AEGIS

Active Earth Geomagnetic Imaging System
_{A real-time, browser-based 3D visualization of Earth's magnetosphere, driven by live space-weather data from NOAA.}

---

AEGIS ray-marches the magnetosphere in a single WebGL2 fragment shader. The magnetopause, bow shock, plasmasphere, ring current, magnetotail current sheet, auroral ovals and substorm reconnection are all recomputed every frame from the solar wind, IMF, Kp and GOES X-ray feeds that NOAA publishes right now. Nothing is pre-baked — the shape you see is the shape of near-Earth space at the moment you load the page.

Demo: https://aegis.sponde.de/

---

What you're looking at

Element	Meaning
Blue/white teardrop	The magnetosphere — Sun-compressed, drawn into a long night-side tail
Warm dayside dome	Bow shock + magnetosheath — shocked, heated, draped solar wind
Arcing strands ("jellyfish")	Geomagnetic field-line shells (McIlwain L = 2…6) — and they flex with the drivers: the dayside bell crushes inward under pressure, the inner shells balloon out as the ring current (Dst) deepens, and the tail draws out under southward Bz
Inner glow	Plasmasphere — cold dense plasma; shrinks inward during storms
Crimson glow hugging Earth (storms)	Partial ring current — noon-tight / midnight-bulged, frozen onto the closed field lines; the Dst signature of a main phase
Tail band	Plasma sheet (Harris current sheet), flapping in real time
Orange spot behind Earth	Near-Earth reconnection X-line (Bz southward)
Polar rings	Aurora ovals — the live NOAA OVATION nowcast; in storm replays, a model oval (Kp/Bz-driven teardrop: fat and deep on the night side, narrow on the day side) that expands equatorward ~30 min after the tail loads (substorm growth phase)
Terminator	Real day/night boundary for the current UTC time

A full annotated walkthrough is in the manuals: English · Deutsch.

---

Quick start

AEGIS is a static site with no build step — plain HTML + ES modules + GLSL. It does need to be served over HTTP (ES module imports and fetch() do not work from file://), and it needs network access to reach the NOAA endpoints.

git clone https://github.com/Kracht/AEGIS.git aegis
cd aegis

# any static file server works; e.g. Python:
python3 -m http.server 8080

# then open http://localhost:8080

Requirements

• A browser with WebGL2 (Chrome/Edge/Firefox/Safari, last few years).
• Internet access for the live NOAA feeds. Offline or if a feed fails, AEGIS falls back to quiet-condition defaults and shows a stale-data warning.

Controls

• C (or the Cam: … label by the FPS counter) toggles free-look: by default the camera flies a slow cinematic arc across the flank (the side-on angle where the storm deformation reads best, never diving down the tail); in free-look you drag to orbit Earth and scroll to zoom (4–45 Rₑ), so you can park on whichever 3/4 angle frames the compression and ring-current inflation. The view stays centered on Earth in both modes.
• F2 (or the Settings [F2] label by the FPS counter) toggles the visual tuning panel — camera orbit/FOV, exposure, gamma, and per-layer intensity.
• F3 (or the Mode: … label next to it) cycles the render mode: Visual (default volumetric scene) → Data (Visual underneath, with a panel of every live uniform — value, units, citation, and the scene feature each drives, including the L1 lag clock) → Physics (SDF line-art of the magnetopause, bow shock and L-shells plus a camera-synced 2D overlay of the mechanism: open-vs-closed field-direction glyphs, draped IMF, and the two reconnection X-lines — the dayside one rides Bz to the equator, the near-Earth neutral line lights minutes later on the lagged tail driver; a schematic, not an MHD solution). The choice persists in localStorage.
• F4 toggles the causal HUD — the two-branch graph of why the scene changed: a fast compression branch (Pdyn → r₀) and a slow storm branch (Bz → reconnection → injection → Dst), drawn as separate tracks because they are independent mechanisms. Nodes light from the live values, edges carry the real propagation delays (the L1 advection clock, the ring-current decay time τ), and hovering a node reveals its governing equation, current value, and citation.
• The transport bar along the bottom replays curated instrument-era storms. Pick Live (NOAA realtime) or a curated event — November 2004, St. Patrick's 2015, Gannon 2024, January 2026, or a high-speed stream — then play / pause / scrub and set the time-acceleration. Selecting a storm reveals the causal HUD automatically; that's where the lag clocks and the branch independence become legible (you can't watch a 7-hour Dst recovery in real time). The high-speed stream is the teaching contrast: it compresses r₀ almost as hard as the November superstorm, yet drives only a tenth of the Dst — compression and storm are not the same thing. During a replay the status panel shows the measured SYM-H beside the modeled Dst, so you can watch the estimate track (or miss) the real ring current. The Dst node in the causal HUD also carries an observed Hp30 ghost (GFZ Potsdam's 30-min planetary Kp-family index, open-ended above 9 so superstorms register their actual intensity); hover the node to see it, and the node's border tints amber when Hp30 crosses the Kp-equivalent storm-onset (≥ 5). The propagation delays are always on view there too — L1 → bow shock and the further aurora growth-phase lag. The bar collapses: click the TIMELINE handle along its top edge to slide it down to a thin strip (and click again to bring it back) — handy for an unobstructed full-scene view. The state is remembered.

---

How it works

NOAA SWPC / GOES / OVATION  ──▶  data-fetcher.js   ──▶  Shue (1997) r₀, α
                                  aurora-texture.js ──▶  polar aurora grids
                                          │
                                          ▼
                            renderer.js  (uniforms, textures)
                                          │
                                          ▼
              fragment.glsl  —  one full-screen triangle,
              volumetric ray march of the whole scene

1. magnetosphere-model.js is the physics core: given a raw solar-wind sample and a clock, it derives the Shue et al. (1997) magnetopause standoff r₀ and flaring exponent α plus the solar-wind dynamic pressure, and integrates the Burton/O'Brien (2000) ring-current equation for a live Dst estimate. It buffers each snapshot into a 90-min history ring and reports the L1-advected quantities at now − 1.5×10⁶ km / v_sw (~55 min @ 450 km/s, ~31 min @ 800 km/s), so the scene shows what the magnetosphere is seeing now — not what the L1 probe just observed. The Dst ODE is closed over the same lag, and a τ≈5 min low-pass eases Kp's 3-hourly bin steps. The model is time-agnostic (no wallclock inside): data-fetcher.js drives it with Date.now() from the live NOAA feed, while timeline-source.js drives the same model with a scrub clock from a curated storm — so replay obeys identical physics, and the causal sequencing emerges rather than being keyframed. Both sit behind data-source.js and are interchangeable through createDataSource(id).
2. aurora-texture.js polls the NOAA OVATION aurora nowcast into two polar R8 textures. These drive the oval in live mode; in replay (no historical OVATION exists) the shader synthesises the oval from the model state — a Gussenhoven (1983) equatorward boundary (~2°/Kp) shaped into a teardrop, fed by a Bz delayed an extra ~30 min beyond the L1 lag so the oval expands after the tail loads (the substorm growth phase).
3. renderer.js compiles the program and pushes per-frame uniforms; the only geometry is a single oversized triangle.
4. fragment.glsl ray-marches emission/extinction through an SDF model of the magnetosphere (96 jittered steps, Reinhard tone map). The field-line shells are not a frozen shape: their deforming warp is driven by the live state the way an empirical field model (Tsyganenko) parametrises its analytic deformation — dynamic pressure (via r₀) pinches the dayside, southward Bz stretches the tail, and the integrated Dst inflates the inner closed shells. Because Dst is the lagged, decaying ODE output, the fast pressure compression and the slow ring-current inflation visibly separate in time.

---

Data sources

All live data is fetched client-side from NOAA's Space Weather Prediction Center (SWPC) — U.S. Government work, public domain.

Feed	Endpoint	Used for
Solar wind magnetic field	services.swpc.noaa.gov/json/rtsw/rtsw_mag_1m.json	Bz, Bt
Solar wind plasma	services.swpc.noaa.gov/json/rtsw/rtsw_wind_1m.json	speed, density
Planetary K-index	services.swpc.noaa.gov/products/noaa-planetary-k-index.json	Kp, G-storm scale
GOES X-ray flux	services.swpc.noaa.gov/json/goes/primary/xrays-1-day.json	flare class
OVATION aurora	services.swpc.noaa.gov/json/ovation_aurora_latest.json	auroral ovals
Hp30 (GFZ)	kp.gfz.de/fileadmin/files_for_gfz_cms/Hp30_ap30_nowcast.txt (via the optional api/hp30_proxy.php)	observed 30-min planetary index — ghost trace beside the modelled aurora driver

Solar-wind measurements come from whichever L1 monitor SWPC marks active in the RTSW feed (SWFO-L1, IMAP, ACE or DSCOVR); X-ray flux comes from the GOES satellites. Hp30 is the Kp-family 30-minute geomagnetic index maintained by Geomagnetic Observatory Niemegk / GFZ Helmholtz Centre for Geosciences (Yamazaki et al. 2024, DOI 10.5880/Hpo.0003, CC BY 4.0); unlike Kp it is open-ended above 9, so the strongest storms register their actual intensity instead of saturating.

Live Hp30 needs a same-origin proxy. GFZ's file server sends no Access-Control-Allow-Origin, so a browser cannot fetch it directly. The repo ships a tiny optional pass-through at api/hp30_proxy.php — drop it on any PHP-capable host alongside index.html and live Hp30 lights up; serve the repo from anything without PHP (Python http.server, GitHub Pages, npx serve) and the source self-disables silently. The replay path is unaffected either way — tools/augment-hp30.mjs bakes Hp30 into the scenario JSONs at build time, so the curated storms always show the index climbing past 9 regardless of how you host the live demo.

The curated-storm replays use real instrument-era data from NASA's OMNI dataset (IMF + plasma time-shifted to the bow-shock nose, 5-min cadence, plus hourly planetary Kp), fetched once via the CDAWeb REST service and bundled as compact JSON under data/scenarios/ (regenerate with tools/build-scenarios.mjs). Halloween 2003 is conspicuously absent: OMNI's upstream monitors were saturated during that superstorm, so a faithful driver-driven replay is impossible — the same "real data only" rule that rules out Carrington 1859. November 2004 (fully covered, comparably deep) stands in.

---

Project structure

index.html            # entry; canvas + status panel mount points
src/
  main.js             # boot + render loop + orbital (terminator) maths
  renderer.js         # WebGL2 program, textures, per-frame uniforms
  data-source.js      # data-source seam + createDataSource(id) factory
  magnetosphere-model.js # time-agnostic physics core (Shue, Dst ODE, L1 lag)
  data-fetcher.js     # live NOAA SWPC ingestion — wallclock driver of the model
  hp30-source.js      # live GFZ Hp30 nowcast poller — observed-ghost trace
  timeline-source.js  # curated-storm replay — scrub-clock driver of the model
  scenarios.js        # curated-storm manifest (shared by source + transport)
  aurora-texture.js   # OVATION nowcast → polar GL textures
  ui.js               # live status HUD (incl. modeled Dst) + EN/DE manual links
  dev-panel.js        # FPS readout + Settings [F2] tuning panel
  render-mode.js      # Visual/Structural/Data/Physics mode controller (F3)
  causal-hud.js       # two-branch causal graph overlay (F4)
  camera-projection.js # JS mirror of the shader camera (projects world→screen)
  physics-overlay.js  # Physics mode: camera-synced field/topology 2D overlay
  camera.js           # auto cinematic orbit ↔ free-look (drag/zoom) controller (C)
  transport.js        # scenario picker + scrub/play/speed bar
shaders/
  vertex.glsl         # full-screen triangle
  fragment.glsl       # the entire scene (ray-marched volumetrics)
textures/             # NASA Blue Marble (monthly) + Black Marble night
data/scenarios/       # bundled real OMNI storm series (replay)
tools/build-scenarios.mjs # offline regenerator for data/scenarios/ (NASA CDAWeb)
tools/augment-hp30.mjs    # offline: appends GFZ Hp30 ghost column to each scenario
api/hp30_proxy.php        # optional CORS pass-through for live GFZ Hp30 nowcast
docs/                 # user manuals (EN / DE)

---

Acknowledgements & references

This project stands on published space-physics models, public NASA/NOAA data, and well-known real-time graphics techniques.

Space-physics & empirical models

• Shue, J.-H., et al. (1997). A new functional form to study the solar wind control of the magnetopause size and shape. J. Geophys. Res., 102(A5), 9497–9511. — magnetopause shape.
• Fairfield, D. H. (1971). Average and unusual locations of the Earth's magnetopause and bow shock. J. Geophys. Res., 76(28), 6700–6716.
• Cairns, I. H., et al. (1995). — bow-shock standoff scaling.
• Harris, E. G. (1962). On a plasma sheath separating regions of oppositely directed magnetic field. Nuovo Cimento, 23, 115–121. — tail current sheet.
• Tsyganenko, N. A. (1995, 2002). Modeling the Earth's magnetospheric magnetic field… J. Geophys. Res. — the empirical, driver-parametrised field deformation (Pdyn, Dst, IMF) that the field-line warp emulates qualitatively.
• Carpenter, D. L., & Anderson, R. R. (1992). An ISEE/whistler model of equatorial electron density in the magnetosphere. J. Geophys. Res., 97(A2), 1097–1108. — plasmapause.
• Burton, R. K., McPherron, R. L., & Russell, C. T. (1975). An empirical relationship between interplanetary conditions and Dst. J. Geophys. Res., 80(31), 4204–4214. — ring-current / Dst equation.
• O'Brien, T. P., & McPherron, R. L. (2000). An empirical phase space analysis of ring current dynamics. J. Geophys. Res., 105(A4), 7707–7719. — the injection / decay parameterisation used for the live Dst estimate.
• Newell, P. T., et al. (2009). — OVATION auroral precipitation model (delivered operationally as NOAA SWPC's OVATION aurora nowcast).
• Gussenhoven, M. S., Hardy, D. A., & Heinemann, N. (1983). Systematics of the equatorward diffuse auroral boundary. J. Geophys. Res., 88(A7), 5692–5708. — the ~2°/Kp equatorward-boundary relation used for the replay aurora oval.
• Cooper, P. I. (1969). — solar declination equation, used to place the day/night terminator.
• Reinhard, E., et al. (2002). Photographic Tone Reproduction for Digital Images. — the c/(1+c) tone-mapping operator.

Real-time graphics techniques

• Inigo Quilez — articles on raymarching distance fields, value noise / fBm, and domain warping (iquilezles.org). The SDF scene model and turbulence are built on these techniques.
• The classic GLSL fract(sin(dot(...)) * 43758.5453) hash, and the sine-free variants surveyed in Dave Hoskins, "Hash without Sine" (Shadertoy).
• The single full-screen-triangle trick for shader-only rendering.

Data & imagery

• NOAA Space Weather Prediction Center — live space-weather data (DSCOVR @ L1, GOES, OVATION). Public domain.
• NASA Visible Earth — Blue Marble Next Generation (monthly) and Black Marble / Earth at Night surface imagery. Credit: NASA Earth Observatory; used with attribution.
• NASA/GSFC Space Physics Data Facility — OMNI (King, J. H., & Papitashvili, N. E.). High-resolution (5-min) and hourly OMNI data, accessed via CDAWeb. doi:10.48322/hkaw-ff03. Public domain — the curated-storm replays.

Any errors in the physical interpretation are mine, not the cited authors'.

---

Disclaimer

AEGIS is an illustrative, schematic visualization for education and outreach. It blends empirical models with artistic interpolation and is not a forecasting or operational tool. The on-screen Dst is a modeled estimate from the Burton/O'Brien coupling, not the official Kyoto Dst index. For authoritative space-weather information see NOAA SWPC.

---

License

MIT © 2026 skracht. Third-party data/imagery (NOAA, NASA) retain their own terms — see the LICENSE notice.