Tracks
Every published track, its version, and how to check it
Each track has a card: the question it asks, where its records come from, what drives each sound, what is simulated rather than observed, what to be careful about, and how to reproduce it byte for byte. A track’s version changes when its music changes; the changelog says when and why. Tracks are grouped into albums. See Methods for how versions, variants and remixes work, and CONTRIBUTING.md to propose a remix.
Water Year Rings
One hydrologic year in two southern British Columbia watersheds, on recorded acoustic instruments. The album’s page.
Granby River, water year 2024
v1.0 · WY2024 (water year, 2023-10-01 to 2024-09-30) · listen
What does a warm El Niño winter sound like on a small mountain river: a year that never fully committed to winter, with broken ice and an early, small freshet?
Place. The Granby River at Grand Forks and its tributary Burrell Creek, with the Billings climate station, in the Kettle River basin of southern British Columbia.
Records.
- Granby River at Grand Forks, daily discharge and ice flag (Water Survey of Canada 08NN002)
- Burrell Creek above Gloucester Creek, daily discharge (WSC 08NN023)
- Billings daily climate (ECCC station 1100)
- Plant productivity (NASA MODIS MOD17A2HGF, Granby valley point)
| Record | What you hear | |
|---|---|---|
| Granby River flow | Piano gets denser as flow rises; contrabass holds the root, cello adds a 5th at high flow. | R |
| Flow vs. 15-yr normal | Sets the harmony’s brightness: below normal leans one mode darker, above normal one brighter. | D |
| Burrell Creek flow | Hand drums: tumba, conga, shaker, and a 3-3-2 log-drum figure at the peak. | R |
| Ice conditions (WSC flag) | Freeze: falling glockenspiel + bowed cymbal. Frozen: crackle and descending singing-ice chirps. Breakup: cymbal swell and rising harp. | R |
| Air temperature, Billings | Colder days make the ice crackle denser; also brightens the harmony across the season. | R |
| Rain | Each rainy day scatters harp notes from the current chord. | R |
| Snowfall | Snowfall scatters soft glockenspiel notes, high up. | R |
| Snow on ground | Deep snow swaps the reverb to a dark, short room: snow absorbs sound. | R |
| Plant productivity (MODIS GPP) | Violins swell with photosynthesis. | R |
| Redband trout spawning | A short rising flute figure. Timing is modelled from temperature and season. | S |
| Mountain whitefish spawning | A falling cello pizzicato figure. Timing is modelled. | S |
| Day length | With temperature, sets how bright the mode is. | D |
Every step from record to sound, with its bounds: the mapping dictionary (tables in tracks/mapping/granby-wy2024/).
Harmony. Key of D, 72 bpm, one bar per three days (122 bars, about 7 minutes). The mode moves along a dark-to-bright ladder, from Aeolian to Ionian, set by day length and temperature, and nudged one step by how the river’s flow compares with its 2010-24 normal; it changes at most one step every eight bars. The river’s phase (dormant, rising, peak, recession, low) chooses a four-chord loop.
Simulated, not observed.
- Redband trout and mountain whitefish spawning windows, modelled from air temperature and season.
- Bird arrival windows (dipper, Swainson’s thrush, bank swallow), typical regional timing, not sightings.
Be careful about.
- Air temperature stands in for water temperature; no Granby stream-temperature record was available.
- The ice record is WSC’s “Ice Conditions” data flag, which marks when ice affected the gauge, not an observation of the whole river.
- Plant productivity is the mean of a 2.3 km block and mixes land covers.
Reproduce. python scripts/pilot_granby_wy2024.py · seed 2024 · locked 2026-09-27 · SHA-256 of the MP3 on this site d047773665713d75… · check it with python scripts/renders.py verify granby-wy2024
Credits. Graham Watt, the question, the rivers and records, listening and direction; Claude Code (Anthropic’s AI coding agent), music theory, mapping and all of the code; Instruments, VSCO 2 Community Edition (Versilian Studios, CC0); birds from Wikimedia Commons (see samples_manifest.csv).
Lineage. variants: frozen, no-perc, slow.
Changelog.
- v1.0 (2026-09-25): First published version.
Okanagan, water year 2024
v1.0 · WY2024 (water year, 2023-10-01 to 2024-09-30) · listen
What does it sound like when a drought year on a regulated lake is also the biggest sockeye return in a decade?
Place. Okanagan Lake at Kelowna and the Okanagan River at Penticton and near Oliver, with the Summerland climate station, in southern British Columbia; and Wells Dam on the Columbia, below the Okanagan confluence.
Records.
- Okanagan Lake at Kelowna, daily level (WSC 08NM083)
- Okanagan River at Penticton, dam outflow (WSC 08NM050)
- Okanagan River near Oliver, daily discharge (WSC 08NM085)
- Summerland CS daily climate (ECCC station 979)
- Plant productivity (NASA MODIS MOD17A2HGF, Summerland point)
- Adult sockeye and water temperature at Wells Dam (Columbia River DART; fetched, not redistributed)
| Record | What you hear | |
|---|---|---|
| Okanagan Lake level, Kelowna | A low F held under everything, louder as the lake fills; cello adds a 5th while it rises. | R |
| Lake vs. 15-yr normal | Sets the harmony’s brightness: below normal leans one mode darker. | D |
| Outflow at Penticton dam | Piano density follows the release. When the dam holds the flow flat, the piano plays strictly, with no variation. | R |
| Okanagan River near Oliver | Soft log drums, conga and shaker. | R |
| Sockeye passing Wells Dam | Marimba runs climb the chord; more fish means more notes, and the runs start higher as the season goes. | R |
| Water temperature at Wells | Above 18 C a violin holds a high note: thermal stress for migrating sockeye. | R |
| Sockeye spawning near Oliver | Low, falling cello pizzicato around mid-October, scaled by run size. Timing is modelled. | S |
| Air temperature, Summerland | Below -8 C: frost crackle. | R |
| Rain | Harp notes from the current chord. Rain vs. snow is split by temperature. | D |
| Snow | Soft, high glockenspiel. | D |
| Humidity (daily min) | Humid days get a long hall reverb; dry days a short, bright room. | R |
| Plant productivity (MODIS GPP) | Violins swell with photosynthesis. | R |
Every step from record to sound, with its bounds: the mapping dictionary (tables in tracks/mapping/okanagan-wy2024/).
Harmony. Key of F, 64 bpm, one bar per three days (122 bars, about 8 minutes). The same mode ladder as the Granby piece, set by day length and temperature and nudged by the lake’s level against its 2010-24 normal. The lake’s phase (winter, filling, full, drawdown) chooses the chord loop, and the piano plays strictly, without variation, while the dam holds the outflow flat.
Simulated, not observed.
- Sockeye spawning near Oliver, timing modelled around mid-October and scaled by the run that produced it.
- Bird arrival windows (western meadowlark, sandhill crane), typical timing, not sightings.
Be careful about.
- The Wells temperature is Columbia River water, pumped from the dam tailrace into the fishway, not Okanagan water.
- Summerland’s separate rain and snow values are missing on most days, so rain and snow are split from total precipitation by temperature (derived).
- Lake level is measured from the gauge’s local datum, not sea level.
Reproduce. python scripts/pilot_okanagan_wy2024.py · seed 1997 · locked 2026-09-27 · SHA-256 of the MP3 on this site f2c2e997c5c33d64… · check it with python scripts/renders.py verify okanagan-wy2024
Credits. Graham Watt, the question, the rivers and records, listening and direction; Claude Code (Anthropic’s AI coding agent), music theory, mapping and all of the code; Instruments, VSCO 2 Community Edition (Versilian Studios, CC0); birds from Wikimedia Commons (see samples_manifest.csv).
Changelog.
- v1.0 (2026-09-25): First published version.
Climate Pair
Sixty-eight years of the global climate record, one year per bar, heard from the Pacific and from the Atlantic. The album’s page.
Climate Pair: Pacific
v2.0 · 1958-2025 (calendar years, 1958-01-01 to 2025-12-31) · listen
Can seventy years of a warming planet be heard as one rising form, with the ocean’s own cycles and storms moving through it?
Place. The global record, heard from the Pacific: its ocean phase and storms, and the Kettle River, which crosses the BC-Washington border near Grand Forks.
Records.
- Mauna Loa CO₂ (NOAA GML)
- Global ocean surface temperature anomaly (NOAA NCEI) and global temperature (NASA GISTEMP v4)
- Ocean heat content 0-2000 m (IAP v4.2; fetched, not redistributed)
- Oceanic Niño Index (NOAA CPC)
- Kettle River near Laurier, annual peak daily flow (USGS 12404500)
- Pacific Decadal Oscillation (NOAA NCEI)
- West Pacific (IBTrACS) and Northeast and Central Pacific (HURDAT2) best tracks
- Sunspot number (SILSO; fetched, not redistributed)
| Layer | Data | What you hear | |
|---|---|---|---|
| Drone and organ pedal | Mauna Loa CO₂, deseasonalized (NOAA GML) | level and brightness rise with ppm; the seasonal CO₂ cycle makes it breathe | R |
| Cellos | global ocean surface temperature anomaly (NOAA NCEI) | chord tones, from one voice to three, louder as it warms | R |
| Contrabass | ocean heat content, 0-2000 m (IAP v4.2) | the deep ocean filling: root, then root and fifth, louder | R |
| Chords and month pulse | Oceanic Niño Index (NOAA CPC) | El Niño lifts the chords, La Niña sinks them; one pulse note a month, louder in strong years | R |
| Mode and pad | Pacific Decadal Oscillation, 3-year mean | Dorian in the warm phase, Aeolian in the cool; the pad brightens with the index | D |
| Drums | West Pacific typhoons (IBTrACS, panned left) and Northeast Pacific hurricanes (HURDAT2, panned right) | tropical storm: bass-drum rub; hurricane: timpani; major: timpani, bass drum and sub; category 5: tam-tam; landfall: cymbal | R |
| Rumble | season’s accumulated cyclone energy (ACE) | a low storm rumble | D |
| Shimmer | sunspot number (SILSO) | high shimmer and glockenspiel sparkles on the 11-year cycle | R |
| High violin | global temperature anomaly (NASA GISTEMP) | a held high note once it passes +0.8 °C | R |
| Gong | large volcanic eruptions and their cooling | a gong; the drone darkens for about two years | D |
| Harp bloom | Kettle River annual peak flow (USGS 12404500) | a spring harp and piano bloom, sized by the peak’s rank | R |
Every step from record to sound, with its bounds: the mapping dictionary (tables in tracks/mapping/climate-pair-pacific/).
Harmony. Key of D, one bar per year (68 bars of 7 seconds, about 8 minutes); the four beats are the seasons and the twelve pulses the months. ENSO sets the chords and the ocean’s multidecadal phase sets the mode.
Be careful about.
- Early storms are undercounted, before routine satellite coverage (about 1966; later in the Northeast Pacific).
- Records end at different times. Interior gaps are interpolated and each record’s last value is held to the end.
- The ocean phases are 3-year means, so their turns sound about two years late.
- The volcanic cooling shape, the ENSO-to-chord table and every loudness curve are interpretive choices, not measurements.
Reproduce. python scripts/climate_pair.py pacific · seeds 1958, 99 · locked 2026-09-27 · SHA-256 of the MP3 on this site 314128ee9c53d2ad… · check it with python scripts/renders.py verify climate-pair-pacific
Credits. Graham Watt, the brief (CO₂, SST, ENSO, PDO and storms; rising, moody without being overly dark), the choice of records, listening and direction; Claude Code (Anthropic’s AI coding agent), music theory, mapping and all of the code; Instruments, VSCO 2 Community Edition (Versilian Studios, CC0), plus synthesizers.
Changelog.
- v2.0 (2026-09-25): First published version. The Kettle River replaced the Granby as the valley link (the Granby record has a 1958-66 gap); West Pacific typhoons (IBTrACS) and ocean heat content (IAP) were added; storm levels rebalanced so the rise is heard.
- v1.0 (2026-09-25): First render, not published.
Climate Pair: Atlantic
v2.0 · 1958-2025 (calendar years, 1958-01-01 to 2025-12-31) · listen
Can seventy years of a warming planet be heard as one rising form, with the ocean’s own cycles and storms moving through it?
Place. The global record, heard from the Atlantic: its ocean phase, hurricanes, the North Atlantic Oscillation and Arctic sea ice.
Records.
- Mauna Loa CO₂ (NOAA GML)
- Global ocean surface temperature anomaly (NOAA NCEI) and global temperature (NASA GISTEMP v4)
- Ocean heat content 0-2000 m (IAP v4.2; fetched, not redistributed)
- Oceanic Niño Index (NOAA CPC)
- Atlantic Multidecadal Oscillation and North Atlantic Oscillation (NOAA PSL, CPC)
- Atlantic hurricane best tracks (HURDAT2)
- Arctic sea-ice extent (NSIDC Sea Ice Index v4, from November 1978)
- Sunspot number (SILSO; fetched, not redistributed)
| Layer | Data | What you hear | |
|---|---|---|---|
| Drone and organ pedal | Mauna Loa CO₂, deseasonalized (NOAA GML) | level and brightness rise with ppm; the seasonal CO₂ cycle makes it breathe | R |
| Cellos | global ocean surface temperature anomaly (NOAA NCEI) | chord tones, from one voice to three, louder as it warms | R |
| Contrabass | ocean heat content, 0-2000 m (IAP v4.2) | the deep ocean filling: root, then root and fifth, louder | R |
| Chords and month pulse | Oceanic Niño Index (NOAA CPC) | El Niño lifts the chords, La Niña sinks them; one pulse note a month, louder in strong years | R |
| Mode and pad | Atlantic Multidecadal Oscillation, 3-year mean | Dorian in the warm phase, Aeolian in the cool; the pad brightens with the index | D |
| Drums | Atlantic hurricanes (HURDAT2) | tropical storm: bass-drum rub; hurricane: timpani; major: timpani, bass drum and sub; category 5: tam-tam; landfall: cymbal | R |
| Rumble | season’s accumulated cyclone energy (ACE) | a low storm rumble | D |
| Shimmer | sunspot number (SILSO) | high shimmer and glockenspiel sparkles on the 11-year cycle | R |
| High violin | global temperature anomaly (NASA GISTEMP) | a held high note once it passes +0.8 °C | R |
| Gong | large volcanic eruptions and their cooling | a gong; the drone darkens for about two years | D |
| Winter wind | North Atlantic Oscillation (NOAA CPC) | band-limited wind, stronger when positive | R |
| Glass voice | Arctic sea-ice extent (NSIDC), from Nov 1978 | breathes with the season and thins as September ice declines | R |
Every step from record to sound, with its bounds: the mapping dictionary (tables in tracks/mapping/climate-pair-atlantic/).
Harmony. Key of A, one bar per year (68 bars of 7 seconds, about 8 minutes); the four beats are the seasons and the twelve pulses the months. ENSO sets the chords and the ocean’s multidecadal phase sets the mode.
Be careful about.
- Early storms are undercounted, before routine satellite coverage (about 1966; later in the Northeast Pacific).
- Records end at different times; the AMO stopped updating in January 2023, and the piece holds its last value to the end. Interior gaps are interpolated.
- The ocean phases are 3-year means, so their turns sound about two years late.
- The volcanic cooling shape, the ENSO-to-chord table and every loudness curve are interpretive choices, not measurements.
Reproduce. python scripts/climate_pair.py atlantic · seeds 1492, 99 · locked 2026-09-27 · SHA-256 of the MP3 on this site f5eff1acdd7babad… · check it with python scripts/renders.py verify climate-pair-atlantic
Credits. Graham Watt, the brief (CO₂, SST, ENSO, PDO and storms; rising, moody without being overly dark), the choice of records, listening and direction; Claude Code (Anthropic’s AI coding agent), music theory, mapping and all of the code; Instruments, VSCO 2 Community Edition (Versilian Studios, CC0), plus synthesizers.
Changelog.
- v2.0 (2026-09-25): First published version. Ocean heat content (IAP) and Arctic sea ice (NSIDC) were added; storm levels rebalanced so the rise is heard.
- v1.0 (2026-09-25): First render, not published.
Meridian Chorus
Thirteen voices on the 100th meridian west, pole to pole, singing one year of daylight, air, sea and ice. The album’s page.
Meridian Chorus, light, air, sea and ice
v0.1 · 2023-24 (solstice to solstice, 2023-12-22 to 2024-12-21) · listen
What does one year sound like along a single line of longitude, from pole to pole, when daylight, air, sea and ice each have a voice?
Place. The 100th meridian west, thirteen voices one every 15 degrees from the North Pole to the South Pole, across land, ocean and ice.
Territory. The land and ice voices stand on Inuit Nunangat, at the Nunavut-Manitoba boundary, in South Dakota, on the Edwards Plateau in Texas, and in Antarctica. Acknowledgements are being prepared and are not yet on the page.
Records.
- Daylight and twilight, computed from solar geometry (solar.py)
- Daily maximum and minimum 2 m air temperature (ERA5 via Open-Meteo, CC BY 4.0)
- Sea surface temperature (NOAA OISST v2.1)
- Sea-ice concentration (NOAA/NSIDC Climate Data Record)
| Layer | Data | What you hear | |
|---|---|---|---|
| light | sun above the horizon and through civil twilight | the sung voice, on the voice’s pitch; it opens from oo to ah as the sun climbs | D |
| air | daily maximum air temperature | the organ’s registration: cold is one dark rank, warm adds the octave and opens the tone | D |
| night air | daily minimum air temperature | the in-breath: a warm night adds a lower, fuller band | D |
| sea | sea surface temperature at the ocean voices | contrabass and cello on a floor below the chorus, D1 to A3, day and night | R |
| ice | sea-ice concentration at the four polar voices | a lid on the sea voice and a high glass shimmer | R |
Every step from record to sound, with its bounds: the mapping dictionary (tables in tracks/mapping/meridian-layer2-wide-long/).
Harmony. D major pentatonic on thirteen distinct notes, D4 to A6 for the sky choir and D1 to A3 for the sea, so no two voices share a pitch. North sings on the left and south on the right. Three days per breath (about 16 minutes), in four movements; every voice centres on the same solar noon, so the chorus breathes together.
Be careful about.
- Air temperature is reanalysis at one point per voice, not a station.
- Each breath plays the mean over its block of days.
- The ear hears the season’s shape, not the value at any one parallel.
Reproduce. python scripts/meridian.py --variant layer2-wide-long · seed 2023 · locked 2026-09-28 · SHA-256 of the MP3 on this site 4cddbaa4110b7f0f… · check it with python scripts/renders.py verify meridian-chorus-layer2:wide-long
Credits. Graham Watt, the question, the meridian, listening and direction; Claude Code (Anthropic’s AI coding agent), mapping, sound design and all of the code; Instruments, VSCO 2 Community Edition (Versilian Studios, CC0); the sung voices are synthesized.
Lineage. remix of meridian-daylight.
Changelog.
- v0.1 (2026-09-28): First published version (layer 2.1: sea on its own floor, by hemisphere).
Meridian Chorus, daylight
v0.2 · 2023-24 (solstice to solstice, 2023-12-22 to 2024-12-21) · listen
What does a year of daylight sound like along one meridian, when every voice’s note lasts as long as its day?
Place. The 100th meridian west, thirteen voices one every 15 degrees from pole to pole.
Territory. See the Meridian Chorus card; acknowledgements are being prepared.
Records.
- Daylight and twilight, computed from solar geometry (solar.py); no measured records
| Layer | Data | What you hear | |
|---|---|---|---|
| note | sun above -0.833° (sunrise to sunset) | organ, strings and sung voice on the voice’s pitch | D |
| twilight | sun between -6° and -0.833° | the note swells in and fades out | D |
| vowel | the sun’s altitude | the sung voice opens from oo to ah | D |
| in-breath | sun below the horizon | a faint filtered-noise inhale, peaking at midnight | D |
Every step from record to sound, with its bounds: the mapping dictionary (tables in tracks/mapping/meridian-daylight/).
Harmony. D major pentatonic, D at both poles and the equator, north above and south below, D2 to D6. A week per 8-second breath (about 7 minutes). At the equinoxes every voice is about twelve hours long; at the solstices one pole holds its note through the whole breath while the other only breathes.
Be careful about.
- Everything is computed from solar geometry; nothing is measured.
- Refraction is a standard constant; local horizons and weather are ignored.
Reproduce. python scripts/meridian.py · seed 2023 · locked 2026-09-28 · SHA-256 of the MP3 on this site 1431374a3ae741e3… · check it with python scripts/renders.py verify meridian-chorus-daylight
Credits. Graham Watt, the question, the meridian, listening and direction; Claude Code (Anthropic’s AI coding agent), mapping, sound design and all of the code; Instruments, VSCO 2 Community Edition (Versilian Studios, CC0); the sung voices are synthesized.
Lineage. variants: long, layer2-wide-long.
Changelog.
- v0.2 (2026-09-27): After a blind two-model review: an even daytime level at every latitude, the pentatonic chord restored, the poles’ breathing fixed. Re-locked 2026-09-28 with unchanged audio.
- v0.1 (2026-09-27): First render.
Superior Ice Year
One winter of ice on Lake Superior, 2025-26, heard region by region from west to east. The album’s page.
Superior Ice Year
v1.0 · 2025-26 (ice season, 2025-11-01 to 2026-06-15) · listen
How does one winter of ice move across Lake Superior, region by region, and how does it compare with the lake’s recent ordinary year?
Place. Lake Superior (Gichigami), cut into eight regions from the Western Arm to Whitefish Bay.
Territory. Anishinaabe territory, including Ojibwe nations with treaty rights there: the Robinson-Superior Treaty of 1850, and the treaties of 1836, 1842 and 1854. The acknowledgement on the piece’s page is a draft, open to correction.
Records.
- Ice concentration, daily by region (NOAA GLERL and CoastWatch Great Lakes Node, GLSEA with NIC ice charts)
- Surface water temperature, daily by region (NOAA CoastWatch ACSPO GLSEA)
- Lake-average ice, 2008-2025 median (NOAA GLERL)
- Air temperature, wind, precipitation, snowfall and shortwave radiation (ERA5 via Open-Meteo, CC BY 4.0)
| Layer | Data | What you hear | |
|---|---|---|---|
| ice | region ice concentration | a bowed-glass tone per region, D5 to F#6 west to east, louder as the ice thickens | R |
| events | 3-day-smoothed daily change in region ice | freeze: a glass ping; breakup: a low drop and crack, each at a random distance (quieter, darker, later, wetter) | D |
| water | region surface temperature | four low string voices | R |
| turnover | lake-mean skin temperature crossing 4 °C | a soft sub drop | D |
| air | lake-mean air temperature | an organ that adds ranks and brightens as it warms | R |
| sun | day length at 47.6° N and shortwave radiation | a choir that swells and opens from oo to ah | D |
| wind | peak wind of the eight regions | a filtered-noise bed gated by gales | R |
| snow | region snowfall | dry unpitched ticks | R |
| rain | region liquid precipitation | soft harp drops | D |
| birds | timing from ice-off and warmth | loon and goose recordings | S |
| ghost | 2008-2025 median lake ice | a quiet detuned glass tone; after the data end it holds a low floor and fades (a symbolic coda), and a soft pulse marks the last day of data | D |
| deep water | deep water near 4 °C under the ice (known physics, not measured here) | a constant low hum | D |
Every step from record to sound, with its bounds: the mapping dictionary (tables in tracks/mapping/superior-ice/).
Harmony. D major pentatonic. Each region is a note, low in the west and high in the east, panned by longitude, so the chord sounding at any moment is the ice pattern on the map. Three seconds per day (11 min 46 s), cut into four movements at ice-defined days: freeze-up, ice-in, long melt, open water. Higher regions are trimmed 0.5 dB a step so the east doesn’t sound louder.
Simulated, not observed.
- Bird timing is symbolic, from each region’s ice-off and warmth; the calls are recordings, not observations.
- The ghost’s coda after the last day of data is symbolic.
Be careful about.
- Regions are wedges by distance from eight anchor points, not basins or depth zones.
- Water temperature is the satellite skin of the top millimetres, not the lake at depth.
- Air, wind and precipitation are ERA5 reanalysis at one point per region, not station observations.
- Freeze and breakup events are derived with fixed thresholds (4 points a day, 3 days apart).
- The ear does not read the map (a listening probe agreed); the page’s picture carries space.
Reproduce. python scripts/icecover.py --variant v1 · seed 2026 · locked 2026-09-29 · SHA-256 of the MP3 on this site 837c4a520d286b39… · check it with python scripts/renders.py verify icecover-superior-ice:v1
Credits. Graham Watt, the question, the lake, listening, and every design choice on the A/B bench; Claude Code (Anthropic’s AI coding agent), mapping, sound design and all of the code; Loon and goose recordings, Jonathon Jongsma (CC BY-SA 3.0, Wikimedia Commons); instruments, VSCO 2 Community Edition (CC0).
Lineage. variants: long, ghost, ice-only, beds, weather.
Changelog.
- v1.0 (2026-09-29): Page 1.1, the picture only; the audio is unchanged. The map becomes one Blender render per day under toggleable overlays, with the data as measured behind a line you drag.
- v1.0 (2026-09-29): The long version as the main track, with the ghost and its coda, a deep-water hum, distance-modelled crackles, staggered events and the pitch-loudness tilt, chosen by GW on an A/B bench. Fixes a v0.2 bug in the ghost’s pitch (summed at too low a precision, it stepped between wrong notes, mostly about 2.5 semitones sharp).
- v0.2 (2026-09-28): After a first listen: fewer, more distant birds; quieter, wetter events; a softer ghost.
- v0.1 (2026-09-28): First render.
Superior Ice Year, shorter
v0.2 · 2025-26 (ice season, 2025-11-01 to 2026-06-15) · listen
The same winter of ice on Lake Superior, at 1.5 seconds per day, for a shorter listen.
Place. Lake Superior (Gichigami), cut into eight regions from the Western Arm to Whitefish Bay.
Territory. Anishinaabe territory; see the Superior Ice Year card.
Records.
- As Superior Ice Year, without the 2008-2025 median.
| Layer | Data | What you hear | |
|---|---|---|---|
| all | as Superior Ice Year | the same nine layers, without the ghost, the coda, the deep-water hum or the v1 crackle and balance changes | D |
Every step from record to sound, with its bounds: the mapping dictionary (tables in tracks/mapping/superior-ice-short/).
Harmony. As Superior Ice Year, at 1.5 seconds per day (5 min 48 s), in one piece.
Be careful about.
- As Superior Ice Year.
- This is the v0.2 render; the v1.0 changes are in the main track only.
Reproduce. python scripts/icecover.py · seed 2026 · locked 2026-09-28 · SHA-256 of the MP3 on this site d1a4ca0ae26d7641… · check it with python scripts/renders.py verify icecover-superior-ice
Credits. Graham Watt, the question, the lake and listening; Claude Code (Anthropic’s AI coding agent), mapping, sound design and all of the code; Loon and goose recordings, Jonathon Jongsma (CC BY-SA 3.0, Wikimedia Commons); instruments, VSCO 2 Community Edition (CC0).
Lineage. remix of superior-ice.
Changelog.
- v0.2 (2026-09-28): After a first listen: fewer, more distant birds; quieter, wetter events.
- v0.1 (2026-09-28): First render.
Flow Regimes
Four very different water years of interior British Columbia, each a drum voice, entering one at a time. The album’s page.
Interior Flow, B-side
v1.0 · WY1988, 1995, 2012, 2018 (four exemplar water years drawn from 1987 to 2018 (not contiguous), 1987-10-01 to 2018-09-30) · listen
What do four very different water years of interior British Columbia sound like when each one is a drum voice, entering one at a time over the same year of snow, melt and warmth?
Place. Interior British Columbia: regional weekly medians over the 61 catchments that have at least 36 complete water years (WY1981 to 2022) and BEC cover (58 to 61 report in each of the four years), from the Borealis catchment-mean forcing set. The four years are the exemplars of a four-group clustering of the 42 years.
Records.
- Streamflow (Water Survey of Canada) and Daymet catchment-mean snow water equivalent and air temperature, from the Borealis dataset (Kovacek, CC BY-SA 4.0, doi 10.5683/SP3/65FXAS), version 2 of 2025-12-02 (R, regional weekly medians, D)
- The four years are k=4 exemplars chosen by Ward clustering on freshet timing, volume, late-summer flow and peak snow (D)
- Instruments, VSCO 2 Community Edition (CC0)
| Layer | Data | What you hear | |
|---|---|---|---|
| year voices | each exemplar year’s weekly regional flow, relative to each catchment’s own mean | timpani (1988, 9-step cycle), conga (1995, 7), log drum (2012, 5), shaker and triangle (2018, 3); a Euclidean pattern whose density and loudness follow the flow, one voice entering per pass | D |
| lead | weekly fall in snow water equivalent (snowmelt) | a harp on a 7-step cell (from pass 3) and a marimba doubling it three steps ahead (from pass 4), denser in faster melt, so they sound only in melt weeks | D |
| filter | snow water equivalent | a low-pass on the hats, shaker and lead: dull under deep snow, open at melt | D |
| mode | weekly air temperature | B natural (Dorian) above 10 °C, B flat (Aeolian) below | D |
| sub | flow floor (lowest weekly flow of the last six weeks) | a sine pedal whose level follows the floor; its note (D, C, B flat) changes at fixed bars | D |
| breakdown | mean level of the sounding voices, from week 38 (late June) | kick and clap drop out in low-flow bars from week 38 (late June) | D |
Every step from record to sound, with its bounds: the mapping dictionary (tables in tracks/mapping/regimes-bside/).
Harmony. A held D pedal; the lead stays in D Aeolian, with B natural (D Dorian) in warm weeks. 124 beats per minute, 16 steps a bar, swing 0.56, a week to a bar, five passes of 52 bars and a 4-bar coda (8 min 42 s). The year voices loop on 9, 7, 5 and 3 steps against the bar and rotate a little every pass.
Simulated, not observed.
- The idiom (odd-length Euclidean cycles against a steady 4/4, swing, micro-timing, additive entries, a filter as form) is evoked from a description of a style, not taken from any record; no melody, sample or voice comes from any recording.
Be careful about.
- The lead is silent until the first melt of pass 3 (4:17), so the piece is percussion and sub for its first four minutes; the marimba doubling joins in pass 4.
- The breakdown happens in every pass, so it marks the season more than the year; the sub’s note changes at fixed bars, not at the freshet.
- Snow and temperature are averaged over the voices that have entered, so the filter in pass 1 follows 1988 alone.
- The years form a continuum more than four clean groups (silhouette about 0.3 at every k from 3 to 7); four is a working choice, not a finding.
- The weekly series are regional medians, so they hide the spread between catchments and the elevation effect, which explains freshet timing better than the BEC zone.
- Claude Code cannot hear; levels were set by measurement (active level against the mix) and Graham has listened. Whose land the catchments are on is not yet mapped; the territory note is pending.
- The scripts that derive data/regimes/weekly.json from the Borealis files and the BEC map are not yet in this repository; the weekly table itself is.
Reproduce. python scripts/regimes_bside.py · seed 124 · locked 2026-09-30 · SHA-256 of the MP3 on this site 26d46a6c69e95eb3… · check it with python scripts/renders.py verify regimes-bside
Credits. Graham Watt, the question, the region and direction; Claude Code (Anthropic’s AI coding agent), mapping, sound design and all of the code; Data, Borealis (Kovacek, CC BY-SA 4.0); instruments, VSCO 2 Community Edition (CC0).
Changelog.
- v1.0 (2026-09-30): First published render.
Arctic Round
Forty-eight years of Arctic sea ice, one ice year per bar, as a set of pieces. The Wheel, a variant, is published first; the Round, the main piece, is in preparation. The album’s page.
The Wheel (a variant on Arctic Round)
v1.0 · Ice years 1979 to 2026 (48 ice years, each from 1 October to 30 September, one bar each, 1978-10-01 to 2026-09-30) · listen
If eight organ notes turn in the same order at the same pace for 48 years, and the Arctic sea ice decides each week how many of them can sound, what does the loss of the ice sound like?
Place. The Arctic Ocean and its marginal seas: the Northern Hemisphere sea ice of the satellite record, taken as one total. Sea ice is part of the homeland of Inuit and of other Indigenous peoples of the Arctic; this piece uses no Inuit knowledge and no Inuit data, and the piece’s page carries the acknowledgement.
Records.
- Sea ice extent, daily, NSIDC Sea Ice Index, Version 4 (G02135; Fetterer and others, doi 10.7265/a98x-0f50) (R; before mid-1987 every other day, and 2 December 1987 to 13 January 1988 is filled by interpolation)
- Sea ice volume, monthly, PIOMAS v2.1, Polar Science Center, University of Washington (Zhang and Rothrock 2003; Schweiger and others 2011) (M, a model; fetched by the script and not republished here)
- Instruments, VSCO 2 Community Edition (CC0)
| Layer | Data | What you hear | |
|---|---|---|---|
| organ wheel | total sea ice extent, weekly, smoothed over three weeks | eight notes (C3 G3 C4 E♭4 F4 G4 B♭4 C5), one a week in a fixed order; extent sets how many sound (8, 7, 6 and 2 notes at 17.5, 12, 7.2 and 3.4 M km²); the highest and lowest drop out first, and a dropped note is a rest | D |
| organ upper octave | open water, 1 − extent / 17.5 M km² | each sounding note gains its upper octave once extent is below about 9.6 M km², louder as the water opens | D |
| cello | ice volume (PIOMAS), sampled six times a year | one bowed note every two months; thick ice is low (from E♭2), thin ice is high (to C5) | M |
| bass | the ice year’s mean volume | a C2 pedal, full above 24 thousand km³ and silent below 14 | M |
| swell | total extent | the cello and bass rise and fall with the year’s ice, down to 15 % of full level at the lowest extent | D |
| room | total extent | the reverb moves from a dry, dark room (ice) to a wide, bright one (open water) | D |
| rubato | week-to-week change in extent | weeks press up to 8 % in freeze-up and drag up to 10 % in melt; every bar stays exactly 16 seconds | D |
Every step from record to sound, with its bounds: the mapping dictionary (tables in tracks/mapping/arctic-wheel/).
Harmony. C minor pentatonic throughout (C, E♭, F, G, B♭), which has no semitone and no tritone, so any subset of the wheel is consonant. 48 bars of 16 seconds, 52 pulses a bar, and a 7-second coda of six October days in which the wheel restarts thin and the cello holds an open fifth (G4 and D5), unresolved; 13 min 19 s with the tail.
Simulated, not observed.
- Ice volume is a model (PIOMAS), not a measurement; the cello and the bass follow it.
- PIOMAS for March to September 2026 repeats 2025, so the 2026 cello register and bass follow that extension.
Be careful about.
- The wheel follows the total extent, not particular seas; no note stands for a region.
- Warming is not in this piece; temperature and CO2 belong to the Round, the main piece of the pair, which is not yet published.
- The number of notes is a chosen piecewise scale of extent, not a physical threshold; a note is silent or sounding, so small changes within a step are not heard in the organ.
- Extent is smoothed over three weeks and volume over two months, so daily noise and short events are not played.
- Claude Code cannot hear; levels were set by measurement and Graham has listened.
- The engine’s data file (which holds the PIOMAS series) is not in this repository;
fetch_arctic.pyrebuilds it, so the lock’s data hash can only be checked after that step.
Reproduce. python scripts/arctic_bside.py --set organ_gain=6.5 --set bass_gain=1.1 --set name=arctic_bside_v2 · seed 2026 · locked 2026-10-10 · SHA-256 of the MP3 on this site db431d72105f462f… · check it with python scripts/renders.py verify arctic-wheel
Credits. Graham Watt, the question, the form (a minimalist round) and direction; Claude Code (Anthropic’s AI coding agent), mapping, sound design and all of the code; Data, NSIDC Sea Ice Index v4 and PIOMAS v2.1; instruments, VSCO 2 Community Edition (CC0).
Changelog.
- v1.0 (2026-10-31): Page update, music unchanged: the player moves while the map loads, load progress shows, and the notes fold.
- v1.0 (2026-10-10): First published render (take 3).
Godbout Outlook
A 15-day weather forecast for Godbout, Quebec, one bar for every six hours, played as a groove. The album’s page.
Godbout Outlook
v1.0 · 16 to 31 October 2026 (a 15-day forecast, 61 steps of six hours, one bar each, 2026-10-16 to 2026-10-31) · listen
What does a 15-day weather outlook sound like when every six hours of the model run is one bar of a groove, with a storm at each end, snow at one and rain at the other?
Place. The north shore of the St Lawrence estuary at Godbout, Quebec. The model point is the village itself, 49.32 N, 67.62 W (the ECCC station’s coordinates), averaged over a 0.25-degree model grid box that takes in the shore and some of the estuary. One frozen run of ECMWF’s AIFS Single (16 October 2026, 12 UTC), 61 six-hour steps, with times in a fixed UTC-5 clock.
Territory. Godbout lies on the north shore of the St Lawrence, in Nitassinan, the Innu homeland, within the traditional territory of the Innu of Pessamit.
Records.
- {‘ECMWF AIFS Single, run of 16 October 2026 12 UTC (CC BY 4.0, https://www.ecmwf.int/en/forecasts/datasets/open-data)’: “2 m temperature, dew point, 850 hPa temperature, cloud cover, sea-level pressure, 10 m and 700 hPa wind, downward shortwave radiation, rain and snow, taken from ECMWF’s open data and interpolated to the point by scripts/fetch_aifs_open.py (R; snow in cm is the model’s snowfall, counted as 1 cm of snow per mm of water)”}
- Sea floor and land for the page’s banner, GEBCO 2025 Grid (doi 10.5285/37c52e96-24ea-67ce-e063-7086abc05f29; public domain, not for navigation)
- Instruments, VSCO 2 Community Edition and Karoryfer Samples Big Rusty Drums, War Tuba and Bear Sax (all CC0)
| Layer | Data | What you hear | |
|---|---|---|---|
| pulse (hats | downward shortwave radiation, a 6 h mean ending at each time stamp | the model’s value is spread along that window’s daylight curve (the sine of the sun’s height, rounded at sunrise and sunset, as in Meridian Chorus), keeping the window’s mean; each hit’s loudness follows the curve: offbeat hats first, then backbeat kicks and shaker, then snare, then sixteenth hats; two soft kicks at night, with the night hand drums taking over; the run’s best moment sets the ceiling | D |
| drum kit | rain (mm) plus snow (cm) in the six hours, over 4, capped at 1 | from 0.2 the full kit (kick, snare, ghost notes); a crash when it first passes 0.5, on the first bar of a day with weight above 0.1, and on the first bar of the song | A |
| log drums and marimba | rain | 3 to 9 notes of 16 steps, more the harder it falls | D |
| glock | snow | 3 to 7 notes of 16, a bell tree when snow starts | D |
| congas | 10 m wind speed and direction | 3 to 9 hits of 16 by speed; the direction picks one of three accent sets (S/E, W, N) | R |
| night hand drums | the sun pulse (its complement) and 10 m wind speed | as the sun gate closes at dusk, a loose heartbeat of low conga, log drum and brushed tambourine: each hit drawn from a probability for its step, nudged off the grid by about 10 ms; wind sets the density from 0.7 to 1.4 times; quiet when rain brings the kit | A |
| cello pizzicato | 700 hPa wind speed | 2 to 8 notes of 16 | R |
| harp counter-line | angle between the 10 m and 700 hPa wind directions (shear) | a third above the harp line under 60 degrees, a fifth to 120, a seventh above | D |
| harp | chord, 2 m temperature, temperature minus dew point | chord tones; an octave higher from 6 C; drier air plays quieter | R |
| chord loop and tuba riff | 2 m temperature and sea-level pressure | 3 C or warmer: D Dorian (i IV VII IV); colder: D Aeolian (i VI iv VII); pressure level drifts the loop up to two scale degrees | D |
| low cello drone | 6 h pressure change | root, fifth and third; a fall of 3 hPa or more turns the third into a fourth | D |
| Bear Sax and sub drop | 6 h pressure change | an octave under the chord root: held through a fall of 3 hPa or more; on a fall of 6 hPa or more it slides down a fifth with a sine sub drop | D |
| timpani | 6 h pressure change of 6 hPa or more | a fall: timpani hit and the tuba stepping fifth to root; a rise: timpani roll and a violin triad | D |
| violin pad | cloud cover | a sustained pad on every second bar with cloud of 20 percent or more | R |
| lead-in and wrap-out | the first and last chords | four bars each of drone, harp and the deep reed, with no kit; they carry no data | A |
Every step from record to sound, with its bounds: the mapping dictionary (tables in tracks/mapping/godbout-village/).
Harmony. 4/4 at 126 bpm, one bar per six-hour step (four bars a day, 61 bars) with 16 steps a bar, swing 0.54 and a small drag per voice. D Dorian or D Aeolian by temperature, one chord a bar from a four-chord loop; every pitched voice takes chord tones, the one diminished triad in each mode is skipped, and the only added colour is the suspended fourth. 2 min 18 s with the lead-in, the wrap-out and the reverb tail.
Simulated, not observed.
- The sun pulse is a modelled curve, not a measurement. The model gives one radiation value per six hours, and the shape inside each window is the clear-sky daylight curve scaled to that value.
- Every value is a model forecast; the run was not compared with what happened.
Be careful about.
- Weather appreciation, not a forecast, and not for route, marine, snowpack or avalanche decisions.
- The model’s grid box covers shore and water, so each value is an average over about 28 by 18 km, not the village alone; the temperature sits near the 3 C line that switches the mode, and a point off Baie-Comeau, 45 km west, gives nearly the same song (the mode differs on 1 bar of the 61).
- Six-hour steps hide a front’s passage; the model ends at 15 days, so the last storm is two bars.
- The thresholds (3 C, 3 and 6 hPa, the sun gates) and the wind-direction accent sets are chosen for musical contrast, not weather criteria.
- The values come from ECMWF’s open data; the saved point series is in data/meteogram/ and the extraction script is scripts/fetch_aifs_open.py.
- Claude Code cannot hear; levels were set by measurement and Graham has listened.
Reproduce. python scripts/meteogram_groove.py data/meteogram/aifs_godbout_village_ecmwf_20261016.json groove-v1-2-village · seed 1016 · locked 2026-10-17 · SHA-256 of the MP3 on this site cf8a123a0ac77f81… · check it with python scripts/renders.py verify godbout-village
Credits. Graham Watt, the place, the genre, the question and direction; Claude Code (Anthropic’s AI coding agent), mapping, sound design and all of the code; Data, ECMWF AIFS open data (CC BY 4.0); banner, GEBCO 2025; instruments, VSCO 2 Community Edition and Karoryfer Samples Big Rusty Drums, War Tuba and Bear Sax (CC0).
Changelog.
- v1.0 (2026-10-17): First published render of the Godbout take (groove v1.2: the lead-in, the wrap-out, the Bear Sax, the daylight-curve pulse and the night hand drums), from the 16 October 12 UTC run read at the village.
2424/461
2,424 lightning flashes over 461 minutes in the Cariboo and Chilcotin, the afternoon and evening of 30 June 2021, heard from a pin you can move. The album’s page.