Transportation noise assessment: road, rail and the differences that matter
Transportation · SoundPLAN Asia · · 8 min read
Transport schemes generate the largest noise-exposed populations of any project type, and they are assessed under more scrutiny than almost anything else, because the affected people are numerous, identifiable and usually already there. Road and rail get grouped together as "transportation noise", which obscures the fact that their sources behave differently enough to change what mitigation works.
Road: a line source governed by three inputs
Road traffic noise is effectively a continuous line source, and its emission is dominated by three things:
- Flow and composition. Total volume matters less than the heavy vehicle percentage. A small proportion of heavy vehicles can dominate the total, which is why an assessment using an undifferentiated daily total is unreliable regardless of how good the propagation model is.
- Speed. The relationship is not linear, and it differs between propulsion and tyre noise. Above roughly 50 km/h tyre-road interaction generally dominates for light vehicles, which is why speed reduction is effective on some links and disappointing on others.
- Road surface. Surface type and its condition are a real variable, not a detail. A surface specified as low-noise and assessed at its as-new performance will not deliver that performance in year eight, and an assessment that quietly assumes it does is overstating the scheme.
Rail: discrete events with a different source geometry
Railway noise is a series of pass-by events rather than a continuous flow, and the source sits at several heights at once:
- Rolling noise from wheel-rail interaction, low down near the railhead. This usually dominates at conventional speeds, and it depends heavily on wheel and rail roughness - a maintenance variable, not a fixed property of the line.
- Traction and equipment noise, higher on the vehicle.
- Aerodynamic noise at high speed, and pantograph noise well above the track.
- Structure-radiated noise on bridges and viaducts, which frequently needs separate treatment and is a common omission.
- Curve squeal on tight radii and braking noise at approaches - localised, tonal, and disproportionately complained about.
Because rail is event-based, an equivalent continuous level alone can under-represent the experience. Many frameworks also require maximum levels or event counts, particularly at night, where a small number of pass-bys can dominate sleep disturbance while barely moving the average.
Why source height decides whether barriers work
A barrier works by breaking the line of sight between source and receiver. That makes source height the governing variable.
Rolling noise sits near the railhead, so a modest barrier close to the track can be very effective against it. Pantograph noise on an electrified line sits several metres higher and is far less affected by the same barrier, which is why high-speed schemes sometimes deliver less than the modelling implied if the pantograph contribution was underweighted. On roads the source is low, so barriers are generally effective - but the receiver height matters just as much, and a barrier that protects a ground floor may do little for a fourth-floor bedroom.
Two geometry issues are underestimated everywhere: end diffraction, where sound wraps round the end of a barrier that was not extended far enough beyond the exposure, and gaps at the base, which degrade performance disproportionately. Both are model-able, and both are cheaper to get right on a drawing than on site.
The output that should drive the decision
For a transport scheme, contours are a poor decision tool because they do not say how many people are affected. Exposure counts do.
The practical consequence is that the assessment should compare options, not report a single scheme. Alignment, vertical profile - cutting, embankment or viaduct - barrier heights and lengths, speed limits and surface specification all produce different exposure counts, and the differences between them are usually far larger than the uncertainty in any one prediction. An assessment that models only the preferred option has skipped the part where the modelling could have changed something.
Six things that predict a defensible assessment
- Traffic data broken down by vehicle category and period, not a single daily figure
- Road surface type and assumed age stated explicitly
- For rail, the rolling stock types, speeds and assumed rail roughness condition
- Bridges and viaducts treated as their own source, not as ordinary track
- Receptors assessed per floor on multi-storey facades
- At least one alternative modelled, so the recommendation has something to be better than
SoundPLAN is used for road, rail, aircraft and industrial noise across Asia and the Middle East, with 50+ international calculation standards including ISO 9613-2 and CNOSSOS-EU. Tell us which national method your scheme must be reported against and we will confirm how it is handled.