How accurate does your elevation need to be?
In our first post we showed that standing high up brings sunrise forward by minutes, and that a terrain lookup often misses a clifftop. That left a practical question we get asked: if I supply the elevation myself, how exact does it have to be?
The answer depends a great deal on how high you are.
The curve is a square root
We computed sunrise at 53°N on the spring equinox for a range of heights and measured how much earlier it came than at sea level.
| Height | Sunrise earlier by |
|---|---|
| 5 m | 28 s |
| 10 m | 40 s |
| 25 m | 1m 04s |
| 50 m | 1m 30s |
| 100 m | 2m 08s |
| 200 m | 3m 00s |
| 500 m | 4m 45s |
| 1,000 m | 6m 44s |
| 2,000 m | 9m 31s |
The distance to the horizon grows with the square root of your height, and the time shift follows it.[1] Four times the height gives twice the shift.
So an error of ten metres means very different things in different places. On a sea wall, the first ten metres are worth 40 seconds. At 200 m, ten metres either way moves sunrise by about five seconds. At 1,000 m it is two.
In practice, if you are near the water, the difference between the beach and the top of the dune matters more than you would guess. If you are on a hill, a figure read off a map contour is plenty.
How two terrain lookups did
We tested two public terrain models against places whose height is well
known. The first is ETOPO5, which has cells about 9 km across and is what
our auto_elevation option reads.[2] The second is ETOPO1, with cells about
1.8 km across,[3] as returned by the sunrisesunset.io API, which reports
the elevation it used.[4]
| Place | Actual | ETOPO5 | ETOPO1 |
|---|---|---|---|
| Cliffs of Moher, Ireland | 214 m | 0 m | 112 m |
| Beachy Head, England | 162 m | 0 m | 13 m |
| Cape Point, South Africa | 249 m | 0 m | 0 m |
| Cabo Girão, Madeira | 580 m | 0 m | 35 m |
| Preikestolen, Norway | 604 m | 0 m | 462 m |
| Látrabjarg, Iceland | 441 m | 0 m | 0 m |
| Zermatt, Switzerland | 1,608 m | 2,393 m | 1,740 m |
| Denver, Colorado | 1,609 m | 1,658 m | 1,614 m |
| Mauna Kea summit, Hawaii | 4,207 m | 1,554 m | 4,032 m |
The finer grid does better on mountains. It gets Denver to within 5 m, Zermatt to within 132 m where the coarse one was 785 m out, and Mauna Kea to within 175 m.
Sea cliffs defeat both. With the finer grid, two of the six still come back
as sea level and two more are under 40 m. A cliff is a step in the terrain,
and any grid cell that straddles the step averages land with water. That is
why our API lets you pass elevation yourself.
If you need to look your height up, there are finer sources. NOAA's ETOPO 2022 has 15 arc-second cells, about 450 m,[5] and the Copernicus DEM goes down to 30 m.[6] In the United States, the Geological Survey will return the height of a single point.[7]
Height above what?
This matters more than the precision of the number.
The correction lowers the horizon by an amount that depends on your height above it. On a sea cliff the horizon is the sea, so your height above sea level is the right figure. That is the case in most of the table.
Denver is not. The city is 1,609 m above sea level, and its horizon to the east is the plain it stands on. Pass 1,609 m and you get a sunrise nearly seven minutes early, computed for a sea that isn't there. The right input for Denver is your height above the surrounding plain, which for most of the city is close to zero. Zermatt is the reverse problem, a valley floor with mountains above it, where the real sunrise is later than any table says.
So ask what you will be looking at when the sun comes up. If it is water or lowland well below you, pass your height above that. If it is ground level with you, pass zero.
Passing your height
curl -H "Authorization: Bearer $THIRDTRAIL_KEY" \
"https://thirdtrail.life/v1/ephemerides/?lat=52.9715&lon=-9.4309&elevation=214&days=1"
The response reports the elevation it applied and where it came from, under
resolved.elevation_m and resolved.elevation_source.
Sources
- Young, A. T. "Dip of the Horizon." San Diego State University. Gives the geometric dip and the commonly used value with refraction, 1.75 arcminutes times the square root of the height in metres. https://aty.sdsu.edu/explain/atmos_refr/dip.html
- National Geophysical Data Center (1988). "5-minute Gridded Global Relief Data (ETOPO5)." NOAA National Centers for Environmental Information. https://doi.org/10.7289/V5D798BF
- NOAA National Geophysical Data Center (2009). "ETOPO1 1 Arc-Minute Global Relief Model." NOAA National Centers for Environmental Information. https://doi.org/10.7289/V5C8276M
- SunriseSunset.io. "Sunrise and Sunset API", parameter `elevation`. https://sunrisesunset.io/api/
- NOAA National Centers for Environmental Information (2022). "ETOPO 2022 15 Arc-Second Global Relief Model." https://doi.org/10.25921/fd45-gt74
- European Space Agency (2022). "Copernicus DEM." https://doi.org/10.5270/ESA-c5d3d65
- US Geological Survey. "Elevation Point Query Service." https://apps.nationalmap.gov/epqs/
Method: sunrise from JPL DE421 with Skyfield at 52.97°N on 20 March 2026,
with the horizon lowered by the geometric dip. ETOPO5 values are our own
lookup. ETOPO1 values are the elevation field returned by sunrisesunset.io
on 15 November 2026. Informational use only, not for navigation or
life-safety decisions.