# 2027 astronomy reference

This directory is a planning reference for Fairbanks, Yellowknife, Reykjavík,
Tromsø, Rovaniemi, Abisko, Hobart and Queenstown. It is not an aurora forecast,
a cloud forecast or a probability of a successful sighting.

`astronomy-2027-eight-cities-daily.v1.csv` contains 365 local observing nights
for each city; `astronomy-2027-eight-cities-monthly.v1.csv` aggregates them.
An observing night starts at local noon and ends at the next local noon, then is
labelled with the start date. That convention keeps the after-midnight portion
with the preceding evening and preserves IANA daylight-saving transitions.

The calculation samples every ten minutes at midpoint. It uses apparent,
topocentric centre altitudes for the Sun and Moon with no atmospheric
refraction, terrain, observer elevation or local-horizon model. The CSV's
cumulative threshold columns mean Sun centre below −6°, −12° and −18°.
In standard terminology, the civil, nautical and astronomical twilight bands
are 0° to −6°, −6° to −12° and −12° to −18°; below −18° is astronomical
darkness. `moon_below` means Moon centre below a geometric level horizon while
the Sun is below −18°. Each boundary has approximately one ten-minute grid
interval of resolution. Moon illumination is the physical illuminated fraction from
Sun–Moon–observer geometry, not a linear phase approximation.

The source coordinates are the [published `city-aacgm-inputs.v1.json`](https://auroraforecast.me/data/research/aurora-2026-11/city-aacgm-inputs.v1.json)
from the prior AuroraMe dataset. The generator is
`web-seo/scripts/research/2026-11-08/build_astronomy_2027.py`; it requires
Python 3.14, Skyfield 1.55, NumPy 2.5.4, pandas 3.1.1 and Matplotlib 3.11.2.
It downloads JPL DE440s from
`https://ssd.jpl.nasa.gov/ftp/eph/planets/bsp/de440s.bsp`; its SHA-256 is in the
manifest. Run it with `--input`, `--output` and `--ephemeris` paths so the
ephemeris remains outside the repository.

The daily illumination mean is uniformly weighted over its ten-minute dark
samples. The monthly field
`unweighted_mean_of_nightly_dark_sample_illumination_pct` is instead the equal
weight mean of daily dark-sample means for observing nights that had at least
one astronomical-dark sample; it is not a pooled dark-hour mean. The
[machine-readable manifest](astronomy-2027-manifest.v1.json) records source and
ephemeris hashes, versions and validation definitions.

Validation: 2,920 daily rows and 96 monthly rows; every row passed
`0 ≤ moon-below ≤ astronomical ≤ nautical ≤ civil ≤ night duration`. The 2027
March full Moon was computed as 10:43:48 UTC; the [US Naval Observatory phase
API](https://aa.usno.navy.mil/api/moon/phases/year?year=2027) reports 10:44 UTC.
The local clock handling is visible in the data: Fairbanks runs 23 hours from
2027-03-13 noon −09:00 to 2027-03-14 noon −08:00 and 25 hours from 2027-11-06
noon −08:00 to 2027-11-07 noon −09:00. As a polar sanity check, the Fairbanks
and Tromsø June 21 rows have zero cumulative values at all three thresholds,
while Hobart retains 11.5 astronomical-dark hours.
The `yellowknife-march-2027-full-moon-window-example.v1.csv` uses grid windows,
not precise moonrise or moonset times.

These data omit clouds, light pollution, actual auroral activity and local
terrain. A zero under the strict −18° criterion does not mean bright aurora is
impossible in twilight.