Humans are extraordinarily sensitive to nocturnal noise. Studies using polysomnography (Muzet, 2007) show that a single 45 dB(A) event during deep sleep can trigger a partial arousal even when it does not wake the subject. The cumulative effect over a night is what turns a "noisy flat" into chronic insomnia.
What the WHO changed in 2018
The 2018 guidelines tightened the recommended maximum night-time exposure from 55 dB Lnight (the old 2009 value) to 45 dB Lnight for road traffic. For rail the recommendation is 44 dB, for aircraft 40 dB. Those are average values over the 23:00-07:00 window - short peaks can be much higher without triggering a warning, which is one reason the standard misses the worst real-world situations.
What it feels like (everyday anchors)
The WHO numbers are long-term Lnight averages outdoors. They are easier to read if you map them to familiar short-term cues, the same way public dashboards such as Noise Map colour their legends. The table is pedagogical, not a field measurement protocol.
| Indicative band | Typical cues near the source |
|---|---|
| around 40 dB(A) | Quiet bedroom in a village, refrigerator in the kitchen. |
| around 50 dB(A) | Very quiet suburb, library reading room. |
| 50-55 dB(A) | Average open-plan office hum. |
| 55-60 dB(A) | Normal conversation at about one metre. |
| 60-65 dB(A) | Lively classroom, piano practice. |
| 65-70 dB(A) | Noisy restaurant; Cerema maps place about 30 % of Parisians near this level on a main boulevard at 100 m at night. |
| 70 dB(A) and above | Vacuum cleaner a few metres away, heavy traffic at the kerb. |
| around 75 dB(A) | Next to an urban ring road. Every +3 dB doubles the acoustic energy. |
The 2018 WHO 45 dB Lnight guideline for road traffic is therefore far below a "chat at one metre" because it describes chronic outdoor exposure while you try to sleep, not a short indoor conversation.
How ClearSpot reads your neighbourhood
We use the strategic noise maps produced under the EU 2002/49/CE directive (Cerema for transport axes, Bruitparif for Ile-de-France sensors). A cell is "not clear" whenever the modelled Lnight exceeds 45 dB(A). The UI lets you switch to the daytime Lden metric if you care more about working-from-home conditions.
Where Bruitparif sensors are present, we use their real measured values instead of the model. That mostly matters in Paris intramuros and the first ring, where real Lnight can exceed the modelled value by 5-8 dB on Friday nights.