Skip to main content

Understanding ISO 9613-2: industrial noise prediction in practice

Industry Standards · SoundPLAN Asia · · 9 min read

ISO 9613-2 is the calculation method most industrial noise assessments in Asia and the Middle East are built on, and it is also the method most often quoted without being understood. It is worth knowing precisely what it computes, what it assumes while doing so, and at what point a result stops meaning anything.

What the standard actually does

ISO 9613-2 predicts the A-weighted sound pressure level at a receiver from one or more sources outdoors. It works octave band by octave band, from 63 Hz to 8 kHz, and it does so by starting from the sound power level of each source and subtracting a series of attenuation terms along the propagation path.

The equation is a subtraction, and every term in it is a separate physical mechanism:

  • Adiv - geometrical divergence. The spreading of energy over an expanding surface as the wave travels. For a point source this is the familiar 6 dB per doubling of distance, and it is usually the largest single term.
  • Aatm - atmospheric absorption. Air itself absorbs sound, strongly at high frequency and negligibly at low. It depends on temperature and relative humidity, which is why the same plant is louder at the fence line on a cool, damp night than the same calculation run at 30 degrees and 70 per cent suggests.
  • Agr - ground effect. Interference between the direct path and the path reflected off the ground. Governed by the ground factor G, where 0 is acoustically hard (water, concrete, compacted sand) and 1 is acoustically soft (grass, cultivated ground). This is the term engineers get wrong most often.
  • Abar - screening. The insertion loss of barriers, buildings and terrain that break the line of sight.
  • Amisc - everything else the standard treats separately: foliage, industrial site scattering, and housing.

The assumption that surprises people

ISO 9613-2 does not predict an average. It predicts levels under meteorological conditions favourable to propagation - that is, downwind from source to receiver, or under a moderate ground-based temperature inversion. In plain terms, it models the wind blowing from the plant towards the neighbour, or a calm, clear night.

This matters commercially. If a regulator asks for the worst realistic case, a favourable-conditions result is the right answer. If a regulator asks for a long-term average, the standard provides the meteorological correction C0 to convert a favourable-condition result into a long-term one, and that correction must be stated and justified rather than quietly applied. Two assessments of the same plant that differ by several decibels usually differ because one applied C0 and the other did not.

Where the method stops being valid

The standard states its own accuracy, and it is narrower than most people assume. The estimated accuracy applies to broadband noise from sources at effective heights between roughly 0 and 30 metres, at distances between roughly 10 and 1000 metres. Within that envelope the standard quotes an accuracy of about plus or minus 3 dB.

Outside it, the method is being extrapolated. Three situations come up constantly on industrial sites in the region:

  • Long distances. Beyond about a kilometre, meteorological variability dominates and a single-number prediction carries a spread far wider than 3 dB.
  • Elevated sources. Flare tips, stack outlets and cooler decks well above 30 metres are outside the validated range, and the ground-effect formulation in particular becomes unreliable.
  • Strong tones. ISO 9613-2 is a broadband method. A compressor or transformer with a dominant tone needs a tonal penalty applied at the assessment stage, according to whichever national or project criterion applies. The propagation calculation will not produce it for you.

The ground factor is where assessments diverge

Of all the inputs, G does the most damage when it is chosen casually. Desert sand that has been graded and compacted for a plant road is not the same acoustic surface as undisturbed sand, and neither is the same as the irrigated ground of an adjacent farm. A model that applies a single G across a whole study area will be optimistic in some directions and pessimistic in others, and the error does not cancel out - it lands on whichever receiver happens to sit downwind.

The practical discipline is to define ground regions explicitly in the model, to state the value used for each, and to test the sensitivity of the controlling receiver to that choice before the report leaves the office.

ISO 9613-2 or CNOSSOS-EU?

The two are not competitors; they answer different questions. ISO 9613-2 is a general propagation method suited to industrial sources and site-specific assessment. CNOSSOS-EU is the common assessment method mandated for strategic noise mapping under the European Environmental Noise Directive, and it carries its own source models for road, rail, aircraft and industry along with its own propagation treatment.

Where a project is an industrial permit, an environmental impact assessment or a plant expansion, ISO 9613-2 is normally what the authority expects. Where the task is city-scale exposure mapping against an END-style framework, CNOSSOS-EU is the method. Projects in the region increasingly meet both - an industrial assessment feeding into a wider strategic map - which is the practical reason for using software that implements them side by side rather than choosing one.

What to check before you trust a result

  1. Are the source sound power levels measured, or taken from a manufacturer's data sheet? If the latter, are they octave band or a single A-weighted figure with an assumed spectrum?
  2. Is the source height the acoustic centre, or the top of the equipment?
  3. Have ground regions been defined, or has one G been applied everywhere?
  4. Has C0 been applied, and is that stated in the report?
  5. Are the controlling receivers at the right height? A ground-floor receiver and a fourth-floor balcony on the same facade can differ by several decibels once screening is accounted for.
  6. Does any source carry a tone or impulsive character that needs a penalty the propagation model will not add?

A prediction that survives those six questions is defensible. One that has not been asked them is a number, not an assessment.

Industrial noise assessment with SoundPLAN · The full standards list · Ask us about a specific project

All articles