Data assimilation of radar reflectivity volumes in a LETKF scheme

Gastaldo, Thomas; Poli, Virginia; Marsigli, Chiara; Alberoni, Pier Paolo; Paccagnella, Tiziana

doi:https://doi.org/10.5194/npg-25-747-2018

Articles | Volume 25, issue 4

https://doi.org/10.5194/npg-25-747-2018

© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.

Special issue:

Numerical modeling, predictability and data assimilation in...

https://doi.org/10.5194/npg-25-747-2018

© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 25, issue 4

Research article

|

07 Nov 2018

Research article |

| 07 Nov 2018

Data assimilation of radar reflectivity volumes in a LETKF scheme

Thomas Gastaldo, Virginia Poli, Chiara Marsigli, Pier Paolo Alberoni, and Tiziana Paccagnella

Download

Final revised paper (published on 07 Nov 2018)
Preprint (discussion started on 26 Apr 2018)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Review: Data assimilation of radar reflectivity volumes in a LETKF scheme', Anonymous Referee #1, 01 Jun 2018
RC2: 'Referee report', Anonymous Referee #2, 18 Jun 2018
EC1: 'Editorial comment', Alberto Carrassi, 18 Jun 2018
- AC1: 'Authors comment', Thomas Gastaldo, 21 Jun 2018
EC2: 'Editorial comment on the 2nd version', Alberto Carrassi, 24 Jul 2018

Peer-review completion

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

AR by Thomas Gastaldo on behalf of the Authors (23 Jul 2018) Author's response Manuscript

ED: Referee Nomination & Report Request started (24 Jul 2018) by Natale Alberto Carrassi

RR by Anonymous Referee #1 (07 Aug 2018)

Suggestions for revision or reasons for rejection

The authors did a great effort to address my comments. I found that the paper has improved, however some important aspects are still unclear.

Major points

-One of the main objectives of this paper is to assess the impact of assimilation of reflectivity data in a convective scale data assimilation system. However all the experiments presented in this paper assimilate reflectivity data either using nudging or using direct reflectivity assimilation. An experiment in which only conventional observations are assimilated would be very useful to properly assess the impact of assimilating reflectivity. If this experiment is performed then it would be possible to evaluate if nudging and direct reflectivity assimilation provides equivalent results or if we are in a case in which the assimilation of reflectivity do not produce a significant impact upon the forecast skill. In the later case it would also interesting to provide some hypothesis about the causes for such behavior (for example if the complex terrain over the study region is significantly enhancing rain predictability reducing the impact of initial conditions at the mesoscale over the short range forecast skill, or perhaps the impact of reflectivity upon forecast skill is limited to lead times under 3 hours and is not clearly seen beyond that time).
The sensitivity experiments will also became more meaningful if the authors can demonstrate that including reflectivity data is producing a positive impact upon forecast skill in this case.

Minor points

Page 3, line 25 ((LHN - Remove the extra parenthesis.

Page 7, Why 3D data is available only for radars over Northern Italy? Is because how data is archived?

Page 8, line 10 observed precipitation

Page 8, line 23 every 3 hours

Page 8, line 26 and

Page 10, line 7 February 5 line 8 February 6.

Page 10, Line 7 On February 5 / new precipitating systems

Page 10, Line 15 Spatially averaged forecasted precipitation is compared against ...

Page 10, Line 29 precipitating systems instead of precipitation nuclei

Page 12, Line 18 reflectivity data is

Page 12, Line 20 Spatially averaged

Page 18, lines 20-21 The experiment assimilating only conventional data is not described in this work. All the experiments assimilate reflectivity either using nudging or direct assimilation. This is explained in the following sentences, but it would be better to state the conclusion in a clearer way from the beginning instead of making this clarification later.

Page 19, For the case study considered in this work... Since the experiments provided in this work shows little sensitivity to the different strategies used to assimilate reflectivity, the lack of sensitivity reported here might be a consequence of the general lack of sensitivity of these forecast to the assimilation of reflectivity.

To further reduce the number of figures, Figure 1 and 2 can be combined if the color scale is changed accordingly.

I still think that including some examples of the precipitation or reflectivity forecast in the paper would be useful. Statistical metrics are powerful and useful, but having such examples also gives a good subjective idea of how good is the fit between the forecast and the observations. This is particularly useful in the context of novel high resolution convective scale data assimilation systems. If the results changes from case to case, then two examples can be provided.

Hide

RR by Anonymous Referee #2 (19 Aug 2018)

Suggestions for revision or reasons for rejection

Based on the evidence presented in this study, it is difficult to draw a solid conclusion about positive impact from assimilating radar reflectivity. I think it is mainly because the authors are focusing on the one-day QPF, which is claimed to be the operational interest. However, it is well recognized by the literature that the impact of radar data is identified for (very) short-term forecast. I would like to suggest the authors that since this work is the new component for the KENDA system, it is essential to identify and justify the impact of the radar data, even though the impact may only last few hours. Also, I’d think that it is important to explain why additionally assimilating radar data doesn’t seem to gain more benefit than applying LHN. Is this because the information is double used or redundant? Based on these concerns, I suggest major revision for this manuscript. My comments are as follows:

 Based on the verification merits used in this study, it is difficult to justify the impact of radar data. Have the authors consider adopting other verification merits such as POD, FAR, TS and BIAS to identify whether there is clearer signal about the data impact? Or, illustrate the results with a case study (from the 4-day assimilation)?
o Discussion about Fig. 4 is vague. Line13, Page11: “Overall, the correspondence of rad60 …. to observations is equal or better than that of conv60 (red)”. Also, by eye, rad60_nolhn is slightly better than rad60. rad60_Bm tends to have the least rainfall among all the rad60-related experiments shown on Fig. 4 but the authors claim that the impact of the additive inflation cannot be judged. To quantify such statement. I suggest to summarize Fig. 4 in terms of RMS error.
o As mentioned in my general comments, I’d think that it is important to explain why additionally assimilating radar data doesn’t seem to gain benefit than applying LHN. Is this somewhat related to the fact that the SRI product uses the radar reflectivity? The authors can illustrate the result with even a case demonstration. Should we expect a large difference between radar data assimilation and LHN with heavy rainfall events?
 Although the authors focus on the impact of radar data on 1-day QPF, it is undeniable that the impact is mostly evident within 6 hours. Therefore, I suggest that the authors should address the impact of radar data on short-range forecast (< 6h).
o It is concluded that the radar data only slightly improve QPF both during the assimilation procedure and for the subsequent forecasts, compared to the assimilation of only conventional data (Line3, Page 18). This sentence should be modified for 1-day QPF.
 Why does rapid update (15-min) with a small observation actually lead to a dry condition over northern Italy (Fig. 9)? Fig. 9 should be also compared with results from rad15 with rad15_roe0.5.

Minor comments:
 Since radar reflectivity is the main focus in this work, the authors should briefly comment or summarize what are the issues/difficulties with assimilation of reflectivity volumes in the introduction.
 “Grater” is often used in the text. But, I think the authors meant “greater”.

Hide

ED: Reconsider after major revisions (further review by editor and referees) (20 Aug 2018) by Natale Alberto Carrassi

AR by Thomas Gastaldo on behalf of the Authors (01 Oct 2018) Author's response Manuscript

ED: Publish subject to technical corrections (22 Oct 2018) by Natale Alberto Carrassi

Cari Autori,

I found that the last version of your manuscript greatly improved and that the main issues arisen by the Reviewers have been satisfactorily addressed.

I am thus glad to accept the manuscript, but have identified some additional very minor points for correction and/or suggestions, that I encourage the authors to address before sending the final version:
- Page 2, Line 10: "...best estimate ..." ---> "....best possible estimate....".
- Page 2, Line 12: "... around the world." A reference, perhaps a review, would suit here.
- Page 2, Line 28-29: "...and..." is used twice in the sentence.
- Page 3, Line 19: "... and observation operator errors." A reference would suit here, among which I suggest Janjić, T., et al. "On the representation error in data assimilation." Quarterly Journal of the Royal Meteorological Society (2017)
- Page 5, Eq. 1: Please make clearer how the relaxation toward the background ensemble is accomplished. This cannot be achieved by looking at Eq. 1, that does not include any term related to the forecast ensemble.
- Page 5, Line 3. It is not clear whether additive inflation is not employed in the current manuscript but employed in KENDA in general, or it not employed at all. Please clarify.
- Page 6, Line 1 (on top of the page): It is not obvious why the back and forth interpolation on a coarsened grid should not affect negatively the analysis. I suggest the authors to smooth a bit this statement or to make fully clear what they meant. Also, what do you exactly mean by "analysis weights" versus "analysis increments", are the former the innovations? If so, I expect that their computation on a coarser grid will smooth out smaller scale feature of the data.
- Page 6, Line 3 (toward the bottom of the page): " ... in the most part of assimilation schemes ..." --> "... in many data assimilation schemes..."
- Page 6, Line 4: '' ...is chosen." A reference would suit here.
- Page 8, Line 33: " ... , experiments on ... " --> " ..., sensitivity experiments on ..."
- Page 9, Line 16: " ... each 3 hours ..." --> "... every 3 hours ...".
- Page 10, Line 15: "... station location ..." --> " ...station locations ...".
- Page 11, Line 28 and 30: Two consecutive sentences starting with "In fact ...".
- Page 11, Line 31: "... representative for ..." --> "... representative of ...".
- Page 25, Line 36: "... making the ensemble too small ..." --> "... potentially making the ensemble spread too small ...".

Best Regards,

Alberto

Hide

AR by Thomas Gastaldo on behalf of the Authors (24 Oct 2018) Manuscript

Short summary

Accuracy of numerical weather prediction forecasts is strongly related to the quality of initial conditions employed. To improve them, it seems advantageous to use radar reflectivity observations because of their high spatial and temporal resolution. This is tested in a high-resolution model whose domain covers Italy. Results show that the employment of reflectivity observations improves precipitation forecast accuracy, but the positive impact is lost after a few hours of forecast.