Radar Reflectivity Assimilation Based on Hydrometeor Control Variables and Its Impact on Short-Term Precipitation Forecasting

Round 1

Reviewer 1 Report

Title: Radar Reflectivity Assimilation Based on Hydrometeor Control Variables and Its Impact on Short-Term Precipitation Forecasting

Authors: Hong Zheng et al.

Summary:

The static background error covariance (B) structure significantly influences the quality of variational hydrometeor analysis. How to construct a B that can better initialize the hydrometeor fields and further improve the atmospheric analysis and its subsequent forecasts remains an open question. This paper is a good attempt to tackle this challenging problem.

With WRFDA, this paper attempts to tackle this problem by including hydrometeors as additional control variables and incorporating the vertical and multivariate error correlations into the static B. With the new-formulated B (H-BEC), the authors further conduct the single-obs test and real-case experiments to evaluate the impact of the new-formulated B on the analysis and forecast. 

Their key findings are:

(1) The static B that includes vertical and multivariate error correlations of hydrometeors can speed up the minimization procedure in the Var system.

(2) Real-observation experiments show that precipitation forecasts initialized from the analysis with their new B are better than those using the B without including hydrometeors. A further investigation of these improved forecasts shows that incorporating the multivariate error correlation between the conventional CVs and the additional hydrometeors plays a dominant role in the enhanced forecasts.

The findings of this study are important and valuable. Considering all these, I suggest minor revisions for this paper.

There are a few places where clarifications are needed, especially in Section “2. H-BEC in variational assimilation framework”. Please find them in the section of my major and minor comments. It should be straightforward to fix them.

Recommendation:

Minor revision

Major comments:

1.         You might want to rewrite the methodology section, Section “2. H-BEC in variational assimilation framework”. The most prominent problems are

(1)       what are your control variables (CVs) exactly? Around L100-L103, you mentioned using U, V, T, RH, PS, and mixing ratios of other 5 hydrometeors. But in your Eq (4), the last column shows that your CVs are U, unbalanced V/T/PS/RH (or actually pseudo-RH?), and unbalanced hydrometeors. Which is correct?

(2)       The matrix describing Up transform in Eq (4) seems inconsistent with the description in the context: L118 “Notably, different from Wang and Wang, the cross-correlations between hydrometeors are ignored due to the nonlinearity of hydrometeors.”: If this statement is true, then, for example, the last row in (4), reg_{10_6} to reg_{10_9} should be zero, right? Otherwise, a nonzero delta_q_g will bring back a nonzero increment to all hydrometeors. If my interpretation is incorrect, can you clarify your statement “the cross-correlations between hydrometeors are ignored”? This also causes me difficulty interpreting your single-obs test (Figure 3). I assume that your observation operator omits q_c. If there is no error correlation among hydrometeors, why does increments for q_c exist?

(3)       It will be good to shorten reg to r or another character in Eq (4).

(4)       L99: “Most studies only use five control variables for v, for example, …”: You might want to rephrase this sentence such as “Most studies for mesoscale weather applications” since this statement is only true for small-scale DA applications. For large-scale applications (e.g., global model applications), streamfunction/velocity potential, or vorticity/divergence are chosen as the CVs.

(5)       L123-124 “Uh defines the … using the Laplacian Method on the EOF modes.”: Uh deals with the horizontal transform. As far as I know (at least from the GSI code), Wu et al. 2002 use the high-order recursive filter instead of the Laplacian Method for spatial filtering in B.

2.         Can you show your observation operator (i.e., equations) for the radar reflectivity? Otherwise, I have difficulty interpreting your single-obs test.

Minor comments:

1.         L92, “it may be” => it can be

2.         Figure 2: the ticks of the right Y axis for the last column are cut off: Need to zoom out this figure. 

3.         Eq (4), the last row, you incorrectly have two reg_{10_5} there.

4.         L118: “The corss-correlation between hydrometeor…”: between => among

5.         L133, “each of the three-dimensional analysis control variables”: “analysis variables” or “control variables” here?

6.         L148: “The balanced part contribution reflects” => “The magnitude arising from the balance part in the physical transform (Up)”

7.         L283: “5.2. The analysis increment in the first assimilation”: add “cycle” at the end

8.         L284: “…in the first assimilation…” add “cycle” after “assimilation”

9.         L343, Figure 13: Can you also the observed reflectivity here?

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Manuscript Number: remotesensing-2088550

 Title:  Radar Reflectivity Assimilation Based on Hydrometeor Control Variables and Its Impact on Short-Term Precipitation Forecasting

Author(s):  Hong Zheng, Yaodeng Chen, Shiwei Zheng, Deming Meng and Tao Sun

 Overall comments:  

In this manuscript, the authors demonstrated the better initialization of hydrometeors using vertical and multivariable correlations of hydrometeor control variables and the more reasonable thermodynamic and dynamical structure of the initial field. I recommend the publication of the manuscript once the following question is answered and addressed in the manuscript.

 Major:

For better results, it is recommended to compare the Contoured Frequency by Altitude Diagrams (CFADs) of radar reflectivity for CNTL, Hydro, Hydro+reg experiments and observations.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors have developed WRFDA-3DVar to enhance the assimilation of radar reflectivities. Four case studies are analyzed to quantify the impact of the radar assimilation on precipitation. The precipitation forecasts improves as a result of the developments. I think the contents of the manuscript should be of interest to Remote Sensing readers. My main comments are 1) to add a simulation without radar assimilation to quantify the benefit of the radar assimilation methodologies presented, and 2) to describe the improvements found by other works assimilating radar observations. Below are more specific comments.

MAJOR COMMENT 1

The authors should add another experiment without assimilating the radar observations. This will clearly illustrate the benefit of assimilating this remote sensing observations and will put in perspective the magnitude of the improvements obtained as a result of the present developments.

MAJOR COMMENT 2

In the introduction you should go beyond methodologies and describe, and quantify if possible, the improvements found by other authors assimilating radar observations.

SPECIFIC COMMENTS

Line 37, It may be better to say "falling hydrometeor information" instead of "hydrometeor information"

Section 2: You should probably describe the observation operator H

Section 2: The methodology from line 106 until the end of the section is hard to follow for non experts. The description is missing key sentences to allow non experts to follow it.

Fig. 1: I do not understand what time series you use to calculate the autocorrelations shown in Figure 1. Please improve the description of Figure 1.

Line 142-145: I do not follow the causality in this sentence. Clarify the impact of the vertical mixing.

lines 193-201: The land surface model reference is missing.

Figure 4. Please describe the different observational sources (the legend in panel a) in the caption or even better in the body of manuscript

Figure 5. What are the '3h fcst' labels?

Line 220: You need to explain what is the 20-cycle.

lines 228-230 and Figure 6. Is the improvement statistically significant? The errors are within the error bars in Figure 6. So one cannot say that one experiment is better than the other. Please describe in the caption how you calculate the error bars

Section 4.3: The description of FSS and BS is very hard to follow. The definition of Pf, Po, NA, NB and NC need to be improved. Also, what is the radius?

Figure 7: Here it would be important to add a column with the no-radar assimilation scores to put in perspective the radar-assimilation improvements shown

Figure 10 and related discussion should go into section 3. This is a methodological aspect that is within the results.

Section 5.4: Can you provide a quantification of the precipitation overestimations?

Line 347-348 can you quantify the precipitation improvements?

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

The authors added the CFAD plots as I suggested. I accept the document in its present form.

Reviewer 3 Report

The authors have considered all my suggestions. I did not have time to go through the details of all the additional material but the manuscript appears to be suitable for publication.