Solstice, earliest sunset and latest sunrise / 冬至、最早日落与最迟日出 / 冬至と日の出入り
Article date and source retrieval date: 2026-10-27

Files in the ZIP
  solar-winter-usno-analysis.py     Complete offline analysis and optional plot
  solar-winter-usno-snapshot.json   Four raw plain-text annual calculation tables,
                                  their SHA-256 hashes and query URLs, and the
                                  seasons / two one-day JSON responses
  solar-winter-usno-results.json    Actual analysis output
  solar-winter-beijing.svg          Original plot made from the supplied data
  solar-winter-usno-README.txt       This file

Run from the extracted directory
  python3 solar-winter-usno-analysis.py > rerun.json

Default analysis uses only Python's standard library and makes no network calls.
The script resolves the snapshot path relative to itself, not the current working
directory. Environment values may differ on another machine; numeric and date
results should agree. This is a fixed analysis for these selected data, not a
general rise/set calculator or parser for every USNO phenomenon/latitude.

Recorded environment
  macOS 27.0.1, arm64, CPython 3.12.14.
  Optional plot: Matplotlib 3.11.2 (installed outside the blog project).
  To redraw, install that version in your own Python environment, then run:
  python3 solar-winter-usno-analysis.py --plot > rerun.json
  This writes solar-winter-beijing.svg beside the script and a PNG preview to
  /tmp/gumi-solar-winter-beijing-20261027.png. The plot shows daily predictions
  displayed to whole minutes, with lines connecting dates; not observations.

Inputs and precision
  Beijing representative point: 39.90 N, 116.40 E, fixed UTC+8.
  Tokyo representative point:   35.70 N, 139.75 E, fixed UTC+9.
  No daylight saving adjustment. These are specified points, not a claim that
  all parts of each city have identical rise/set times.
  Annual tables for 2026 and 2027, 365 valid dates each.
  Analysis window: 2026-11-01 through 2027-02-28 inclusive: 120 dates per city.
  All tied displayed minute extrema are retained. No hidden seconds are inferred.
  D = set - rise and M = (rise + set) / 2 use displayed endpoint readings.
  M is an arithmetic midpoint, not a general exact transit-time calculation.
  Independent endpoint rounding can cause non-monotonic one-minute differences
  close to the shortest-day plateau. Do not rank exact day lengths from that.

Actual checks
  Four original table text hashes and 365 rows for each city-year.
  Fixed-width columns preserve blanks for invalid short-month dates.
  120 consecutive winter dates per city across the year boundary.
  Two Beijing one-day API responses agree with table rise/set values and the
  arithmetic midpoint at the reported minute (2026-12-10 and 2026-12-22).
  Beijing change, Dec10 to Dec22: rise +8 min, set +4 min, duration -4 min,
  midpoint +6 min; both midpoint/duration identities checked.
  UTC solstice converted to 2026-12-22 04:50 UTC+8 / 05:50 UTC+9.
  Earliest displayed sunset before, and latest displayed sunrise after, the
  local solstice date in both selected winter windows.
  The script stops on an assertion failure. It does not independently validate
  USNO ephemerides, measure refraction/weather, or model terrain/observer height.

Primary data
  All exact query URLs and retrieval timestamp are in the source snapshot.
  Table service (choose year, coordinates, sunrise/sunset, fixed UTC offset):
  https://aa.usno.navy.mil/data/RS_OneYear
  API documentation:
  https://aa.usno.navy.mil/data/api.html
  Seasons response (UTC):
  https://aa.usno.navy.mil/api/seasons?year=2026&tz=0&dst=false

Primary explanations read 2026-10-27
  USNO: Sunrise and Sunset Times Near the Solstices
  https://aa.usno.navy.mil/faq/rs_solstices
  USNO: The Equation of Time
  https://aa.usno.navy.mil/faq/eqtime
  USNO: Rise, Set, and Twilight Definitions
  https://aa.usno.navy.mil/faq/RST_defs
  USNO: The Seasons and the Earth's Orbit
  https://aa.usno.navy.mil/faq/seasons_orbit
  NAOJ: Earliest/latest sunrise and sunset
  https://www.nao.ac.jp/faq/a0104.html
  NAOJ: Changing solar transit times and Equation of Time sign convention
  https://www.nao.ac.jp/faq/a0107.html
  NAOJ: Solar terms (select 2026; Japanese standard time)
  https://eco.mtk.nao.ac.jp/cgi-bin/koyomi/cande/phenomena_s.cgi

E is defined as apparent solar time minus mean solar time in the article.
Ignoring sub-minute time-scale details, solar noon in local clock minutes is
approximately 720 + 60*UTC_offset_hours - 4*east_longitude_degrees - E_minutes.
The longitude-only mean noon baseline is 734.4 min (12:14:24) for the selected
Beijing point, 701 min (11:41) for the selected Tokyo point. Neither is a
prediction of actual solar transit on every date.
