Sources, methods, credits, and AI use
NASA imagery and scientific context
- Aerosols in the Air, NASA SVS 5552, created by NASA GMAO and the Scientific Visualization Studio; visualizers Joseph V. Ardizzone and Helen-Nicole Kostis, scientific consulting Lesley Ott.
- The movie depicts August–September 2024 GEOS modeled aerosols.
- Our regional crop stitches the dateline and covers 170°E–150°W, 15–25°N.
- The embedded eruption footage is M. Zoeller / USGS Hawaiian Volcano Observatory, September 19, 2024, public domain, and is not a matched frame from the NASA movie.
- Coastline geometry comes from Hawaiʻi Statewide GIS, based on USGS Digital Line Graphs.
From pixels to music
- Our software checks the NASA movie five times per second to decide which musical notes to play. It uses 368 images from across the movie.
- We count a pixel as green when it is bright enough and its green component is much stronger than its red or blue components. This helps us pick out green aerosol patterns from the background.
- We divide the image into four equal sections from left to right. Each section controls one string note: C3, E3, G3, or C4. A note plays when green covers more than 5% of its section. The section containing the strongest small green patch also plays its note.
- We follow up to three green shapes and give each a number and a choir sound: alto, soprano, or bass and tenor together, singing “ah.” A shape’s position chooses one of four notes. Its voice gets quieter as the green fades. If we briefly lose track of a shape, its voice continues for a short time before stopping.
- Each held note uses a smoothed sustain loop, with a 0.12-second attack and 0.18-second crossfade on identity or pitch changes.
- We first downloaded NASA’s numerical GEOS-FP aerosol data and tried to recreate the green sulfate patterns in the movie. We tested different ways to display the values and could reproduce some broad patterns, but we could not establish an exact visual match or align individual movie frames with the data’s dates and times.
- The data includes sulfate aerosol optical thickness, or AOT. AOT describes how much airborne sulfate reduces light passing through a column of atmosphere by scattering and absorbing it. It is not the sulfate concentration at ground level.
- For this website, we chose to sample colors directly from NASA’s movie instead. Our software identifies green shapes and follows their patterns to trigger music that stays in step with the imagery. These image-derived features describe rendered colors, not numerical sulfate measurements.
Music, recordings, and licenses
- Choir samples: Epic Choir v1.0 by thepok.
- Choir samples derived from VocalSet (Wilkins, Seetharaman, Wahl, Pardo 2018), CC BY 4.0, VocalSet source dataset.
- The choir pack segments, repitches, extends, level-matches, and combines VocalSet recordings with simulated hall processing.
- The 16 loop-ready WAVs select the steady 0.8–2.7-second portion of each choir source, blend the last 0.30 seconds into the beginning, and retain a 1.6-second sustain loop.
- Browser playback adjusts G3/G4 by one semitone as in the source experiment. Loop source/output manifest. Full attribution · License text · CC BY 4.0.
- No source-author endorsement is implied.
Technology and accessibility
- This version uses semantic HTML, CSS, native video, Canvas, and Web Audio, with no external runtime library, account, analytics, or saved preferences. Asset provenance and hashes.
- Music starts only when requested; starting another player pauses the first.
- Use time sliders with arrow keys or Home/End.
- Musical activity is available as text without continuous playback announcements.
- Reduced-motion mode holds a still while music remains available; pausing and stepping through time updates the view.
- We aim to make this experience usable with screen readers, keyboard navigation, zoom, and color-independent text.
Typefaces
AI use
- OpenAI Codex assisted with research, writing and editing text, building the website, and coding the identification of green pixels and tracking of visual patterns. It also helped implement the musical rules, document the methods, and run automated checks.
- Humans, not AI, created the pixel-to-music concept and wrote the project video script.