Every populated place inside the path of the total solar eclipse of 2 August 2027, with how long totality lasts there and what that costs against the longest totality anywhere on the path. Places within 100 km outside the path are listed too, with how much of the sun is covered there.
Local circumstances are computed from the Besselian elements published by NASA GSFC, following the method in the Explanatory Supplement to the Astronomical Ephemeris (1974) and Meeus, Elements of Solar Eclipses (1989). Nothing is traced from a map.
Computed with ΔT = 76.0 s, the value NASA publishes for this eclipse, based on Morrison and Stephenson (2004). ΔT is not perfectly predictable and shifts the path east or west: an older NASA map for this eclipse used 71.7 s, a difference of about 2 km of longitude. The figure used is stated here and in the dataset, so you can tell which one these numbers assume.
The Moon has mountains, which move the path limits by 1–2 km and duration by 1–3 s against the smooth-limb figures used here. Limb-corrected predictions are usually published 12–18 months before an event.
These are different points. Greatest eclipse is where the shadow axis passes
closest to the centre of the Earth; greatest duration is where totality lasts
longest. For this eclipse they are 0.6 s and about 215 km apart. The site shows
how long totality lasts at each place, not how much it falls short. That gap
is in the CSV download as seconds_sacrificed, measured against
greatest duration, which is the honest denominator.
Times are local clock time at each place: the computed moment of maximum eclipse, less half the duration for the start and plus half for the end, in the place’s time zone from GeoNames. They are worked out when the dataset is built, with the rules of the IANA time zone database, so they include summer time: Egypt is on UTC+3 in August 2027, not UTC+2. Morocco moved to permanent UTC+0 on 20 September 2026, so Moroccan times here are an hour earlier than in sources made before then. Times are truncated to the minute, and the Moon’s mountains can move the start and end by a second or two.
All distances are great-circle. Circle membership is by distance, not by testing points against the drawn circle polygon, which would make places on the edge flicker in and out as you zoom. The radius is straight-line, and straight lines are not journeys: 120 km along the Nile and 120 km into the Western Desert are very different drives.
The computation is checked against independently published values before anything downstream is trusted.
| Check | Published | Computed |
|---|---|---|
| Greatest eclipse on the shadow axis | ~0 | 1.0e-5 |
| Sun altitude at greatest eclipse | 81.7° | 81.69° |
| Duration at greatest eclipse | 6m 23s | 382.5 s |
| Luxor | 6m 20s | 380.4 s |
| Countries crossed | 9 and Gibraltar | 9 and Gibraltar |
Sudan appears in some lists of countries this eclipse crosses. It does not: where the path is at Sudan’s longitudes the centre line is near 25–27°N, and Sudan’s northern border is at 22°N.
From GeoNames cities1000, filtered to populated places and
administrative seats. Sections of a town are left out, since GeoNames lists
them as places of their own and they would list one town twice. Every country
is searched, and the eclipse calculation decides which places are in totality.
Coverage is uneven by country: 117 Spanish places inside
the path against 18 Libyan and 8 Somali. That is a fact about who edits
GeoNames, not about where towns are. A place missing here is missing from
GeoNames.
Population figures come from GeoNames as published and are not always the town: for some Egyptian entries they are the surrounding district, which is why Esna appears larger than Luxor.
Every place in the same GeoNames list that is within 100 km of the edge of totality but not inside it: 1,061 places in 13 countries. They see a partial eclipse only. They are on the map as hollow dots and join the list when you search or drop a pin. They have no cloud, stays or ranking, and are in a CSV of their own.
For each one the site shows how much of the sun is covered at the peak, and the local time of the peak. The share covered is the part of the sun’s disc hidden by the Moon, by area. At 99% the sky dims, but the day stays bright and the corona does not appear: only totality shows it.
Two distances are given: to the nearest point on the edge of totality, and to the nearest point on the centre line, where totality is longest. Both are straight-line distances, measured to the path as drawn on the map, and are not road distances.
A partial eclipse is never safe to look at without eclipse glasses or a proper solar filter, even at 99%. Only totality itself is safe to look at with the naked eye.
ERA5 total cloud cover, 1985–2024, sampled at each place’s local eclipse hour on the eclipse date and the day either side. That is 120 observations per place, not 120 independent years, since consecutive days in one August are correlated.
ERA5 runs on a 0.25 degree grid, about 28 km. That is coarser than the terrain here: one cell can span the Nile valley and the desert plateau beside it, which do not have the same weather. Read these numbers as regional.
Checked against Jay Anderson’s satellite-derived figures at eclipsophile.com: Tarifa computes at 23.8% against his 26%, close agreement from an independent method and dataset. Luxor computes at 3.1% against his 0.7%, and Melilla at 25.3% against his 39%; both read high for the grid-resolution reason above. For a specific site Anderson is the better source. His tables are not reproduced here because they state no licence.
Climatology is not a forecast. A real forecast for 2 August 2027 will exist about a week beforehand and will be worth more than any of this.
An approximate count of somewhere you could sleep within 25 km and within 80 km, from OpenStreetMap. It answers one question: can you stay near here at all. Anything countable as a stay is included, whether it calls itself a hotel, a guest house, a hostel, an apartment or a hut. Campsites are counted separately.
It counts establishments, not rooms or beds. A 200-room hotel and a three-room guest house each count as one. Real room counts are not available: only 7% of these features carry a room count in OpenStreetMap, so any total built from them would be mostly guesswork.
No prices. This site holds none and scrapes none.
A count is what OpenStreetMap has mapped, not what exists, and says nothing about whether any room is free on the day.
The counts differ enormously by country: 28,463 places to stay in Spain against 1,314 in Libya. Much of that is how much tourism each country has. To check how much is mapping, hotels mapped in OpenStreetMap are compared with each country’s official count of hotel establishments. Hotels are used because they are the one kind of stay that both OpenStreetMap and every tourism ministry count.
| Country | Mapped | Official | Ratio | Coverage | Official source |
|---|---|---|---|---|---|
| Spain | 12,320 | 16,951 | 73% | high | INE, August 2026, provisional |
| Morocco | 2,748 | 1,223 | 225% | high | HCP, 31 December 2024 |
| Algeria | 1,078 | 1,423 | 76% | high | UN Tourism Statistics Database, 2024 |
| Tunisia | 735 | 665 | 111% | high | ONTT, 2022 |
| Libya | 318 | none | unknown | No official count since 2008, before the conflict. | |
| Egypt | 1,370 | 1,207 | 114% | high | UN Tourism Statistics Database, 2022 |
| Saudi Arabia | 1,834 | 3,020 | 61% | medium | GASTAT, Q2 2026 |
| Yemen | 672 | none | unknown | No official count since the war began; the last is from 2010. | |
| Somalia | 163 | none | unknown | No official hotel count found. | |
| Gibraltar | 6 | 5 | 120% | high | Statistics Office, 2025 |
Coverage is high when OpenStreetMap has at least 70% as many hotels as the official count, medium from 35% to 70%, low below 35%, and unknown where no credible official count exists. Above 100% is normal where the official count leaves out unclassified hotels. The comparison is rough, because each country defines a hotel a little differently:
The rating is national. It cannot tell whether the mapped hotels sit in the
cities or are spread evenly, so a small town’s count can still miss
places that exist. And a 25 km radius around a village next to a city sweeps
up the city’s hotels, so several places near Luxor show almost identical
counts. The figures and their sources are in data/hotel_registers.json,
and scripts/build_coverage.py recomputes the ratings.
From the UK Foreign, Commonwealth & Development Office, fetched when the dataset was built and shown with the date the FCDO last reviewed it. The advisory is attached to the country group, never to individual places, and nothing is hidden or reordered because of one. Where the FCDO names the specific regions an advisory covers, those are listed, because “parts of the country” without saying which parts is not useful.
On the map, a country under an advisory against all travel is shaded: red when the advisory covers the whole country, pink when it covers parts. The FCDO publishes those parts as names, not outlines, so the pink covers the whole country. Country outlines are from Natural Earth, in its UK point-of-view edition to match the source of the advisories.
The full dataset is on the map page as two CSV downloads: the places in totality, and the places within 100 km outside the path. Corrections are visible in the git history.
The dataset, both CSV files included, is ODbL 1.0. Use it, change it, build on it, including commercially; if you publish a database derived from it, publish that under ODbL too and carry the attributions above. It is ODbL rather than something more permissive because the lodging counts come from OpenStreetMap, which is share-alike.
The code that produces it is AGPL 3.0.