the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 07 Aug 2026
Derivation of the multi-Doppler 3D wind field during the 2023 WesCon-WOEST campaign in southern England
Abstract. Convection drives high-impact weather events but remains challenging to represent accurately in modern convection-resolving weather prediction models. Robust evaluation and development of these models therefore require detailed observations of convective dynamics, with enough data for statistical application. In this study we present the first multi-Doppler radar 3D wind analysis in the UK. Five radars used in the WesCON-WOEST field campaign are used with an updated version of the PyDDA algorithm. This includes a newly implemented "ORIGAMI" unfolding algorithm based on model winds, a combined VAD initialisation and a cloud-top boundary condition. The resulting dataset provides a complete 3D wind grid with 1 km horizontal and 500 m vertical resolution every 10 minutes over the 3 month field campaign. Validation against two radar wind profilers show the horizontal wind components have RMSE better than 2.4 ms−1, with a small bias in the westerly component, u. Statistical comparison to aircraft flight data shows similar distribution, but underestimates the strongest up- and downdrafts, as a result of not resolving the scale of the smallest updrafts. This data provides important dynamical information on convective processes and offers a valuable resource for evaluating and improving convection-resolving models.
- Preprint
(15132 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-4631', Anonymous Referee #1, 29 Aug 2026
-
RC2: 'Comment on egusphere-2026-4631', Anonymous Referee #2, 04 Sep 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-4631/egusphere-2026-4631-RC2-supplement.pdf
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 165 | 52 | 12 | 229 | 9 | 9 |
- HTML: 165
- PDF: 52
- XML: 12
- Total: 229
- BibTeX: 9
- EndNote: 9
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript documents the first multi‑Doppler 3‑D wind dataset over the United Kingdom, built upon the PyDDA variational framework with several important methodological upgrades: the new ORIGAMI model‑guided velocity dealiasing algorithm, multi‑radar composite VAD initialisation, and a cloud‑top upper boundary condition to suppress spurious vertical velocities above cloud layers. The 1‑km horizontal/500‑m vertical, 10‑min‑cycled three‑dimensional wind product derived from five radars during the WesCon‑WOEST field campaign represents a valuable new observational resource for evaluating convection‑permitting numerical weather prediction (NWP) models. The work includes comprehensive validation against radar wind profilers (RWP), FAAM aircraft in‑cloud measurements, and qualitative comparison against high‑resolution CAMRa RHI retrievals.
Overall, the study is well‑motivated and fits well within the scope of Atmospheric Measurement Techniques, as it describes both algorithm modifications and a complete, campaign‑scale observational dataset intended for public archiving. The manuscript is generally clearly structured; however, several key methodological descriptions, validation interpretations, and dataset documentation details require substantial improvement before acceptance. Therefore, a major revision is therefore warranted.
Major comments
1. The ORIGAMI algorithm is a core novelty of this work, using NWP‑simulated radial velocities to select the optimal Nyquist fold for observed radar radial velocities. While the four‑step workflow is summarised, critical implementation details remain absent. For example: (1) What spatial/temporal interpolation is applied to map WRF model winds onto each radar’s polar coordinate system? (2) What tolerance or threshold is used to flag cases where model winds are too discrepant from observations for reliable dealiasing? The paper mentions misplacement of intense convective cells in WRF can degrade ORIGAMI performance, yet no quantitative diagnostic is provided to identify such problematic volumes in the final dataset.
2. The manuscript correctly identifies spatial‑temporal scale mismatch as the primary driver for PyDDA’s underestimation of peak updrafts and downdrafts compared with FAAM aircraft and CAMRa RHI products. However, the discussion conflates three distinct error sources: intrinsic smoothing from the 1 km × 1 km × 500 m grid resolution, temporal smoothing from the multi‑minute radar sampling window, and limitations within the variational cost‑function setup. At present, it is difficult for readers to disentangle how much of the vertical‑wind bias arises from each factor. I strongly suggest the authors explicitly separate these sources of uncertainty in Section 7: Discussion. Where possible, add quantitative estimates: for instance, comment on what horizontal scale of convective updrafts can realistically be resolved given the radar sampling geometry and chosen grid resolution. Explicitly state these resolution‑derived limits within the dataset metadata description so future users can appropriately interpret updraft statistics computed from this product.
3. The manuscript presents RWP‑derived wind profiles extending down toward the surface in Figure 4b, with no explicit indication of the instrument’s low‑altitude blind‑zone. For typical L‑band (1290 MHz) boundary‑layer radar wind profilers, the hardware blind zone / minimum usable height commonly lies in the range of 300–500 m AGL, as documented in Guo et al. (2023, https://doi.org/10.1002/qj.4474). In this work, the deployed PCL1300 RWPs were operated in low‑mode with the first range‑gate centred at 75 m AGL (gate spacing 75 m). However, the lowest gates are frequently contaminated by ground clutter, transmitter leakage and low signal‑to‑noise ratio, so valid meteorological winds are not always reliable at these very lowest levels even if raw gates exist. This important limitation is not visualised or annotated in Figure 4b. This is physically unrealistic and may mislead readers.
The authors can revise the main‑text description to clarify: “While the PCL1300 RWP samples starting at 75 m AGL in low‑mode, the lowest ~several hundred metres often suffer from ground‑clutter contamination and low SNR; therefore, only data passing strict SNR quality flags are treated as reliable for comparison, consistent with typical L‑band RWP performance described in Guo et al. (2023).”
Besides, please check other figures (e.g. Figure 6 time‑height cross‑sections) to ensure that the RWP low‑altitude quality‑control thresholds are either plotted as masked regions or clearly described in captions.
4. The sensitivity tests across 16 combinations of cm and co support the final choice cm=4096, co=0.01. The bootstrap significance testing shows metrics are relatively insensitive across a broad parameter space when ≥3 radars observe a grid cell. However, the manuscript provides little practical guidance for other researchers who wish to apply this modified PyDDA workflow to other radar networks. It remains unclear whether these weight values are transferable to other campaigns or are tuned specifically for the WesCon‑WOEST radar configuration.
5. Reduced retrieval quality below 2 km is attributed jointly to surface w=0 boundary condition, mass‑continuity constraints, and poor multi‑radar beam coverage close to the surface. However, the relative contribution of observational geometry versus algorithm constraints is not quantified.
Minor comments
Typographical and notation issues
Figure improvements
Text clarity
Terminology