the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 12 Aug 2024
Enhancing turbulent fluctuation measurement with tailored wind lidar profilers
Abstract. This study separately investigates the impact of an enhanced sampling rate and reduced probe length on turbulent measurements using the Vaisala WindCube v2.1 lidar profiler, in comparison to the commercially configured WindCube v2.1. In the first experiment, a tailored lidar sampled four times faster than the standard setup. In the second experiment, a tailored lidar employed a 15 m probe length, compared to the commercial configuration’s 23 m. The study offers a detailed analysis of how these changes affect various aspects of wind measurement, including mean wind speed, standard deviation, velocity spectra, noise level, integral length scale, and dissipation rate. Increasing the sampling rate improves turbulence measurement without affecting mean wind speed measurement. However, a slight reduction in data availability was observed compared to the commercial configuration. Reducing the probe length results in higher standard deviation values compared to the commercial configuration, but this comes at the expense of increased noise levels, making it unclear whether the higher standard deviations are due to the energy of smaller eddies or noise. Additionally, the reduced probe length configuration exhibited a high bias in mean wind speed measurement and had a limited impact on other turbulence metrics. These findings suggest that the best improvement for turbulence measurement with the WindCube lidar profiler is achieved through an increased sampling rate.
- Preprint
(1760 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed (peer review stopped)
-
RC1: 'Comment on wes-2024-93', Anonymous Referee #1, 19 Aug 2024
General comments:
In the the manuscript, two modified prototype versions of the WindCube vertical profiling pulse wind lidar are compared to the default WindCube v2.1. The authors look at one version featuring four times faster sampling rates with accordingly shorter accumulation times and a second version with reduced pulse length resulting in shorter probe volumes. They assess the impact of the modifications on data availability, mean wind speed, noise level, standard deviations of turbulent velocity fluctuations, integral length scale of turbulence, velocity spectra and dissipation rate.The study is of high importance for the field because profiling wind lidar suffer from their limited ability to measure turbulence accurately. Therefore, every effort to increase this ability is sought after. The experimental setup, modifications to the lidar units, and methods for assessement of the results are overall well described.
In the overall acceptable introduction and the good and well-written description of the methods, Thiebaut et al. should be more accurate in some of their claims (see specific comments). The experiment with the WindCube with reduced probe length is only 4 days long. This very short trial length minimizes the value of the conclusions drawn from it and is a major drawback for the study. The manuscript would benefit from a more thorough explanation of the theory behind the tested modifications. For example, it remains unclear how the increased sampling rate of the modified lidar leads to a reduction of noise, although every LoS measurement will have a higher potential for noise if accumulation times are reduced. The conclusions section would benefit from some more reflection on theoretical considerations and practical implications.
I recommend to reconsider the manuscript for publication in WES after major revisions.
Specific comments:
l. 26: Considering a beam diameter of 3cm, only a probe length of around 17km fills up a probe volume 12 cubic meters. Thus, the statement of lidars providing averages of "up to several dozen cubic meters" appears to be a vast exaggeration. Such a statement should be backed up by references.l. 38: The ZX 300, one of the two market-leading wind profilers, requires one second to complete a full scanning circle. The statement that "Lidar profilers require several seconds to complete a full scanning circle" is therefore wrong and must be corrected.
l. 80: "the WindCube lidar collects data at each location in 1 second" is wrong because it samples 5 beams within 4 seconds. So, on average the data accumulation time per beam direction cannot be higher than 0.8 sec.. And in practice some dead time for swinging the beam must be accounted for.
l. 89: The authors should not report an "improvement" in the method section. Instead they can write e.g., about the "modification".
l. 89: Accumulation time is reduced by 70% to which value (0.3*0.8s=0.24s)? In line 88 1Hz and 4Hz, although there are five beams to be sampled (1Hz*5=>0.2s per beam). So, either the 4Hz or the 70% reduction are wrong.
l. 93: Please describe why you chose 30-min intervals, instead of 10-minutes commonly used in wind industry.
l. 109: It is unclear why a "50% reduction in pulse duration" leads "to a reduction in the probe length from 23m to 15m". Due to the linear relationship, the reader would expect that a 33% reduction in pulse duration leads to this reduction in probe length. Please clarify.
l 111: The short trial duration of 4 days is a major drawback of the study. Please explain why it was not possible to perform a longer experiment and how the short trial durations impeded the study.
2.1.1 & 2.1.2: Please elaborate on the effect of both modifications from a theoretical standpoint. Why did you choose these modifications. What is the trade-off between duty cycle, accumulation time and sampling rate? What are the technological limitations? Why is the default configuration of the WindCube different?
l. 125: Please mention the special role of the vertical beam for this study. Many of the results shown are based on the fifth beam only.
l. 129: Mean velocity does not have a standard deviation, but the fluctuations superimposed on the mean velocity and there is no mean velocity across the wind propagation. Please reformulate.
l. 130: Please add an Eq. for the velocity rotation procedure.
l. 176: Noise is not(!) due to relative motion between the source and the observer, otherwise it would be the signal. Please rewrite.
l. 239: The absence of reduction in 100% data availability appears to be insignificant and conclusions on the effect of reduced probe length cannot be drawn from the available data. A four day period with 100% availability is not representative for a commercial measurement campaign in which the data availability of the commercial lidar is <100%. This must be reflected on in the text. What are your expectations from theory? Does the increased sampling rate reduce the duty cycle, so that there is less total measurement time?
l. 242: Please name the first and last measurement height (40m and 200m) to help the reader. Please consider to show how the 0.5% reduction is distributed along the vertical profile.
ll. 244-249: IEC (2017) does not prescribe any "Best Practice" criteria. Where do the 1% and 1.5% thresholds come from?
l. 253: How does the relative deviation of 1.3% between prototype and commercial unit relate to typical measurement deviations between two random commercial lidar units? Can the results be attributed to the modifications with certainty?
ll. 266-268: Slopes below unity and positive intercepts are always expected for standard linear regression with randomly scattered data. Consider using Deming regression to assume identical random errors for the prototype and commercial lidar.
Fig. 6: Please rearrange the box plots to maintain the u,v,w/b5 order.
Fig. 8: Consider the add a titles to the figure.
Fig. 9: The legend entries seem to be wrong. "Fit - Com. lidar" ranges all the way to the highest frequency.
Fig. 9: Consider to merge the two sub figures into one, since only the fitted line is different and could be compared more easily in one plot.
Fig. 9: The distribution of frequency bins is different for the com. and the pro. lidar. Consider to create logarithmically spaced frequency bins for both lidars and average the spectral energy within these bins. This would make the comparison easier.
l. 295: Why are the spectral plots for reduced probe length not shown? Same for l. 305.
3.4: The noise assessment would benefit from a figure.
Fig. 11: It is unclear why the "Laminar flow" curve is considered laminar flow although it contains the highest total spectral energy of all four examples. Where would laminar flow come from in field experiments?
ll. 342-347: 0.5% reduction in data availability are not a "slight" reduction. And 1-2% deviation in mean wind speed are significant and in the order of magnitude of the total measurement uncertainty of profiling wind lidars. The authors clarify that a deviation in wind speed beyond 1.5% prohibits certification (l. 248). Here they call it a "very slight bias". I understand that the manuscript focuses on turbulence estimates, still the deviations in mean wind speed must be analyzed equally carefully.
l. 350: The WindCube with increased sampling rate has a probe length of 28m that effectively acts as a low-pass filter. Still, the authors assume that an increase in sampling rate from 0.25Hz to 1Hz per beam leads to a "greater sensitivity to smaller-scale fluctuations". With a mean wind speed of e.g. 7m/s the sampling rates correspond to eddy sizes of 28m and 7m respectively. The authors have to explain why they still think, that the higher standard deviation for all three turbulent velocity components can be caused by the beneficial aspects of an increased sampling rate.
l. 399: As commented before, the effect of increased sampling rate on mean wind speed appears significant and also the impact on data availability seems to be stronger than "slight".
Technical corrections:
l. 71: "the mea(n) wind speed"l. 147: "pair of (parallel) beams"
l. 167: "each...subset( )"
l. 259: "illustrated (further / in the form of scatter plots) in Fig. 4."
l. 274: "the vertical (profile) of the mean..."
l. 276: "Notably, the (deviations) were..."
l. 283: "scheme(s)"
l. 193: "three (times) higher"
l. 331: "configuration (with) increased..."
l. 334: "not exceed(ing) 5.1m/s"
l. 381: "potential(ly) improved"
Citation: https://doi.org/10.5194/wes-2024-93-RC1 - RC2: 'Comment on wes-2024-93', Anonymous Referee #2, 06 Nov 2024
Status: closed (peer review stopped)
-
RC1: 'Comment on wes-2024-93', Anonymous Referee #1, 19 Aug 2024
General comments:
In the the manuscript, two modified prototype versions of the WindCube vertical profiling pulse wind lidar are compared to the default WindCube v2.1. The authors look at one version featuring four times faster sampling rates with accordingly shorter accumulation times and a second version with reduced pulse length resulting in shorter probe volumes. They assess the impact of the modifications on data availability, mean wind speed, noise level, standard deviations of turbulent velocity fluctuations, integral length scale of turbulence, velocity spectra and dissipation rate.The study is of high importance for the field because profiling wind lidar suffer from their limited ability to measure turbulence accurately. Therefore, every effort to increase this ability is sought after. The experimental setup, modifications to the lidar units, and methods for assessement of the results are overall well described.
In the overall acceptable introduction and the good and well-written description of the methods, Thiebaut et al. should be more accurate in some of their claims (see specific comments). The experiment with the WindCube with reduced probe length is only 4 days long. This very short trial length minimizes the value of the conclusions drawn from it and is a major drawback for the study. The manuscript would benefit from a more thorough explanation of the theory behind the tested modifications. For example, it remains unclear how the increased sampling rate of the modified lidar leads to a reduction of noise, although every LoS measurement will have a higher potential for noise if accumulation times are reduced. The conclusions section would benefit from some more reflection on theoretical considerations and practical implications.
I recommend to reconsider the manuscript for publication in WES after major revisions.
Specific comments:
l. 26: Considering a beam diameter of 3cm, only a probe length of around 17km fills up a probe volume 12 cubic meters. Thus, the statement of lidars providing averages of "up to several dozen cubic meters" appears to be a vast exaggeration. Such a statement should be backed up by references.l. 38: The ZX 300, one of the two market-leading wind profilers, requires one second to complete a full scanning circle. The statement that "Lidar profilers require several seconds to complete a full scanning circle" is therefore wrong and must be corrected.
l. 80: "the WindCube lidar collects data at each location in 1 second" is wrong because it samples 5 beams within 4 seconds. So, on average the data accumulation time per beam direction cannot be higher than 0.8 sec.. And in practice some dead time for swinging the beam must be accounted for.
l. 89: The authors should not report an "improvement" in the method section. Instead they can write e.g., about the "modification".
l. 89: Accumulation time is reduced by 70% to which value (0.3*0.8s=0.24s)? In line 88 1Hz and 4Hz, although there are five beams to be sampled (1Hz*5=>0.2s per beam). So, either the 4Hz or the 70% reduction are wrong.
l. 93: Please describe why you chose 30-min intervals, instead of 10-minutes commonly used in wind industry.
l. 109: It is unclear why a "50% reduction in pulse duration" leads "to a reduction in the probe length from 23m to 15m". Due to the linear relationship, the reader would expect that a 33% reduction in pulse duration leads to this reduction in probe length. Please clarify.
l 111: The short trial duration of 4 days is a major drawback of the study. Please explain why it was not possible to perform a longer experiment and how the short trial durations impeded the study.
2.1.1 & 2.1.2: Please elaborate on the effect of both modifications from a theoretical standpoint. Why did you choose these modifications. What is the trade-off between duty cycle, accumulation time and sampling rate? What are the technological limitations? Why is the default configuration of the WindCube different?
l. 125: Please mention the special role of the vertical beam for this study. Many of the results shown are based on the fifth beam only.
l. 129: Mean velocity does not have a standard deviation, but the fluctuations superimposed on the mean velocity and there is no mean velocity across the wind propagation. Please reformulate.
l. 130: Please add an Eq. for the velocity rotation procedure.
l. 176: Noise is not(!) due to relative motion between the source and the observer, otherwise it would be the signal. Please rewrite.
l. 239: The absence of reduction in 100% data availability appears to be insignificant and conclusions on the effect of reduced probe length cannot be drawn from the available data. A four day period with 100% availability is not representative for a commercial measurement campaign in which the data availability of the commercial lidar is <100%. This must be reflected on in the text. What are your expectations from theory? Does the increased sampling rate reduce the duty cycle, so that there is less total measurement time?
l. 242: Please name the first and last measurement height (40m and 200m) to help the reader. Please consider to show how the 0.5% reduction is distributed along the vertical profile.
ll. 244-249: IEC (2017) does not prescribe any "Best Practice" criteria. Where do the 1% and 1.5% thresholds come from?
l. 253: How does the relative deviation of 1.3% between prototype and commercial unit relate to typical measurement deviations between two random commercial lidar units? Can the results be attributed to the modifications with certainty?
ll. 266-268: Slopes below unity and positive intercepts are always expected for standard linear regression with randomly scattered data. Consider using Deming regression to assume identical random errors for the prototype and commercial lidar.
Fig. 6: Please rearrange the box plots to maintain the u,v,w/b5 order.
Fig. 8: Consider the add a titles to the figure.
Fig. 9: The legend entries seem to be wrong. "Fit - Com. lidar" ranges all the way to the highest frequency.
Fig. 9: Consider to merge the two sub figures into one, since only the fitted line is different and could be compared more easily in one plot.
Fig. 9: The distribution of frequency bins is different for the com. and the pro. lidar. Consider to create logarithmically spaced frequency bins for both lidars and average the spectral energy within these bins. This would make the comparison easier.
l. 295: Why are the spectral plots for reduced probe length not shown? Same for l. 305.
3.4: The noise assessment would benefit from a figure.
Fig. 11: It is unclear why the "Laminar flow" curve is considered laminar flow although it contains the highest total spectral energy of all four examples. Where would laminar flow come from in field experiments?
ll. 342-347: 0.5% reduction in data availability are not a "slight" reduction. And 1-2% deviation in mean wind speed are significant and in the order of magnitude of the total measurement uncertainty of profiling wind lidars. The authors clarify that a deviation in wind speed beyond 1.5% prohibits certification (l. 248). Here they call it a "very slight bias". I understand that the manuscript focuses on turbulence estimates, still the deviations in mean wind speed must be analyzed equally carefully.
l. 350: The WindCube with increased sampling rate has a probe length of 28m that effectively acts as a low-pass filter. Still, the authors assume that an increase in sampling rate from 0.25Hz to 1Hz per beam leads to a "greater sensitivity to smaller-scale fluctuations". With a mean wind speed of e.g. 7m/s the sampling rates correspond to eddy sizes of 28m and 7m respectively. The authors have to explain why they still think, that the higher standard deviation for all three turbulent velocity components can be caused by the beneficial aspects of an increased sampling rate.
l. 399: As commented before, the effect of increased sampling rate on mean wind speed appears significant and also the impact on data availability seems to be stronger than "slight".
Technical corrections:
l. 71: "the mea(n) wind speed"l. 147: "pair of (parallel) beams"
l. 167: "each...subset( )"
l. 259: "illustrated (further / in the form of scatter plots) in Fig. 4."
l. 274: "the vertical (profile) of the mean..."
l. 276: "Notably, the (deviations) were..."
l. 283: "scheme(s)"
l. 193: "three (times) higher"
l. 331: "configuration (with) increased..."
l. 334: "not exceed(ing) 5.1m/s"
l. 381: "potential(ly) improved"
Citation: https://doi.org/10.5194/wes-2024-93-RC1 - RC2: 'Comment on wes-2024-93', Anonymous Referee #2, 06 Nov 2024
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 225 | 68 | 16 | 309 | 13 | 13 |
- HTML: 225
- PDF: 68
- XML: 16
- Total: 309
- BibTeX: 13
- EndNote: 13
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1