the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 05 Mar 2026
Country-wide rainfall estimates from a commercial microwave link network in Belgium
Abstract. Accurate quantitative precipitation estimation (QPE) at high spatiotemporal resolution remains challenging despite advances in observational technology. This study presents the first comprehensive evaluation of rainfall retrievals from a commercial microwave link (CML) network in Belgium, examining whether CML-derived rainfall can complement existing dense rain gauge and weather radar networks. We analyze four intense summer rainfall events in 2023 using over 2800 microwave (sub)links operating across frequencies from 10 to 85 GHz. Through systematic sensitivity experiments, we assess the impact of optimizing the processing procedures. Our results demonstrate that careful processing of CML data is essential: a novel outlier filtering algorithm, radar-based wet-dry classification, rainfall-intensity-dependent wet-antenna correction, and fitting local drop size distributions from three disdrometers substantially improve rainfall retrievals. The optimized CML-derived rainfall estimates match or exceed the performance of a state-of-the-art radar-gauge merged product compared to a dense rain-gauge network, particularly over urban areas with dense high-frequency link coverage, like the Brussels-Capital Region. These findings provide strong evidence that integration of CML information into multi-source precipitation products could yield substantial improvements in high-resolution QPE, particularly for urban hydrological applications and extreme-event monitoring.
- Preprint
(13020 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-457', Anonymous Referee #1, 08 Apr 2026
-
RC2: 'Comment on egusphere-2026-457', Anonymous Referee #2, 10 Jun 2026
General comments
The paper adresses a relevant scientific questions within the scope of HESS: utilisation of opportunistics data, i.e. commercial microwave link (CML) data, to estimate rainfall at the scale of a country.
The title is well suited to the content of the article.
The summary should be slightly modified as the following wording is confusing: “whether CML-derived rainfall can complement existing dense rain gauge and weather radar network”. This is not estimated. And the following wording is excessive: “These findings provide strong evidence that integration of CML information into multi-source precipitation products could yield substantial improvements in high-resolution QPE”. This is not demonstrated in the paper.
The paper compares the quality of quantitative precipitation estimations (QPEs) provided by various sources, using measurements from a network of rain gauges as a reference for actual precipitation. The sources of QPEs are:
- an operational state-of-the-art radar-gauge merged product
- an interpolated product based on rain gauges only, estimated by ordinary kriging procedures from the pycomlink software
- seven versions of CML estimates, obtained using various combinations of data processing options provided by the pycomlink software.
Since the data processing relies on the free software pycomlink, there are not really any new concepts or methods, apart from the use of disdrometer data to estimate specific attenuation-rainfall rate relationships (although this has already been done).
The data describes four extreme rainfall events of 24h or 48h (i.e. 7 days), all within a period of 38 days in summer 2023, which is a quite limited diversity in hydrometeorological situation.
The overall presentation is well structured and clear, the language is fluent, and references are appropriate. Nevertheless, several improvements need to be made (see specific comments). Particularly, we noted some inconsistencies in the assertions which are sources of confusion and need to be clarified (see specific comments). Also, some criteria should be presented more explicitly (with formulas and definition of the values compared), and in general, the authors should verify that all abbreviations or variable names are correctly defined before use. May be a short sub-section could be useful in section 3 (Method) to define the validation methodology and the calculation of the criteria used. Some figures are not very easy to read, and we provided some other suggestions in technical corrections.
Specific comments
1) Several inconsistencies
First confusion in 2.4 Radar Data
Radar data is described as an operational state-of-the-art radar-gauge merged product. In Journée et al. (2023) we find: “Instantaneous rain rates are obtained every 5 min, corresponding to the full 3D radar scan. The rainfall accumulation over 5 min is obtained by computing the movement of precipitation using optical flow techniques.”
But in the 2.4 section, the authors write “(aggregating the three 5-minute values, without taking into account advection over a 5-minute interval)”. This is confusing because:
- it is not necessary if the advection is already done in the 5-minute radar QPE computing.
- if the radar-based 15-minute QPE not include advection, this QPE is not state-of-the-art.
Please clarify.
Second confusion about best WAA correction
In the summary and in section 5 (line 440), rainfall-intensity-dependent wet-antenna correction is presented as the best option, which is in accordance with several results in literature, but in section 4.3 (line 371) the authors write “RII appears to have the worst metrics of the three wet-antenna correction method” (so why to us it in RID version?). Please clarify. Remark: the RII “correlation” is the best for “all cases” except radar and RID (table 5).
Third confusion
The validation use two related criteria:
- The coefficient of determination (R²)
- The “correlation” (r)
Please clarify if r is the Pearson correlation coefficient, and clarify the calculation of r and R² (formula, values used), and explain the large differences of values between R² and r² when you compare two sets of rainfall values.
2) Several Remarks
2.1 Please define a “sublink” (page 4, line 81).
2.2 Could you explain the impact of ATPC on attenuation estimation? (page 4)
2.3 Wet-dry classification: could you explain what you think about estimating standard variation of attenuation using 8 values? The authors write “The advantage of this method is that generally no calibration is needed” I am surprised that this criterion can be used without calibration, because thresholds must be used. Please explain.
2.4 WAA line 227: the decrease in WAA after the end of the rain: many references indicate that this decrease is highly variable, depending on various factors (including weather conditions).
Line 232: Could the authors verify the 14 dB value for an upper limit for WAA? (I cannot access the Leijnse reference)
2.5 In section 4.1 and figure 4 : Could the authors explain the impact of an ordinary kriging on the validation criteria, particularly the e-folding radius? Reminder: RAD-QPE is the best radar-gauge spatialization, and a simple ordinary kriging of rain gauges measurements is not a state-of-the-art product. It would be interesting to know if the kriging used the same variogram or not for the different estimations.
2.6 In table 5: Please, define the bias formula: “CML estimate minus reference” or “reference minus CML estimate”?
Line 371 and figure 5: RII results surprises me. Please, verify the determination R² values and the values compared for each criterion (see above remarks).
Line 375: Remark: The authors write “It should be mentioned that the constant value of 2.3 dB could in principle be reduced in RCI and RTI, particularly during light rain, to allow more light rain to be detected”. Have you a concrete method to do that?
Figure 7: same remark as before: How R² and r values have been calculated?
Table 6: could you more explain the differences between “mean rate” and “mean intensity”? With or without zero values?
Remark lines 446-447: The authors write “It should also be mentioned that considerable uncertainty on the wet-antenna correction exists due to the fact that both antennas of a CML are not necessarily wet at the same time, particularly during scattered shower.” Why not use radar data for that?
Line 465: The authors write “our results provide evidence that existing multimodal rainfall products could be substantially improved through integration of CML information that can be accessed in real time.” No. They haven't tried to merge CML data with other data, so the results not really “provide evidence” on the matter.
Technical corrections
For equation (4) and (5) , define Di, ND, NV
line 258 “polarisation pol” and not “polarisation p”
Table3. Can the authors provide the Overeem et al. (2016a) values?
Figure 5: In the legend: “and (b-i) all permutation”, may be (b-h)?; Title graph RII and RTI seems inverted; Is it possible to better define the 453 compared values?
Figure 6: overlapping lines make the figures not very legible. Some lines are not visible. Try to improve it.
Figure 10: Please, change the colour of dots (not legible), particularly blue/green and red/purple
Citation: https://doi.org/10.5194/egusphere-2026-457-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 976 | 534 | 83 | 1,593 | 73 | 125 |
- HTML: 976
- PDF: 534
- XML: 83
- Total: 1,593
- BibTeX: 73
- EndNote: 125
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
My review is in the attached file.