doi.org/10.1175/BAMS-84-8-1041]), although more recently there has been increased focus on its use for studying convective cloud processes (e.g. Feng et al. 2022 [https://doi.org/10.1175/MWR-D-21-0237.1])
L126: It would be informative to state why this improved makes a difference, both in terms of the tracking of DCCs and the study of convective processes
L130: It would be clearer to state that these thresholds have been chosen by the authors, and link with the rest of the paragraph about why they have been chosen, and what storm features/intensity they correspond to
L153: Initiation of the tracked cell, convection and precipitation are not necessarily the same. It is expected for there to be some delay between the initiation of convection and the radar reflectivity reaching the detection threshold. It would be better to be clear that the measured initiation is the first time of detection.
L173: This median nearest neighbour distance seems very small. Is this the distance between the centre of each feature or the edges? Can neighbouring cells be clearly defined with this separation on a 1km grid? It may be helpful to include a figure of an example of the gridded Z field with detected features to show this clearly.
L182: A 36-minute lifetime should correspond to 4, rather than 3, radar scans (inclusive of start and end time)
L190: Text says 1,130 cells, but table 1 says 672 cells. Is there a difference in the selection criteria here?
L244: Giangrande et al., 2016 mentions two different classification schemes (Steiner et al. 1995 & Giangrande et al., 2013). It would be helpful to provide a brief description of the scheme in the paper, rather than simply providing a reference.
L273 I think it would be clearer if cells with areas > 500 km2 were excluded in the criteria listed earlier, and figure 3 shown without their influence.
L304: I would like to see this result included in the main paper, rather than the supplementary materials. The difference in the lifetime and organisation of cells between the wet and dry seasons is a nice result to show from this analysis. Possibly also shown cell area distributions, as this is also mentioned as being different in the text
L325: These are interesting results. It would also be interesting to show if there is a difference in average cell lifetime for the different diurnal cycle bins
L497: These areas do not seem to match those shown in figure 3
L622: It may also be worth noting the relative lifecycle of the anvil cloud compared to the core. i.e. we only expect to see the anvil start to dissipate after the core has dissipated, and so while the core is in its dissipating phase the anvil is still mature and therefore expected to continue to have cold Tb
L645: Figure 17c appears to show that Tb reached a maximum during the mature phase for cells with ETH between 10 and 12 km. Are the classification by Tb and ETH for each bin in b and c respectively just based on the measurements of those properties during that lifecycle bin? If so, that may confound the analysis shown here. For example, only the most intense cells are likely to have an ETH > 10km during the developing phase, and so the average Tb of these cells is expected to be cold. However, more, less intense, cells may reach the 10 km ETH level during the mature phase, which could make the average Tb warmer simply because the distribution of the intensity of sampled cells has changed. It would be clearer if cells were classified by the minimum Tb and maximum ETH observed over their entire lifetime, rather than during the lifecycle stage, for figure 17b and c respectively. However, while this should be possible for Tb, I’m not sure if this is possible with ETH provided by the RWP. Perhaps the SIPAM measurement could be used instead to provide ETH?
L649: Cases with Tb this low, particularly during the mature stage, seem like they could be due to a misidentification or mislocation of the GOES-13 imagery. Have the authors looked at what is shown in the GOES imagery here (i.e. are there colder pixels near to but not at the collocated location).
L651: Use of the 241 K threshold dates all the way back to Maddox, 1980 [https://www.jstor.org/stable/26221473], but it was selected on account of the -32C contour being provided by the GOES-1 cold Tb enhancement algorithm rather than a specific physical basis and has stuck ever since.
Figure 1: The cell tracks plotted in a/c are very unclear. It may be clearer to plot a “heatmap” with all of the tracks plotted in the same colour but with translucency. Including an arrow or similar to show the direction of each track would also be informative. In addition, the river looks like a cell track at first glance. Showing the location of the SIPAM would also be informative
Figure 7: Mention in the caption that the 18-00 LT bin is not shown due to the small number of samples as mentioned in the text.
Table S1: Please format these more clearly as dates
Citation: https://doi.org/10.5194/egusphere-2023-2410-RC1 -
RC2: 'Comment on egusphere-2023-2410', Anonymous Referee #2, 18 Dec 2023
Review of egusphere-2023-2410: Lifecycle of Updrafts and Mass Flux in Isolated Deep Convection over the Amazon Rainforest: Insights from Cell Tracking
Summary: In this manuscript, the authors calculate a variety of statistics as a function of storm lifetime and season from a database of tracked convective cells from GoAmazon2014-15 observations. The authors present a commendable set of different methods, observations and statistics to provide additional insights about convective dynamics, how they vary as a function of storm lifecycle and season, and how their results fit in with recent results from their research group. Given the complexity of their analysis and dataset, there were many aspects of their analysis that warrant additional description and details to properly understand what their data sample represents and the subsequent results that are being presented. Similarly, many of the figures / analyses require additional descriptions to interpret and understand what is being presented. Therefore, while the work fits well within ACP’s focus and could add insights for understanding convective storm dynamics, I am not able to fully assess the science results without a better understanding of the methods.
Major comments, questions, and/or concerns:
Combining MAO data with storms. If a storm is within 20 km of the MAO site at any point in its lifecycle, the RWP data from MAO are used and attributed to a storm lifecycle stage based on the tracking. The average distance between the feature center and MAO site was 8.5 km, with 70% of feature positions being with 10 km of the MAO site (L252). Whether the storm updraft goes directly over the site versus if its center is ~10 km away from the site might make a significant difference in the RWP data (i.e., Z, W) that are being associated with the storm (i.e., profiles of Z and W at the center of the storm updraft region will be different from the edges of the storm). This seems like a large source of uncertainty that is not discussed. For example, are there biases in the distance of the updraft center to what was sampled by MAO with respect to the different lifecycle stages and seasons, and if so, how do these biases impact the results in the manuscript? Perhaps, showing Figure 1 as a heat map both for all storms and for storms at the different stages and seasons would help address this concern.
Tracking storms with a coarse time-step. The authors focus on tracking isolated cells with a 12-minute radar data frequency at 2 km AGL. Could there be instances in the dataset where a storms dissipates, and a new one forms in similar location within the 12-minute time step, which is therefore tracked as a continuous storm? This seems like a large source of uncertainty that isn’t discussed.
- Splits/mergers. What happens if a storm splits into two storms or merges with a different storm? Could these events be present in the dataset? This is related to concerns about the coarse time step for tracking, and how confident the authors can be about the determination of lifecycle stages.
Using 2 km reflectivity. The authors use 2 km reflectivity for tracking and defining feature centers (L127). Why do they use this altitude, as opposed to a column maximum reflectivity that may be more accurately located with the updraft? Have they tried using different altitudes? For storms that may not be precipitating heavily, could the 2 km reflectivity with a 30 dBZ minimum threshold miss some of the beginning and ending stages of a storm lifetime? Similarly, for sheared storms, the 2 km location may not accurately describe the updraft location?
p(w) seems to play an important role in the calculation of vertical mass flux and transport but is placed in the supplement without much discussion of how it is calculated and what it means. The authors do include a brief explanation of p(w), but it was not clear (L460-462). Can the authors spend more time describing this, given its importance in their calculations of mass flux?
- L448-450: The authors state that 2 min periods of RWP observations are selected, and this is based on an assumed median updraft width of 1 km and average propagation speed. I do not understand how these variables are related to the 2 min period of RWP observations that are chosen, and found this sentence confusing? Can the authors make this clearer?
- L452: The authors state that the average updraft/downdraft are weighted by the probability of sampling an updraft/downdraft during this time period (p(w)). However, the authors do not provide a clear description of how p(w) is calculated and place the figure of p(w) in the appendix (Figure S3). Can the authors provide a clearer description of how p(w) is calculated, and place this figure in the main manuscript, given its importance in calculating mass flux?
Additional Methodology Clarifications:
L164: How exactly are isolated convection (“ISO”) days determined? The authors point to Giangrande et al., 2023, but should provide the description here, given that this is defining the dataset used in this study.
The authors state that isolated DCCs observed on “ISO” days are used for subsequent analysis. How are these isolated DCCs chosen? I am especially curious given that some 500 km2 storms seem to make it into their dataset (i.e., L274).
It seems like many of the tracked storms last 3 or 4 time steps, but how can the authors classify these cells into their 5 lifecycle bins?
How big is the radar domain and is this constrained by using 2 km CAPPIs, given that the radar beams will increase in altitude away from the radar?
Can storms that initiate outside the radar domain move into the domain? If so, how would the authors know its lifecycle stage?
“Average Z:” Many times in the manuscript the authors talk about average Z, and sometimes I was confused whether they are referring to the 2 km CAPPI Z used in the tracking, Z from the SIPAM radar at different altitudes, or the Z from the RWP observations. Can the authors make this clearer throughout the manuscript?
L234: Does “at least 10 consecutive cloud echoes” mean consecutive in time or height?
L366: When the authors use the term “Mature” DCCs (here and elsewhere in the manuscript), are the authors referring to data in lifetime bin 3 or data in lifetime bins 2-4, since bins 2 and 4 are defined as “early mature” and “late mature?”
L556: The authors state that “time 0 represents the timestep when the indicated lifecycle stage of the DCC was identified using tobac.” I am having a hard time interpreting what this means. What if there are more than one radar timesteps for a given lifecycle stage? In that case, which time is used?
L556: Is the equivalent potential temperature from MAO used even when cells are >20km away from MAO? If so, can what is happening at MAO be accurately related to the storm dynamics from the respective storm?
L609: What Tb measurement is being used in this analysis? Is it the Tb measurement at MAO, at the feature center, or some average over the entire feature?
Figure 1 (L762): Is this data for all tracked cells, or only the cells that go into the analyses presented in the remaining figures?
Figure 2 (L768): Are “other DCC days” the same as ACE days as described in the text? If so, can the authors use the same terminology?
Figure 2: (L770): Why is there a propagation speed threshold here of 0.5 m/s? I don’t believe this was mentioned in the text?
Figure 3: For this (and all) boxplot figures, what is being shown? For example, do the bars that extend out represent the max/min values or some other value? A clearer description of that is being shown in the boxplots is necessary.
Figure 16: Is this showing the mean or median equivalent potential temperature?
L826: Is there a typo here? Should Figure 8 be Figure 13?
There is an inconsistency of Figure S3 caption to what is stated in the figure image (within 1 min versus within 2 mins).
Other Suggestions:
L69: The NASA Investigation of Convective Updrafts Mission (e.g., van den Heever, 2021; Stephens et al., 2020 Prasanth et al., 2023) is planning to do this, so it would probably be good to mention this here, since this research has implications for this mission as well.
Many of the results involve comparing results between the different seasons, which show up in different figures and makes comparing them quite difficult. Can the authors combine figures effectively, so that differences between results can be more clearly assessed? This is particularly the case for Figures 12-14 and 8-10.
In terms of assessing whether their smaller sample had sampling biases, what is the reason for using the nearest neighbor distance as a metric, as opposed to metrics (i.e., reflectivities, ETHs) that may be more telling of the types/intensities of storms being sampled? It seems like it would be important to understand if their sample represents both weak and strong storms, compared to the possible population of storms?
L457: For calculating mass flux, the authors assume the storm area is constant as a function of height, based on the area of the storm at 2 km AGL? It seems like the authors potentially have the data to validate this assumption (i.e., SIPAM radar data), or can find supporting literature for this assumption?
This manuscript is already quite lengthy, and while all the analyses are interesting, perhaps some areas (i.e., S3.3) are less relevant to the authors focus on updrafts and can be removed. This may allow them to better describe their other analyses within a more reasonable space.
References:
van den Heever, S. (2021). NASA selects new mission to study storms, impacts on climate models. NASA Earth. Retrieved from https://www.nasa.gov/press-release/nasa-selects-new-mission-to-study-storms-impacts-on-climate-models
Stephens, G., van den Heever, S. C., Haddad, Z. S., Posselt, D. J., Storer, R. L., Grant, L. D., et al. (2020). A distributed small satellite approach for measuring convective transports in the Earth's atmosphere. IEEE Transactions on Geoscience and Remote Sensing, 58(1), 4–13. https://doi.org/10.1109/TGRS.2019.2918090
Prasanth, S., Haddad, Z. S., Sawaya, R. C., Sy, O. O., van den Heever, M., Narayana Rao, T., & Hristova-Veleva, S. (2023). Quantifying the vertical transport in convective storms using time sequences of radar reflectivity observations. Journal of Geophysical Research: Atmospheres, 128, e2022JD037701. https://doi.org/10.1029/2022JD037701.
Citation: https://doi.org/10.5194/egusphere-2023-2410-RC2 - Splits/mergers. What happens if a storm splits into two storms or merges with a different storm? Could these events be present in the dataset? This is related to concerns about the coarse time step for tracking, and how confident the authors can be about the determination of lifecycle stages.
-
AC1: 'Author responses to referee comments', Siddhant Gupta, 10 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2410/egusphere-2023-2410-AC1-supplement.pdf
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2410', Anonymous Referee #1, 09 Dec 2023
The authors present novel, insightful and pertinent research into the relation between convective intensity, lifecycle, season and convective mass flux using a combination of RWP and geostationary satellite observations with a cloud-tracking approach. This research is highly relevant as an improved understanding of how convective-scale processes impact convective mass flux, and how these processes change over the convective lifecycle, is vital to understanding their connected to larger scale changes such as anvil area and lifetime. In addition, the study is performed using data from the Amazon region, where deep convective dynamics are important but factors affecting convective-scale processes are less well understood.
In general I find the paper to be insightful and provide useful information, and the results are generally well described. Clear conclusions are reached, and these are linked to prior research. However, I have several concerns regarding how the lifecycle stages of the tracked DCCs are defined. These lifecycle stages are crucial to the further analysis of the paper. In addition, while the results are generally described clearly in the text, the figures are less clear. I would like to see improvements in how they are presented, and the inclusion of additional figures to show how features are detected using the radar data. My recommendation is that the article be reconsidered after major revisions.
Major comments:
Section 2.4: It is very unclear exactly how the lifetime stages are defined. There are several references to the stages being “determined” by Z, cell area or their tendencies, but it seems from L208-211 that the actual definition of the five-lifetime stages is quintiles of the total cell lifetime. Is this correct? The labelling of these bins is then justified by the average Z and area of features within each lifetime quintile. L213 justifies the third quintile as being the mature stage as it has the highest average Z and area, although figure 3c shows that the 4th bin has the largest average area. Have the authors considered whether the different lifecycle stages make up the same proportion of cell lifetime in all cases? In particular, are longer-lived cells expected to have the same proportions as shorter-lived cells?
The second major issue with section 2.4 is a potential sampling bias on the lifecycle bins for cell lifetime due to the similar number of bins as the average number of observations of most features. In particular, the most common observed lifetime of 36 minutes means that there are four radar scans of the tracked cell. In this case, therefore, there will always be one lifecycle bin that is never classified for cells of this length. Based on a quintile normalisation, that will be bin 3 or the “mature” stage. If we assume that longer lived cells are also more intense, then we would expect to have a bias in the Z of “mature” DCC observations because short-lived DCCs are never classified into bin 3. Is this the case, and if so, could the authors look into whether this does affect the analysis? I am not sure how to resolve this issue, other than reducing the number of bins or increasing the minimum length of cells selected for analysis.
Figures: In general I found that the figures were not very clear, particularly from figure 8 onwards. Use of similar colours for the lines in figures 8 onwards makes them difficult to distinguish compared to the line colours used in figures 6 & 7. In addition, multiple different lines are plotted in the same charts without using different line styles (e.g. dashed, dotted) to distinguish them. In many plots, “lifetime bin” is used as the x axis. It would be clearer to a reader looking first at the figures if this was changed to “lifetime stage” with labels “developing”, “early mature”, “mature” etc. In addition, several of the figures show the same properties but for the wet or dry seasons (figures 9 & 10, 14 & 14). It would be easier to compare these cases if both wet & dry season plots were included on the same figure. In addition, I think it would help the reader to understand the features being tracked if an example of the gridded Z field with tracked features was included as an additional figure.
Minor comments:
L35: It may be informative to include a more recent reference showing that km-scale convective resolving models do not fully resolve this issue
L71: Jeyaratnam et al., 2020 missing from references
L77: Minor: The acronym for ‘Manacapuru, Brazil’ is listed here as T3/MAO, but MAO is used throughout the text. It would be clearer simply to list the acronym as MAO here.
L111: How much is the gridded Z affected by low-level convection and congestus at 2km?
L120: For reproducibility it would be good to state which version of tobac was used. Also, if a newer version of tobac is used than that presented in the 2019 paper, it may be appropriate to also cite Sokolowsky & Freeman et al. 2023 (https://egusphere.copernicus.org/preprints/2023/egusphere-2023-1722/, in review)
L124: It would be good to also discuss previous radar cloud tracking, possibly noting that much of the application of cloud tracking with cloud radar data has focused on nowcasting rather than weather (e.g. Wilson et al. 1998 [https://doi.org/10.1175/1520-0477(1998)079%3C2079:NTASR%3E2.0.CO;2], Keenan et al. 2003 [https://doi.org/10.1175/BAMS-84-8-1041]), although more recently there has been increased focus on its use for studying convective cloud processes (e.g. Feng et al. 2022 [https://doi.org/10.1175/MWR-D-21-0237.1])
L126: It would be informative to state why this improved makes a difference, both in terms of the tracking of DCCs and the study of convective processes
L130: It would be clearer to state that these thresholds have been chosen by the authors, and link with the rest of the paragraph about why they have been chosen, and what storm features/intensity they correspond to
L153: Initiation of the tracked cell, convection and precipitation are not necessarily the same. It is expected for there to be some delay between the initiation of convection and the radar reflectivity reaching the detection threshold. It would be better to be clear that the measured initiation is the first time of detection.
L173: This median nearest neighbour distance seems very small. Is this the distance between the centre of each feature or the edges? Can neighbouring cells be clearly defined with this separation on a 1km grid? It may be helpful to include a figure of an example of the gridded Z field with detected features to show this clearly.
L182: A 36-minute lifetime should correspond to 4, rather than 3, radar scans (inclusive of start and end time)
L190: Text says 1,130 cells, but table 1 says 672 cells. Is there a difference in the selection criteria here?
L244: Giangrande et al., 2016 mentions two different classification schemes (Steiner et al. 1995 & Giangrande et al., 2013). It would be helpful to provide a brief description of the scheme in the paper, rather than simply providing a reference.
L273 I think it would be clearer if cells with areas > 500 km2 were excluded in the criteria listed earlier, and figure 3 shown without their influence.
L304: I would like to see this result included in the main paper, rather than the supplementary materials. The difference in the lifetime and organisation of cells between the wet and dry seasons is a nice result to show from this analysis. Possibly also shown cell area distributions, as this is also mentioned as being different in the text
L325: These are interesting results. It would also be interesting to show if there is a difference in average cell lifetime for the different diurnal cycle bins
L497: These areas do not seem to match those shown in figure 3
L622: It may also be worth noting the relative lifecycle of the anvil cloud compared to the core. i.e. we only expect to see the anvil start to dissipate after the core has dissipated, and so while the core is in its dissipating phase the anvil is still mature and therefore expected to continue to have cold Tb
L645: Figure 17c appears to show that Tb reached a maximum during the mature phase for cells with ETH between 10 and 12 km. Are the classification by Tb and ETH for each bin in b and c respectively just based on the measurements of those properties during that lifecycle bin? If so, that may confound the analysis shown here. For example, only the most intense cells are likely to have an ETH > 10km during the developing phase, and so the average Tb of these cells is expected to be cold. However, more, less intense, cells may reach the 10 km ETH level during the mature phase, which could make the average Tb warmer simply because the distribution of the intensity of sampled cells has changed. It would be clearer if cells were classified by the minimum Tb and maximum ETH observed over their entire lifetime, rather than during the lifecycle stage, for figure 17b and c respectively. However, while this should be possible for Tb, I’m not sure if this is possible with ETH provided by the RWP. Perhaps the SIPAM measurement could be used instead to provide ETH?
L649: Cases with Tb this low, particularly during the mature stage, seem like they could be due to a misidentification or mislocation of the GOES-13 imagery. Have the authors looked at what is shown in the GOES imagery here (i.e. are there colder pixels near to but not at the collocated location).
L651: Use of the 241 K threshold dates all the way back to Maddox, 1980 [https://www.jstor.org/stable/26221473], but it was selected on account of the -32C contour being provided by the GOES-1 cold Tb enhancement algorithm rather than a specific physical basis and has stuck ever since.
Figure 1: The cell tracks plotted in a/c are very unclear. It may be clearer to plot a “heatmap” with all of the tracks plotted in the same colour but with translucency. Including an arrow or similar to show the direction of each track would also be informative. In addition, the river looks like a cell track at first glance. Showing the location of the SIPAM would also be informative
Figure 7: Mention in the caption that the 18-00 LT bin is not shown due to the small number of samples as mentioned in the text.
Table S1: Please format these more clearly as dates
Citation: https://doi.org/10.5194/egusphere-2023-2410-RC1 -
RC2: 'Comment on egusphere-2023-2410', Anonymous Referee #2, 18 Dec 2023
Review of egusphere-2023-2410: Lifecycle of Updrafts and Mass Flux in Isolated Deep Convection over the Amazon Rainforest: Insights from Cell Tracking
Summary: In this manuscript, the authors calculate a variety of statistics as a function of storm lifetime and season from a database of tracked convective cells from GoAmazon2014-15 observations. The authors present a commendable set of different methods, observations and statistics to provide additional insights about convective dynamics, how they vary as a function of storm lifecycle and season, and how their results fit in with recent results from their research group. Given the complexity of their analysis and dataset, there were many aspects of their analysis that warrant additional description and details to properly understand what their data sample represents and the subsequent results that are being presented. Similarly, many of the figures / analyses require additional descriptions to interpret and understand what is being presented. Therefore, while the work fits well within ACP’s focus and could add insights for understanding convective storm dynamics, I am not able to fully assess the science results without a better understanding of the methods.
Major comments, questions, and/or concerns:
Combining MAO data with storms. If a storm is within 20 km of the MAO site at any point in its lifecycle, the RWP data from MAO are used and attributed to a storm lifecycle stage based on the tracking. The average distance between the feature center and MAO site was 8.5 km, with 70% of feature positions being with 10 km of the MAO site (L252). Whether the storm updraft goes directly over the site versus if its center is ~10 km away from the site might make a significant difference in the RWP data (i.e., Z, W) that are being associated with the storm (i.e., profiles of Z and W at the center of the storm updraft region will be different from the edges of the storm). This seems like a large source of uncertainty that is not discussed. For example, are there biases in the distance of the updraft center to what was sampled by MAO with respect to the different lifecycle stages and seasons, and if so, how do these biases impact the results in the manuscript? Perhaps, showing Figure 1 as a heat map both for all storms and for storms at the different stages and seasons would help address this concern.
Tracking storms with a coarse time-step. The authors focus on tracking isolated cells with a 12-minute radar data frequency at 2 km AGL. Could there be instances in the dataset where a storms dissipates, and a new one forms in similar location within the 12-minute time step, which is therefore tracked as a continuous storm? This seems like a large source of uncertainty that isn’t discussed.
- Splits/mergers. What happens if a storm splits into two storms or merges with a different storm? Could these events be present in the dataset? This is related to concerns about the coarse time step for tracking, and how confident the authors can be about the determination of lifecycle stages.
Using 2 km reflectivity. The authors use 2 km reflectivity for tracking and defining feature centers (L127). Why do they use this altitude, as opposed to a column maximum reflectivity that may be more accurately located with the updraft? Have they tried using different altitudes? For storms that may not be precipitating heavily, could the 2 km reflectivity with a 30 dBZ minimum threshold miss some of the beginning and ending stages of a storm lifetime? Similarly, for sheared storms, the 2 km location may not accurately describe the updraft location?
p(w) seems to play an important role in the calculation of vertical mass flux and transport but is placed in the supplement without much discussion of how it is calculated and what it means. The authors do include a brief explanation of p(w), but it was not clear (L460-462). Can the authors spend more time describing this, given its importance in their calculations of mass flux?
- L448-450: The authors state that 2 min periods of RWP observations are selected, and this is based on an assumed median updraft width of 1 km and average propagation speed. I do not understand how these variables are related to the 2 min period of RWP observations that are chosen, and found this sentence confusing? Can the authors make this clearer?
- L452: The authors state that the average updraft/downdraft are weighted by the probability of sampling an updraft/downdraft during this time period (p(w)). However, the authors do not provide a clear description of how p(w) is calculated and place the figure of p(w) in the appendix (Figure S3). Can the authors provide a clearer description of how p(w) is calculated, and place this figure in the main manuscript, given its importance in calculating mass flux?
Additional Methodology Clarifications:
L164: How exactly are isolated convection (“ISO”) days determined? The authors point to Giangrande et al., 2023, but should provide the description here, given that this is defining the dataset used in this study.
The authors state that isolated DCCs observed on “ISO” days are used for subsequent analysis. How are these isolated DCCs chosen? I am especially curious given that some 500 km2 storms seem to make it into their dataset (i.e., L274).
It seems like many of the tracked storms last 3 or 4 time steps, but how can the authors classify these cells into their 5 lifecycle bins?
How big is the radar domain and is this constrained by using 2 km CAPPIs, given that the radar beams will increase in altitude away from the radar?
Can storms that initiate outside the radar domain move into the domain? If so, how would the authors know its lifecycle stage?
“Average Z:” Many times in the manuscript the authors talk about average Z, and sometimes I was confused whether they are referring to the 2 km CAPPI Z used in the tracking, Z from the SIPAM radar at different altitudes, or the Z from the RWP observations. Can the authors make this clearer throughout the manuscript?
L234: Does “at least 10 consecutive cloud echoes” mean consecutive in time or height?
L366: When the authors use the term “Mature” DCCs (here and elsewhere in the manuscript), are the authors referring to data in lifetime bin 3 or data in lifetime bins 2-4, since bins 2 and 4 are defined as “early mature” and “late mature?”
L556: The authors state that “time 0 represents the timestep when the indicated lifecycle stage of the DCC was identified using tobac.” I am having a hard time interpreting what this means. What if there are more than one radar timesteps for a given lifecycle stage? In that case, which time is used?
L556: Is the equivalent potential temperature from MAO used even when cells are >20km away from MAO? If so, can what is happening at MAO be accurately related to the storm dynamics from the respective storm?
L609: What Tb measurement is being used in this analysis? Is it the Tb measurement at MAO, at the feature center, or some average over the entire feature?
Figure 1 (L762): Is this data for all tracked cells, or only the cells that go into the analyses presented in the remaining figures?
Figure 2 (L768): Are “other DCC days” the same as ACE days as described in the text? If so, can the authors use the same terminology?
Figure 2: (L770): Why is there a propagation speed threshold here of 0.5 m/s? I don’t believe this was mentioned in the text?
Figure 3: For this (and all) boxplot figures, what is being shown? For example, do the bars that extend out represent the max/min values or some other value? A clearer description of that is being shown in the boxplots is necessary.
Figure 16: Is this showing the mean or median equivalent potential temperature?
L826: Is there a typo here? Should Figure 8 be Figure 13?
There is an inconsistency of Figure S3 caption to what is stated in the figure image (within 1 min versus within 2 mins).
Other Suggestions:
L69: The NASA Investigation of Convective Updrafts Mission (e.g., van den Heever, 2021; Stephens et al., 2020 Prasanth et al., 2023) is planning to do this, so it would probably be good to mention this here, since this research has implications for this mission as well.
Many of the results involve comparing results between the different seasons, which show up in different figures and makes comparing them quite difficult. Can the authors combine figures effectively, so that differences between results can be more clearly assessed? This is particularly the case for Figures 12-14 and 8-10.
In terms of assessing whether their smaller sample had sampling biases, what is the reason for using the nearest neighbor distance as a metric, as opposed to metrics (i.e., reflectivities, ETHs) that may be more telling of the types/intensities of storms being sampled? It seems like it would be important to understand if their sample represents both weak and strong storms, compared to the possible population of storms?
L457: For calculating mass flux, the authors assume the storm area is constant as a function of height, based on the area of the storm at 2 km AGL? It seems like the authors potentially have the data to validate this assumption (i.e., SIPAM radar data), or can find supporting literature for this assumption?
This manuscript is already quite lengthy, and while all the analyses are interesting, perhaps some areas (i.e., S3.3) are less relevant to the authors focus on updrafts and can be removed. This may allow them to better describe their other analyses within a more reasonable space.
References:
van den Heever, S. (2021). NASA selects new mission to study storms, impacts on climate models. NASA Earth. Retrieved from https://www.nasa.gov/press-release/nasa-selects-new-mission-to-study-storms-impacts-on-climate-models
Stephens, G., van den Heever, S. C., Haddad, Z. S., Posselt, D. J., Storer, R. L., Grant, L. D., et al. (2020). A distributed small satellite approach for measuring convective transports in the Earth's atmosphere. IEEE Transactions on Geoscience and Remote Sensing, 58(1), 4–13. https://doi.org/10.1109/TGRS.2019.2918090
Prasanth, S., Haddad, Z. S., Sawaya, R. C., Sy, O. O., van den Heever, M., Narayana Rao, T., & Hristova-Veleva, S. (2023). Quantifying the vertical transport in convective storms using time sequences of radar reflectivity observations. Journal of Geophysical Research: Atmospheres, 128, e2022JD037701. https://doi.org/10.1029/2022JD037701.
Citation: https://doi.org/10.5194/egusphere-2023-2410-RC2 - Splits/mergers. What happens if a storm splits into two storms or merges with a different storm? Could these events be present in the dataset? This is related to concerns about the coarse time step for tracking, and how confident the authors can be about the determination of lifecycle stages.
-
AC1: 'Author responses to referee comments', Siddhant Gupta, 10 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2410/egusphere-2023-2410-AC1-supplement.pdf
Peer review completion
Post-review adjustments
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 318 | 116 | 21 | 455 | 32 | 13 | 15 |
- HTML: 318
- PDF: 116
- XML: 21
- Total: 455
- Supplement: 32
- BibTeX: 13
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 243 | 54 |
| China | 2 | 30 | 6 |
| Brazil | 3 | 25 | 5 |
| Germany | 4 | 25 | 5 |
| France | 5 | 22 | 4 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 243
Scott E. Giangrande
Thiago S. Biscaro
Michael P. Jensen
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(2310 KB) - Metadata XML
-
Supplement
(429 KB) - BibTeX
- EndNote
- Final revised paper