# Unified Layers — astronomical model notes

Release 119 · reviewed 29 September 2026.

This is a local, educational orientation model supporting a voluntary reflective journey. Its physical directions are tied to an explicit reference time and observer location. The later imaginative scenes are not physical measurements. Changing a scene, gesture or animation does not change the reference time, observer coordinates or planetary state.

## Changes from the supplied kit

The original EarthModel used low-accuracy fitted elements, represented Earth by the Earth–Moon barycenter (EMB), omitted a physical Moon state, and used UTC as an approximate dynamical and rotation argument. The replacement preserves the application API while calling the retained System State Atlas kernel.

Within the carried interval it uses NASA/JPL numerical samples for **Earth center (399), Moon (301), Sun (10), Mercury (199), Venus (299), Mars (499), and giant-planet system barycenters (5–8)**. EMB (3) is also retained. The local Sun and Moon directions subtract the observer's WGS84 position. TT, TDB, UT1 and UTC remain separate.

Outside the ephemeris interval a visible approximate-guide status replaces the sampled-data status. That fallback is a separate calculation, not extrapolated Horizons data. Its Earth object remains EMB and it has **no Moon state**.

## Retained data and bounds

The unchanged pack is **jpl-horizons-2026-6h-r1**, retrieved for release 118 on 27 September 2026. It retains **1,461 six-hour epochs for 11 targets**, from **2026-01-01 00:00 TDB through 2027-01-01 00:00 TDB**. Its source requests were geometric Cartesian states: ICRF axes, solar-system-barycenter origin, kilometers, kilometers per second, TDB time, and no observer/light-time corrections.

Ten target responses identify **DE441**; Mars center identifies **mar099**. Do not describe the pack as a uniformly DE441 body-center solution. Giant-planet names identify planetary systems because the target IDs specify system barycenters. [JPL Horizons API](https://ssd-api.jpl.nasa.gov/doc/horizons.html), [Horizons frames](https://ssd.jpl.nasa.gov/horizons/manual.html#frames)

Cubic Hermite interpolation uses position and velocity only between available samples. Release 118 withheld the intermediate three-hour samples: 1,460 tests per target, 16,060 total. Maximum observed position differences were **78.37 m for Mercury, 19.76 m for the Moon and 0.355 m for Earth**. These compare interpolation with values from the same underlying solution. They are not observational uncertainty estimates or guarantees at every instant.

The approximate guide uses the published JPL fitted elements. Its upper boundary is in TDB, so the latest accepted UTC input is **2049-12-31T23:58:50.814Z** under the future-leap assumption. This prevents extrapolating the final approximately 69 seconds of that UTC year. [JPL approximate positions](https://ssd.jpl.nasa.gov/planets/approx_pos.html)

Replay bounds use a one-millisecond inset from sampled endpoints for numerical edge safety: **2025-12-31T23:58:50.818Z** through **2026-12-31T23:58:50.815Z**. The difference from midnight calendar labels is the TDB–UTC offset. A 36-day illustration is shortened at a boundary; it cannot silently cross between samples and the guide. Replay does not mutate the reference snapshot.

## Time and Earth orientation

The kernel uses historical TAI–UTC, adds **32.184 s** to TAI for TT, and applies the approximate periodic NAIF model for TT–TDB. For 2026, TAI–UTC is **37 s**. Beyond 2026, holding 37 s is an explicit projection, not a known future schedule. JavaScript dates cannot represent the leap-second label 23:59:60. [IERS Bulletin C 72](https://datacenter.iers.org/data/16/bulletinc-072.txt), [NAIF time documentation](https://naif.jpl.nasa.gov/pub/naif/toolkit_docs/C/req/time.html)

USST appears in the snapshot as project profile **usst-tcb-j2000/1**: seconds of TCB coordinate time since JD 2451545.0 TCB. The ephemeris remains TDB-compatible. This declared coordinate label is not a universal proper clock, new physical law, or universal simultaneity. [IAU 2006 Resolution B3](https://iauarchive.eso.org/static/resolutions/IAU2006_Resol3.pdf)

Earth orientation uses IAU 2006 frame bias/precession, ERFA-derived IAU 2000B nutation with P03 adjustments, a truncated equation of equinoxes, UT1 rotation and polar motion. This reduced model is not the complete SOFA apparent-place pipeline. The inherited ERFA license must accompany redistribution. [ERFA source](https://github.com/liberfa/erfa), [SOFA](https://www.iausofa.org/2023-10-11c)

The retained orientation table has 368 daily rows with observation/prediction status, from the official USNO mirror of IERS Bulletin A. Predicted UT1 or polar motion causes prediction labeling. Outside the table, **UT1 = UTC and polar motion = 0** are explicit assumptions. Celestial-pole offsets dX/dY and length-of-day corrections are omitted. [USNO IERS series](https://maia.usno.navy.mil/ser7/finals2000A.all)

Release 118 compared the kernel with full ERFA IAU2006/2000A at 15 epochs. With identical EOP inputs, the largest direction discrepancy was 0.000410 arcsecond; the approximate TT–TDB discrepancy was 26.554 microseconds. These inherited algorithm comparisons neither establish phone pointing accuracy nor remove clock, EOP, location or orientation uncertainty.

## Observer, direction and velocity

Latitude and longitude are WGS84 geodetic coordinates. Optional height is ellipsoidal meters; the UI defaults to zero. The local basis is **east, north, up**. Bearings increase clockwise from true north, elevations from the geometric tangent horizon.

The topocentric calculation subtracts Earth center from Sun/Moon in common ICRF axes, rotates into Earth-fixed axes, subtracts the WGS84 observer position, then resolves the result into local east/north/up. Refraction, terrain, solar limb, horizon dip, light time, aberration and deflection are absent. A below-horizon marker does not establish optical visibility or invite looking at the Sun.

At a geographic pole spin speed is zero and spin bearing is undefined. Vertical targets also have undefined azimuth. Longitude at a pole supplies a conventional local meridian. Undefined values remain null rather than inventing north.

Earth's orbital velocity is heliocentric and uses actual Earth-center velocity when sampled data are active. Local spin uses a uniform adopted Earth angular rate. Solar Galactic velocity is a separate adopted convention. The optional combined velocity is a Newtonian vector sum after compatible rotations, not an absolute velocity or relativistic result. Gaze and pointing directions receive rotations only, never velocity additions.

## Galactic and regional references

The pinned Astropy v4.0 Galactocentric convention has center RA **266.4051°**, declination **−28.936175°**, solar distance **8.122 kpc**, solar height **20.8 pc**, and solar velocity **(12.9, 245.6, 7.78) km/s** in its adopted axes. Galactic-center direction, conventional Galactic north, and solar velocity direction are distinct. These fixed parameters neither integrate a future orbit nor claim to be the newest measurements. [Astropy Galactocentric frame](https://docs.astropy.org/en/stable/coordinates/galactocentric.html)

The Universal-stage marker adopts **ESO image eso9954c's field center**, toward Norma/Great Attractor: RA **16h 14m 22.57s**, declination **−60° 52′ 6.63″**. The ESO page prints these coordinates but does not explicitly identify their equinox/epoch. This application treats them as an **adopted ICRS/J2000-like regional direction**. That serves an illustrative regional marker, not precision astrometry or a uniquely determined Great Attractor center. [ESO image record](https://www.eso.org/public/images/eso9954c/)

Large-scale peculiar flow differs from a sky bearing or Galactic orbit. Tully and colleagues' Laniakea analysis reconstructs a flow watershed after subtracting mean expansion and long-range flows. The procedural scene does not ingest that catalog or integrate its flows. It claims neither a measured velocity field nor a future trajectory and supplies no center of the Universe. [Tully et al., 2014](https://arxiv.org/abs/1409.0880)

Multiversal, Infinite, Unified and the return scene are artwork and project symbolism. They have no physical bearing.

## Rendering contract

Legacy names remain available. Additions:

- **solarSystem(ms)**: reference time, scales, solar states, Moon–Earth vector and source status.
- **bodyState(index, ms)**: indices 0–7 select Mercury through Neptune; giant planets retain system-barycenter identity.
- **orbitalPoint(index, ms)**: actual position when covered, approximate guide otherwise. An explicit eccentric anomaly requests a separate fitted-ellipse helper for legacy compatibility; it must never be labeled a sampled NASA trajectory.
- **velocity(index, ms)**: heliocentric J2000-ecliptic km/s.
- **replayEpoch(referenceMs, offsetMs)**: numeric UTC timestamp clamped to the reference source regime.
- **snapshot.modelStatus**: label, body identity, interval, EOP status, future-UTC caveat and limits.
- **snapshot.moon**: topocentric direction/distance when available; otherwise null.
- **moonRelativeEclAU**: Moon minus Earth in J2000-ecliptic AU; otherwise null.
- **sunGeocentricFixed**: geometric illumination direction for the globe, distinct from the observer's solar direction.

Scene body positions are heliocentric **J2000-ecliptic AU**, rotating compatible ICRF states by fixed obliquity 84381.448 arcseconds. Raw positionKm/velocityKmS keep the snapshot's stated origin: barycentric for NASA samples, heliocentric for the guide. Names do not replace frame metadata.

Directions now carry speed:null; velocity targets retain km/s. This corrects the old unit-direction magnitude being available as a misleading speed. Rendered Moon separation and body sizes may be enlarged only with labels and direction preservation.

## Validation

Run **node tests/model-tests.cjs**. The adapter suite passes **22 cases**: browser/Node consistency, retained samples, origin translation, topocentric Sun/Moon, WGS84, lunar parallax, transformations, time scales, EOP labels, poles, height, bounded replay, fallback endpoints and invalid inputs. It writes research/model-tests.json.

These tests cover the adapter and retained-data application. They do not independently reimplement AtlasAstro. Its release-118 review and numerical comparisons remain a separate evidence layer. Reuse here is not a new NASA retrieval.

The local data hash inventory and the two newly reviewed regional sources are in research/model-data.json. Model calculations do not browse, send location to services, contact NASA, or generate astronomical values with an AI.