Improving the precipitation accumulation analysis using lightning measurements and different integration periods

Gregow, Erik; Pessi, Antti; Mäkelä, Antti; Saltikoff, Elena

doi:https://doi.org/10.5194/hess-21-267-2017

Articles | Volume 21, issue 1

https://doi.org/10.5194/hess-21-267-2017

© Author(s) 2017. This work is distributed under
the Creative Commons Attribution 3.0 License.

Special issue:

Rainfall and urban hydrology

https://doi.org/10.5194/hess-21-267-2017

© Author(s) 2017. This work is distributed under
the Creative Commons Attribution 3.0 License.

Articles | Volume 21, issue 1

Research article

|

11 Jan 2017

Research article |

| 11 Jan 2017

Improving the precipitation accumulation analysis using lightning measurements and different integration periods

Erik Gregow, Antti Pessi, Antti Mäkelä, and Elena Saltikoff

Download

Final revised paper (published on 11 Jan 2017)
Preprint (discussion started on 17 Mar 2016)

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'hess-2016-113 - Improving the precipitation accumulation analysis using radar-, gauge- and lightning measurements - by Gregow et al.', Anonymous Referee #1, 17 Apr 2016
- AC1: 'Reviewer 1 - Authors answers', Erik Gregow, 26 May 2016
RC2: 'Review of "Improving the precipitation accumulation analysis using radar-, gauge- and lightning measurements"', Hidde Leijnse, 17 Apr 2016
- AC2: 'Reviewer 2 - Authors answers', Erik Gregow, 26 May 2016
RC3: 'Review of hess-2016-113 by Gregow et al.', Anonymous Referee #3, 17 Apr 2016
- AC3: 'Reviewer 3 - Authors answers', Erik Gregow, 26 May 2016
RC4: 'Review - Improving the precipitation accumulation analysis using radar-, gauge- and lightning measurements', Susana Ochoa Rodriguez, 18 Apr 2016
- AC4: 'Reviewer 4 - Authors answers', Erik Gregow, 26 May 2016

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

ED: Reconsider after major revisions (23 Jun 2016) by Peter Molnar

AR by Erik Gregow on behalf of the Authors (23 Jun 2016) Author's response Manuscript

ED: Referee Nomination & Report Request started (23 Jun 2016) by Peter Molnar

RR by Anonymous Referee #3 (28 Jun 2016)

Suggestions for revision or reasons for rejection

I was reviewer # 3 for the initial submission of the paper.

I appreciate the improvement of the readability of this paper on a relevant topic. However I still think that the analysis in its current state does not enable to conclude on the interest or not of the use of LDA. The authors basically chose not to address all the serious concerns raised by me (and the other reviewers) with regards to the fact that the methodology suggested does not enable to reach a conclusion (either positive or negative that is not the point) with regards to the use of LDA. I understand that not everything is possible with operational data, but still think that more should be done before publication. See comments below with some suggestions of potential analysis with available (from my understanding) data. The innovative aspects of the paper should also be stressed more clearly.

Detailed comments:

1) Introduction
- Some improvement, but not really a state of the art on assimilation of LDA for QPE. This would also help in understanding the novelties brought by the paper.
- l. 108-111 : this is a bit disappointing for a scientific paper... although the argument is partly valid, I still believe that more analysis can be done with the data available (see suggestions below).

2) Observation and instrumentation
Please add a subsection on the event / period (summer / 25 days / 4 days) selection to clearly present them (moving the paragraph from the method section).

3) Methods
- l.181 : “uses statistical methods”, as already pointed out in the previous review, this formulation is not appropriate for a scientific paper, please be more specific.
- l. 227-228 : please clarify, how the quantiles are computed with regards to altitude (computed for each altitude). How is done the variable Quartile rad-Lig profiles. Since it is mentioned as a new method introduced by the authors, it should be more discussed, also in the results section.
- l. 232-234 and figure 3.b : there seems to be somehow two regimes (at least one point -the one with the smallest number of strokes) seems isolated.
- l. 238-241 : I still find it weird to chose simply maximum and was not convince by the answers of the authors. It seems that this rule also applies when you are close from the radar where radar data is much more reliable. Why not testing other merging schemes that would for example take into account the distance to the radar or the attenuation. I guess that you have access to both radar mosaic and lightning maps separately (given that it is what is plotted in Fig. 8) ? This could add some novelties to the paper.
- l. 244 : please remind the unit of Z (mm6 m-3)
- 3.3.1 : as already pointed out, it is weird to have a 3.3.1 and no 3.3.2...
- 3.4 : what is the interest of presenting results for periods without lightning ? l.312 : remind here why only 7 independent stations are used (as explained in the answer to reviewers)

4) Results
- l. 335-345 : it is frustrating that no lightning occurred over the 7 independent stations during the 25 selected events... Given that it seems some lightning was recorded for some events (since an improvement is noticed), why not selecting a sub-set of event sfor which lightning was observed over the independent stations and discuss the results for them.
- l. 359-360 : comments on Fig 6 : “The 3'rd Quartile gives the largest overestimate, over the whole accumulation scale”, it seems rather obvious no ? An explanation on the variable Quartile profile is missing.
- Section 4.2 : please give more explanations and and also comments on the curves of Fig. 7.

5) Discussion and conclusion (by the way add a number for this title section)
- l.374 : “at least leave them as unaffected as possible”, somehow weird formulation as a goal...
- l. 375-378 : this is indeed an interesting point but not demonstrated.... it could be done with the available data (i.e. not re-running past events) by considering scores for stations located only far (testing different distances) from the radars (ex. : stations with distance to the radar equal to 0-50 km, 50-100km, 100-150km...).
- l. 404-406 : “3rd quartile gives higher impact....”, I may be confused, but it is somehow surprising the median does not work better on average (was the estimation correct). Please discuss this point more

Hide

RR by Anonymous Referee #1 (01 Jul 2016)

RR by Hidde Leijnse (08 Jul 2016)

ED: Reconsider after major revisions (18 Jul 2016) by Peter Molnar

AR by Erik Gregow on behalf of the Authors (05 Sep 2016) Manuscript

ED: Referee Nomination & Report Request started (03 Oct 2016) by Peter Molnar

RR by Anonymous Referee #3 (20 Oct 2016)

RR by Anonymous Referee #1 (30 Oct 2016)

Suggestions for revision or reasons for rejection

I am the anonymous referee #1 of the previous revision and I’ll mainly focus on the response to my previous revision. The authors now put extensive efforts in improving the analysis and providing more interesting results, nevertheless, in my opinion, the manuscript still needs to be improved before being published, mainly because the results are not clearly presented. See my major comments below:

1. First, I’d like to answer the last of the author’s response: “We think there is value of publishing/sharing this work and by that, reach other workers in the same field, doing similar developments and possible help others to start from an advanced development point, using radar-lightning information. One of the main result is that lightning is useful in areas of no radar data! If there is no radar, LAPS accumulation process has no way to continue, it need the reflectivity field. Therefore the lightning information can be crucial, it is there to complement the radar. There are areas (and even countries) which does not have radars, in souther latitudes the lightning is more frequent and here the LAPS-LDA model could be of even more importance, than in for example Finland”. This is very interesting to know and motivates part of the study, but the reader does not necessarily know it. I highly recommend to include it in the introduction (motivation for this study), to highlight it in the proper sections of the discussion, and to recall it in the conclusions (see also point 2). For example, lines 205-211 could be moved to the introduction for this purpose.

2. The “verification periods” are not presented and used clearly. From what I understood from section 2.4, the authors use 5 different verification periods: (a) summer 2015; (b) summer 2016; (c) 25 days during summer 2015; (d) 25 days with “scaled” data during summer 2016 and (e) 4 days with high lightning occurrence during summer 2014. I suggest to organize the periods more clearly, for example numbering them (like I did here.
- How is the data in (d) scaled? “Manually” is not enough to understand, did the authors select a maximum distance from the lightning? What about time?
- When is (d) used? In line 193 the authors introduce (d): “hereafter called scaled dataset”, but the word “scaled” is then used again only once, in line 395. I believe that (d) is used also in Table 1, am I right?
- In the presentation of the results (section 4.1) the use of these periods is not clear: what periods are presented in Table 1? The caption says that the results are presented for (c) and (d) but the text is not clear on this.
- Lines 362-367 say that the verification on (a) and (b) gives no significant impact, so the results are not presented. SO, why are these verification periods included? I imagine the reason is what discussed in the point 1 above, so I highly recommend to discuss this here and to present Fig. 4 with this purpose.
- Fig 4: is the figure from the verification period (a)? or from (c)? Is this the figure the authors refer to in lines 362-367? Perhaps some statistic value could provide the same message more clearly.
- What verification periods are used for the integration period length analysis (sec 4.2, Fig. 7)? This information is not provided to the reader.

3. Results are presented not clearly. They need to be reorganized and their presentation needs to be improved.
- Line 369: what does “shows a biased result”?
- Why are some results presented via table and statistical metrics and other via figures? This is not clear and does not allow to compare between them. Results from the figures are qualitative, since no information on the regression lines, on the bias or other statistic is provided.
- Line 385: “most of the scores are improved”, the improvements observed from the tables are actually minor (the highest being RMSE for the dependent dataset of 2016 decreased by 6.3%), some words on the statistical significance of the improvements or on the idea expressed in the point 1 above could be useful.
- Line 394-396: these lines are more useful at the beginning of this paragraph, in order to introduce the results.
- Table 1 presents different verification periods (period (c) and (d) of point 2 above, if I understood correctly) and different schemes, over different verification datasets. This makes difficult to understand if the differences in performance are due to the different method, to the dataset or to the verification period.

Minor comments:
- Abstract: there is no need of introducing the acronyms in the abstract, if they are not used therein.
- Lines 49-54: not clear
- Line 60: the acronym is not used anymore so there is no need to introduce it
- Lines 71-75 would fit better in the methods
- Lines 105-119: this part is a bit verbose
- Line 112: I’d use “convective” rather than “extreme”
- Lines 147-153: the names of the radars are not reported in the map; I suggest either using the same symbol as the maps or more generic words
- Line 164, 181, 194: LDA, LAPS, FMI-LAPS LDA have not been introduced yet
- Line 180-182: move to section 3.2
- Line 260-262: please provide more details
- Lines 285-286: move to introduction
- Line 291: “accumulating” rather than “by summing up over the”
- Lines 308-309: gauges are not mentioned here but, if I understood correctly, represent a key aspect of the Rand-B method
- Lines 337-339: move before in the paragraph
- Line 342: before in the text the author said that no beam blockage affects the area
- Fig 5 and 6: use the same words for corresponding features in the figures

Hide

ED: Publish subject to minor revisions (further review by Editor) (16 Nov 2016) by Peter Molnar

AR by Erik Gregow on behalf of the Authors (24 Nov 2016) Manuscript

ED: Publish as is (17 Dec 2016) by Peter Molnar

AR by Erik Gregow on behalf of the Authors (19 Dec 2016)

Short summary

A new lightning data assimilation method has been implemented and validated within the Finnish Meteorological Institute – Local Analysis and Prediction System. Lightning data do improve the analysis when no radars are available, and even with radar data, lightning data have a positive impact on the results. We also investigate the usage of different time integration intervals: 1, 6, 12, 24 h and 7 days, where the 1 h integration time length gives the best results.