London Underground has spent approximately £7-10 million a year on rail grinding, partly to remove the rail roughness that turns passing trains into a rumble inside nearby homes, and the grinding budget reached £9 million in 2025/26 (TfL, Tube Noise Update to the Customer Service and Operational Performance Panel 2022; Mayor’s Question Time, 16 July 2026). Underground rail is now a mass exposure: metro services run in 202 cities on 247 networks serving more than 1 billion people (UITP, Global Metro Figures 2024). The engineering fixes work, but unevenly: in one Chinese subway tunnel a steel-spring floating slab cut resonant tunnel-wall vibration by 15 to 22 dB, yet overall tunnel-wall vibration by only 5 dB (Zhou et al., JAABE 2025). We aggregated data from the US Federal Transit Administration, Transport for London, Crossrail, HS2, UITP and peer-reviewed field studies from Sweden, Greece, China, Iran, Malaysia and Poland. Airborne traffic noise is covered in our noise pollution statistics; this roundup follows the energy that travels through soil and structure instead.
TL;DR
- Residential impact threshold for frequent train events: 72 VdB vibration and 35 dBA rumble (FTA Transit Noise and Vibration Impact Assessment Manual 2018).
- Elizabeth line design limit for homes: 40 dB LAmax,S, with extra mitigation effort at 35 dB (Crossrail Learning Legacy 2018).
- HS2 treats indoor ground-borne noise above 45 dB LASmax in homes as a significant effect (HS2 Environmental Statement SV-001-000 2017).
- Freight trains raised the adjusted odds of vibration annoyance 5.3-fold within 100 m, passenger trains 1.7-fold (Maclachlan et al., IJERPH 2018).
- Depressive-symptom scores rose 0.171 points per VdB in a building beside Tehran Metro Line 6 (Sadeghi et al., Noise and Health 2026).
- About 40% of a Beijing heritage district exceeded the 70 dB daytime vibration limit near Metro Line 4 (Su et al., Sensors 2024).
- Rail vibration on a corrugated metro curve varied by up to 20 dB between train passages (Chen et al., Buildings 2025).
- Steel-spring floating slab track cut resonant tunnel-wall vibration by 15 to 22 dB (Zhou et al., JAABE 2025).
- Resilient wheels lowered total tunnel-wall vibration by only 1.6 dB in simulation (Sun et al., Frontiers in Physics 2025).
- Replacing London’s Pandrol Vanguard fasteners with Delkor costs £8,200 per metre (TfL Tube Noise Update 2022).
- Northern line residents lodged more than 280 noise and vibration complaints from 2018 to February 2022 (Barnet Post 2022, reporting TfL figures).
1. The US Yardstick: FTA Impact Criteria in VdB
Ground-borne vibration is the shaking that a train sends through the soil and into a building’s foundations. It arrives with a second effect, ground-borne noise: the low rumble that vibrating walls and floors radiate into a room, which is why a resident can hear a train they cannot see. The US Federal Transit Administration measures the shaking in VdB, a velocity level in decibels referenced to 1 micro-inch per second, and its 2018 manual sets the impact criteria below.
The criteria scale with how often trains pass, not only with how strong each pass is. A home on a line with more than 70 events a day gets a limit 8 dB stricter than one with fewer than 30 (80 - 72 = 8), which is the manual’s way of pricing repetition. The practical stakes show in field data: a six-storey building beside Tehran Metro Line 6 measured 77 dB on its ground floor, 5 dB over the 72 dB residential limit the authors applied (Sadeghi et al., Noise and Health 2026). Light rail sits at the gentler end: in Malaysian measurements, LRT trains produced the lowest vibration of the services tested.
| Metric | Value | Source |
|---|---|---|
| Residential (Category 2) thresholds for frequent / occasional / infrequent events: vibration, and ground-borne noise (rumble heard indoors) | 72 / 75 / 80 VdB, and 35 / 38 / 43 dBA | FTA Transit Noise and Vibration Impact Assessment Manual 2018 |
| Concert halls, TV studios and recording studios: vibration and ground-borne noise thresholds | 65 VdB and 25 dBA | FTA Transit Noise and Vibration Impact Assessment Manual 2018 |
| Event-count definitions: frequent / occasional / infrequent | More than 70 / 30 to 70 / fewer than 30 events per day | FTA Transit Noise and Vibration Impact Assessment Manual 2018 |
| Maximum vibration, ground floor of a building beside Tehran Metro Line 6 | 77 dB | Sadeghi et al., Noise and Health 2026 |
| Malaysian light rapid transit (LRT): peak particle velocity measured 4 to 13 m from the track vs the 0.8 mm/s limit | Most measurements below 0.8 mm/s | Rafidah et al., Scientific Reports 2025 |
Context note: the FTA manual is dated September 2018. Most recent available data: FTA, 2018. Malaysian high-speed trains at 80 to 160 km/h, by contrast, mostly surpassed 0.8 mm/s even at 7 m (Rafidah et al. 2025). Full tables are in the FTA Transit Noise and Vibration Impact Assessment Manual.
2. Design Limits Around the World: Crossrail, HS2, China, Sweden and Malaysia
British rail projects put the rumble at the centre of their limits. Crossrail, now the Elizabeth line, and HS2 both set limits in dB LAmax,S: the highest A-weighted sound level inside a room during a train pass, measured with the “slow” meter setting. HS2 adds a vibration dose value (VDV), a cumulative measure that grows with both the strength and the number of vibration events over a day or night.
The striking pattern is convergence at the quiet end. Concert halls get a 25 dB ground-borne noise limit on Crossrail and on HS2 (Crossrail Learning Legacy 2018; HS2 Environmental Statement 2017), matching the 25 dBA US value in section 1. For homes the two British schemes frame the line differently: Crossrail designed to 40 dB but committed to extra mitigation effort from 35 dB, while HS2 treats 35 dB as the edge of “negligible” and 45 dB as the start of a significant effect.
| Metric | Value | Source |
|---|---|---|
| Crossrail (Elizabeth line) ground-borne noise design criterion for homes, offices, hotels, schools and hospitals, and the level in homes from which reasonable endeavours to add mitigation apply | 40 dB LAmax,S, and 35 dB LAmax,S | Crossrail Learning Legacy, Cobbing, Webb and Bewes 2018 |
| Crossrail criteria for theatres and concert halls / sound recording studios | 25 dB / 30 dB LAmax,S | Crossrail Learning Legacy 2018 |
| HS2 indoor ground-borne noise in homes: negligible / significant effect | Below 35 dB / above 45 dB LASmax | HS2 Environmental Statement SV-001-000, 2017 |
| HS2 residential vibration significance threshold, daytime / night-time | VDV 0.8 / 0.4 m/s^1.75 | HS2 Environmental Statement SV-001-000, 2017 |
| China GB10070-1988 indoor vibration limits for homes near a subway, day / night | 70 dB / 67 dB | Su et al., Sensors 2024 |
| Sweden recommended maximum for railway vibration in homes (weighted RMS) | 0.4 mm/s | Maclachlan et al., IJERPH 2018 |
| Malaysia Department of Environment residential vibration limits, day / night | 0.8 to 1.6 mm/s / 0.4 mm/s | Rafidah et al., Scientific Reports 2025 |
Context note: these units are not interchangeable. VdB is a velocity level, the Chinese dB values are acceleration levels, LAmax,S is indoor sound, and VDV is a dose. Compare thresholds within a system, not across systems. Sources: Crossrail Learning Legacy paper and the HS2 sound, noise and vibration methodology appendix. Most recent available data: Crossrail, 2018; HS2, 2017.
3. Human Response, Annoyance and Health
The research on people is thinner than the research on track, and much of it comes from mainline rail rather than metros. Sweden’s EpiVib study analysed 6,894 respondents out of 35,011 invited (Maclachlan et al., IJERPH 2018). Its clearest finding is that the type of train matters. Freight trains raised the adjusted odds of vibration annoyance within 100 m 5.3-fold (95% CI 3.8 to 7.4), against 1.7-fold (95% CI 1.2 to 2.6) for passenger trains, and freight annoyance reached up to 400 m.
Newer metro work points at sleep and concentration. A 2026 study measured 31 residential buildings beside an underground metro in Tehran, 18 of them with documented complaints, and found attention and sleep the most affected mental-health domains (Sadeghi et al., Sustainable Cities and Society 2026). A companion case study used the PHQ-9, a standard nine-question depression screening questionnaire, and found scores rose with vibration level. These are associations from one city, not proof of cause.
| Metric | Value | Source |
|---|---|---|
| Approximate threshold of human perception of ground-borne vibration | About 65 VdB | FTA Transit Noise and Vibration Impact Assessment Manual 2018 |
| Swedish EpiVib respondents analysed / invited | 6,894 / 35,011 | Maclachlan et al., IJERPH 2018 |
| Adjusted odds of vibration annoyance within 100 m: freight trains vs passenger trains | 5.3 (95% CI 3.8 to 7.4) vs 1.7 (95% CI 1.2 to 2.6) | Maclachlan et al., IJERPH 2018 |
| Distance to which freight-train vibration annoyance extends | Up to 400 m | Maclachlan et al., IJERPH 2018 |
| Norway: respondents highly annoyed / annoyed at about 0.5 mm/s | 10% / 40% | Klaeboe and Fyhri, cited in Maclachlan et al., IJERPH 2018 |
| Tehran: change in PHQ-9 depressive-symptom score per VdB of vibration (adjusted) | 0.171 (95% CI 0.026 to 0.316; P = 0.0207) | Sadeghi et al., Noise and Health 2026 |
| Tehran: mean score of the sleep-problem item (0 to 3 scale) | 2.81 | Sadeghi et al., Noise and Health 2026 |
| Tehran: residential buildings measured beside the metro / with documented complaints | 31 / 18 | Sadeghi et al., Sustainable Cities and Society 2026 |
Outlier note: in the same Tehran building, ground-borne noise alone was not a statistically significant predictor of depressive-symptom scores (P = 0.0829), while vibration was. Sources: Maclachlan et al. 2018, Sadeghi et al. 2026, Noise and Health and Sadeghi et al. 2026, Sustainable Cities and Society. Most recent available data for the Swedish survey: Maclachlan et al., 2018. Low-frequency exposure from other sources shows similar annoyance debates, as our wind turbine infrasound statistics show.
4. What Makes a Line Rumble: Rails, Wheels, Distance and Floors
Much of the variation in metro vibration starts at the wheel-rail contact. Transport for London names rail corrugation, a regular wave pattern worn into the running surface of the rail, as the principal cause of Tube noise, and says it is more prevalent on lines with automatic train operation, where every train brakes and accelerates at the same spots (TfL Tube Noise Update 2022). The federal rough-wheel penalty in the table below is bigger than any of the federal track-treatment credits listed in section 5.
Field data confirm the volatility. On a 350 m radius metro curve in South China, rail vibration differed by up to 20 dB from one train passage to the next (Chen et al., Buildings 2025), with corrugation only 20 to 30 µm deep. Distance and height work in the other direction: in Beijing, the zone of severe exceedance ended at 27 m from the street above Metro Line 4, and in Tehran each floor up shaved about 2 dB.
| Metric | Value | Source |
|---|---|---|
| Increase in vibration from rough wheels or rails, extreme cases | Up to 20 dB | FTA Transit Noise and Vibration Impact Assessment Manual 2018 |
| Variation in rail vibration between train passages on a corrugated 350 m radius curve | Up to 20 dB | Chen et al., Buildings 2025 |
| Rail corrugation depth measured on that curve | 20 to 30 µm | Chen et al., Buildings 2025 |
| Measured vibration attenuation per floor, Tehran residential building | 1.99 VdB per floor | Sadeghi et al., Noise and Health 2026 |
| Beijing heritage district near Metro Line 4: share of area above the 70 dB daytime limit / 67 dB night limit | About 40% / about 60% | Su et al., Sensors 2024 |
| Beijing: distance band with severe exceedance / with no exceedance | Within 27 m / 63 m and beyond | Su et al., Sensors 2024 |
Context note: the Beijing subway vibration energy concentrated around 60 Hz at the surface and 40 to 80 Hz indoors. Sources: Chen et al. 2025 and Su et al. 2024. Mass timber buildings raise their own structure-borne transmission questions, covered in our CLT flanking transmission statistics.
5. Track Mitigation: What Floating Slabs and Resilient Parts Deliver
The headline number for any vibration treatment depends on where it is measured. The federal design credits in the table below are planning figures for a floating slab (a concrete track bed resting on springs or rubber bearings) and for softer track components. Field tests show the trade-offs behind them. In one Chinese subway tunnel, a steel-spring floating slab cut resonant tunnel-wall vibration by 15 to 22 dB but overall tunnel-wall vibration by only 5 dB versus a monolithic roadbed, while the slab itself vibrated 10 dB more and low-frequency forces rose in the 1 to 20 Hz range (Zhou et al., JAABE 2025).
Treatments on the train side look weaker. A 2025 simulation of resilient wheels, which put a rubber layer between rim and hub, found a total tunnel-wall reduction of 1.6 dB. Under-sleeper pads tested in a Polish laboratory cut subgrade acceleration by up to 40 to 50%, but field verification is still pending.
| Metric | Value | Source |
|---|---|---|
| Design adjustment for floating slab / ballast mats / high-resilience fasteners | -15 / -10 / -5 dB | FTA Transit Noise and Vibration Impact Assessment Manual 2018 |
| Steel-spring floating slab vs monolithic roadbed: tunnel-wall vibration / resonant peaks | 5 dB lower / 15 to 22 dB lower | Zhou et al., Journal of Asian Architecture and Building Engineering 2025 |
| Same comparison: slab vibration / transverse wheel-rail force | 10 dB higher / 47% higher | Zhou et al., Journal of Asian Architecture and Building Engineering 2025 |
| Athens Metro Lines 2 and 3: floating slab insertion loss at crossovers, 40 to 125 Hz band | About 20 dB | Vogiatzis and Kouroussis, International Journal of Rail Transportation 2015 |
| South China metro curve: tunnel-wall peak acceleration, untreated section vs floating slab section | About 3 m/s² vs 0.05 m/s² | Chen et al., Buildings 2025 |
| Beijing: simulated reduction in maximum vertical vibration level from a steel-spring floating slab | 6.93 to 10.81 dB | Su et al., Sensors 2024 |
| Resilient wheel vs steel wheel: total Z-weighted tunnel-wall vibration (simulation) | 70.7 vs 72.3 dBZ | Sun et al., Frontiers in Physics 2025 |
| Under-sleeper pads: reduction in subgrade acceleration (laboratory) | Up to 40 to 50% | Chudyba et al., Scientific Reports 2026 |
Outlier note: the South China figures compare two different track sections with different corrugation, so the ratio (3 / 0.05 = 60 times) is not a controlled insertion loss. The Athens result is from 2015; most recent available data for that system: Vogiatzis and Kouroussis, 2015. Sources: Zhou et al. 2025, Vogiatzis and Kouroussis 2015, Sun et al. 2025 and Chudyba et al. 2026.
6. London Underground and the Elizabeth Line: The Price of Quiet
London shows both ends of the problem: a network TfL described in 2022 as 159 years old, which has to retrofit, and a new railway that could design vibration out. TfL’s 2022 board briefing describes 1,107 km of track maintained with only two rail grinding machines, with each noisy site needing regrinding every six months to a year. Fixes can also backfire: the Pandrol Vanguard resilient fastener, fitted from 2016 ahead of the Night Tube to cut noise in homes, trapped energy in the tunnel and sped up corrugation, and swapping it for the Delkor system costs £8,200 per metre (TfL Tube Noise Update 2022).
Crossrail went the other way, buying isolation before opening: it kept standard track for most of its bored central tunnels, reserved high-attenuation track and floating slabs for the most sensitive locations, and padded every ground-borne noise prediction with a conservative +10 dB allowance (Crossrail Learning Legacy 2018).
| Metric | Value | Source |
|---|---|---|
| TfL spending on rail grinding, maintenance plus noise reduction, five years to 2022 | Approximately £7-10m a year | TfL Tube Noise Update 2022 |
| London Underground rail grinding budget, 2025/26 vs 2024/25 | £9m vs £8.1m | Mayor’s Question Time, 16 July 2026 |
| Noise-related rail grinding, year to December 2022 vs six months to March 2021 | Over 30,000 m vs 10,355 m | TfL Tube Noise Update 2022; TfL letter to Assembly Members 2021 |
| Cost of replacing Pandrol Vanguard fasteners with Delkor | £8,200 per metre | TfL Tube Noise Update 2022 |
| Northern line residential noise and vibration complaints, 2018 to February 2022 | More than 280 | Barnet Post 2022, reporting TfL figures |
| Crossrail central-section track: standard / high-attenuation / floating slab (2 km elastomeric + 1.3 km steel-spring) | 37.8 km / 2.9 km / 3.3 km | Crossrail Learning Legacy 2018 |
| Buildings above Crossrail tunnels assessed for ground-borne noise | 400 | Crossrail Learning Legacy 2018 |
| Earliest date an acoustic rail grinding machine, still to be trialled, could join TfL’s grinding contract | Late 2027 | Mayor’s Question Time, 16 July 2026 |
Context note: TfL also says complaints spiked when Night Tube services began in 2016. The Crossrail paper reports design predictions, not post-opening measurements. Sources: TfL Tube Noise Update 2022, Mayor’s Question Time minutes, July 2026, TfL letter to Assembly Members, March 2021 and Barnet Post. For how other transport noise is regulated around communities, see our aircraft noise exposure statistics.
Summary: Subway and Metro Rail Ground-Borne Vibration by the Numbers
| Metric | Value | Source |
|---|---|---|
| Cities / networks with metro service | 202 / 247 | UITP Global Metro Figures 2024 |
| US residential vibration impact threshold, frequent events | 72 VdB | FTA Transit Noise and Vibration Impact Assessment Manual 2018 |
| Crossrail ground-borne noise design criterion for homes | 40 dB LAmax,S | Crossrail Learning Legacy 2018 |
| HS2 significant-effect threshold for ground-borne noise in homes | Above 45 dB LASmax | HS2 Environmental Statement SV-001-000, 2017 |
| HS2 residential vibration significance threshold, night | VDV 0.4 m/s^1.75 | HS2 Environmental Statement SV-001-000, 2017 |
| China indoor vibration limits near subways, day / night | 70 dB / 67 dB | Su et al., Sensors 2024 |
| Sweden recommended maximum for railway vibration in homes | 0.4 mm/s | Maclachlan et al., IJERPH 2018 |
| Adjusted odds of annoyance within 100 m, freight trains | 5.3 | Maclachlan et al., IJERPH 2018 |
| PHQ-9 score change per VdB, Tehran Metro Line 6 building | 0.171 | Sadeghi et al., Noise and Health 2026 |
| Tehran buildings measured / with complaints | 31 / 18 | Sadeghi et al., Sustainable Cities and Society 2026 |
| Beijing heritage district area above the 70 dB daytime limit | About 40% | Su et al., Sensors 2024 |
| Train-to-train variation in rail vibration on a corrugated curve | Up to 20 dB | Chen et al., Buildings 2025 |
| Floating slab: reduction in resonant tunnel-wall vibration | 15 to 22 dB | Zhou et al., JAABE 2025 |
| Athens floating slab insertion loss at crossovers | About 20 dB | Vogiatzis and Kouroussis 2015 |
| Resilient wheel total vibration reduction (simulation) | 1.6 dB | Sun et al., Frontiers in Physics 2025 |
| Under-sleeper pads: subgrade acceleration reduction (laboratory) | Up to 40 to 50% | Chudyba et al., Scientific Reports 2026 |
| London Underground rail grinding budget 2025/26 | £9m | Mayor’s Question Time, 16 July 2026 |
| Cost of replacing Pandrol Vanguard fasteners | £8,200 per metre | TfL Tube Noise Update 2022 |
| Northern line residential noise and vibration complaints, 2018 to February 2022 | More than 280 | Barnet Post 2022, reporting TfL figures |
Methodology and Sources
- FTA, Transit Noise and Vibration Impact Assessment Manual, FTA Report No. 0123 (September 2018): impact criteria, perception threshold, source and mitigation adjustments.
- Cobbing, Webb and Bewes, Control of Railway Induced Groundborne Noise and Vibration from the UK’s Crossrail Project, Crossrail Learning Legacy (2018): IP D10 criteria, track-form lengths, prediction allowance.
- HS2 Ltd, Environmental Statement Volume 5, SV-001-000 sound, noise and vibration methodology (July 2017): ground-borne noise and VDV criteria.
- UITP, Global Metro Figures 2024: networks, cities and people served (data to end of 2023).
- Maclachlan et al., Annoyance in Response to Vibrations from Railways, IJERPH (2018): Swedish EpiVib survey, Swedish and Norwegian reference values.
- Sadeghi et al., Noise and Health (2026): Tehran Metro Line 6 building, floor attenuation, PHQ-9 association.
- Sadeghi et al., Sustainable Cities and Society (2026): 31-building Tehran metro study.
- Su et al., Sensors (2024): Beijing Metro Line 4 heritage district measurements and GB10070 limits.
- Rafidah et al., Scientific Reports (2025): Malaysian DOE vibration limits and rail measurements.
- Chen et al., Buildings (2025): rail corrugation on a metro small-radius curve.
- Zhou et al., Journal of Asian Architecture and Building Engineering (2025): steel-spring floating slab vs monolithic roadbed.
- Vogiatzis and Kouroussis, International Journal of Rail Transportation (2015): Athens Metro floating slabs.
- Sun et al., Frontiers in Physics (2025): resilient wheels.
- Chudyba et al., Scientific Reports (2026): under-sleeper pads.
- TfL, Tube Noise Update, Customer Service and Operational Performance Panel (6 December 2022): grinding spend, causes, Pandrol Vanguard costs.
- London Assembly, Mayor’s Question Time minutes, 16 July 2026: 2024/25 and 2025/26 grinding budgets.
- TfL, Tube noise update letter to Assembly Members (15 March 2021): grinding over the previous six months.
- Barnet Post (27 June 2022): Northern line complaint count, reporting TfL figures.
- Data watch: TfL’s 2022 slide gives £8,200 per metre and 6.7 km of Pandrol Vanguard replacement at circa £59.9m, but 6,700 m x £8,200 = £54.9m, so we cite only the per-metre figure. The Northern line complaint count comes from a local newspaper reporting TfL figures, not a TfL dataset we could read directly. Units differ across sources (VdB velocity, Chinese acceleration dB, LAmax,S indoor sound, VDV dose, mm/s), so thresholds cannot be ranked against each other. The Swedish survey covers mainline railways, including freight, not metros. The South China tunnel-wall comparison crosses two track sections, and the floating slab study also found increased low-frequency (1 to 20 Hz) forces. The Tehran health findings come from one building and one city. The FTA manual (2018), HS2 criteria (2017) and Crossrail paper (2018) are older than three years but remain the current published versions we could verify.
Last updated: October 18, 2026. We update this roundup quarterly, and the next refresh is expected when TfL issues its next six-monthly Tube noise update to the London Assembly and UITP publishes its next Global Metro Figures edition.