eclipses.app

Methodology

How we compute real eclipse visibility from your location

We combine topocentric astronomical computation, a 30 m digital elevation model and a calculation of which clouds cross your line of sight to the Sun to answer one simple question: will you actually see it from where you plan to be?

The problem

Official sources publish the path of totality: the band within which the Sun is completely covered by the Moon. It is an exact astronomical calculation, but it ignores local terrain and atmosphere.

For the 12 August 2026 eclipse, the path crosses the north of the Iberian Peninsula and ends grazing Mallorca. On paper, 166 Spanish cities lie within it. In practice, in the vast majority the Sun will be less than 5 degrees above the horizon during totality — an “eclipsed sunset”. And a real topographic horizon (a mountain to the west, buildings, vegetation) can block the Sun before the event ends.

Our analysis checks those 166 cities against the real topographic horizon (DEM, not the geometric one): in 149 of them totality is visible in full above the terrain, plus 3,533 more viewpoints in the same case. That is exactly what a naive reading of the path map does not tell you: where the terrain blocks the Sun.

That difference, between “being inside the path” and “actually seeing it”, is what we compute at eclipses.app.

We precompute 560 locations — 394 municipalities with more than 20,000 residents plus 166 districts of the largest cities, out of Spain’s 8,132 municipalities — and 18,878 viewpoints. For any other point, the simulator calculates just as accurately.

Our approach

1.Topocentric astronomical computation

We start from JPL's high-precision planetary ephemerides to place the Sun and Moon, and we solve the eclipse geometry point by point across the Earth's surface, with corrections for parallax and for the real elevation of your location — not at sea level.

From there we derive the eclipse contacts at your location: C1 (start of partial), C2 (start of totality if applicable), peak, C3 (end of totality) and C4 (end of partial), with their UTC and local timestamps.

To turn that geometry into a clock time, we need to know exactly how much the Earth has rotated — a value that is measured and published daily, and that cannot be predicted years in advance. We always use the latest available measurement rather than a figure set long ago.

2.Timing precision

Our goal is to provide the most accurate calculation of contact times you can find. To do that we solve the eclipse geometry with high-precision ephemerides, apply the most recently published Earth rotation value, calculate at the real elevation of your point, and — where we have the data — correct for the real relief of the Moon's limb.

You may find differences of a few seconds between our times and those from other published sources, including official platforms and commonly used applications. The usual cause is not the geometry, on which we all agree very closely, but the Earth rotation value: many of those publications were generated years ago with an estimate of that value that we now know is outdated, and they have not been recalculated since.

That offset does not translate one-to-one to the contact time. The Earth rotates while the shadow advances, so a few seconds' difference in rotation becomes slightly more or slightly less time depending on the geometry of each eclipse: the difference is not the same in 2026 as in 2027 or 2028.

For the same reason, the level of detail is not identical across the three eclipses. We prefer to say what each calculation includes rather than present them all as equivalent.

3.30 m Digital Elevation Model

The key question is not “is the Sun geometrically above the horizon?” but “is it above the real topographic horizon?”. To answer it we need to know what surrounds the observer.

We use Copernicus DEM GLO-30 tiles, with 30-metre resolution (≈0.0003°) covering the entire Iberian Peninsula, the Balearic Islands and the Canary Islands. At each point we trace 1,440 rays around the point —one every 0.25°— and follow each one out to 100 km, apply corrections for Earth’s curvature and atmospheric refraction, and obtain the maximum elevation angle the Sun would have to clear to be visible at each azimuth.

The difference (delta) between the Sun’s altitude and the topographic horizon’s altitude is what we call margin over the real horizon.

4.Clouds in your line of sight to the Sun

During the eclipse the Sun sits very low over the horizon — between 1° and 8° depending on the location — so what decides whether you see it isn't the cloud right above you, but whatever crosses the straight line between you and the Sun, sometimes far away. We follow that line point by point, not just the vertical column of air over the observer.

That's why we look out to 250 km, not just a few dozen. A low cloud layer right next to you can sit entirely above that line — at 5 km, with the Sun at 1°, the line runs at under 90 metres — and still cut straight through it much further out, around 56 km. A high cloud can block the Sun from nearly 240 km away: looking only nearby would miss it.

We discard phantom cloud: on mountainous terrain, some of the cloud data the models report corresponds to a height that actually falls below the ground itself. In a sweep across northern Spain, around two out of every three low-cloud readings turned out to be exactly that, and they're excluded from the calculation.

And when several layers overlap along that line we don't simply add them up: the same cloud seen several times in a row counts once, so a partially cloudy sky doesn't turn, through naive arithmetic, into an almost fully covered one.

5.Historical cloud-cover climatology

The calculation above needs a real weather forecast, which is only reliable up to about 16 days before the event. For dates further out — and for the “historical cloud cover” block you see on every city page — we fall back on climatology: ECMWF's ERA5 reanalysis (1991-2020 baseline) tells us the historical probability of clear skies at that time of day and month, from years of real data — not a forecast for the specific day, but the average of what usually happens.

The two coexist on purpose: climatology gives a stable reference available from day one; the line-of-sight calculation above kicks in as 12 August approaches, replacing that average with an increasingly precise forecast.

Lunar limb and Baily's beads

Beyond the terrestrial horizon, we model the real relief of the lunar limb — the Moon's irregular edge, with mountains and valleys — to place Baily's beads accurately. We rely on JPL's DE440 planetary ephemerides, lunar orientation through NASA's SPICE/NAIF kernels, and lunar topography from NASA LRO LOLA and SELENE/Kaguya (SLDEM2015).

With this data, C2 and C3 contact times are corrected for the real limb profile — not a perfect lunar disk — and Baily's beads appear in the real valleys of the lunar edge, where the last sunlight filters through the mountains.

This correction is available for the 2026 and 2027 eclipses. For 2028 it is not yet computed: it is a long calculation, point by point across the entire path, and we prefer to say where we apply it and where we don't rather than present all three eclipses as equivalent. We will publish it when it is finished.

The solar corona and prominences we show in the simulator are an illustrative representation of the phenomenon, not a prediction: their exact shape cannot be known in advance.

How we simulate Baily's beads →Real lunar-limb relief (LRO/Kaguya), precision ephemerides and a physical photographic exposure: what the eclipses.app eclipse simulator computes, and from which data.

0-100 score

We combine the above factors into a single number between 0 and 100. A score > 80 means excellent conditions across all dimensions: totality, ample margin over the real horizon, significant duration and favourable weather. A score < 40 means it’s worth looking for another nearby spot: either because the Sun will be almost on the ground, because a mountain silhouette will hide the event from you, or because clouds historically cover the area in August at that hour.

The score combines four factors: eclipse type (total > annular > partial), margin over the real topographic horizon, duration of the phenomenon from your point, and climatology. The exact formula is not published to prevent SEO gaming, but the breakdown is always accessible via API with an authorised account (contact for academic access).

Honest limitations

What the score does not model:

  • Forecast, not a guarantee — close to the date we compute which clouds cross your line of sight using the latest weather model, but it's still a forecast: it changes from one model run to the next, so it's worth checking again close to 12 August. Far from the date we use historical climatology, which is an average of past years, not a prediction for the exact day.
  • Local air conditions — haze, suspended dust, humidity. They affect the appearance of the low Sun, not computed.
  • Nearby obstacles — an adjacent building, a tree, a streetlight ruin the field of view and do not appear in the 30 m DEM.
  • Unplanned events — wildfires, Saharan dust, exceptional contrails.

These limitations are not negligence, they are honesty about the model: the best way to reach the spot with confidence is to check the forecast 24 h beforehand and have a 60-90 minute drive alternative ready. For PRO users, the app includes a meteorological plan B: three alternative locations within a reasonable radius, with real-time updated forecast.

Open data

We publish:

  • Press kit with the 2026 eclipse summary + JSON dataset by autonomous community.
  • Public API to verify our calculations against external sources (no commercial use).
  • Score map in raster format covering the entire eclipse corridor.

Validation

We check our times against independent implementations, written with a different method and without sharing code with ours: they agree within the same second. We also collect timed observations from people who were there, because a real measurement is the only way to verify the lunar relief model. If you timed a contact, you can tell us. Did you time the eclipse?

Are you a journalist or researcher?

If you need:

  • Access to the full dataset (CSV/JSON without gating)
  • Verification of a specific calculation before publishing
  • Visual examples to illustrate an article
  • Technical citations with attribution

Write to us at [email protected] or check the press kit at /prensa. We answer within 48 hours.