Sub-Daily Rainfall Intensity Extremes: Evaluating Suitable Indices at Australian Arid and Wet Tropical Observing Sites

Round 1

Reviewer 1 Report

L45-48: So, the SDII is the mean daily  precipitation of wet days.

L48: “Locations with different degrees of rainfall intermittency may exhibit different intensities when raining, even if their SDII values are the same.” Not only. Different type of intermittency is one case. The other is simply different distributions for wet-day precipitation that have the same mean but other characteristics differ, e.g, different variance and skewness.

L124: use “:” after include.

L79-90: We tend to forget than when we use extreme values distribution (GEV) to describe annual maxima series or the GP distribution of POT time series that these distributions are limiting laws. So, in cases where the parent distribution is e.g., stretched exponential or has a lognormal tail the is essentially no convergence. I think this point should be mentioned.

L209-215: This has to do with the sample variability which causes variation in all statistics. The Q95 or any Q from a smaller sample has larger variability compared with one estimated by a larger sample. That’s all.

L220-223: For daily tails see e.g., the study of Serinaldi and Kilsby (2014) mentioning the bias and for hourly precipitation Papalexiou et al. (2018) showing the bias in identifying correctly the tail and the effect of the quantile choice.

Section 2.2. Please provide a histogram of the nonzero values for both stations and all durations studied. It’ll be nice to see the general shape of the probability density function that might also reveal issues in the data.

L254: this is the average of wet days or the total average including zeros. If it is the second it’s better to provide the mean of nonzero values of every month and the probability dry in another graph or at the same as a value e.g., above each bar. The strong seasonality is typical due to the seasonality of the probability dry and not of intensity.

Section 3.2. Could the author provide some physically based reasoning on this diurnal cycles? Obviously this related to regional weather patterns.

L300: “The shortest ADs yield 300 the least loss of intensity information, and 5 min AD data are explored here for that reason.” Not sure about that. It depends. If the instrument can not catch the intensity variations and then most values at 1 sec will be the same which does not offer a clear picture of the process at this scale (that’s why disdrometer are preferred). That’s why asked the author to provide histograms of the values.

Section 3.3: Again, please clarify if these values refer to the unconditional or conditional rainfall (wet days).

Fig 6. Does the author include the zero here or the smaller values is observed in some cases up to 80%. These graphs are not informative. It is much better to show the Exceedance probability vs. intensity in log-log scale. It might reveal indications of power laws or sub exponential tails as expected.

Summarizing, this is straight-forward analysis and it could be useful mainly because it uses fine scale resolution data. However, the author misses the opportunity to perform a more detailed analysis and focus on very basic methods. For example, I’d be really interested to see some investigation on the distributions that can be used and on the autocorrelation structure. In my opinion the paper needs major revisions before it is considered for publication. The author also must make a clear distinction between zero and nonzero values. Precipitation is a process with mixed-type marginal distribution (please see the afore mentioned paper). Also, the title is misleading. Which exactly are the indices that have been evaluated? There isn’t any kind of evaluation and comparison of typically used indices. The author has to better clarify the scopes of the paper. I would suggest major revisions.

Refs.

Papalexiou, S.M., AghaKouchak, A., Foufoula‐Georgiou, E., 2018. A Diagnostic Framework for Understanding Climatology of Tails of Hourly Precipitation Extremes in the United States. Water Resources Research. https://doi.org/10.1029/2018WR022732

Serinaldi, F., Kilsby, C.G., 2014. Rainfall extremes: Toward reconciliation after the battle of distributions. Water Resour. Res. 50, 336–352. https://doi.org/10.1002/2013WR014211

Author Response

Please see the attachment

Reviewer 2 Report

Minor Points:

L.18: Change to “, an index (R_Q95) …”

L.20: Either write out ‘accumulation duration in Abstract or change to ‘accumulation duration (AD)’ in Abstract.

In MANY places in the paper I would change to ‘5-min’ instead of ‘5 min’.  There are numerous places in the paper where a number precedes 5 min, which becomes causes some confusion.  I prefer 5-min in most cases.

L.42: “a correspondingly diversity of data …“ is a very awkward phrasing.

L.203: This should be “the Revised Universal Soil Loss Equation (RUSLE)’ as stated in reference [74].

Figure 1: ‘intensities’ is misspelled in the title for the lower left figure.

Figures 1/2: The Y-axes for Figures 1 and 2 should match for easier between-site comparisons. Upper left: OK; Upper right: both should range from 0-25; Lower left: Both should range from 0-16; Lower right: Both should range from 0-10.    The Y-axis labels should also match – either mean intensity (mm h-1) or intensity (mm h-1). [intensity to match Figures 3 and 4]

 Figures 3/4: The Y-axes for Figures 3 and 4 should match for easier between-site comparisons. Upper left: both should range from 0-40; Upper right: both should range from OK; Lower left: OK; Lower right: OK.   

L.291: Change to ‘… intensity is only about …’

L.297: Does the author mean Figure 5?

L.306: Change to ‘4%’ for consistency.

Table 1: Bold either all or none of the table headings.

Table 1: Last column of data: Give all values to only one decimal place.

Table 2: Bold either all or none of the table headings.

L.404/405: Change to either ‘relative infrequency of observations’ or ‘relatively infrequent observations’

References: Journal article titles are given in lower case for all but the first word in the title except for references: [18, 44, 45, 54, 59 and 70].  Be consistent.

References: All journal names are written fully except for references: [26, 29, 41, 57, 59, and 65]. Be consistent.

References: Spell out Catena correctly in reference [73].

675: remove comma before journal title.

 

Comments to Author:

The paper is very well-written and the literature is fully reported upon.  If anything, the Introduction is a bit too discursive.  I would try to reduce its length a bit by focusing more directly on ideas and published research most directly tied to study objectives and goals.  A somewhat more focused Introduction would help the reader zero in on the study goals and objectives.  Even the references illustrate this point.  The vast majority of the references are cited only a single time, when multiple times for a single reference is more common when the work of previous research pertaining to one’s research goals and objectives are emphasized.  This is a minor point, but the Introduction could be improved upon by greater focus.  Similarly, the Discussion is a bit too far-ranging in places. 

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

Review for:

Sub-Daily Rainfall Intensity Extremes: Evaluating Suitable Indices at Australian Arid and Wet Tropical Observing Sites

This paper investigates the mean rainfall intensity observed within different temporal windows at two observation sites in Australia and the contribution of intensity to annual total rainfall. The author first provides a comprehensive and useful overview of rainfall intensity measurements in the context of extremes and climate change. Using high temporal resolution (1s) rain gauge measurements, the seasonal and diurnal cycles of rainfall at different temporal aggregation periods are investigated followed by a statistical examination of the top percentiles of the distribution of rainfall intensities compared to the 95^thpercentile of the annual total. The paper highlights that examining the contribution of rainfall intensities to R_Q95, the 95^thpercentile of the total annual rainfall, is a more robust approach to examine extremes compared to using the upper percentiles of rainfall intensity. Understanding the temporal characteristics of rainfall intensity and timing in relation to total rainfall, as stressed in the paper, is an important topic which needs to be considered by studies of rainfall variability and change. The manuscript is generally well written however in places it is difficult to clearly understand results and points made and some of the analysis lacks depth which would be desirable to improve the manuscript before it is published.

 

 

General comments:

The statistical analysis applied to the data in Section 3.1 and 3.2 is basic – only the meanrain-rate is presented at each AD, thus the seasonal and diurnal cycles don’t really show anything new. Other characteristics and properties of this data could be analysed to add more insight and weight to the findings – for example, basic stats such as variance and/or standard deviation of each hour in the diurnal cycle (e.g. is the afternoon peak in the diurnal cycle a more variable feature) and examination of rainfall intensity given rainfall events >0 may provide a clearer picture of the rainfall distribution in each AD.

 

The author mentions that intermittency/duration of rainy events is important, but does not investigate the dependence of these characteristics on the diurnal and seasonal cycles and their impact and contribution to Q95 and R_Q95. Since the data analysed has such a rare high temporal resolution it would be insightful to show how the intermittency/event duration contributes to the observed seasonal and diurnal cycles, Q95 and R_Q95. For example, the frequency and duration of heavy rain events compared to longer-lived moderate rainfall events is likely to vary between the sites and may change seasonally and diurnally with variability in the type of precipitation systems (e.g. frontal precipitation in winter, convective precip in summer). What time of day do rainfall events contribute most to the annual total, for example? This information will also help the physical interpretation of the percentile analysis in Sect 3.3.

 

It is argued that examining rainfall intensity contributions to the 95^thpercentile of the total annual rainfall (index R_Q95), is a more robust way to investigate rainfall extremes than applying a quintile threshold directly to the intensity distribution. It was difficult to clearly understand this point from reading the manuscript, given the way the methodology and results were described. In particular, the definition of this index and the methodology applied to investigate intensity contributions must be more clearly defined and referred to in a much more consistent manner throughout the manuscript– see some specific comments below.

 

Many times numerical results are quoted in the text but are not presented in a figure which makes the reading arduous in places, including the discussion section. Some additional figures would make the results much easier to visualise (e.g. Lines 302-307 could be presented by a histogram).

 

 

Specific comments

Abstract: Define metrics in the abstract instead of stating their acronyms, i.e., SDII (line 14), ‘Q95’ and ‘Q99’ (Line 15, I know they are the 95^thand 99^thpercentile – better to explicitly state this), AR_Q95 (line 18) and AD (Line 20) are not defined.

 

Line 197: define ETCCDI, RX1day, RX5day, R10mm and R20mm

 

Line 238: total rainfall is stated at the two sites – does this represent the total rainfall during the 9.5 and 3.3 years of the record at each site? Please clarify

 

Line 263: assuming that the 790 rain days corresponds to the total number of rain days that were observed at MM during the 9.5 years of observations – how do you define a rain day, > 0mm? Please clarify

 

Lines 270-272: the reason that the amplitude declines with increasing ADs is just because the data are averaged over a longer time period? I also don’t see a strong damping of the seasonal variability with increasing ADs although, of course, the magnitude of the mean rainfall decreases with increasing AD.

 

Section 3.2: It’s dangerous to group all months/seasons into one mean diurnal cycle figure, as the diurnal cycle may change seasonally. I suggest investigating this further.

 

Line 301: why not analyse the 1 second data? That would provide no loss of intensity information.

 

Line 302: Assuming ‘with rain’ means any values > 0 mm h? Please clarify in the text.

 

Line 306: “Only 1364 6 min ADs have” – should this not be 5 minADs?

 

Lines 302-310: it would be much clearer to present the 5min AD rainfall distribution during ‘rain’ using histograms and box-whisker plots.

 

Lines 325-326: It’s not clear how the data is aggregated into intensity classes – For example for all 5 min AD intensities between say 1-2 mm h, are you essentially totalling the rainfall amount during this AD intensity observed during the 9.5 years of observations at MM?

 

Lines 336: Could you show the contribution of different rainfall intensities to ‘hours of rain’ and fraction of total at each observing site as a figure?

 

Line 341: Assuming you mean ‘the intensity above which 5% of the annual total rainfall’? Please clarify in the text

 

Lines 344-358 and Table 2: When referring to Table 2, can you also refer readers to Figure 6, as Figure 6 shows some of the values presented in Table 2.

 

Line 346: When referring to Table 2 on line 346 it’s stated that “R_Q95 and Q95 of the 5 min AD data reveal quite different characterisations of extreme intensities”. However I do not see any reference to R_Q95 and Q95 made in Table 2.

 

Lines 371-374: It’s stated that knowledge of the diurnal and seasonal cycle of rainfall intensity is important for understanding Q95, but the data in Sect. 3.3 is pooled to look at the different quintiles during all months and all hours of the day. Is it possible to show the impact of the diurnal and seasonal cycle on understanding variations in the upper quintiles of the distribution in Section 3.3?

 

Lines 379-381: It is stated here that “the Q95 of 5 min AD rainfalls is ~ 15 mm h-1. This does not mean that 5% of the rain exceeds this intensity. Rather, Q95 of the 5 min AD data means that 5% of the AD values exceed ~ 15 mm h-1.” I’m confused exactly how the definitions of intensity and AD values in the two sentences are actually different. More clarification is needed.

 

Lines 382-383:  Given that up to a quarter of the total rainfall exceeds the Q95 intensity, the suitability of this value as a measure of 'extreme' can be questioned.Needs clarification - a quarter of total rainfall during whole observing period/what time period? A quarter of the total rainfall observed at a certain intensity?

 

Lines 401-402: To compute R_Q95you not only need knowledge of the annual total but also knowledge of the intensity distribution and therefore to compute R_Q95 you need high-temporal resolution rain-rate data, which is quite rare. I therefore don’t think it is very straightforward to compute as is stated in the text. It has to be computed from data such as 5 min AD rainfall observations.

 

Line 424-425: R_Q95is defined clearly here as the 95^thpercentile of the total annual rainfall amount. However, it was not clearly introduced previously in the data processing section on Line 249. I am confused of the definition of R_Q95and since this index is integral to the rest of the results, its definition needs better clarification much earlier on in the manuscript. It wasn’t until here that I began to understand the point of the manuscript.

 

Line 469: Do you mean diurnal variability? I don’t see an analysis of the annual variability in AD intensities anywhere.

 

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The author has improved the manuscript. I would suggest publication.

Author Response

There were no comments from this reviewer

Reviewer 2 Report

1) 'Intesities' is still misspelled in LL Figure 1 graphic.  Not a game changer.

2) L.361: Should 'mean' be replaced by 'meant'?

All of the suggested comments have been addressed.

Author Response

There were no comments from this reviewer

Reviewer 3 Report

Many of my main 'general comments' have either been addressed in other papers by the author or will be conducted in a future publication. The reviewer has addressed most of my other comments adequately, providing clarification where was requested. While I’m generally happy with most changes, I have several minor comments that should be addressed before publication:

 

Lines 262-263: The statement ‘FG …, lacking any marked seasonality of rainfall ‘ is not true. The results in Section 3.1, Figure 2 show clear seasonality at FG - wetter in October to March, drier during other months. This statement should be corrected.

 

Section 3.3 The 5-min and unaggregated ITT intensity data: full period of record: The result descriptions in this section are still difficult to follow and clearly understand why some points are being made. I suggest more effort is made to coherently and explicitly describe the rational of looking at the distributions, the main points from each result and how the results link with one another. Here are some examples/suggestions:

Lines 354-363: What point is being made by stating some of these statistics? E.g. is it just to highlight that there aren’t many ‘heavy’ rain events? Or is it just to give an overall picture? Please explicitly state in text.

Lines 365-369: It would be clearer to explicitly state at the start of this paragraph that you are now examining the impact of temporal aggregation on the intensity information. I don’t really understand that this is what your aim is until after the results are described. It would also be helpful to explicitly state that ‘This is examined in Figure 6 which compares the intensity distribution from 5min AD data with that at 60-min.’

Line 394-395: what is ‘useful information’ about the intensity distribution? If this was explicitly defined it would make it clearer why the aggregation of rainfall data to intensity classes is being done and is important to look at.

 

Line 359: ‘their total duration is equivalent of ~4.7 days of rainfall’ – Is there a reason for providing total duration here? If so, it should be stated, otherwise it’s not clear what benefit this statement adds and it should probably be removed.

 

Lines 414-415: It’s not actually explicitly shown/mentioned in the results before this sentence how ‘The data from MM and FG suggest that the intensity above which 5% of the rainfall is delivered might provide a suitable index for characterising intense rainfall.’ If you are going to state this point you need to point the reader to plausible evidence before doing so.

 

Lines 415-416 – Table 2 should be introduced in this sentence because this table is where this ‘hypothesis’ mentioned in Lines 414-415 is actually being tested.

 

Lines 416: Some typos – should be ‘‘5% of the rainfall’ means 5% of the total in any record’ . Maybe clearer to say, ‘For the definition of R_Q95, 5% of the rainfall means 5% of the total in any record.’

 

Lines 416-418: Suggest moving this sentence to directly after the one on lines 414-415 (more clearer). Then referring to ‘This is evaluated further in Table 2’ after it.

 

Line 426: Table 2. Is it possible to add (R_Q) to the percentile of rain depth header and (Q) to the percentile of 5 -min ADs header? It would make it clearer what this table is showing when comparing these results with the statements of R_Q95 and Q95 in the text.

 

Discussion section 4: This section is long, with many paragraphs and it is difficult to keep track of the key points and coherency between points made. It might be helpful, for example, to add sub-section headings to the discussion to make it clearer what points are being made there.

 

Line 450: Please provide a reference to back up the following statement if possible, ‘This is perhaps unsurprising, given the greater convective activity during those times.’

 

Lines 462-463: Can this statement be made even more clearer saying something like: ‘This does not mean that 5% of the instantaneous rain rate exceeds this intensity. Rather, Q95 of the 5-min AD data means that 5% of the accumulated rainfall within a 5 minute period exceeds ~ 15 mm h-1.’? In response to my comment on clarifying this sentence it was stated that: The point being made is that rain intensities derived from AD data - even 5-min AD data - are not true intensities. It would also be made even more clearer if that statement (or one along the same lines) was explicitly added/mentioned here.

 

Lines 495-496: I don’t understand how ‘the use of long records may hinder the detection of secular change in rainfall intensities.’ Surely the use of long rainfall records, e.g. spanning several decades, is extremely advantageous to detect secular change, whereas a short record would be very limiting and not sufficient to obtain a credible detection of secular change.

 

Should paragraphs on Lines 513-517, 519-521, 523 -535, 537-540 be merged since they are addressing the same issue? Would make it clearer that the same point is being addressed.

 

Lines 523-535, referring to MM is quite arduous reading repeating many numerical results already described in the Results section – can this be condensed to make the discussion point clearer?

 

Lines 550-552: Similar to previous point, it may make more sense to only state percentages instead of number of events (ie. At MM, 63% of ADs with rain have the value 0.2 mm…) – would be more concise and easier to follow.

 

For Figure 6 I suggest limiting the histogram x-axis limit to 30mm/h - there are no visible data points for any bins greater than 30 mm/h so no point in showing them making the y-axis in the histograms a log-scale so that the higher rain rate the percentage contributions are clearer since the distributions are dominated by the first bin making it difficult to see the exact contributions of the larger rain rates  Fixing the figure caption so that it correctly describes the figure (caption currently refers to a cumulative distribution)

 

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 3

Reviewer 3 Report

Thank you for addressing my comments.