the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 12 Sep 2025
A dual-frequency (W-band and G-band) radar optimal estimation framework to retrieve drizzle properties more accurately
Abstract. High-resolution cloud radar observations are generated from a large eddy simulation of drizzling marine stratocumulus. These observations are then used to investigate dual-frequency measurements combining W-band (94 GHz) and G-band (239 GHz), a pairing that offers unique sensitivity to early-stage drizzle and small liquid water paths by exploiting the differential backscatter and extinction signatures of hydrometeors. An optimal estimation framework is implemented to retrieve key drizzle microphysical properties from the simulated observations. We demonstrate that the synergies of a nadir-looking W-band and G-band radar system can result in more than one order of magnitude reduction in the uncertainty of the estimated drizzle mass mixing ratio, number concentration, and mass-weighted mean diameter compared to W-band only observations. The methodology can be applied to W-band and G-band airborne observations to improve drizzle estimation. Furthermore, we show that these reductions in uncertainty can be attainable from a spaceborne platform with mission architecture and radar parameters realizable with current technology.
- Preprint
(2385 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-4248', Anonymous Referee #1, 13 Oct 2025
-
RC2: 'Comment on egusphere-2025-4248', Anonymous Referee #2, 14 Oct 2025
The paper develops an optimal-estimation retrieval that ingests simultaneous W-band (94 GHz) and G-band (239 GHz) radar observables to retrieve drizzle mixing ratio (qr), number concentration (Nr), and mass-weighted mean diameter (Dr) in marine stratocumulus. Using a VOCALS-based LES model simulations, the dual-frequency setup reduces posterior uncertainties by over an order of magnitude versus W-only; an information-content analysis shows >10× gain when priors are weak. A notional spaceborne W+G configuration (2-m antenna, −21/−30 dBZ sensitivities, ~50 m vertical, ~1–1.5 km along-track) still yields substantially lower errors than W-only despite beam-mismatch and clutter effects.
The paper is clearly written and makes a strong case for multi-frequency radar measurements to better constrain global cloud microphysics. A few additions would further strengthen the manuscript:
1. Quantifying retrieval errors from synthetic tests
Because the evaluation relies on synthetic data, please include quantitative error metrics. For example, show joint histograms (or density scatter plots) of retrieved vs. model “truth” for the key variables, and report bias, RMSE, and correlation.2. Non-uniform beam filling (NUBF) and related biases
NUBF can strongly affect attenuation and thus DFR/PIA, especially when horizontal microphysical gradients are sharp. Please discuss this in more detail and, where possible, compare retrievals to model fields averaged to the radar footprint (in addition to high-resolution truth- point 1) to quantify NUBF-induced biases.3. Optimal-estimation variable choice and prior statistics
OE assumes (approximately) Gaussian statistics in the state space. Since number concentration Nr spans orders of magnitude, a log-transformed variable (e.g., log10 Nr) may be more appropriate. Please consider retrieving in log-space and show the a-priori distributions (from the model) as 2D histograms for the retrieved pairs to demonstrate approximate normality.4. Liquid cloud contribution and differential attenuation
The current state vector appears to exclude liquid cloud parameters. Please clarify how cloud water is treated in the forward model. If cloud contributions to reflectivity/attenuation are neglected, differential attenuation below cloud base could introduce systematic biases; quantifying or bounding this effect would be valuable.Minor points
1. Figures 5 & 7: If only diagonal elements are non-zero, consider plotting the diagonal as a line (or bar) rather than a full matrix for clearer readability.
2. Formula (17): I believe this should use the determinant of the relevant matrix, with S_x^{−1} (not a fraction notation). Please check and correct the expression.
Citation: https://doi.org/10.5194/egusphere-2025-4248-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,325 | 29 | 11 | 1,365 | 12 | 17 |
- HTML: 1,325
- PDF: 29
- XML: 11
- Total: 1,365
- BibTeX: 12
- EndNote: 17
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of"A Dual-Frequency (W-Band and G-Band) Radar Optimal Estimation Framework to Retrieve Drizzle Properties More Accurately" by Socuellamos et al.
This manuscript presents a dual-frequency (W-band and G-band) radar optimal estimation framework aimed at improving the accuracy of drizzle property retrievals. The authors use large-eddy simulations (LES) of stratocumulus clouds and apply both single- and dual-frequency radar retrievals to assess the retrieval performance. The results show that the synergy of dual-frequency observations yield more accurate drizzle property retrievals than single-frequency approaches.
The paper provides valuable insights into the potential of G-band radar for drizzle retrievals and its future applicability in satellite missions. However, I find that the manuscript would benefit from clearer presentation, more thoughtful discussion, and more careful proofreading. A detailed list of the comments is shown below.
Major Comments
Minor and Technical Comments
Figure 2a: Please explain why the visible optical depth increases again around 80 km.
Line 148: Define the acronym “ARM”.
Line 179: Is the path-integrated attenuation computed only within the cloud or precipitation layer? If not, please specify the vertical range over which the attenuation is integrated.
Line 185: Are the retrievals limited to drizzle properties, or are cloud properties also retrieved? Given G-band’s stronger attenuation, I expect the differential attenuation to be more sensitive to cloud droplets.
Line 215: This sentence read odd, please revise for clarity.
Line 237: This sentence is unclear; please rephrase. Also, are cloud and precipitation properties retrieved separately? If so, do they share the same DSD function?
Line 254: Clarify which “first profile” is referenced and its starting point.
Lines 255–265 (Figure 6): The description of sub-diagonal and off-diagonal elements does not match the four-panel figure layout. The color bar combines two units, making it unclear which applies to the Jacobian matrix. If keeping Figure 6, clearly describe each quadrant and explain its relationship to Eq. 14.
Figure 3: The virtual comparison between the simulated and observed radar reflectivity (Figure 3 and Figure 4) is not sufficient to show the representation of the simulated radar variables. Please provide additional details (such as LWP) for better comparison. Also, please provide the G-band radar specifications (frequencies, bandwidths) and explain if any calibration procedures are performed. Finally, in Figure3c, the observed dual-frequency ratio does not monotonically increase from cloud base to top, which may suggest reflectivity calibration issues. Please clarify.
Equation 5: Specify the particle size range used in the DSD integration. Similarly, clarify the integration range used for attenuation calculations.
Line 300: Indicate where the 0th, 10th, and 100th profiles are located, and describe their cloud/precipitation structures.
Figure 8: I didn’t not notice any discussions regarding on differences across Figures 8a–8c, particularly the large edge values in Fig. 8c and their possible causes.
Line 305: Most covariance elements appear negative (red); please confirm and explain.
Line 318: The text mentions small uncertainty near the surface, but Fig. 9c shows smaller uncertainty near the cloud top, please clarify.
Line 321: Please specify the origin of the diameter range used.
Line 348: The text refers to “PDF of relative errors,” but Fig. 11’s title is “uncertainty.” Please ensure consistency.