Dual-Doppler Radar Characterization during the Krummendeich Field Experiment: A Campaign Overview and Performance Assessment

Jordan, Arianna Marie; Hung, Lin-Ya; Wolken-Möhlmann, Gerrit; Gottschall, Julia

doi:10.5194/wes-2026-114

Preprints

https://doi.org/10.5194/wes-2026-114

Preprints

28 Jul 2026

| 28 Jul 2026

Status: this preprint is currently under review for the journal WES.

Dual-Doppler Radar Characterization during the Krummendeich Field Experiment: A Campaign Overview and Performance Assessment

Arianna Marie Jordan, Lin-Ya Hung, Gerrit Wolken-Möhlmann, and Julia Gottschall

Abstract. The Krummendeich campaign in northern Germany was a field experiment designed to test dual-Doppler radar (DDR) technology for resolving flow fields. The campaign took place at the WiValdi research wind farm containing two operating turbines. Throughout the measurement period, a comprehensive observational network consisted of DDR, scanning and profiling lidars, a microwave radiometer, a meteorological mast, and laser disdrometers. Capitalizing on these datasets, the study herein offers an overview of the campaign while demonstrating the utility of the DDR method as a wind measurement tool for wind energy applications, specifically. This is done through validations against a co-located Doppler scanning lidar, assessments of data availability across meteorological conditions, and analyses of the method's capability to resolve flow features, i.e., turbine wake signatures. The site's temperate maritime climate, marked by frequent precipitation and wind direction-dependent stability regimes, provided a naturally diverse atmospheric environment for testing the method across conditions relevant to central European wind energy deployments. Validation showed great agreement with scanning lidar across most height levels, with discrepancies mainly occurring at near-surface levels. Hydrometeor presence strongly governed data availability, where precipitating periods sustained near-complete domain coverage. While radar returns were occasionally affected by artifacts even during otherwise well-covered scans, the method successfully captured both low-level jet and turbine wake within its flow field. Together, these results performance verification contribute to a continuously growing body of evidence supporting DDR as a viable and scalable tool for wind energy research and industry, with plans for deeper analysis and additional fieldwork to follow.

Received: 26 Jun 2026 – Discussion started: 28 Jul 2026

Competing interests: At least one of the (co-)authors is a member of the editorial board of Wind Energy Science.

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Arianna Marie Jordan, Lin-Ya Hung, Gerrit Wolken-Möhlmann, and Julia Gottschall

Status: open (until 26 Sep 2026)

Post a comment Subscribe to comment alert

RC1:
'Comment on wes-2026-114', Anonymous Referee #1, 14 Aug 2026 reply
The study contributes to already existing Dual-Doppler Radar (DDR) experiments with a new measurement campaign. The main focus is testing whether the DDR setup holds up for a smaller domain than were used in the previous studies and explore its applicability in the wind turbine wake research. Generally, the findings are split into two groups, adding to one of those research directions. First, the authors confirm that the DDR results lie within the expectations and do not contradict recent similar studies. Second, the examples of wakes captured or failed to be captured by DDR are presented, giving a good overview of how the wake measurements may be hindered. The authors mostly refrain from the specific recommendations for the DDR setup, as the primary goal was to confirm whether the DDR measurement of wake features or atmospheric events like low-level jets were possible at all.

Since the study builds upon the known and well-researched DDR method, it is naturally expected that the reader is already familiar with its basics. If not, they may check the detailed description in one of the cited articles. Still, I noticed an occasional lack of technical details specific to the current study which I will list in minor suggestions.

The plan and goals are laid out nicely, focusing on the features that will be explored in the measured data. The findings set a foundation for further research in the compact DDR setup.
Specific comments:
I see a discrepancy in how the problem of insect backscatters is constantly highlighted as affecting the measurements prior to filtering and how little information is presented on the exact scale of this influence. Judging from the remarks, filtering the noise from biological backscatters may warrant a separate article, but there are a couple of points which I would like to see clarified:
Line 130: A reproducibility question: Which filters were used? If any research group encounters a similar issue, would it be easy for them to remove the insect noise with conventional methods, like median filter? If the filters are completely new, how were they validated?

What percentage of the data was affected by the insects and had to be filtered eventually? How filtering the noise caused by the insect backscatter compares to filtering by other criteria?

Lines 235-236, Fig. 8: If I understand correctly, the discrepancy between pre-filtered DDR profiles and reference data behaved similar to the profiles shown in the cited articles. Still, seeing the pre-filtered profiles for the Krummendeich dataset would give an idea on how the initial results were similar to the previous studies that, presumably, did not filter insect backscattering, and how much the filtered improved the agreement to the reference data.

In general, the only metric of the effect of insect noise is found at line 364, in Discussion, which is quite late, considering that this was an important part of the data processing and even required a special procedure.

Lines 131-134: I would like a clarification here to reduce ambiguity: are the resolutions of the coarse and fine grid also unknown? If the grid resolutions are known and only the processing details are hidden, then the values should be provided for reproducibility. It would be also interesting to see a schematic on how coarse and fine grid relate to the points scanned by the radars -- how many lidar points, on average, define a Cartesian grid point. This question is tackled later in Discussion (lines 380-384), although without exact values, so it is confusing why the grid size was not brought up earlier to build the later discussion upon it.

Figure 2. j) What is the cause of the DDR fine grid data being absent after September 2025, while the coarse grid data are still present? I could have missed that in the text.

Line 174: A minor question: which range of the bulk Richardson number is considered here as neutral conditions?

Figure 11. Having the wind direction series plotted would be helpful to understand Fig. 12: the fields for T2-T5 are quite noisy to be certain whether they show a wake at about 280⁰ wind direction, or is it a measurement fluctuation, while the wake is lost to the noise.

Line 349 and Fig.12. Clarification needed: What is exactly referred to in the part 'certain wind directions that avoid the bulk of the artifacts can be targeted (in the Fig. 12 example'? Is it the T1 timestamp that produced good quality scans, so this wind direction can be considered? Or wind directions for all azimuths that were not affected by the artifacts as seen in the T2-T5 columns? Again, as in the previous comment: knowing the wind direction for this period would helpful for the interpretation.

Technical corrections
Line 112: a typo in the experiment period. That should be 19 March 2025–9 January 2026, I suppose?

Figure 2
a) Channel A is almost overlaid by Channel B, so this panel conveys little on whether the channels mostly agree or have discrepancies and whether it was necessary to plot both channels. Since the figure focuses on the data availability, the scale makes it mildly ambiguous on what happens near 0 mm/hour. Is it: zero values are not plotted but there are numerous instances of near-zero precipitation, which are shown in the plot. Or: any line means recorded data even if the value is 0 mm/hour, no lines -- no data from both disdrometer channels. I would presume the first, based on how the precipitation is usually visualised. If it is the latter, then the remark would be appreciated.

c) Why does a non-centered colour bar goes down to -50⁰C, if the temperatures seem to stay within 0-20⁰? The colour bar range should be adjusted.

Figure in general: from the position of the minor ticks, I infer that the minor ticks mark the weeks and the grid lines separate the days – is that understanding correct? A remark on that in the caption would be nice.

Figure 2 is referred to only once as a general reference. Pointing out to specific panels when the respective instrument is mentioned on page 7-8 would be nice.

Figure 3: A personal preference, but I would try to adjust the radial range to be the same for both wind roses to make the comparison fair at a glance.

Figure 6: Mean Absolute Difference (in the subplot title and main text) and Mean Absolute Deviation (in the caption) are different metrics. Which one is actually plotted? Please, check that correcting the term does not affect the reasoning.

Figure 10. Although the plot is simple and remains readable, the image quality is very low compared to other figures. Some other images are also pixelated, but not to that extent. If it is planned to update all figures with vectorised versions or better resolution in the final version, then disregard this comment.

Figure 11: The colour bars for panels (b) and (c) are missing the quantity label, although it is present for panel (a) colour bar.

There were several instances of missing words or unclear phrasing. I might have not caught all of them, so I recommend another round of proofreading to ensure that no important information was accidentally omitted during the editing.
Line 63: 'Moreover, how DDR performance varies with the prevailing atmospheric regime is necessary, especially in environments that differ from Oklahoma’s semi-arid climate' -- the part in bold may be incomplete.

Line 259: Unclear line or editing artifact here: 'Seasonally, there were major differences in the actual fractional values: precipitating conditions in the colder months, foggy conditions yielded much higher levels of availability than warmer months, and vice-versa with precipitating conditions, reflecting the relevance of the types of hydrometeor droplets having an effect'. The first mention of 'precipitating conditions' looks out of place.

Lines 262-271: please, double-check the paragraph, as it lacks proofreading compared to the rest of the text. Particularly, the following sentence reads as incomplete: 'As such, Figure 10 portrays the relationship between DDR data availability, where the rain particle number (Fig. 10a), the rain intensity used in Fig. 9 (Fig. 10b), and the kinetic energy of the droplets'. Is it just listing the data series for comparison (then 'where' should be removed) or is a part of the sentence after 'where' missing?

Line 391: 'Further, the DDR method is that strong data availability requires substantial hydrometers for the radar beams to scatter off' -- either a word is missing after 'is' or 'is' should be replaced with a verb.

Line 442: 'At the heights in Fig. 7, the scanning lidar was seen to measure well, the were well aligned with r between 0.93 and 0.95...' is a noun missing between 'the' and 'were'?

Line 444: 'Note that the rain-exclusion filter applied in the profiling–scanning lidar comparisons of is intentionally not applied here' -- another instance of a missing text

There is an inconsistency in referring to the panels in figure captions. For example, Fig. 1, 3 refer to panels as left-right, while Fig. 2, 5, 7 use letter indices, Fig. 6 refers to panels without any indices even though they are marked as a-b. I leave that to the editor to decide whether this has to be corrected for the consistency or can be kept as it is.

Reply
Citation: https://doi.org/10.5194/wes-2026-114-RC1
RC2:
'Comment on wes-2026-114', Anonymous Referee #2, 09 Sep 2026 reply
These same comments have been uploaded in PDF form in case the formatting of this text box is not conducive or removes some formatting.
The manuscript details a field experiment in dual-Doppler radar (DDR) instrumentation, a well-established remote sensing technology for obtaining wind energy-relevant wind measurements. In this case, DDR is applied to a smaller domain than that seen in the American Wake Experiment (AWAKEN), which serves as a sort of quasi-predecessor or precedent. The study aims to determine whether DDR was, for this domain, capable of resolving spatially coherent flow features (e.g., wakes, low-level jets) more effectively than point measurements like those from lidars or met masts. To make this determination, DDR measurements are (a) quantitatively compared against lidar measurements under a variety of wind directions, sky conditions, and precipitation states, and (b) plotted and visually analyzed for the presence of various wake features. The authors note that DDR generally agrees well with coterminous lidar measurements, and show evidence that wake-related features can be seen in DDR-generated fields.
The project’s goals are articulated well, though there are a few areas that feel underexplained or could benefit from more thorough investigation.
Clarifying questions/unclear areas

L55: Who does ‘they’ refer to? One or multiple of the mentioned publications?

L61: Why is uncertainty amplified over larger areas? Later in the manuscript it is implied that the degree of spatial averaging and widening beam width are the cause, but it would be good to specify it (or any other relevant explanation) here as well.

L72: “Scanning lidar observations with well-established performance and uncertainty characteristics” – could the authors provide citation(s) to previous studies or other resources that describe these well-established characteristics?

L112: I assume this should read 19 March 2025 - 9 January 2026?

L120: Regarding descriptions of the radar scanning strategies, we can infer azimuths from the figures (though it would still be good to specify them), but the elevation angles are only referenced in quantity (“up to 14”) and cannot be inferred from any figures. Were there not consistently 14 elevation angles as the use of “up to” implies, and if not, why? Which elevation angles were used? Since the comparison is made that the AWAKEN sectors are larger and feature lower elevation angles, I assume the elevation angles here are known. If you can, please provide them.

L124-125: “To favor vertical wind fields over a more localized domain” – does this mean vertical profiles of horizontal wind, or values for the actual vertical wind w? The phrasing as it is now makes it seem like the latter, but no analysis in this study is done with vertical components, so I assume it is the former. On the topic, some mention of the vertical component’s role or absence in the study should be made, i.e., whether instruments measured it or not, and, if they did, why it was excluded from the study’s analysis.

L129-131: The conclusion that the increased wind speeds/distorted radial velocities are solely attributable to insects, and the editing steps implemented therein to remove them, feels a bit rushed. In fact, later this claim is qualified: “the removal of influence supposedly generated by insects…” (L235). I suggest including some examples of relevant literature on these signals and their behavior, and some information about the process used to remove them, in this section, as well as some metric or image that can demonstrate a marked difference in the before and after.

L136: “The ratio of the number of bins that pass the filter” – which filter is being referred to here? The aforementioned insect filter? If so, I would specify it again, since some space exists between talking about filtering and talking about coverage factor calculation.

L136-137: Regarding the overlapping of bins from both radars, is any location where data overlap exists between the two radars usable? Other studies that use dual-Doppler strategies typically list a range of ideal crossing angles between the two radars’ beams that are usable, so if that restriction is imposed here as well, it should be noted. Also, do the radars feature any range blanking zones?

L153: How does the scanning lidar have a range gate length of 50 m but an overall range of 100 m to 3825 m? 3825 is not a multiple of 50.

L163: From what source is the regional climatology defined?

L201: The manuscript states that errors “increase notably above 150 m”, but in Figure 6 they actually decrease for both metrics (RMSE and MAD) and for both radar grids (coarse and fine) between 150 and 200 m. The error only begins increasing at 200 m, and we can only say so for the coarse grid, because the fine grid coverage stops. I suggest revisiting this figure to make sure it, and its interpretation, are correct.

L225-226: How many of the nearest DDR grid points are used? In the Figure 2 caption, “a single grid point encompassing the location of the lidar” is used. Are these two referring to different things? A bit more information on the spacing of the radar bins, as well as the probe volume of the lidar, would be good to augment the trustworthiness of the comparison.

L242: Should Table 2 be Table 1? Table 1 contains the radar availability data, where Table 2 contains the list of (presently, all non-radar) instruments and corresponding notes.

L242: How is ‘samples’ defined here? Bins, scans, volumes?

L263-264: It is not a fundamental characteristic of X-band radars that they require hydrometeors to return a sufficiently strong signal; they can operate and obtain measurements in clear air environments, and in fact do so in this experiment as evidenced by nonzero availability during clear-air periods in Figure 9a-d I and II. The data availability may be reduced without the presence of hydrometeors, but data are still obtained. This statement is thus too absolute, and is itself contradicted in L273. It is also worth noting that fog, which is composed of hydrometeors, virtually eliminates all valid return, so using hydrometeors as a blanket tracer to guarantee high availability is a bit too broad and should be moderated by a reminder statement on the effects of fog. Out of curiosity, was the same effect observed during periods of very light precipitation (i.e., drizzle)?

L275: If I am interpreting the figure correctly, the jumps noted in the manuscript at the 0.1 and 0.4 particle count and rain intensity threshold do not exist. If these values are referring to the y-axis, the step function does not touch them in either of the top two panels in Figure 10. If they are referring to the x-axis, no change in the step function exists at 0.1 or 0.4 in the top panel, and in the bottom panel only a marginal, steady change occurs at 0.1 while a small peak occurs at 0.4. If I have interpreted the figure or text incorrectly, it should be made more clear what features are being referred to, and what exactly ‘thresholds’ means here.

L277: As in the comment above, no gradient in availability fraction occurs around 1; counting logarithmic ticks, it looks like this actually occurs around 0.5 J m-2 hr-1. Am I missing something?

L285: Are ‘fraction values’ and ‘availability’ synonymous? If so, limiting the number of names by which that quantity is referenced would improve readability.

L285: What is meant by ‘artificial’ data? Could you provide an example(s)?

L304-305: Should ‘Figure 11a and 11b’ be ‘Figure 11b and 11c’? Subplots (b) and (c) are the ones that show MWR data.

L308: Which of T1-T5 correspond to before, during, and after LLJ formation? From the images alone it is hard to tell. It looks like maybe T3 and T5 are during, so T2 and T4 might be before, but what is T1? It would be good to state outright so readers do not assume incorrectly.

L317: It is interesting that despite the SCADA data being unavailable/showing no generated power, there are wakes evident in Figure 12f and 12k. Since these turbines are within the authors’ “home domain”, so to speak, maybe more information is available to describe why this was the case?

L331: Should 2025-06-06 be 2025-06-06? It is labelled as such in Figure 13.

L333: As above, should 2025-07-23 be 2025-07-22?

L334: “A more broadened and deepened velocity deficit” – relative to what? The right center figure looks smaller and less intense than the right top figure, so this description is a bit confusing.

L349: If certain wind directions were indeed to be omitted to avoid radar artifacts, how seriously would the climatological representativeness assumption be violated? Since stability is, as is written here, highly dependent on wind direction in this geographic location, fully removing some wind directions could consequently remove certain conditions from the dataset, could it not? Might this make it less true to the Northern Germany climate? A statement or statistic on this sensitivity would be helpful.

L385-386: “temporal resolution of the radars” – this is not touched upon in the manuscript or in Table 2, so some information on this would be helpful. What was the revisit time for a full volume, how long did a single scan take, what was the azimuthal scan speed, etc. It could be added to Table 2, since information on the radars is conspicuously missing there.

Figures

F1: The bird’s eye view is denoted as top left, but there is no top or bottom or row structure in this figure. Also, the caption mentions plural disdrometers, but only one disdrometer location is illustrated on the map.

F2a: It appears Channel A may be largely overlapped by Channel B; is there some way to also show the values obtained by Channel A more visibly? Perhaps by splitting the two traces?

F2c: Only a small portion of this colorbar is used, and negative values do not appear anywhere. Can the color axis be shrunken to reveal more detail?

F2: Some explanation of the reason for gaps in some traces would be helpful. Heavy precipitation, lack of tracers, instrument downtime, etc.

F2: The x-axis label lists [Month in 2025], though the data extend into 2026. Could the axis ticks be changed to include month and year to avoid this issue?

F2: What does it mean for DDR scans to have been “derived from a single grid point encompassing the location of the lidar”? How can a single point encompass something? Maybe it would be better to explain this from the perspective of the DDR interpolated Cartesian grid spacing and the physical volume it represents with respect to the location of the lidar and its measurement volume..

F4: Other images with panels are lettered, so this one should be as well. The caption can be augmented that way, so that all necessary information in the figure can be obtained from the caption alone.

F5: Since panels (a) and (b) have parentheses in the figure, they should in the caption as well.

F6: This figure appears before being referenced in the text, unless I missed an earlier reference somewhere.

F6: Could grid lines be added, or made darker?

F7: This figure appears before being referenced in text.

F9: Plots (a)-(f) on the left and the corresponding set on the right are not discernible as being from the warm season and cold season by the figure alone and rely on the caption, which could be misleading. Could they be re-labeled to avoid confusion?

F9: The definition of ‘samples’ and ‘availability’ is a bit hard to follow in this caption as well. Are ‘samples’ bins, rays, sectors, volumes? A singular statement on how ‘availability’ is defined would also be helpful.

F10: Lettered subpanels (a)-(c) are referenced in the caption, but do not exist in the figure itself.

F10: The meaning of the colored vertical lines is a bit difficult to decipher. What do these signify? They’re referred to as ‘exact quantity thresholds’, but what of? How were these thresholds computed?

F10: It seems a bit odd to conduct most of the study’s comparison with the fine grid and therefore focus the discussion on the fine grid, but show these stepwise curves for the coarse grid. Are the corresponding plots for the fine grid available, even if they show similar results?

F10: The word ‘both’ is used, even though there are 3 subplots.

F11: It looks like the LLJ max is centered around 400 m AGL, above the fine grid. Is there a reason the coarse grid was not used to corroborate these measurements, if it reached that altitude?

F11: I would add titles to each subplot, even if they are defined in the caption, so that the figure can be visually parsed on its own.

F11b: As with F2c, only a small portion of the colorbar is used. Could a different colorbar provide more detail?

F13: I would add letters to each subpanel.

F13: Where is the red dot on the left center overhead figure?

Tables

T1: The caption and data are not intuitive to follow. For coverage factors greater than or equal to 8, indicating data coverage of 80-90%, what does the listed percentage (much lower than 80%) mean? Does the table suggest that, e.g. 10.29% of radar volumes met the 80% coverage threshold at 60 m in the coarse grid? The phrase “coverage out of the full spatial domain” is a little bit difficult to follow or conceptualize; giving more explanation or a hypothetical example would be useful.

T2: The Resolution/notes column for WiValdi Met Mast/Wind Vane/Vane-2: 88 ends with a trailing comma. Was information left out?

T2: Is there a reason the radars and their associated spatiotemporal frequencies are not included in this table?

T2: What are the sampling frequencies/temporal resolution of the sonic anemometers, cup anemometers, and wind vanes?

There are a number of errors in conventions resulting in incorrect grammar, missing information, or unclear points. The manuscript should be re-proofed in full to make sure these types of errors are caught.

L14 (these results performance verification)

L21-22 (This gap in understanding grows increasing critical to address as wind)

L32 (Doppler wind lidars has been used) - subject/verb agreement

L40-41 (under a rapid return times) - articles/pluralization

L60 (during AWAKEN, however, those) - comma splice

L63 ([an understanding of?] how DDR performance varies with the prevailing atmospheric regime is necessary)

L153 (a range gate lengths) - articles/pluralization

L258-261 (Seasonally…effect) - unclear/hard to follow, are words or punctuation missing?

L264 (to require requires hydrometeors)

L267-268 (As such, Figure 10…droplets) - unclear/hard to follow, are words missing? The sentence could be rephrased to be clearer.

L272 (approximately 50% and 70%) - is this 50% for one radar and 70% for the other, or should the range read 50% [to] 70%?

L284 (condition) - should this be [condition’s]?

L287-288 (Since…turbine blockage) - a verb is missing somewhere in this passage

L290 (so it recommended to)

L324 (though they be offset) - should this be [though they should be offset]?

L360-361 (overestimation in DDR samples were very limited) - subject-verb agreement

L391 (Further, the DDR method is that strong)

L407-408 (One strong…radar scans) – sentence fragment

L442 (the were)

L444 (lidar comparisons of is intentionally) - is something missing after of?

Make sure acronym usage is consistent across the manuscript.

L437-438 (which the dual-Doppler radar…scanning lidar and DDR)

Make sure unit conventions are consistent across the manuscript.

m vs. m AGL

m/s vs. m s-1, sec. vs. s (especially check Table 2)

Other notes

L134: The D in Dual does not need to be capitalized

L151: I didn’t see the acronym VAD defined earlier in the manuscript.

L174: Bulk is not capitalized here but is in the caption of Figure 5, but not here. I’ve seen both styles, but a consistent one should be chosen.

L198: There are two spaces between Figure and 6.

L331-335: There is a tense shift in this paragraph from present (shows) to past (captured). I would, upon completion of the manuscript, read over to make sure a consistent, intentional tense is used.

Reply
Citation: https://doi.org/10.5194/wes-2026-114-RC2

Arianna Marie Jordan, Lin-Ya Hung, Gerrit Wolken-Möhlmann, and Julia Gottschall

Viewed

Total article views: 420 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
255	112	53	420	68	61

HTML: 255
PDF: 112
XML: 53
Total: 420
BibTeX: 68
EndNote: 61

Views and downloads (calculated since 28 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	94	29	6	129
Aug 2026	114	50	10	174
Sep 2026	47	33	37	117

Cumulative views and downloads (calculated since 28 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	94	29	6	129
Aug 2026	114	50	10	174
Sep 2026	47	33	37	117

Viewed (geographical distribution)

Total article views: 370 (including HTML, PDF, and XML) Thereof 370 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 10 Sep 2026

Short summary

Wind energy needs accurate airflow tools around turbines. We tested dual-Doppler radar at a research wind farm in Germany alongside lidar (a point-based wind sensor). The method matched lidar well at most heights and mapped wind shadows behind turbines, including wake meandering. Data quality depended on rainfall, with rainy periods giving the best coverage and fog the worst. This information is useful for implementation of dual-Doppler radar as a wind measurement tool for the industry.


Total:	0
HTML:	0
PDF:	0
XML:	0