Ground motion emissions due to wind turbines: observations, acoustic coupling, and attenuation relationships

Gaßner, Laura; Ritter, Joachim

doi:https://doi.org/10.5194/egusphere-2022-108

Preprints

https://doi.org/10.5194/egusphere-2022-108

Preprints

12 May 2022

| 12 May 2022

Ground motion emissions due to wind turbines: observations, acoustic coupling, and attenuation relationships

Laura Gaßner and Joachim Ritter

Abstract. Emissions from wind turbines (WT) cover a wide range of infrasound and ground motions. Some of these emissions are perceived by local residents and can become a source of disturbance or annoyance. To mitigate such disturbances, it is necessary to better understand and, if possible, suppress WT-induced emissions. Within the project Inter-Wind we record and analyze ground motion signals in the vicinity of two wind farms on the Swabian Alb in Southern Germany, simultaneously with acoustic and meteorological measurements, as well as psychological surveys done by cooperating research groups. The investigated wind farms consist of three and sixteen WTs, respectively, and are located on the Alb peneplain at 700–800 m height, approximately 300 m higher than the municipalities. The main aim is to better understand reasons why residents may be affected from WT immissions, based on interdisciplinary data, methods, and expertise.

Known ground motions include vibrations due to eigen modes of the WT tower and blades, and the interaction between the passing blade and the tower, causing signals at constant frequencies below 12 Hz. In addition, we observe signals in ground motion recordings at frequencies up to 90 Hz which are proportional to the blade-passing frequency. We can correlate these signals with acoustic recordings and estimate sound pressure to ground motion coupling transfer coefficients of 3–16.5. Sources for these emissions are the WT generator and possibly the gearing box. The identification of such noise sources can help to design potential counter-measures in order to increase the public acceptance of WTs. The measurements in the municipalities indicate that WTs are perceived more in the location where the wind farm is closer to the municipality (approx. 1 km). However, there is also a major railway line which produces higher vibration and infrasound signal amplitudes compared to the WTs.

Along the measurement lines the decay rate of the WT-induced ground motions is determined for a damping relation proportional to 1 / r^b. We find frequency-dependent b-values for different scenarios at our geological setting of Jurassic limestone on marl, sandstone, and Quaternary deposits. These damping relationships can be used to estimated emissions in the far- field and to plan mitigation strategies.

Received: 28 Mar 2022 – Discussion started: 12 May 2022

Laura Gaßner and Joachim Ritter

Status: final response (author comments only)

RC1:
'Comment on egusphere-2022-108', Anonymous Referee #1, 14 Jun 2022

The topic addressed by the authors is of scientific and practical interest, as they analyze emissions from the presence and operation of wind turbines.

The authors present the results of their analysis in a clear, concise manner and with the use of valid and known methods.

However, most of the results presented and discussed are based on the analysis of the amplitude decay with distance of the vibrations caused by wind turbines. This has already been presented in numerous publications so I do not find what is novel in this discussion.

What is novel in the paper is on the one hand the relation to the experiences of the residents, and on the other hand the analysis and comparison with acoustic signals. The authors should give much more emphasis and more space to these two aspects. The energy decay should be an adjunct to support this discussion and the conclusions about the problem with the residents.

The language needs to be improved. The structure of the abstract should be improved: both in form and language. In the first three lines there is too much repetition of the word emissions. It is not explained whether the words immission and emission are used interchangeably or whether they have any distinctive meaning. There are problems of wording or use of words such as "measurements in the resident" that need to be improved before the manuscript is accepted. Sometimes it is said (compare figure XX) but it is not said what it is to be compared with

Citation: https://doi.org/10.5194/egusphere-2022-108-RC1
- AC1: 'Reply on RC1', Laura Gassner, 11 Oct 2022
  
  The language needs to be improved. The structure of the abstract should be improved: both in form and language. In the first three lines there is too much repetition of the word emissions. It is not explained whether the words immission and emission are used interchangeably or whether they have any distinctive meaning.
  - We have changed the second sentence of the abstract to reduce the repetition of the word “emissions” and introduced the term “immission” here.
  - It is not apparent to us where the language of our manuscript is suggested to be in need of improvement.
  There are problems of wording or use of words such as "measurements in the resident" that need to be improved before the manuscript is accepted.
  - No such wording as “measurements in the resident” is used in our manuscript.
  Sometimes it is said (compare figure XX) but it is not said what it is to be compared with.
  - We have exchanged “compare” for “see” in some references for clarity.
  
  Citation: https://doi.org/10.5194/egusphere-2022-108-AC1
RC2:
'Comment on egusphere-2022-108', Anonymous Referee #2, 01 Jul 2022

The authors present the results of an experiment aimed at registering and analyzing seismic and acoustic emissions from wind turbines. While the applied methods are not new, the results provide important insights into the seismic and accoustic emissions of wind turbines, especially with regard to the desired expansion of renewable energy sources. The manuscript is clearly structured and presents the measurements as well as the derived results in a comprehensible way. I recommend its publication after minor revisions which take account for following remarks.

Seismic stations were placed in the vicinity of two wind farms in southern Germany. Additional infrasound sensors were installed at selected locations to register the accoustic emissions simultaneously. Using PSD spectra, the authors first analyze the characteristics of the registered seismic signals close to the sources (i.e., wind turbines). They identify frequency-independent signals that are assigned to the natural oscillations of the wind turbines, as well as highfrequency components that correlate with the rotation rate of the wind blades or the generator. The results are presented in an understandable way and are convincingly documented by the illustrations.

In the following section ("Signals at place of immission"), the authors discuss signals which have been recorded in the village of Kuchen, in about 1 km distance from the closest windfarm. Obviously, some residents feel disturbed by the wind farm and one of the project goals lies on how residents experience the WT emissions. Unfortunately, it then turns out that the recordings in Kuchen cannot be evaluated, because here the wind turbine signals are completely covered by the vibrations of a nearby and heavily frequented railroad line. Therefore, I would shorten this section significantly and also remove Fig. 8 since, in my opinion, it does not contribute anything to the targeted questions.

Another aspect that has not been addressed as often as seismic emissions is the measurement of infrasound signals in the vicinity of wind farms. At a few sites, both, seismometers and acoustic sensors were installed, and the authors compare the registered ground motion with the infrasound signal. For certain frequencies, the PSDs of ground motion and infrasound exhibit an astonishing agreement, so that the same sources (blade passing, generator rotation) are supposed for both signal types .

From the amplitude ratio of infrasound and ground motion, the authors derive the so-called coupling transfer coefficient at different frequencies. The results are well documented by text and figures, however, I miss the discussion of the significance of this parameter, which is not very common in seismology. How is the mechanism of the transmission of airborne sound to ground motion, are there any insights on this? Or, which information can be derived from this parameter, what does this coupling factor depend on, e.g. properties of the subsurface? This section should be expanded accordingly.

One of the most important results of this manuscript is the presentation of the b-values describing the spatial decay of the seismic wind turbine emissions. There are already numerous publications on this topic, but the results vary quite strongly depending on e.g. the number of wind turbines, wind farm geometry, or geological conditions. I think each further experiment can help to bring systematics into the results and to better understand the emitted seismic wave field of a wind farm. Table 3 gives a nice compilation of published b-values. However I would also like to see a graphical representation, which enables a better and quick overview.

To calculate the b-values, the authors use both, absolute PSD amplitudes and relative PSD values. The comparison shows that the relative PSD method results in somewhat more stable and more reliable b-values, particularly if the registration of wind turbine emissions is superimposed by transient noise. I would appreciate if this aspect would be discussed in more detail, as it could leed to unified measurement rules with which comparable b-values can be obtained in future.

In the discussion section there are some statements, which are not proofed by the presented results, concerning the range of the emitted wind turbine signals. At page 17, line 289 the authors write that the signals can be observed "over distances of several kilometers". However, no PSD recorded at several km distance from a wind turbine is shown in this manuscript.

some technical comments:

Fig. 1: at WF Lauterstein there are 3 white WT symbols which means that these WTs were "not studied".

How is it possible to exclude them from the measuments?

page 12, line 182: ... The closest recording station (IW02F, 20m distance) ...

due to Fig.1 and Fig.8 the closest station is IW02G, IW02F is at 80m distance to the railroad track

Fig.8: I would sort the legend entries by distance

page 12, line 183: you should use just one unit for the PSD amplitudes throughout the manuscript (same in text and figures): either dB relative to 1 (m/s)**2/Hz or dB relative to 1 (micrometer/s)**2/Hz

page 13, lines 192, 193: I think ground motions and acoustics were registered only at 2 resident sites simultaneously (IMC-B1 and IMC-B2)

page 13, lines 199, 200; "Both, ground motion and acoustic, spectra contain clear signals ..."

move the comma to "Both, ground motion and acoustic spectra, contain clear signals ..."

page 13, lines 217, 218: please add the appropriate unit to the C_AS values

page 14, Fig.10, caption: "Signals with frequencies 32×BPF can be observed in all three data sets."

Please mark the addressed signals as you did in Fig. 9.

page 17, line 286: "At sites with 150 m to 1900 m distance to the nearest WT ..."

Obviously you are talking about WF Lauterstein. You should name it here.

page 18, line 287: "Measurements at different wind farms ...", the same as above. This is WF Tegelfeld.

page 22, lines 329 330: "Thus, recordings with seismometers and high sampling rates may be used in the future to monitor infrasound signals in the near-field of WTs."

You cannot monitor infrasound with seismometers, can you?

Citation: https://doi.org/10.5194/egusphere-2022-108-RC2
- AC2: 'Reply on RC2', Laura Gassner, 11 Oct 2022
  
  In the following section ("Signals at place of immission"), the authors discuss signals which have been recorded in the village of Kuchen, in about 1 km distance from the closest wind farm. Obviously, some residents feel disturbed by the wind farm and one of the project goals lies on how residents experience the WT emissions. Unfortunately, it then turns out that the recordings in Kuchen cannot be evaluated, because here the wind turbine signals are completely covered by the vibrations of a nearby and heavily frequented railroad line. Therefore, I would shorten this section significantly and also remove Fig. 8 since, in my opinion, it does not contribute anything to the targeted questions.
  - We think it is a significant finding that the train signals are of such high amplitudes that the WT emissions are masked but the residents nevertheless claim to be disturbed rather by the WTs than the trains. Therefore, we would like to keep Figure 8 in the manuscript.
  From the amplitude ratio of infrasound and ground motion, the authors derive the so-called coupling transfer coefficient at different frequencies. The results are well documented by text and figures, however, I miss the discussion of the significance of this parameter, which is not very common in seismology. How is the mechanism of the transmission of airborne sound to ground motion, are there any insights on this? Or, which information can be derived from this parameter, what does this coupling factor depend on, e.g. properties of the subsurface? This section should be expanded accordingly.
  - We have added a passage to the manuscript to clarify this point (lines 220-225).
  One of the most important results of this manuscript is the presentation of the b-values describing the spatial decay of the seismic wind turbine emissions. There are already numerous publications on this topic, but the results vary quite strongly depending on e.g. the number of wind turbines, wind farm geometry, or geological conditions. I think each further experiment can help to bring systematics into the results and to better understand the emitted seismic wave field of a wind farm. Table 3 gives a nice compilation of published b-values. However I would also like to see a graphical representation, which enables a better and quick overview.
  - We added a graphical representation to the manuscript (Fig. 16) and simplified Table 3.
  To calculate the b-values, the authors use both, absolute PSD amplitudes and relative PSD values. The comparison shows that the relative PSD method results in somewhat more stable and more reliable b-values, particularly if the registration of wind turbine emissions is superimposed by transient noise. I would appreciate if this aspect would be discussed in more detail, as it could lead to unified measurement rules with which comparable b-values can be obtained in future.
  - We have added a passage to the manuscript to clarify this point (lines 292-294).
  In the discussion section there are some statements, which are not proofed by the presented results, concerning the range of the emitted wind turbine signals. At page 17, line 289 the authors write that the signals can be observed "over distances of several kilometers". However, no PSD recorded at several km distance from a wind turbine is shown in this manuscript.
  - We clarified this point in the manuscript (lines 306-308) and added a reference.
  Fig. 1: at WF Lauterstein there are 3 white WT symbols which means that these WTs were "not studied". - How is it possible to exclude them from the measurements?
  - The adjacent WF is of course responsible for part of the recorded signals, nevertheless, we have no operating data from those turbines. We rephrase to “no operating data available”.
  page 12, line 182: ... The closest recording station (IW02F, 20m distance) ... - due to Fig.1 and Fig.8 the closest station is IW02G, IW02F is at 80m distance to the railroad track
  - This is correct and was changed accordingly.
  Fig.8: I would sort the legend entries by distance
  - Done.
  page 12, line 183: you should use just one unit for the PSD amplitudes throughout the manuscript (same in text and figures): either dB relative to 1 (m/s)**2/Hz or dB relative to 1 (micrometer/s)**2/Hz
  - The spectra have been changed to (m/s)^2/Hz, the spectrograms are kept with (µm/s)^2/Hz to match the respective time series shown in units of µm/s.
  page 13, lines 192, 193: I think ground motions and acoustics were registered only at 2 resident sites simultaneously (IMC-B1 and IMC-B2)
  - The names IMC_B1 and IMC_B2 refer to the indoor and outdoor acoustic measurements, respectively. The names remained the same for each resident site. This is now explicitly explained in the text (lines 196-197).
  page 13, lines 199, 200; "Both, ground motion and acoustic, spectra contain clear signals …" - move the comma to "Both, ground motion and acoustic spectra, contain clear signals ..."
  - Done.
  page 13, lines 217, 218: please add the appropriate unit to the CAS values
  - Done.
  page 14, Fig.10, caption: "Signals with frequencies 32×BPF can be observed in all three data sets." - Please mark the addressed signals as you did in Fig. 9.
  - Done.
  page 17, line 286: "At sites with 150 m to 1900 m distance to the nearest WT …" - Obviously you are talking about WF Lauterstein. You should name it here.
  - Here we discuss both wind farms, closer stations (from 150 m) are at WF Tegelberg, stations up to distances of 1.9 km are at Lauterstein.
  page 18, line 287: "Measurements at different wind farms ...", the same as above. This is WF Tegelfeld.
  - Same as before, both wind farms are discussed.
  page 22, lines 329 330: "Thus, recordings with seismometers and high sampling rates may be used in the future to monitor infrasound signals in the near-field of WTs." - You cannot monitor infrasound with seismometers, can you?
  - Changed to “signals in the frequency range of infrasound”.
  
  Citation: https://doi.org/10.5194/egusphere-2022-108-AC2
RC3:
'Comment on egusphere-2022-108', Anonymous Referee #3, 09 Sep 2022

In the submitted paper with the title "Ground motion emissions due to wind turbines: observations, acoustic coupling, and attenuation relationships", the authors present results of two measurement campaigns addressing the topic of seismic noise signals from wind turbines, which has been discussed in the seismology community for several years. The authors use established methods to present the results of the analyses of seismic signals from WTs in an appropriate manner. Moreover, they combine data from seismic recordings with acoustic data, which is a new approach in this context. Both the structure and the general form of this work are presented in an understandable and comprehensible way. I recommend minor revisions in terms of content and form before publication. Detailed comments, corrections, and recommendations are provided below, which may overlap with the comments of the other two reviewers.

Abstract:

Line 19: typo – … can be used to estimate …

1. Introduction:

Line 25: 10 times the total height of the WT (to distinguish between hub height and total height)

2. Measurements:

Line 66: For each campaign… - Where did you place the seismic sensors inside the WT tower? In the center of the foundation or at the outermost edge inside the tower? With regard to the comparison of amplitudes presented later in the study, this should be explained briefly to ensure comparability.

3. Ground motion signals:

Line 129: I agree that the train traffic signals are dominantly visible below 10 Hz, but they can also be detected at higher frequencies over the entire displayed frequency range (fig 3d).

Line 140: Again, where did you place the sensor? (see comment above)

Fig 5: This figure is a bit confusing and it is difficult to assess the shown results. I recommend using four different colors to represent the four sensor data and removing the distance arrow. Please add the distance information in the legend.

3.1 Signals at place of emission:

Line 161 – 165: How exactly did you determine the maximum value of a 10-min segment? Did you simply identify the maximum absolute value of all values within a 10-min segment? How can you ensure that this value is due to WT vibrations and not to any other noise sources? I recommend to calculate the 95 % or maybe 99 % - percentile of all absolute values within each 10-min time window in order to get a more statistically representative value.

3.2 Signals at place of immission:

Fig 8: Please sort the legend labels by distance between seismic station and WT.

3.3 Comparison to acoustic signals:

Line 211 – 221: This part needs more details for clarification. Again, how did you calculate the max values for seismic and acoustic data (see comment above)? Does your method correspond to the method described in Novoselov et al. 2020 (they used the maximum of the envelopes of the respective signals)? What is the outcome of this analysis and what can we learn from these results?

Fig 11: I recommend to use different colors for each dataset of a given frequency and to include the multiples (32x, 46x and 64x) of the BPF in the legend.

4. Amplitude decay:

Fig 13 and Fig 14: Please include the distance information for each station in the legend.

Fig 16 and related text: I think it doesn’t make sense to calculate amplitude decay curves and b-values from PSD data that are obviously not dominated by WT induced signals. How representative are the mean PSD spectra (Fig 14 right side) for periods without WT operation? I do not believe that this short period of only 2.5 h reliably represents the local noise level at the station sites. I recommend to remove Fig 16 and the corresponding text and to keep Fig 14 to describe why a decay value calculation for wind farm Lauterstein is not reasonable at this point.

Most figures: I recommend to change the y-axis labels to a unified unit of “PSD in db rel. to 1 (m/s)²/Hz” for all figures showing PSD spectra. From a seismological point of view, it would also be nice if you could include the global models for high and low noise after Peterson (1993).

Citation: https://doi.org/10.5194/egusphere-2022-108-RC3
- AC3: 'Reply on RC3', Laura Gassner, 11 Oct 2022
  
  Abstract: Line 19: typo – … can be used to estimate …
  - Done.
  1. Introduction:
  Line 25: 10 times the total height of the WT (to distinguish between hub height and total height)
  - Done.
  2. Measurements:
  Line 66: For each campaign… - Where did you place the seismic sensors inside the WT tower? In the center of the foundation or at the outermost edge inside the tower? With regard to the comparison of amplitudes presented later in the study, this should be explained briefly to ensure comparability.
  - Sensors were placed at the outermost edge of the tower interior. A comment was added to clarify the exact location of the sensor (lines 68, 96, and 142-143).
  3. Ground motion signals:
  Line 129: I agree that the train traffic signals are dominantly visible below 10 Hz, but they can also be detected at higher frequencies over the entire displayed frequency range (Fig 3d).
  - True, the text is changed accordingly.
  Line 140: Again, where did you place the sensor? (see comment above)
  - A comment was added to clarify the exact location of the sensor.
  Fig 5: This figure is a bit confusing and it is difficult to assess the shown results. I recommend using four different colors to represent the four sensor data and removing the distance arrow. Please add the distance information in the legend.
  - Done.
  3.1 Signals at place of emission:
  Line 161 – 165: How exactly did you determine the maximum value of a 10-min segment? Did you simply identify the maximum absolute value of all values within a 10-min segment? How can you ensure that this value is due to WT vibrations and not to any other noise sources? I recommend to calculate the 95 % or maybe 99 % - percentile of all absolute values within each 10-min time window in order to get a more statistically representative value.
  - Yes, maximum amplitudes are taken per 10-min segment. It is a very quiet location and rarely transient signals are seen in the data which would not be connected to the WT operation. We checked that time windows are mostly free of these transient signals. By using approx. 360-550 time segments, we think trends seen are valid even if one or two time windows with signals not connected to WT operation would be included.
  3.2 Signals at place of immission:
  Fig 8: Please sort the legend labels by distance between seismic station and WT.
  - Done.
  3.3 Comparison to acoustic signals:
  Line 211 – 221: This part needs more details for clarification. Again, how did you calculate the max values for seismic and acoustic data (see comment above)? Does your method correspond to the method described in Novoselov et al. 2020 (they used the maximum of the envelopes of the respective signals)? What is the outcome of this analysis and what can we learn from these results?
  - We have added a passage to the manuscript to clarify this point (lines 220-231).
  Fig 11: I recommend to use different colors for each dataset of a given frequency and to include the multiples (32x, 46x and 64x) of the BPF in the legend.
  - Done.
  4. Amplitude decay:
  Fig 13 and Fig 14: Please include the distance information for each station in the legend.
  - Done.
  Fig 16 and related text: I think it doesn’t make sense to calculate amplitude decay curves and b-values from PSD data that are obviously not dominated by WT induced signals. How representative are the mean PSD spectra (Fig 14 right side) for periods without WT operation? I do not believe that this short period of only 2.5 h reliably represents the local noise level at the station sites. I recommend to remove Fig 16 and the corresponding text and to keep Fig 14 to describe why a decay value calculation for wind farm Lauterstein is not reasonable at this point.
  - We decided to move all graphics and text related to amplitude decay estimation at wind farm Lauterstein to the appendix.
  Most figures: I recommend to change the y-axis labels to a unified unit of “PSD in db rel. to 1 (m/s)²/Hz” for all figures showing PSD spectra.
  - Done.
  From a seismological point of view, it would also be nice if you could include the global models for high and low noise after Peterson (1993).
  - The NLNM/NHNM is not shown because it includes only frequencies of up to 10 Hz which is rarely suitable for the frequency range shown in this manuscript.
  
  Citation: https://doi.org/10.5194/egusphere-2022-108-AC3

Laura Gaßner and Joachim Ritter

Viewed

Total article views: 459 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
323	114	22	459	8	9

HTML: 323
PDF: 114
XML: 22
Total: 459
BibTeX: 8
EndNote: 9

Views and downloads (calculated since 12 May 2022)

Month	HTML	PDF	XML	Total
May 2022	94	22	7	123
Jun 2022	57	8	2	67
Jul 2022	22	5	0	27
Aug 2022	14	7	0	21
Sep 2022	24	15	4	43
Oct 2022	37	20	6	63
Nov 2022	19	5	0	24
Dec 2022	6	5	0	11
Jan 2023	9	2	1	12
Feb 2023	15	15	0	30
Mar 2023	12	4	2	18
Apr 2023	9	2	0	11
May 2023	4	4	0	8
Jun 2023	1	0	1

Cumulative views and downloads (calculated since 12 May 2022)

Month	HTML	PDF	XML	Total
May 2022	94	22	7	123
Jun 2022	57	8	2	67
Jul 2022	22	5	0	27
Aug 2022	14	7	0	21
Sep 2022	24	15	4	43
Oct 2022	37	20	6	63
Nov 2022	19	5	0	24
Dec 2022	6	5	0	11
Jan 2023	9	2	1	12
Feb 2023	15	15	0	30
Mar 2023	12	4	2	18
Apr 2023	9	2	0	11
May 2023	4	4	0	8
Jun 2023	1	0	1

Viewed (geographical distribution)

Total article views: 419 (including HTML, PDF, and XML) Thereof 419 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 06 Jun 2023

Short summary

In this work we analyze signals emitted from wind turbines. They induce sound as well as ground motion waves which propagate through the subsurface and are registered by sensible instruments. In our data we observe when these signals are present and how strong they are. Some signals are present in ground motion and sound data, providing the opportunity to study similarities and better characterize emissions. Furthermore, we study the amplitudes with distance to improve the prediction of signals.


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