How CANM turns a sound source into a predicted level at a receiver: the propagation chain, the equations behind each term, and the data each one draws on. Written for environmental practitioners, planners and engineers who need to know what the number means and where it came from.
Developed by Nobert Mrema and Eduward Gumbwa, Environmental Engineers - Clanert Sustain.
CANM predicts outdoor environmental noise. You describe a site - where the sources are, how loud they are, what stands between them and the people around them - and it computes the sound pressure level that reaches every point of a grid and every receiver you place, then rates those levels against the limits that apply.
The prediction is an octave-band implementation of ISO 9613-2, with the extensions CNOSSOS-EU adds for meteorology, the mean ground plane and lateral diffraction. Every attenuation term is the equation the standard gives, evaluated at the eight octave centre frequencies from 63 Hz to 8 kHz and summed on an energy basis. The rest of this page is what those equations are and what feeds them.
There is nothing to install. The map, the drawing tools and the live preview run in the browser; the octave-band solve runs on the server and returns a raster. A site you build stays on the server against your project, so you can close the tab and come back to it.
A site begins with its area, not its sources. The area decides three things at once: what the noise map covers, what gets pulled in from OpenStreetMap, and the extent every result is reported over. Draw it before anything else and the rest of the model has somewhere to live.
Making it larger costs time quadratically - twice the side is four times the grid - so draw the area you need to report on, not the area you can see. You can change it at any point; the model is not fixed by it.
Buildings and roads come from OpenStreetMap over the area you drew. Buildings matter more than people expect: on one real site they were worth 4.4 dB of screening on average across the grid. Where a building has no height tagged, one is inferred from its levels or its footprint and marked as estimated, so you can see which heights are real.
A source is a point, a line or an area, each with a sound power level and a height. Roads can be driven from traffic flow, speed and heavy-vehicle percentage rather than a sound power you have to work out yourself. Anything that does not run continuously gets a duty cycle, which is a 10 log10 f correction rather than a guess.
Two things compute, and it helps to know which one you are looking at:
The preview and the full solve use the same engine, so the shape agrees; the full solve resolves more of it.
Solve time is driven by the number of grid points times the number of source points that survive the screening cull - not by how complicated the site looks. Measured on real sites:
| Site | Sources | Grid points | Full solve |
|---|---|---|---|
| Market site, 250 m search radius | 5 | 10 000 | 49 s |
| Market site, full search radius | 5 | 10 000 | 80 s |
| Quarry, plant and traffic | 64 | 10 000 | 280 s |
If a solve is slower than you want, the two controls that matter are the size of the calculation area and the screening search radius in Site settings. Reducing the radius stops distant sources being tested against every building; it changes the answer only where those distant sources actually contributed.
Edits save as you make them - there is no save button to forget. A site holds its area, its geometry, its sources and its settings; each calculation you run is stored alongside it as a run, with the results and the inputs that produced them.
Every run records the method revision that computed it. When a change to the engine moves a computed level, that revision changes, and any run made before it is flagged as out of date wherever it appears. The results are still there and still say what they said; the flag tells you they were not produced by the method now in force. Re-run the calculation before issuing anything from it.
A scenario is the same site with changes applied - a barrier added, a machine quieter, a source switched off. It does not copy the site, so the baseline cannot drift underneath it. Solve a scenario and you can put its result next to the base case and read the difference directly, which is the number a mitigation decision actually turns on.
The map exports as a PNG with the legend and, if you want it, a scale bar. What goes on it is yours to set: buildings outlined or filled, sources shown with their types, roads drawn in their own colours, and each of those layers placed above or below the noise raster. A map where the buildings are hidden behind the colour is a map that cannot be checked.
Receiver results export as CSV - one row per receiver, with the level, the limit it was compared against and the margin. That is the form that goes into a spreadsheet or an appendix without being retyped.
A run produces a report that carries the map, the receiver table, the charts, the assessment against the limits in force, and every input the calculation used. The inputs are part of the report on purpose: a predicted level without the assumptions behind it cannot be reviewed by anyone.
A single number at a receiver is not a result you can defend. These are the ways to open one up.
Every receiver can show the calculation behind it: which sources reached it and how much each contributed, and how many decibels each attenuation term took off the way - divergence, atmospheric absorption, ground, screening, vegetation. It answers most questions about a surprising level on its own, and it is the first thing to look at before assuming something is wrong.
It also shows what a change would be worth. A barrier added where buildings already break the line of sight buys very little: on one real site a 5 m screen was worth 19 dB in open ground and 0.43 dB in a dense block, because the buildings were already doing the work.
A scenario result can be read against the base case receiver by receiver, so the output of a mitigation study is a difference in decibels at named places rather than two maps to eyeball.
Buildings, barriers, terrain and sources drawn at their real heights, updating with the 2D map. Screening is a three-dimensional problem - whether a barrier breaks a path depends on source height, barrier height and receiver height together - and a plan view hides exactly the dimension that decides it.
Level against distance, contribution by source, and the octave spectrum at a receiver. The spectrum is worth reading whenever a result is unexpected: a level dominated by the 63 Hz band behaves nothing like the same level dominated by 1 kHz, and only the spectrum shows the difference.
A source is defined by its sound power level LW - how much acoustic energy it emits, independent of where you stand. That is the only description that travels; a measured pressure level belongs to the point it was measured at.
Most equipment is characterised by a reading rather than a sound power level. CANM converts one to the other, but the measurement surface must be stated - it is worth 3 to 6 dB and is the most common error in the step.
| Measured over | C, dB | When it applies |
|---|---|---|
| Hemisphere over a reflecting plane (ISO 3744 / 3746) | 8 | Source standing on the ground or a hard floor - the usual case |
| Full sphere, free field | 11 | Source suspended clear of any surface, or an anechoic room |
| Quarter space - source against a wall | 5 | Source against a wall |
| Eighth space - source in a corner | 2 | Source in a corner |
A source on the ground already benefits from the ground reflection, so it needs 3 dB less power to produce the same reading than one suspended in free field. That is the same 3 dB that appears in the ground term in section 5, seen from the other end.
Roads, conveyors and yards are not points. CANM splits them into incoherent point sub-sources - the technique CadnaA, SoundPLAN and CNOSSOS-EU all use internally - with each sub-source carrying the power of the length or area it represents, so total radiated energy is conserved exactly.
A traffic stream is converted to sound power per metre of carriageway from flow, speed and heavy-vehicle percentage, then discretised like any other line source.
Calibrated so 1,000 veh/h at 50 km/h with 10% heavy vehicles reads about 65 dB(A) at 10 m over a hard carriageway - the standard planning figure, and what a meter at the kerb shows.
A prediction inherits the quality of its inputs, so each source records how its LW was obtained. The report prints it beside the value.
LAeq is an energy average over the assessment period, so a source that runs for part of it contributes proportionally less. A generator running twenty minutes in an hour is 7.8 dB below the same generator running continuously.
A railway is described by its service, not by a sound power somebody already knows. Rolling noise dominates in the normal speed range and rises more steeply with speed than road traffic, because wheel and rail roughness excitation grows faster than tyre noise.
An area source is flat by default, which is wrong for a quarry face, a stockpile or a tip. Those radiate over a height range, and the height decides what a pit rim actually screens: the top of a 20 m face can be in clear view while its foot is fully shielded.
Setting a vertical extent stacks the sampling grid into layers and splits the energy between them, so the total radiated power is unchanged. Measured on a face in a 15 m pit: modelled flat it reads 26.2 dB(A) at the receiver; with 20 m of extent, 39.1 dB(A). That 13 dB is the difference between no issue and an issue, and a flat source cannot represent it.
Every level CANM reports is built the same way. A source radiates a known sound power in each octave band; the sound loses energy on its way to the receiver through a series of attenuation terms; what is left is summed across bands and across every source and every path. Nothing is interpolated from a look-up table or scaled off a nomogram.
Sound spreading from a point source covers an ever-larger sphere, so intensity falls with the square of distance. This is the largest term in most calculations and the only one that is exact.
That gives exactly 6.02 dB per doubling of distance from a point source. A line source behaves differently: summing its sub-sources produces 3 dB per doubling, because the sound spreads cylindrically rather than spherically. CANM does not special-case this - it emerges from the summation, which is a useful check that the discretisation is right.
Air itself absorbs sound, strongly at high frequency and hardly at all at low frequency. It is why distant traffic sounds like a rumble: the top of the spectrum is gone. The coefficient depends on temperature and humidity in a way that is not monotonic, so CANM implements the full formulation rather than interpolating a table.
Sound reaching a receiver has travelled two routes: directly, and via a bounce off the ground. What the two do when they meet depends entirely on what the ground is made of.
This replaced a flat −3 dB. The 1996 edition credited hard ground with the full 3 dB everywhere, including a metre from the machine, where the reflected path is much longer than the direct one and no coherent reinforcement happens. The 2024 edition added Kgeo to describe that, so the gain now reaches 3.01 dB across a site and fades to under 1 dB at arm's length. Beyond about 25 m the two agree to a tenth of a decibel, so nothing at ordinary receiver distances moved.
The two regimes work in opposite directions, which is the part worth understanding:
| Surface | G | Behaviour |
|---|---|---|
| Water / ice | 0.0 | Reflective - reinforces |
| Paving, asphalt, concrete | 0.0 | Reflective - reinforces |
| Compacted bare ground / hardstanding | 0.1 | Reflective - reinforces |
| Gravel, quarry, industrial yard | 0.3 | Mixed |
| Mixed hard and soft ground | 0.5 | Mixed |
| Sparse grass / dry soil | 0.7 | Mixed |
| Grass, farmland, vegetation | 1.0 | Porous - attenuates |
| Forest floor / dense planting | 1.0 | Porous - attenuates |
A barrier or building does not block sound. It forces it to bend over the top edge, and lets some through the fabric, and lets some round the ends. All three arrive together, and CANM computes all three.
Earlier versions of CANM used the Kurze-Anderson fit to Maekawa's curve here. That is a respectable curve but not the one ISO 9613-2 specifies, and it ran 2 to 3 dB per band optimistic against Formula (18) - always crediting a barrier with more than the standard allows. It also had no meteorological term at all, so a low barrier was credited with screening at distances where none survives.
Sound also passes through the construction, attenuated by its transmission loss R (section 7), and around its ends by lateral diffraction. Each is a separate route to the same receiver, so they combine on an energy basis:
A building screens over its highest effective edge. For a pitched roof that is neither the eaves nor the ridge: a ridge is a line rather than a wall, so sound bending over it sees less than the full ridge height. CANM adds a fraction of the roof rise to the eaves.
| Roof shape | Fraction of rise added | Typical rise, if unstated |
|---|---|---|
| Flat | 0.0 | - |
| Gabled / pitched | 0.7 | 0.28 × height |
| Hipped | 0.5 | 0.25 × height |
| Pyramidal | 0.4 | 0.3 × height |
| Mono-pitch / skillion | 0.6 | 0.18 × height |
| Domed | 0.4 | 0.3 × height |
| Not stated | 0.0 | - |
Everything above describes one path. Put the receivers in a line behind the screen and the formulae describe a shadow, and it does not behave like the one a wall casts in front of a light. Two differences decide whether a barrier is worth building, and both fall out of Formula (18) rather than being added to it.
The shadow is deepest at the screen and softens behind it. z is an extra distance measured against the direct path, so as the receiver retreats both grow and their difference does not keep pace. The same screen below is worth 18.5 dB at 5 m and 12.6 dB at 320 m: 5.9 dB given up for nothing but standing further back. It is why moving a barrier towards the source, or towards the receiver, beats making it taller.
Low frequencies bend around it; high ones do not. λ is in the denominator, so the same z buys far more attenuation at 8 kHz than at 63 Hz. Close behind the screen that is 19.9 dB against 8.0 dB, a spread of 11.9 dB across the spectrum. This is the single most common surprise after a barrier is built: it removes the clatter and leaves the hum, and the complaint that it “did not work” usually means it worked on the part nobody was listening to.
| Receiver, behind the screen | 63 Hz | 125 Hz | 250 Hz | 500 Hz | 1000 Hz | 2000 Hz | 4000 Hz | 8000 Hz | Overall |
|---|---|---|---|---|---|---|---|---|---|
| 5 m | 8.0 | 10.4 | 13.0 | 15.8 | 18.7 | 19.9 | 19.9 | 19.9 | 18.5 |
| 10 m | 6.9 | 8.9 | 11.2 | 13.8 | 16.6 | 19.4 | 19.9 | 19.9 | 17.3 |
| 20 m | 6.1 | 7.7 | 9.7 | 12.0 | 14.6 | 17.4 | 19.9 | 19.9 | 15.9 |
| 40 m | 5.6 | 6.9 | 8.6 | 10.6 | 13.1 | 15.8 | 18.6 | 19.9 | 14.5 |
| 80 m | 5.3 | 6.4 | 7.9 | 9.7 | 12.0 | 14.6 | 17.4 | 19.9 | 13.6 |
| 160 m | 5.1 | 6.2 | 7.5 | 9.2 | 11.4 | 13.9 | 16.7 | 19.5 | 13.0 |
| 320 m | 5.0 | 6.0 | 7.2 | 8.9 | 11.0 | 13.5 | 16.2 | 19.0 | 12.6 |
Insertion loss in dB, computed on page load from the same screening term the solver uses, for a 5 m screen with the source 30 m in front of it at 2 m and receivers at 1.5 m. The overall figure is the A-weighted difference between the screened and unscreened sums, not an average of the bands - decibels do not average. Your own site will differ; the shape of the effect will not.
Diffraction is geometry; transmission is materials. A 4 m timber fence and a 4 m concrete wall have identical geometry and very different performance, and this is where that difference lives.
The limp mass law assumes a panel with no stiffness. Real constructions manage about 5 dB per doubling rather than 6, and every panel has a coincidence dip at a frequency set by its thickness and material. Both make the figure slightly optimistic for thin stiff sheets such as profiled steel or glazing, so a measured R from an EN 1793-2 test report always takes precedence and can be entered directly.
Note that absorption and transmission are independent. Since R = 10 log10(1/τ) with τ = 1 − α, a material with α = 0.9 has a transmission loss of only 10 dB. High absorption does not mean high insulation - which is why CANM keeps surface mass and absorption as separate fields and never infers one from the other.
| Construction | kg/m² | 63 | 125 | 250 | 500 | 1000 | 2000 | 4000 | 8000 |
|---|---|---|---|---|---|---|---|---|---|
| Close-boarded timber fence, 18-22 mm | 12 | 11 | 17 | 23 | 29 | 35 | 41 | 45 | 45 |
| Timber, 45 mm tongue-and-groove | 25 | 17 | 23 | 29 | 35 | 41 | 45 | 45 | 45 |
| Profiled steel sheet, 1.5 mm | 12 | 11 | 17 | 23 | 29 | 35 | 41 | 45 | 45 |
| Transparent acrylic / polycarbonate, 15 mm | 18 | 14 | 20 | 26 | 32 | 38 | 44 | 45 | 45 |
| Timber frame, mineral wool absorptive face | 30 | 19 | 25 | 31 | 37 | 43 | 45 | 45 | 45 |
| Absorptive metal cassette, perforated face | 28 | 18 | 24 | 30 | 36 | 42 | 45 | 45 | 45 |
| Precast concrete panel, 100 mm | 240 | 37 | 43 | 45 | 45 | 45 | 45 | 45 | 45 |
| Brick or dense blockwork, 100 mm | 200 | 35 | 41 | 45 | 45 | 45 | 45 | 45 | 45 |
| Gabion basket, 500 mm stone fill | 700 | 45 | 45 | 45 | 45 | 45 | 45 | 45 | 45 |
| Earth bund, grassed | 2000 | 45 | 45 | 45 | 45 | 45 | 45 | 45 | 45 |
| Custom / from test report | 25 | 17 | 23 | 29 | 35 | 41 | 45 | 45 | 45 |
Transmission loss R in dB per octave band, Hz. Typical published figures for the construction named; a specific product's EN 1793-1 (absorption) and EN 1793-2 (airborne insulation) test data always takes precedence.
A facade opposite a source reflects sound back across it. CANM models this by the image-source method: each reflecting surface creates a mirrored source behind it, which is then propagated through the full chain like any other source and energy-summed with the direct path.
Reflections are what make street canyons and courtyards louder than an open site, and they are the reason an absorptive facing is specified on barriers facing each other across a road.
Foliage attenuates the part of the path that runs through it - and only where the planting is dense enough to block sight completely. A line of trees you can see between attenuates nothing, however wide the belt.
| Path through foliage | 63 | 125 | 250 | 500 | 1000 | 2000 | 4000 | 8000 |
|---|---|---|---|---|---|---|---|---|
| Under 10 m | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| 10-20 m, total dB | 0.0 | 0.0 | 1.0 | 1.0 | 1.0 | 1.0 | 2.0 | 3.0 |
| 20-200 m, dB per metre | 0.02 | 0.03 | 0.04 | 0.05 | 0.06 | 0.08 | 0.09 | 0.12 |
ISO 9613-2 Table 1. Bands in Hz. No further attenuation is credited beyond 200 m of path.
These are small numbers, and deliberately so. At 1 kHz a dense belt gives 0.06 dB per metre, so 30 m of woodland buys about 2 dB - near the limit of what is perceptible. Planting is valuable for many reasons; it is not an effective acoustic screen, and a model that suggested otherwise would send people down an expensive dead end.
| Cover | Density factor | Effect |
|---|---|---|
| Dense woodland / forest | 1.0 | Attenuates the path, and softens the ground beneath |
| Open woodland, scattered trees | 0.4 | Attenuates the path, and softens the ground beneath |
| Scrub, bush, thicket | 0.6 | Attenuates the path, and softens the ground beneath |
| Hedge or planted screen | 0.8 | Attenuates the path, and softens the ground beneath |
| Grassland, meadow, crops | 0.0 | No canopy to pass through - changes the ground factor only |
Wind and temperature gradients bend sound. Downwind - and under a temperature inversion - rays curve downwards, filling in shadow zones and carrying sound further. Upwind they curve upwards and leave an acoustic shadow. This is refraction by the atmosphere, not the wind physically carrying sound: wind speed changes the speed of sound by about 1%, which is negligible, but the wind gradient with height is not.
One honest limit. ISO 9613-2 and CNOSSOS treat upwind as the homogeneous case - never better than still air. Everest and Pohlmann note that it occasionally can be: sound travelling upwind is held above the ground, which spares it the ground losses it would otherwise take at grazing incidence. CANM follows the standards here, so a strongly upwind receiver over soft ground may read very slightly conservative.
Conditions can be averaged over a chosen date range and assessment period rather than read off a single forecast hour - which matters because temperature and wind effects "might combine to produce a larger effect, or tend to negate one another… results are unpredictable".
Wind direction is averaged as a vector, weighted by speed, never as a plain number. Directions are angles on a circle: a week of northerlies scattered around 350° and 10° averages arithmetically to 180° - due south, the exact opposite of what happened. The constancy of that mean is reported alongside it, because a mean direction taken from a wind that boxed the compass is not a direction at all. The proportion of hours actually blowing towards the receiver is counted separately, and that fraction is the p the equation above wants.
Ground is not flat, and where it is not, it screens. An earth bund is a barrier made of ground; a road in a cutting is screened by its own side slopes; a plant at the bottom of a quarry is screened by the rim. All three fall out of one mechanism rather than three special cases.
The ground profile is sampled along each source-receiver path. Where the terrain rises above the straight line of sight, the highest such point becomes a diffracting edge and is treated with the same Formula (18) construction as a built barrier - so a bund and a fence of equal effective height behave the same, which is what a designer choosing between them needs.
Terrain can be fetched for the study area rather than drawn. CANM samples a lattice over the site and asks an elevation service for each post: SRTM GL1 v3 at about 30 m first, falling back to Copernicus GLO-90 at about 90 m. Which one answered is recorded on the grid and stated in the report, because 30 m and 90 m are not the same claim.
The lattice is never finer than the data. A 400 m site is sampled on about a 13 × 13 grid, because SRTM has one post every 30 m and anything finer would be interpolation dressed as measurement. Both services cap a request at 100 points, so a full site is 41 requests and about a minute - it runs in the background and the panel reports which batch it is on.
What SRTM is not. It is an InSAR product and it measures the reflective surface: over woodland it sits somewhere in the canopy, and over dense settlement it is pulled up by the roofs. Those are not errors, they are what the data is - but a canopy read as ground becomes a bund that screens. Imported terrain is a starting point to check against local knowledge, not a survey. Nor can a 30 m post spacing carry a 4 m bund: designed features are drawn, and the DEM carries the landform they sit in.
Contours are drawn under the noise map, with the index lines carrying their height above sea level. They sit under it because the levels are the result and the ground is the context those levels fall on - the result is never drawn over.
Elevation is set per ground zone, positive for a bund or plateau and negative for a cutting or pit. Raising or lowering the whole site changes nothing - that is a datum shift, not acoustics - and a test pins it, because if it ever stopped being true every terrain result would be suspect.
One result that surprises people and is nevertheless correct: a source standing only a metre or two above a wide plateau is screened by its own plateau when the receiver is far below, because the sight line drops into the ground before it reaches the edge. You cannot see the foot of a cliff from the middle of the clifftop, and sound cannot either. Lift the source clear of the surface and the screening goes.
Sound levels are logarithmic, so they never simply add. Contributions combine as energy:
Two consequences worth internalising. Two equally loud sources give +3 dB, not double. And a source 10 dB below the total contributes under a tenth of the energy, so silencing it entirely changes the result by less than 0.5 dB - which is why the report shows each source's share and what removing it would actually buy. Treating a source that was never controlling is the most common and most expensive mistake in mitigation design.
The ear is far less sensitive at low frequency, so band levels are A-weighted before summing to a single dB(A) figure.
| Band, Hz | 63 | 125 | 250 | 500 | 1000 | 2000 | 4000 | 8000 |
|---|---|---|---|---|---|---|---|---|
| A-weighting, dB | -26.2 | -16.1 | -8.6 | -3.2 | 0.0 | 1.2 | 1.0 | -1.1 |
IEC 61672. Note the −26.2 dB at 63 Hz: low-frequency energy is heavily discounted, which is why a broadband dB(A) can hide a rumble that occupants still complain about. CANM keeps the unweighted band levels for exactly that reason.
A predicted level means nothing until it is set against a period and a limit. These are regulatory questions, independent of the calculation - the same prediction is judged against different numbers depending on who is asking.
| Period | Window | Measured | Represents | Counts as |
|---|---|---|---|---|
| Spot measurement (minutes) | from the start time | as recorded | - | taken from the start time |
| 1 hour | from the start time | 1 h | 1 h | taken from the start time |
| IFC daytime, full period (07:00-22:00, 15 h) | 07:00-22:00 | 15 h | 15 h | day |
| IFC night, full period (22:00-07:00, 9 h) | 22:00-07:00 | 9 h | 9 h | night |
| IFC daytime, 8 h sample (within 07:00-22:00) | 07:00-22:00 | 8 h | 15 h | day |
| IFC night, 8 h sample (within 22:00-07:00) | 22:00-07:00 | 8 h | 9 h | night |
| END daytime (07:00-19:00) | 07:00-19:00 | 12 h | 12 h | day |
| END evening (19:00-23:00) | 19:00-23:00 | 4 h | 4 h | taken from the start time |
| END night (23:00-07:00) | 23:00-07:00 | 8 h | 8 h | night |
| 24 hours | from the start time | 24 h | 24 h | taken from the start time |
Where “measured” is shorter than “represents”, the survey is a sample of the period and is recorded as partial rather than presented as a full-period figure.
IFC EHS splits the day at 07:00 and 22:00. The EU Environmental Noise Directive splits it three ways (07:00-19:00 / 19:00-23:00 / 23:00-07:00). The two do not line up - 22:00 to 23:00 is night under IFC and evening under END - so CANM labels each period with the framework it belongs to rather than merging them. Surveys shorter than a full period, such as an 8-hour sample inside the daytime window, are first-class options and are recorded as partial.
| Receptor | Day, dB(A) | Night, dB(A) |
|---|---|---|
| Residential, institutional, educational | 55 | 45 |
| Industrial, commercial | 70 | 70 |
IFC / World Bank General EHS Guidelines (2007), Table 1.7.1. The stated metric is the one-hour LAeq; a level averaged over a longer period is a proxy for it and is labelled as such. The guidelines also cap the increase over background at 3 dB at the nearest off-site receptor.
Selectable per project. The values below are Table 1.1 (general environment) and Table 1.8 (civil aviation) of TZS 932:2017, Acoustics – General tolerance limits for environmental and occupational noise, second edition, which cancels and replaces TZS 932:2007.
| Facility class | Day, dB(A) | Night, dB(A) | Clause |
|---|---|---|---|
| Hospital, aged care, sanatorium, learning institution, office, conference room, public library, recreational site | 52 | 42 | Table 1.1 |
| Residential building | 55 | 45 | Table 1.1 |
| Mixed: commercial, residential, entertainment, place of worship | 55 | 45 | Table 1.1 |
| Residential with industry, small-scale production or commerce | 60 | 50 | Table 1.1 |
| Industrial area | 70 | 60 | Table 1.1 |
| Nearest residence or office to an aerodrome | 60 | 50 | Table 1.8 |
| Within an aerodrome building | 70 | 60 | Table 1.8 |
Two differences from IFC matter in practice. The day starts an hour earlier: TZS 932 defines day as 06:00–22:00 and night as 22:00–06:00, against IFC's 07:00. And the industrial night limit is 60 dB(A), not 70 – IFC applies 70 to both periods, so a site judged against the wrong scheme can pass by 10 dB. The standard sets absolute limits and does not cap the increase over background the way IFC does, so that criterion is not applied unless it is added deliberately and attributed to whichever document it came from. Minimum measurement duration is 8 hours for each of day and night.
A standard is not only its tables of period averages, and the tables CANM scores are not all of TZS 932. These are carried in the report rather than left out, because a reader who sees five tables assessed will reasonably assume those were all of them.
| Clause | Applies to | Metric | Why it is not scored |
|---|---|---|---|
| Table 1.4 | Mines, quarries and surrounding areas | Lmax, dB(C) | Not from the propagation model - a blast is an impulsive event whose energy sits below the lowest octave band the engine resolves, and it is judged as a peak rather than an energy average. It is answered instead by the airblast model (Blasting), which predicts peak overpressure from the cube-root scaled distance after USBM RI 8485. Because Table 1.4 is written in dB(C), that site must be monitored in dB(C) and its own site factors fitted: the published coefficients predict an unweighted peak, and the two differ by tens of decibels for the same shot. |
| Table 1.2 | Factory and workplace exposure | LAeq against daily and weekly duration | Occupational exposure is what a worker receives over a shift, measured with a personal dosimeter. It is not a level at a fixed position, so a propagation model cannot produce it. The 75 dB(A) compound limit in the note under that table IS checked, because it applies at a position. |
| Table 1.3 | Impact or impulsive noise | Lmax, dB(A), against a permitted count per day | A limit on peak level and the number of events, which needs the count and the peak of each blast or impact. This model produces an energy average, not an event count. |
| Tables 1.5, 1.6, 1.7 | Public address systems, places of worship, vehicles | LAeq, dB(A) | These apply to a specific device or vehicle rather than to a site, and they are assessed by measuring that device. Model one as a source if you need to, then compare against the table yourself. |
Table 1.4 is the one to watch. Its title is mines, quarries and surrounding areas, so it looks like the table a quarry assessment needs – but it is Lmax in dB(C), and this model predicts LAeq in dB(A). Those are a different weighting and a different statistic, and no conversion between them exists that does not depend on the spectrum and crest factor of the event. Comparing a predicted dB(A) average against a 109 dB(C) peak limit gives a comfortable pass that means nothing at all. It has to be measured with a meter set to C weighting.
On whether a calculated result is admissible at all: Annex B.4 lists what a compliance report must contain, and item (g) covers “measurements or calculations” while item (i) requires the calculation method to be stated. A predicted assessment is contemplated by the standard, provided the method is named.
Two standards are shipped; the rest of the world is not, and guessing a limit for a regulation nobody here has read would put an invented figure into an assessment that gets defended in public. So the third scheme ships deliberately empty, with the same seven facility classes as the populated ones so that a five-row or six-row national table can be transcribed rather than rounded to the nearest two. Enter the values, the day and night boundary, and the name and clause of the regulation they come from; the report cites exactly what was entered. A class left blank is reported as having no stated limit rather than being judged against a borrowed one, and where a shipped standard is varied – by a permit condition, say – the report names the departure and prints both the published and the applied value.
ISO 1996-2 specifies a Class 1 sound level meter (IEC 61672-1) for its reference and engineering methods. A Class 2 instrument is nominally ±1.4 dB at 1 kHz against ±1.1 dB for Class 1, with tolerances widening at the ends of the range, and many Class 2 meters have no octave filters at all. CANM records the class alongside every reading, and a broadband-only survey is used in full for the broadband comparison.
Everything above predicts a level. BS 4142:2014+A1:2019 rates one: it takes the sound under investigation, adds a correction for how objectionable its character is, and compares the result against the background that would be there anyway. The output is an excess, not a level.
Around +10 dB indicates a significant adverse impact and around +5 dB an adverse impact, depending on context - which the standard says repeatedly and in terms. These are indications, not thresholds.
Two details that decide whether an assessment is a BS 4142 assessment at all. Its night period runs 23:00-07:00, not from 22:00 as IFC does, so the hour between is night under one and day under the other. And its reference time interval is 15 minutes at night, one hour by day: a night level averaged over eight hours is not a rating level, and averaging that long hides the short loud events the interval exists to catch.
A prediction on its own answers the wrong question. Nobody hears the specific sound in isolation - they hear it on top of everything already there.
Everest and Pohlmann put the consequence plainly, about a studio but true everywhere: there is little point demanding a low level from one system when intrusion from elsewhere is already higher than it. Controlling a source well below the background is money spent on something nobody can hear.
This is also why two tests are needed rather than one. The same plant assessed at three receivers: beside a quiet lane it raised the environment 6.7 dB and failed the increase test while passing the absolute limit; beside a road it raised it 0.01 dB and failed the absolute limit - but that limit was already breached by the existing traffic, not by the plant. Only having both separates those.
One trap worth stating outright: LA90 and LAeq are not interchangeable. LA90 is the quiet floor exceeded 90% of the time; LAeq is the energy average and sits several decibels above it in any varying environment. Energy-summing onto LA90 understates the environment; using LAeq as a BS 4142 background overstates it. Each does one job here and they are never swapped.
Everything above is written for a strategic map: a district, a corridor, receivers hundreds of metres from the source. Modelling one industrial site is a different job, and three things change.
Site settings carry a study scale. It is a modelling decision, not a performance setting: it decides how finely an extended source is discretised, and therefore whether it still behaves as one at the distances your receivers actually sit.
| Strategic | Detailed | |
|---|---|---|
| Sub-points along one line source | 24 | 240 |
| Sub-points over one area source | 80 | 400 |
| Screening search radius | 500 m | 250 m |
| Largest raster side | 100 | 160 |
A 200 m conveyor split into 24 sub-points is a string of beads 8 m apart. A boundary position 3 m away sees the beads rather than the conveyor, and the level it gets depends on whether it happens to sit opposite one. On the raster that is worth about 1 dB a few metres out and nothing at 60 m - measurable, and in the direction that flatters the site if you leave it coarse.
Discrete receivers are already safe from this: the line is subdivided adaptively where a receiver is close to it, whatever the budget. It is the map near the plant that needs the finer setting.
This is the one that moves results most, and it is easy to miss because nothing on screen looks wrong without it. ISO 9613-2 clause 6 is
| Where the machine sits | Ω | Dc |
|---|---|---|
| Free field, nothing alongside | 4π | 0 dB |
| Against one surface - a building wall | 2π | +3 dB |
| In an edge, two surfaces | π | +6 dB |
| In a corner, three surfaces | π/2 | +9 dB |
A fan bolted to a wall is 3 dB louder at every receiver than the same fan standing in the yard, before any propagation happens at all. Nothing downstream recovers that, because it was never in the sound power. On a site where half the plant is against buildings, leaving every source in free field under-predicts the whole model.
ISO 9613-2 notes that where the source sits closer to the reflector than a quarter wavelength the superposition is coherent and the correction can reach twice these values. At 63 Hz a quarter wavelength is about 1.4 m, so a machine bolted hard to a wall can exceed +3 dB in the low bands. That is not modelled; where it matters, enter a measured directivity instead.
The Layers panel reports any receiver standing closer to a source than the sub-sources it is modelled with, and any receiver inside the 1 m distance floor where the level is held rather than computed. Both are normal on a site study and both have the same two remedies: raise the study scale, or move the assessment position to where a measurement could actually be taken.
ISO 9613-2 is an engineering method built on point sources. Very close to a large machine there is no point source to be had, and no setting in this tool changes that - a facade 2 m from a fan intake is a near-field problem, not a propagation one.
First-order reflections are off by default because on a strategic map they cost a great deal and change little. Between the buildings of a single site they are part of the answer: a yard enclosed on three sides behaves nothing like an open one. Site settings turn them on, and the order controls how many surfaces a ray may bounce off.
Two British Standards that answer questions ISO 9613-2 does not. Neither is a propagation method - the engine still predicts the level. What they add is what the level means.
Construction noise has no fixed limit, and it should not have one: a level that is unremarkable beside a motorway is intolerable in a quiet village. Annex E therefore sets the threshold from the existing environment. Measure the ambient level, round it to the nearest 5 dB, and read off the category:
| Rounded ambient | Category that applies |
|---|---|
| Below the Category A value | Category A |
| Equal to the Category A value | Category B |
| Above the Category A value | Category C |
The rounding is not cosmetic. An ambient of 63 dB is below the 65 dB Category A value but rounds to it, so Category B applies - a 5 dB difference in the threshold decided entirely by the rounding step. Reading the category off the unrounded level is the commonest slip in the method.
Annex E indicates significance. It is not a consent limit: a section 61 consent or a local authority limit is set separately and may be tighter.
The method above is the standard’s. The threshold values shipped are the published Category A minima of 65 / 55 / 45 dB for the day, evening and night periods, with B and C at +5 and +10 dB. They are editable, and the report cites whatever is used - confirm them against Table E.1 in your own copy of the standard before issuing an assessment.
Table 4 gives guideline values for dwellings, and every one of them is an indoor level. The model predicts outdoors, so applying the table directly to a prediction would fail every dwelling in the country and mean nothing. The facade has to come off first:
| Activity | Location | 07:00-23:00 | 23:00-07:00 |
|---|---|---|---|
| Resting | Living room | 35 dB LAeq,16h | - |
| Dining | Dining room/area | 40 dB LAeq,16h | - |
| Sleeping (daytime resting) | Bedroom | 35 dB LAeq,16h | 30 dB LAeq,8h |
The blanks are meaningful. Table 4 sets no night value for a living or dining room because they are not slept in, and that gap must not be filled with the bedroom figure.
The window decides most assessments. A facade level of 58 dB gives 25 dB indoors behind closed standard glazing and 40 dB with the window open - a pass and a clear fail from the same prediction. NOTE 5 to Table 4 is explicit that if closed windows are relied on, there must be alternative ventilation that does not compromise the facade.
NOTE 7 allows the internal targets to be relaxed by up to 5 dB where development is considered necessary or desirable despite external levels above the WHO guidelines. That is a planning judgement rather than an acoustic one, so the relaxation is entered explicitly and stated in the report.
For gardens and patios, 7.7.3.2 gives a desirable external level of 50 dB LAeq,T with an upper guideline of 55 dB “which would be acceptable in noisier environments”. The clause also says these are not achievable everywhere and that development should not be prohibited on them alone - so CANM reports which band a level falls in rather than passing or failing it.
Table 4 covers noise without a specific character (7.7.1). A distinguishable tone, an irregular source, or strong low-frequency content may warrant lower values than the table gives.
Aircraft noise is computed by a different engine, not by a variation on the one described above. Every heading in this section exists because something about aircraft breaks an assumption the ground method depends on.
The ground method starts from a sound power level and subtracts attenuation. None of that framing survives contact with an aeroplane. The source moves, so what a resident experiences is an event and the quantity that matters is its integral - sound exposure level - not an instantaneous value. The source strength changes along the path, because thrust does. And there is no usable sound power level at all: aircraft noise is tabulated empirically as noise-power-distance curves, measured level against slant distance at a given engine power. An NPD curve already contains the source strength, the spherical spreading, the atmospheric absorption of a reference atmosphere and the event duration, bundled together. You cannot take it apart, and you must not add those terms again.
CANM implements the segmentation method of ECAC.CEAC Doc 29, 4th Edition, Volume 2, which is the European statement of the method ICAO Doc 9911 recommends.
Put a sixty-metre building between the runway and a receiver and the contour does not change by so much as a hundredth of a decibel. This is the most common thing to be reported as a bug in an aircraft model, and it is not one. Doc 29 says so twice:
The second is the stronger reason. The curves were measured, and normalised, over ground with nothing standing on it. Subtracting a screening term from them would not be adding missing physics - it would be correcting the data away from the condition at which they are defined. A barrier drawn on the map still works; it works on the engine entitled to apply it, which is the ground one.
Taxiing, engine testing and auxiliary power units are excluded from the air noise procedure by Doc 29 §1.1, which directs them instead to an industrial propagation model. That is the other engine in CANM, and in it barriers, earth bunds, noise pens and buildings all attenuate normally. If run-ups or ground movements are in scope, model them as ground sources on the same site - and read the two results separately, because they are separate assessments.
Doc 29 §2.4.1 defines the aerodrome reference altitude as the altitude of the nominal ground plane on which the contours are drawn. So an aerodrome 55 m above sea level is not quieter for it: its neighbours are 55 m up as well, and z = 0 is at both. The runway’s height above the ground plane is therefore normally zero, and is non-zero only where a runway genuinely stands above the rest of the field. The aerodrome altitude is recorded separately, for the record, as §2.4.1 requires.
Terrain is treated as pseudo-level per §2.4.5 and §3.3.4: receivers are taken to lie on that plane unless a height above it is entered. Doc 29 notes this matters most under approach tracks, where aircraft are low.
An aircraft contour is an aggregate of many movements over a day, weighted by time of day. It is not an LAeq from steady plant and cannot be added to one, compared with one, or read against a limit written for one. Each index is banded on its own reporting thresholds, which is why the legend names the index rather than saying “dB(A)”:
| Index | What it is | Bands |
|---|---|---|
| Lden | Day-evening-night level; evening +5 dB, night +10 dB | Directive 2002/49/EC reporting thresholds, 55 dB then 5 dB steps |
| Ldn | Day-night level; night +10 dB | 5 dB steps; 65 dB is the threshold in common use |
| LAeq day / night | Equivalent level over the stated period | 5 dB steps; the night scale starts lower, at the 2002/49/EC Lnight threshold of 50 dB |
| SEL | Sound exposure level of the loudest single movement | Its own scale from 75 dB - a single event runs 30 to 50 dB above any daily index |
| NAT | Number of movements above a stated LAmax | A count, not a level, so it is banded as one |
These bandings are fixed by the metric and are not offered in the banding chooser, which applies to ground maps only. Doubling the movements adds 3 dB to any of the energy-based indices.
Doc 29 §1.1 is explicit that the methodology applies only to long-term average noise exposure and “cannot be relied upon to predict with any accuracy the absolute level of a single aircraft movement”. A contour is an average day, not a peak day, and not a single flight.
Where a movement is flown by a type with no published noise-and- performance data, a reference aircraft stands in for it. Every result that used one says so, on screen and in the report: substitutes are adequate for scoping and are not adequate for an assessment that will be issued.
Blasting is modelled separately from every other source in CANM, in its own panel and by its own method. That separation is deliberate and it is worth understanding before you use it, because the two models answer different questions and their numbers are not interchangeable.
Quarry and mining practice has its own vocabulary. The panel uses it because that is what appears on a blast record you will be handed, but none of it is assumed knowledge.
| Term | What it means |
|---|---|
| Shot | One blast event. A set of drilled holes, loaded with explosive and fired together as a single operation. Also called a round, or simply a blast. A quarry might fire one shot a fortnight, and each one lasts under a second. |
| Delay | Holes in a shot are not fired at the same instant. Detonators separate them by a few milliseconds so the rock breaks in sequence. Each delay is effectively its own small explosion. |
| Charge per delay | The explosive weight that detonates within one delay window, not the total loaded into the shot. This is the number the model needs. A 500 kg shot fired on twenty delays behaves like twenty 25 kg shots, which is the entire purpose of delay detonators. |
| Stemming | Inert material packed into the top of a loaded hole to contain the gases. Poor or short stemming vents the blast to air and is the most common cause of an unexpectedly loud shot. |
| Confinement | How well the charge is contained. A fully buried charge is far quieter in air than the same charge on an exposed quarry face, which is quieter again than one detonated in the open. The panel calls this the blast type. |
| Airblast, or overpressure | The pressure wave the shot pushes into the air. It is what this model predicts. Distinct from ground vibration, which travels through rock and is measured separately in mm/s. |
| Scaled distance | Distance divided by the cube root of the charge per delay. Two shots at the same scaled distance produce about the same overpressure, which is what lets monitoring at one distance and charge predict another. |
Everything else in CANM is an energy average of a continuous sound, resolved into octave bands from 63 Hz upward and propagated by ISO 9613-2. A blast fits neither half of that description. It is a single impulsive peak rather than an average, and the bulk of its energy sits below the lowest band the engine resolves. Running it through the propagation chain would produce a number, and the number would mean nothing.
Airblast is instead predicted from the cube-root scaled distance, the standard empirical form in USBM RI 8485 and the ISEE Blasters’ Handbook:
P = a · (D / W1/3)−b
where D is the distance to the receiver, W the charge per delay, and a and b are site factors depending on confinement and terrain. CANM ships published values of a and b for each blast type, and can fit your own from monitoring.
This is the part that most often goes wrong, and CANM refuses rather than guesses.
The published coefficients predict an unweighted, linear peak. TZS 932:2017 Table 1.4, the table for mines and quarries, is written in dB(C). For the same shot those two differ by tens of decibels: United States practice caps the linear peak at 133 dB and the C-weighted peak at 105 dB(C) for the same blasting operation. Comparing a linear prediction against a dB(C) limit gives a comfortable pass that means nothing at all.
So the model will not convert between weightings, and will not score a prediction against a limit written in a different one. When the two do not match it returns a refusal naming the mismatch instead of a verdict. The route to a Table 1.4 answer is to monitor that site in dB(C) and fit its own coefficients, which is what the method asks for in any case, since a and b are site factors rather than universal constants.
Screens drawn on the map are not applied to an airblast prediction, and the panel says so whenever any exist on the site. There are two reasons, and both are structural rather than a limitation waiting to be lifted.
First, there is nowhere to put the correction. The scaled-distance relationship has no screening term. A peak overpressure is the height of a shock front arriving at an instant, not a sum of energy across octave bands, so a band-by-band insertion loss cannot be applied to it at all. Any reduction inserted here would be invented rather than computed.
Second, the physics does not support one. Barrier attenuation depends on the path difference measured in wavelengths. The dominant energy in airblast sits between roughly 1 Hz and 20 Hz, and at 10 Hz the wavelength is about 34 m. A 4 m bund or a row of houses is a small fraction of a single wavelength, so the wave diffracts around it instead of casting an acoustic shadow. The Fresnel number for that geometry is of the order of 0.08, which is not a screen in any meaningful sense. A barrier that works well against plant noise at 500 Hz does almost nothing against a blast.
Buildings still matter to a blasting assessment, but as receptors rather than as screens. Structural response, window rattle and the damage criteria are about what the pressure wave does on arrival, not about what it did on the way.
Mitigation for blasting is designed into the shot. It is not built between the shot and the receiver.
Airblast from one shot depends on distance and charge alone. There is no directivity, no screening and no ground term, so a contour of equal overpressure is a circle centred on the shot. CANM draws it as one rather than as a computed surface, because a rendered raster would imply the method knows something about the ground between the shot and the receiver when it does not.
The red rings are the limits applying on this site; the dashed rings are guide levels, listed with their radii in the panel. A receiver outside a red ring is below that limit. The radius is the honest output here: it turns a decibel limit into the separation distance the operation needs, which is the number a management plan or a permit condition can be written around.
Every figure in this section is either quoted from the standard that fixes it, or computed by CANM on this page load. Nothing here is a stored number, because a stated accuracy that is not re-derived is one that goes quietly out of date the first time a coefficient is touched.
| What is modelled | Verified against | Result |
|---|---|---|
| Point, line, area and volume sources | Energy conservation: the sub-source cloud must radiate exactly the sound power entered | Exact to ±0.00 dB, and independent of how finely the source is discretised |
| Divergence, ground, screening, air absorption, foliage | The ISO 9613-2 formulae, evaluated independently | Each term reproduces its formula; Agr matches Formula (14), hard ground gives −3 dB, screening obeys Formulae (16)/(17) |
| Overall outdoor propagation | ISO 9613-2:2024 Clause 9, Table 4 | ±1 to ±3 dB depending on height and distance — see below |
| Aircraft noise | ECAC Doc 29 Volume 3 reference cases, which publish the answers in advance | ±1.08 dB mean, ±1.91 dB worst case, 11 of 11 receptors within 2 dB |
The standard’s favourable case is downwind propagation - a wind of roughly 1 to 5 m/s blowing from source to receiver, measured 3 to 11 m above the ground - or a moderate ground-based temperature inversion. Outside that window the favourable branch is being used past what it was derived for, which matters most at long range in strong winds.
ISO 9613-2 gives the expected accuracy of the broadband A-weighted level it predicts, as a function of mean source-to-receiver height and distance:
| Mean height h | to 100 m | to 1000 m |
|---|---|---|
| 0 < h < 5 m | ±3 dB | ±3 dB |
| 5 ≤ h < 30 m | ±1 dB | ±3 dB |
That is the method’s own figure, and it assumes the inputs are right. In practice the sound power level you put in dominates: a source level 3 dB out moves every receiver it reaches by 3 dB, which is larger than most of the propagation terms being argued over.
Four qualifications the standard attaches to Table 4, which matter when the number is quoted in an assessment:
Aircraft noise is not covered by ISO 9613-2 and has its own benchmark. ECAC Doc 29 Volume 3 publishes reference cases with the answers worked out in advance, so an implementation can be checked rather than taken on trust. CANM runs both of them — a departure and an arrival — and compares against every published receptor:
| Measure | Result |
|---|---|
| Mean absolute deviation | ±1.08 dB |
| Worst single receptor | ±1.91 dB |
| Receptors within 1 dB | 5 of 11 |
| Receptors within 2 dB | 11 of 11 |
Two further properties are worth as much as the deviation itself. The result is stable to 0.003 dB across an eightfold change in internal segment size, so it does not depend on a setting nobody documented. And the model reports what it could not compute: traffic with no flight profile, and receivers whose level rests on the part of the method still being closed, are named rather than absorbed into the total.
The full receptor-by-receptor comparison, and twelve coefficient checks against points where the standards fix the answer, are on the aviation modelling and validation page. It recomputes on every load, at aircraft method revision 2026-08-25-a.
The same guidance the ? button shows beside each tool in the editor, gathered here. It is written once and rendered in both places, so a tool and its instructions cannot drift apart.
A polygon that says what the ground is made of and what height it sits at. Two tools in one: surface type, and terrain.
Step by step
What goes wrong
A pit is not a special object - it is ground at a lower elevation. The rim then screens whatever is below it.
Step by step
What goes wrong
A screen that forces sound to bend over its top edge. It works by geometry, not by a dB rating.
Step by step
What goes wrong
A source spread along a line. Split into sub-sources so it radiates like a road rather than a point.
Step by step
What goes wrong
A single machine, radiating a known sound power. The building block everything else is made of.
Step by step
What goes wrong
A source spread over a surface - a yard, a stockpile, a working face, a building facade radiating from inside.
Step by step
What goes wrong
A solid block that screens, reflects and blocks sound. On a built-up site it does more work than any barrier you add.
Step by step
What goes wrong
The place an assessment is made. Attach measurements to it, and the prediction can be checked and combined with reality.
Step by step
What goes wrong
The strip every flight path is measured from. It makes no noise itself: it fixes where a take-off starts rolling and where a landing touches down.
Step by step
What goes wrong
The ground track an aircraft follows to or from the runway. The contour is a smear along this line, so its shape is this line.
Step by step
What goes wrong
The boundary the noise map is computed over, and the extent every result is reported across.
Step by step
What goes wrong
| Standard | Role in CANM |
|---|---|
| ISO 9613-2 | Core outdoor propagation method: divergence, ground effect, screening, foliage |
| ISO 9613-1 | Atmospheric absorption, evaluated at exact octave centre frequencies |
| CNOSSOS-EU (2015/996) | Dual homogeneous/favourable computation, mean ground plane, lateral diffraction |
| ISO/TR 17534-4 | Equivalent heights, retrodiffraction, exact band centre frequencies |
| ISO 3744 / 3746 / 8297 | Determining source sound power levels |
| ISO 1996-1 / 1996-2 | Describing and measuring environmental noise; baseline surveys and validation |
| IEC 61672-1 | Sound level meter classes, and A-weighting |
| EN 1793-1 / 1793-2 | Barrier absorption and airborne sound insulation |
| BS 4142:2014+A1:2019 | Rating and assessing industrial and commercial sound; character corrections and the comparison against background |
| IFC EHS Guidelines (2007) | Assessment periods and noise limits, Table 1.7.1 |
| DIN 18005-2, NF S 31-130, 2002/49/EC | Noise map colour banding schemes |
The equations are those of the standards listed in the standards index above. Where a standard offers more than one route, the clause taken is named at the relevant section of this page rather than left to be inferred.
| Data | Source |
|---|---|
| Buildings, roads, land use | OpenStreetMap, via the Overpass API |
| Place search | Nominatim |
| Road routes for line sources | OSRM |
| Temperature, humidity, wind | Open-Meteo |
OpenStreetMap data is © OpenStreetMap contributors, available under the Open Database License.
The banded map palettes follow DIN 18005-2, NF S 31-130 and the END (2002/49/EC) scheme, alongside the CANM default. A custom palette can set its own bands and colours where a client or a regulator specifies them.
Maps are drawn with Leaflet over OpenStreetMap tiles. The solver is NumPy; report figures are Matplotlib. Icons are Lucide.
CANM - Clanert Acoustic Noise Model. Developed by Nobert Mrema and Eduward Gumbwa, Environmental Engineers, Clanert Sustain.
