← InstrumentHow it works
Can a place retain its existing meanings while taking on new ones through sound? This instrument leaves that question open. Listeners bring memories, associations, and histories that environmental data alone cannot capture.
Choose a place and time
Choose a named place from the results, or confirm coordinates and an IANA timezone. The default is a stated point at Clemson University. Search sends the place you type, not your device location. Live mode uses the latest acceptable records; a past moment uses ERA5 reanalysis. Nonexistent spring clock times are rejected, and repeated autumn hours use the first occurrence.
Different kinds of evidence
Observed values are instrument reports at a stated place and time. Model estimates describe an area and can refer to current or past conditions, as well as forecasts. Calculated sunlight and moon values are approximate local computations. NOAA tides are predictions, not measured water height. KCEU wind is a nearby airport observation, not a campus sensor. Hartwell is a reservoir gauge near Anderson, not water depth at the chosen point.
Sources and limits
Current weather: Open-Meteo / NOAA GFS Global. Past weather: Open-Meteo / ECMWF ERA5 / Copernicus. Fine particles: Open-Meteo / CAMS Global, a coarse area model. Pollen: CAMS Europe where valid grass, ragweed, or birch estimates exist; no US substitute is invented. USGS provides supported US lake and stream records. NOAA supplies supported coastal reference-station tide predictions. Traffic, trains, and lightning are not connected. Aircraft uses the free experimental ADSB.lol public API within 10 nautical miles (18.5 km). Checks target 30 seconds while visible; shared caching, request coalescing, a conservative daily source budget, provider Retry-After and backoff can delay them. Data © ADSB.lol contributors, ODbL 1.0. Incomplete coverage and two-minute position expiry apply.
The composition is chosen
Sunlight changes the main register and brightness. Clouds and humidity shape harmonic intervals and texture. Wind changes stereo flow. Rain adds sparse soft notes. Pressure sets restrained bass. Temperature changes a more audible middle register. Supported water, particle, moon, and pollen values add their own layers. These are artistic mappings, not natural musical meanings or environmental warning thresholds. Similar conditions may sound related; place names do not determine the music.
Time and missing values
The page checks sources every five minutes, but shared source caches update less often: airport wind at least ten minutes, current model and lake data fifteen minutes, rivers thirty minutes, air and pollen one hour, and tides six hours. Inventory and named-place searches can be held for a day. The record time and download time are distinct. Old or failed records become unavailable and fade out; actual zero stays zero. Public source budgets or upstream limits may make a feed temporarily unavailable.
Listening and saving
Nothing plays until Listen. The composition develops continuously with a seeded sequence, a six-second entrance, and a short Stop fade. Solo temporarily overrides that layer’s mute, suppresses the others, and restores earlier choices when cleared. Volume starts at 35% and extends to 180%, with smooth gain and bounded output. Start with device volume low. Saving exports the current data and settings to your browser. WAV export renders the current conditions; it is not a recording of the whole live session. Current-format saved files load in the browser; older formats require conversion in the local edition.
Public edition
Visuals are hidden. This page publishes only the instrument and its public data adapter; it does not publish local caches, private saved sessions, internal QA reports, or the separate cables preview. Source requests are restricted to named providers and validated parameters. Shared caching and a conservative daily source budget reduce traffic across visitors; the shared KV counter is a best-effort safeguard, not an atomic billing cap.
Attribution and use
Open-Meteo keyless access is for noncommercial use, including qualifying public research and non-profit websites without advertising or subscriptions. Its documented limits apply. Weather and air-model attribution includes NOAA NCEP, ECMWF ERA5, CAMS, and Copernicus as appropriate. Please follow the linked provider terms and attribution requirements. This is an exploratory research instrument, not a weather, health, navigation, or safety service.
Open-Meteo terms · Weather documentation · Air and pollen documentation · USGS · NOAA airport data · NOAA tides · Aircraft provider policy
Optional exact composition rules
Engine 5, revision relational-variation-5.0. Seed 17417; the same saved conditions, settings, sample rate, and duration reproduce the same rendering. Timings below are artistic decisions, not measured environmental rhythms. Missing sources contribute no voice. Real zero remains distinct from unavailable.
Sun elevation −18–75° maps to a pad root of 65.406392 × 2^(2.4 × clamped fraction) Hz, about 65–345 Hz. Pad intervals depend on available cloud and humidity fractions. Upper interval spans 12–19 semitones. Cloud fraction selects one of two composed chord palettes; seeded changes occur every 24–48 seconds, easing over 15 seconds. Daylight increases second and third partials. Independent voice swells have rates 0.009 + 0.003 × voice index + 0.014 × normalized wind Hz. Temperature −10–40°C maps 98 × 2^(2.8 × clamped fraction), about 98–683 Hz, with a rounded FM voice whose index is 0.15 + 1.8 × dry fraction. Temperature's valid three-hour tendency adds up to ±2% slow drift.
Pressure 980–1040 hPa maps 55–82.5 Hz, amplitude 0.012. Three-hour pressure change influences slow breathing. Lake 640–670 ft maps 110–165 Hz with a soft 3:2 partial. Streamflow logarithmically maps 0–1000 ft³/s to 330–440 Hz, with a gently modulated water tone. Cloud has a quiet doubled partial; the amplitude follows cloud fraction. These tones have separate slow windows, leaving room for one another.
Sparse phrases use the available temperature register, or the light register if temperature is missing. Cloud selects an open [0,2,7,9,14,19] or softer minor [0,3,7,10,14] semitone palette. Phrase spacing is 5 + 13 × (1 − normalized wind) + 6 × cloud seconds. Humidity sets duration from 5–12 seconds. Each phrase has a sine-squared entrance/ending and quiet second partial; at most four overlap. No place-name seed or fabricated reading determines identity.
Wind maps 0–15 m/s to stereo motion and airy noise. Daylight sets its low-pass cutoff from 180–5380 Hz. Humidity darkens the veil. Rain retains at most three soft six-second grains with 0.6-second entrances; real zero creates none. Fine particles, lunar approximation, tide predictions, and European pollen retain quiet independent textures.
Aircraft retains three nearest usable airborne positions. Distance 0–18.52 km maps amplitude 0.018–0; bearing sets equal-power pan; barometric altitude 0–40000 ft maps 196–392 Hz. Composed offsets are 0,2,5,2 semitones in eight-second sections, with up to four-second entrances and eight-second endings. IDs preserve phrase age across updates. No trajectory or Doppler is inferred. Provider backoff precedes the 30-second target, and positions expire after two minutes.
Live and WAV use this same engine. Master volume remains 0–180%, initially 35%. The final limiter is linear through ±0.85 and rolls off below ±0.98. Source parameters ease over 15 seconds; explicit mutes use 0.15 seconds. Export has a six-second ending fade. These limits bound numerical output; subjective listening levels depend on the playback system. Earlier saved data can be loaded, but the engine revision is recorded because a newer composition sounds different.