One number from 0 to 100 for any address, built from public data about what makes noise nearby. This page explains what goes in, where it comes from, how it is combined, and how we check it. Every number in a report can be traced to a source.
Each address is scored on seven things, separately, each on a 0–10 scale:
| Axis | What counts | Weight |
|---|---|---|
| Aircraft | Flights actually tracked overhead, how many a day and how high | 22% |
| Resident reports | Ratings from verified residents within 100 m | 22% |
| Venues & crowds | Pubs, bars, clubs and restaurants and how late they trade; malls, markets, beaches | 18% |
| Road traffic | How many vehicles a day, how close, and what stands in between | 16% |
| Rail & trams | Timetabled trains and trams a day, freight lines, level crossings | 12% |
| Industry | Registered industrial premises and ports | 10% |
| Power lines | High-voltage transmission lines, substations and data centres | 6%, only when one is nearby |
Each axis adds up the sources within reach and squashes the total onto its 0–10 scale, so a second source of the same size adds less than the first. A road, a rail line or a flight path is one thing even when the map shows it as many pieces: the loudest piece sets the level and the rest add a little.
We only use data we are licensed to use and can point to. Nothing is invented: where a source does not exist for an area, the report says "estimated" and explains what the estimate is.
| Axis | Source | Where it is estimated |
|---|---|---|
| Aircraft | Live aircraft positions from the adsb.fi community network, sampled every two minutes around 33 airports and built into flight corridors per airport over weeks. Each corridor is drawn along the flight most like the others in it, its height at every point comes from all the flights that used it, and its near end runs on to the runway threshold it lines up with (the airport's centre where it lines up with none) because a two-minute sample last sees a jet a few kilometres out. Below that the geometry is the runway's, not the data's: an arrival is drawn on the extended centreline for the final 8 km at a 3° glide, a departure straight ahead for 2 km at an 8% climb, and the corridor then curves out to the first point the data does see rather than cutting a straight line across the suburbs between. Parallel runways, such as Sydney's, get separate corridors: a flight is matched to the runway whose extended centreline it is on. Departures are grouped by the runway end they leave from and the direction they head, even when the first sample catches them after their turn, because take-offs fan out and grouping them by exact line split them into groups too small to keep. Aircraft report their height above sea level, so heights are taken above the airport's own ground: Canberra's airport sits 575 m up, and reading sea-level heights there put every flight 575 m too high. Smaller airports need fewer flights a day to form a corridor (2% of the airport's daily movements, and never fewer than 20 flights in all), so a regional airport's routes are not dropped for being quieter than Sydney's. Helicopters and light aircraft form their own, quieter corridors rather than being mixed in with jets. The ground side counts too: each runway carries the movements that use its ends as a source at ground level (the take-off roll, the landing, reverse thrust), and each apron, outlined from OpenStreetMap, carries the airport's movements in proportion to its area at half weight (taxiing, engine starts, auxiliary power units), and each taxiway carries them in proportion to its length at the same half weight. Noise made on the ground is much quieter to the side than along the runway, because it reaches you along a grazing path that the ground itself absorbs, so a house beside a runway is given up to 5 dB less than one off the end at the same distance | Corridors depend on volunteer receivers, which cover the capitals well and some regional airports thinly or not at all. Where the live feed sees less than 40% of an airport's published movements, or its corridors keep less than half of what it does see, the report uses an estimate instead and says so: the runway's extended centreline from OpenStreetMap, at the standard 3° approach slope, carrying the airport's published daily movements shared across its runway ends. Measured corridors replace the estimate as tracks accumulate. Defence bases do not report to that table and military jets rarely broadcast positions, so Defence bases are built from Defence's own published figures: Williamtown from the F-35A environmental impact statement and BITRE's airline movements, Amberley from the Super Hornet and Growler noise management plan's 7,700 movements a year, Tindal from the F-35A statement, Pearce from its noise plan's 140 military movements a week, and Edinburgh, East Sale, Richmond, Nowra and Oakey from stated estimates where Defence publishes no total. The fast-jet share at each is carried as its own corridor and weighted 15 dB above an airliner, the conservative end of Defence's published noise levels |
| Road traffic | Government traffic counts: Transport for NSW, VicRoads, Queensland TMR, Main Roads WA, SA Department for Infrastructure and Transport, placed on OpenStreetMap road geometry | Roads with no published count use a typical volume for their class and say so. Tasmania and the Northern Territory publish no counts |
| Rail & trams | Published timetables (GTFS) from Transport for NSW, Public Transport Victoria, TransLink, Transperth, Adelaide Metro and Transport Canberra, counted per corridor on a normal weekday, on OpenStreetMap track geometry | Freight lines, and passenger lines outside south-east Queensland, Perth and Adelaide that have no public timetable, use a typical figure |
| Venues | Liquor licence registers in NSW and Victoria, including licensed closing times | Other states publish no register, so venues come from OpenStreetMap with mapped opening hours, labelled as such |
| Crowds | OpenStreetMap malls, retail strips, markets, attractions and beaches, with hour-of-day profiles measured from the City of Melbourne's pedestrian counters | Beaches have no counter; their profile follows daylight |
| Industry | The National Pollutant Inventory of registered facilities | — |
| Power lines | NSW Spatial Services' surveyed transmission lines and substations in NSW; Geoscience Australia's national transmission lines elsewhere; OpenStreetMap substations, lines and data centres for whatever those lack | Surveys outlive lines. When a line has been put underground or taken down for a new estate, all three sources can still carry it overhead (Whitlam in the ACT, where Evoenergy undergrounded the Molonglo 132 kV line and removed 18 towers from 2022); such stretches are cut from every source against the public record. The national set also draws some buried cables as a straight chord between substations and marks them overhead; a line of two kilometres or more with fewer than 1.5 vertices per kilometre is a chord, not a line, and is dropped. And a 132 kV line on steel poles down a suburban street, with a house every 30 m along it, is not a tower easement: lines with 15 or more mapped roofs per kilometre within 20 m are weighted at half and the report says so |
| Addresses | G-NAF, the Geoscape national address file, every state and territory | — |
| Planned development | NSW Planning Portal development applications: what is proposed nearby, its cost, status and exhibition dates. Barely moves today's score; drives the change alerts, with the exhibition end as the objection deadline | NSW only; other states publish no equivalent feed |
Where a government survey and OpenStreetMap describe the same thing, the survey wins and OpenStreetMap fills the gaps. Every source is named in the report next to the number it produced.
Feeds are refreshed on their own schedules, from every two minutes for aircraft to every few months for road geometry. If a feed stops answering, its last good data stays in place and the report keeps working; the operators are alerted when any feed has gone quiet for longer than twice its normal schedule.
Noise fades with distance, but not all the same way. A road or a railway is a line, so its sound spreads sideways and drops about 3 dB each time you double the distance. A pub or a factory is a point, and drops about 6 dB per doubling. The model uses the standard formulas from road and rail noise assessment rather than guesses.
Traffic volume is counted the way decibels work: doubling the traffic adds about 3 dB, it does not double the noise. Buildings in the way matter a great deal. Within about 25 m of a road's centreline nothing is between you and it. One row of houses in between takes off 6 dB, coming in gradually over the 30 m past the frontage line rather than all at once, since the first few metres are side passages and front gardens, not a wall; two or more rows take off 12 dB. Those figures come from the UK road noise standard, which gives about 6 dB for a row with the gaps real streets have, US highway barrier guidance and measured front-versus-back-row differences. Motorways sit in wide reserves, so their screening starts further out.
Where building footprints are loaded, the report stops inferring rows from distance and measures them: it draws the line from the address to the nearest point on the road, and four more lines to points 75 and 150 m along the road either side, and counts the buildings actually standing on each. A building has to sit at least 3 m across the line to count, so a clipped corner is not a wall. Sound arrives from the whole stretch, so the five lines are averaged as energy: one open driveway among four screened lines gives back about 4 dB of the 6, not all of it. None on any line means open ground even at 60 m; one is the row; two or more the deep figure. The same measurement applies to rail corridors, with rail's own row figures. Footprints are loaded for the whole country from Microsoft's open building footprints (every roof, traced from imagery, February 2026 release), and re-checked against Microsoft's index every quarter so tiles they have re-traced are reloaded and new estates get their roofs; the map tiles keep the distance rule, and every report says which method it used.
Several roads within earshot combine like energy, not arithmetic: the loudest sets the level and the others add a small amount. Without this, an ordinary grid suburb would read far louder than anyone living in it experiences.
| Source | Effective reach | Basis |
|---|---|---|
| Collector street | ~45 m | ~60 dB(A) at the kerb; the first row of houses all but ends it |
| Main road | ~60 m | ~65 dB(A) |
| Arterial | ~100 m | ~70 dB(A) starting level |
| Motorway | ~250 m | ~78 dB(A); audible well past 250 m even when screened |
| Passenger rail | ~130 m | US transit reference of 88 dB(A) per pass-by at 30 m |
| Freight rail | ~300 m | Longer trains, low-frequency rumble, night running |
| Pub or venue | ~60 m | Point source, 6 dB per doubling |
| Stadium | ~800 m | PA and crowd; complaints are documented beyond 1 km |
| Heavy industry | ~350 m | Plant hum carries at night |
| Transmission line | ~100 m | Corona hum in damp weather; the easement itself is the point |
The same logic as roads, anchored to the US Federal Transit Administration's method. The number of trains counts logarithmically, so doubling the service adds about 3 dB. No timetable lists freight, so freight comes from a table of named corridors with movements a day from the operators' own figures: the Hunter Valley coal chain at about 128 coal train movements a day between Muswellbrook and the Newcastle terminals from ARTC's corridor strategy; the Goonyella, Blackwater, Moura and Newlands coal systems in central Queensland and the West Moreton line, derived from Aurizon's and Queensland Rail's published tonnages; the interstate lines between the capitals, the Illawarra and Main West lines, the Adelaide Hills and Avon Valley lines and the port approaches in each city, from ARTC corridor figures where they exist and marked as estimates where they do not. Each corridor is traced onto the track and carried on top of the passenger timetable, and the report names it separately from the passenger line, with its source. The timetable supplies that number for every passenger corridor, so a suburban main line with 500 movements a day and a branch with 100 read differently. The first row of houses removes 4.5 dB and two or more rows 12 dB, measured from building footprints along the corridor the same way as for roads; a track in a cutting is quieter again. Level crossings are the exception: the horn is elevated and directional, so it is not screened. Freight adds about 4 dB for train length and night running. A level crossing adds a horn penalty over a few hundred metres, because horns are 15–20 dB louder than the train and are sounded on purpose. Trams are lighter and quieter and are weighted accordingly.
Aircraft noise depends on the straight-line distance to the aircraft, combining how far to the side it passes and how high it is. Full weight applies within about 400 m of that line; beyond it the weight falls off, more gently than a single aircraft would because a corridor stands for a spread of real flights a kilometre or so wide, not one line. The number of flights a day counts logarithmically, like trains. Corridors are built from real tracked flights, so a flight path that moves with the wind is counted where the aircraft actually went. The distance and count settings were set against Airservices Australia's published Sydney N70 chart, which shows how many flights a day exceed 70 dB at each point in the city.
What the 0–10 bar means for aircraft: about 10 flights a day passing close overhead reads 3 to 4, which is the edge of the N70 chart's lowest band; 25 a day reads about 5; 60 a day reads about 7; the stacked approach corridors nearest Sydney Airport read 8 to 9. A jet at 1,500 m is still clearly audible, so a suburb 20 km out under an arrival stream registers at the low end rather than at zero, even though few of those flights exceed 70 dB.
An average would let five quiet axes bury one loud one, and nobody experiences noise as an average. A home under a flight path is loud no matter how quiet its street is. Research on people exposed to several noise sources at once finds that annoyance is driven mainly by the loudest source, with the others adding only modestly. The score is built that way.
M₃ is a weighted average that already leans toward the loud axes. The final pull means an address with a maxed-out aircraft axis reads 90 or more regardless of its street.
Not all decibels annoy equally. The World Health Organization's 2018 guidelines found people are markedly more bothered by aircraft noise than by road traffic at the same sound level, and slightly less by railway noise. When an axis is already loud, aircraft gets a bonus of up to one point in that final pull and rail a small discount.
The bands are: 0–39 genuinely quiet, 40–59 some noise, 60–79 noticeably noisy, 80 and above loud.
Noise at 2 am is not noise at 2 pm. Every source carries an hour-by-hour profile, shown in the report. Venue weighting rises steeply with closing time, so a 1 am licence carries about four times the weight of a 10 pm one. On top of the base score you can choose a persona, such as shift worker, working from home, family or night owl, which re-weights the hours for how you live. The data does not change, only the lens.
Residents rate only their own address: a number and fixed tags, no free text, one rating per phone-verified account. The phone has to be at the address: within 25 m of it plus whatever accuracy the phone reports, widened by the footprint of a tower or townhouse row that shares the address point, never past 150 m. Ratings pool within 100 m of an address, so what counts is the immediate neighbours, not a street three blocks over. A rating counts from day one but never alone. Every rating is pulled toward what the other sources say, so one voice on a street the data calls quiet moves the number modestly and more voices earn more weight. A rating far outside what its neighbours report is capped rather than averaged in. Owners and landlords can dispute the pooled ratings on an address.
The model is tested against data it was not built from, and the results are kept.
Every report carries a confidence grade, high, medium or provisional, based on how many independent sources contributed, how long the aircraft observation window is, and whether residents have rated the street. There is no demonstration data anywhere in the system: an axis either has live data or says it has none.
It is a statistical estimate for information only. It is not an acoustic survey, it does not measure sound at your window, and it is not a substitute for visiting a property at the times of day that matter to you. Coverage is currently Sydney, Canberra, Melbourne, Brisbane, the Gold Coast, the Sunshine Coast, Perth, Adelaide, Hobart and Darwin. The score and this methodology are free and are never paywalled.