Highway Widening Model: A Sub-3 dB Change Called 'No Impact'
A transportation agency models noise for a proposed widening of a four-lane arterial. Existing traffic is 20,000 vehicles/day at 80 km/h; the design year is 32,000 vehicles/day. The model predicts a worst-case residential facade Leq(day) of 64.2 dBA existing and 65.1 dBA build, and the draft environmental document concludes 'an increase of less than 3 dB, therefore no substantial noise impact and no abatement required.' The receptor is a two-story house; the modeled receiver height is 1.5 m above local ground. The project also repaves the existing lanes, replacing open-graded friction course with dense-graded asphalt, and adds a free-flow right-turn ramp whose centerline passes 120 m from the same house. The model used default ground absorption (G = 0.5) for both scenarios and was not checked against any field measurement. Nearby monitoring at a comparable facade gives an Ldn 3 dB higher than the model's existing case.
1. Critique the 'less than 3 dB, therefore no impact' conclusion. Identify each source of bias and its direction.
Read the worked answer
Several independent biases all run in the direction of understating the build condition:
- Pavement change. Replacing open-graded friction course with dense-graded asphalt typically raises car and truck pass-by noise by roughly 2–4 dB at highway speeds. A model that ignores the surface change omits this entirely.
- Receiver height. 1.5 m is a ground-floor position. Upper-story windows near 5 m see different ground and barrier attenuation and are commonly 1–3 dB higher for a source 100 m or more away; a two-story dwelling should be modeled at the upper floor as well.
- New ramp. A free-flow right-turn ramp at 120 m is an added source with its own volumes and speeds and was not included.
- Ground absorption. Default G = 0.5 for both cases is a place-holder; hard surfaces such as pavement and compacted fill sit closer to G = 0, which raises levels by several decibels at this distance.
- No validation. Monitored Ldn is already 3 dB above the modeled existing case, so the model under-predicts and that bias carries forward.
Finally, 0.9 dB is smaller than typical model uncertainty (roughly ±3 dB), so a null prediction is being treated as proof of no effect - a logical error independent of the omitted inputs.
2. What analysis should be done before the conclusion is finalized?
Read the worked answer
Rebuild the model:
- Correct the pavement type, using a documented pavement adjustment or a model that supports one.
- Model each dwelling at 1.5 m and at 4.5–5 m, and report both.
- Derive a site-specific ground factor from land cover rather than using the default.
- Add ramp traffic as a separate source with its own volumes, speeds and truck percentages.
- Validate against the nearby monitoring by correcting the traffic inputs - volumes, speeds, truck mix, source heights - and document every adjustment; do not apply an arbitrary offset to force agreement.
- Run sensitivity cases on the uncertain inputs (truck percentage, pavement adjustment, ground factor) and report a range rather than a single number.
- Compare the results with the program's actual impact criteria (an absolute threshold and/or a 3 dB increase criterion), citing the specific criterion applied.
If the corrected change remains small, that is a finding supported by analysis; the current document does not support it.
3. How should the predicted impacts and their uncertainty be reported?
Read the worked answer
Report a corrected build level and a stated range, for example a central estimate in the high 60s dBA at the upper facade together with the sensitivity bounds, and identify which inputs drive the range. State plainly that the existing-case under-prediction of about 3 dB limits confidence in absolute levels. State that the predicted change is of the same order as model uncertainty and therefore cannot by itself establish 'no impact'.
List all assumptions - pavement adjustment, ground factor, receptor heights, ramp volumes - so a reviewer can reproduce the numbers. Recommend post-construction measurement at the facade to confirm the prediction and to support any future abatement decision, and make clear that 'no abatement required' cannot rest on an unvalidated model that omits a pavement change and a new ramp.