Zero Shadow Day

The problem

Indonesia straddles the equator, so every place in the country gets two days a year when the sun passes straight overhead at noon. On those "zero shadow days", a vertical stick casts almost no shadow. BMKG, the national weather agency, publishes the dates every year, city by city. Almost nobody outside specialists knows it happens, why, or how to show it.

I built Zero Shadow Day for three groups. Residents can find their own city's two dates and see why they fall where they do. Teachers and students get a classroom activity: pair with a school a few hundred kilometres away and repeat Eratosthenes' 2,200-year-old measurement of the Earth. Anyone else gets quick facts on solar noon, sun height and shadow direction.

What I built

Everything comes from latitude, longitude and date

The app has no data feed, no API key, no database and no backend. Earlier projects of mine had stalled on data licensing, and a project with no data dependency has nothing to license and nothing to go stale. I wrote the sun-position engine, the shadow geometry, the date search, the Eratosthenes module and the 3D scene myself. There is no astronomy, date, timezone or 3D library.

I checked the maths before building any screen

The first milestone had no interface at all, only the sun-position maths checked against published astronomical tables. A common shortcut formula for the sun's position can be off by half a degree. Near the tropics, that turns into a date several days wrong that still looks plausible. I also wrote the shadow geometry with no shared code from the sun engine, and a test enforces that. An error then shows up as two calculations disagreeing.

A shadow never quite vanishes, so I report a window

The sun's disc is about half a degree wide, so a shadow shrinks to a minimum and never reaches nothing. The app always gives a time window, sized from the sun's apparent width on that date. A curious user with a stick outside could disprove a claim of one exact second. To find the dates, I search each year's candidate days for the shortest noon shadow. Near the tropics, where the two dates merge into one at the solstice, this works with no special case.

Where BMKG and I disagree, the tests say why

I checked the app against 76 dates from BMKG's official tables for 2025 and 2026, covering 38 provincial capitals. Seventy-five match exactly, all 76 fall within a day, and times agree to under five seconds. The one miss is Pekanbaru 2026: I name 22 March, BMKG names the 21st. The city is 0.53° from the equator, so the sun crosses overhead between two noons. Four Papua rows in BMKG's 2025 table appear to list only their first time. A test derives that from the physics.

The shadow is the readout

The main view is an animated 3D stick and shadow for any place and date. The shadow's length and direction are the numbers, not a picture beside them. I built the 3D scene by hand to keep the whole app at about 110 KB of JavaScript, under a 150 KB budget. A general 3D engine would have taken the whole budget on its own.

Result

Zero Shadow Day is live in Indonesian and English as a static site, with your coordinates staying in the browser and no network requests at all. It shows your two dates with the sun's highest moment to the second, draws shadow-tip curves and the figure-eight analemma, and sweeps the overhead sun across the islands through the year. Eratosthenes mode finds a measurement partner elsewhere in the archipelago and shows your error against the true circumference, because the gap is the lesson. Any view can be shared as a link, with dates exportable to a calendar file. I built the first version in two days in August 2026, then returned at the end of September for a further pass that redesigned the visuals.