Not a precise science, part 3

(… continued from Not a precise science, part 2, posted on 9 Oct 2015).

The effect that wind speed and direction can have upon train pass-by noise levels measured at a distance from the track is well illustrated by another scatter diagram that is presented in the Environmental Statement (ES), and which is reproduced below (see footnote 1).

(Source: HS2 Ltd)

(Source: HS2 Ltd)

The data underlying this diagram were obtained from the same measurement programme on TGV Atlantique train pass-bys that I mentioned in part 1, but appear, on the face of it, to exhibit wider noise level variations than the scatter diagram that I reproduced in part 1. The two diagrams are, however, not directly comparable. In the first place the two employ different quantities for the abscissas: in the current diagram this is distance from the track, whereas the diagram in part 1 uses noise level. This current diagram also removes the train pass-by speed variability by “normalising” all values to that resulting from a pass-by speed of 300km/hr. Also, very significantly, the diagram in part 1 employs only data captured downwind, or in “positive wind”, conditions, where the wind direction is from noise source to receptor. The current diagram displays both downwind measurements, indicated by small squares, and upwind data, indicated by plus signs, so that we can compare the effect of wind direction.

The ES advises that “the data for upwind and downwind conditions starts to segregate” at “larger” distances from the noise source (see footnote 2), and this effect can be most clearly seen on the scatter diagram at a distance of 200 metres, which is the greatest distance for which both upwind and downwind data have been plotted – I have marked the relevant line of data with a red arrow (the abscissa is drawn with a logarithmic scale).

The ES does not provide the raw data values, nor has a regression analysis been carried out in this case, but I have estimated the noise level corresponding to each of the points plotted along the two-hundred metre line from the graphical representation. This “reconstructed” data sample allows the assessments presented in the ES to be judged.

The ES declares that “the spread observed under downwind conditions was smaller than that for upwind conditions” (see footnote 2). My data sample confirms this to be the case; the downwind spread is 14dB and the upwind spread is 17dB. The overall spread between the loudest downwind measurement in my sample and the quietest upwind measurement is a remarkable 25dB, so petitioners have been spot on when noting that wind direction can have a very noticeable impact on noise level.

The ES also observes that the “measured mean difference in the TGV data … due to wind direction at a distance of 200m from the track is over 10dB” (see footnote 3). This comment applies specifically to when the train noise is expressed as a pass-by SEL (sound exposure level), but my data sample shows it also to be the case when the noise is measured as a LpAF,Max; my calculation from values estimated from the scatter diagram is that the upwind mean value is 68.9dB LpAF,Max and the downwind mean is 79.4dB LpAF,Max (see footnote 4). The ES adds (see footnote 5):

“Differences of 15dB have also been observed in other research at receivers 1km away from a source due to such effects.”

So perhaps this gives some confirmation that the “segregation” of downwind and upwind measurements becomes greater as the observer moves even further away from the source.

The claim in the ES that “the HS2 prediction method is representative of downwind conditions” (see footnote 6) appears to be borne out by the curves plotted on the scatter diagram in red. The unbroken red line shows the results obtained by applying the HS2 prediction method over similar source to receptor paths. The broken line is also derived from this method, but omitting any reduction in noise level due to ground absorption; the ES advises that this is the practice when employing the propagation model for paths where screening, such as earth bunds or noise barriers, is present near to the source (see footnote 7).

Both of the red curves cut the 200-metre abscissa above 80dB, whereas the mean that I have estimated for the downwind data is 79.4dB. There are, of course, downwind data points with levels higher than the prediction, but as we saw in part 1 this can be expected when propagation modelling, and the excursions from the mean seen here are no greater than we encountered in part 1.

I haven’t calculated the sample means for the 400-metre and 800-metre abscissas, but the broken red line is clearly above the mean at the former and close to the mean at the latter. The unbroken line is around the mean at 400 metres, but drops well below the mean at 800 metres, so it would appear that the HS2 propagation model only really holds good at extended distances from the source if the calculation of ground absorption loss is omitted.

The ES contains an important caveat to this consideration of wind effects (see footnote 6):

“… the spread around the predictions can be partly attributed to variations in the sound emission levels across trains and measurement sites.”

In other words, it is not possible to separate out wind effects from other known unknowns when carrying out tests, so the above scatter diagram does not solely reflect the impacts of wind direction. Nevertheless, it is obvious that wind direction is a significant determinate in the variations in noise level that may be recorded at receptors.

(To be continued …)


  1. This diagram is Figure 18 in Annex D2 to Appendix SV-001-000 in Volume 5 of the ES.
  2. See paragraph 1.3.24 in Annex D2 to Appendix SV-001-000 of the ES.
  3. See paragraph 1.3.25 in Annex D2 to Appendix SV-001-000 of the ES.
  4. For an explanation of the concept of sound exposure level please refer to my blog Suffering from exposure (posted 18 Jun 2011).
  5. See paragraph 1.3.26 in Annex D2 to Appendix SV-001-000 of the ES.
  6. See paragraph 1.3.28 in Annex D2 to Appendix SV-001-000 of the ES.
  7. See paragraphs 1.3.16, 1.3.17 and 1.3.27 in Annex D2 to Appendix SV-001-000 of the ES.



2 responses to this post.

  1. Posted by LesF on October 15, 2015 at 9:13 am

    You haven’t yet mentioned the fact that the design speed for HS2 is 400km/h, while the intention is to run the trains at 360km/h initially, compared with 300km/h in the graph you show. The differences may seem marginal but wind resistance increases as the CUBE of speed. Your own blogs show that aerodynamic noise and pantograph noise increase more than pro-rata (I’m tempted to use the word “exponentially” but I’m told that isn’t right) therefore we need to know the sound pressure levels at the actual running speed for each location, bearing in mind that trains will slow for tunnels, junctions and towards termini.
    A friend who is an acoustician describes his profession as a “dark art”. This suggests to me that an acoustician can “prove” anything the person paying him wants to be proved.


    • Yes Les you are right to point this out, but I don’t think that the impact of wind direction on sound propagation will be affected in any way by the speed of the train. Whilst the graph, being produced from TGV measurements, is constructed for a speed of 300km/hr I would expect it to be representative of other speeds.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: