How Accurate Are Sunrise and Sunset Times — and Why Do Sources Disagree by a Minute?

Look up tomorrow’s sunrise on two different websites and there is a decent chance the times will not match. One says 6:12 AM, the other says 6:13 AM. Your phone’s weather app may offer a third answer. None of them are broken. The disagreement comes from something more interesting: sunrise and sunset are not perfectly knowable events. They are predictions built on a standard set of assumptions, and the real atmosphere does not always cooperate.

This article explains what a published sunrise time actually claims, where the genuine uncertainty comes from, and why to-the-second precision is an illusion.

What “Sunrise” Formally Means

The first surprise for most people is that sunrise has an official geometric definition, and it is more specific than “the sun comes up.”

Sunrise is the moment the upper limb of the sun — its top edge — appears to touch the horizon. Not the center of the sun’s disk, and not the moment the full disk clears the horizon. The top edge, just kissing the horizon line. Sunset is the mirror image: the moment the upper limb disappears below the horizon.

Two corrections are baked into that definition:

• The sun’s semi-diameter. The sun is not a point of light; its disk spans about 32 arcminutes, so the distance from the center to the edge is roughly 16 arcminutes. Because we time the upper edge rather than the center, sunrise happens when the sun’s center is still 16 arcminutes below the horizon.
• Atmospheric refraction. Air bends light. Near the horizon, sunlight passes through a long, dense slice of atmosphere, and that bending lifts the sun’s apparent position by about 34 arcminutes — slightly more than the sun’s entire apparent diameter. When you watch a sunrise, the geometric sun is actually still fully below the horizon. You are seeing an image bent over the edge of the Earth.

Add those together and you get the standard convention used by almost every almanac, observatory, and calculator: sunrise and sunset occur when the sun’s center is 50 arcminutes below the ideal horizon — equivalently, at a zenith angle of 90°50′ (90 degrees 50 arcminutes from straight overhead).

That 90°50′ figure is the shared foundation of essentially all published sun times. It is also, quietly, an approximation.

The Refraction Problem: Why ±1–2 Minutes Is the Honest Answer

The 34-arcminute refraction value is a standard-atmosphere assumption. It corresponds to typical temperature and pressure at sea level. Real air is never exactly standard.

Refraction near the horizon depends on the temperature and density structure of the atmosphere along the entire light path — hundreds of kilometers of air between you and the horizon. Cold, dense air bends light more; warm, thin air bends it less. Temperature inversions, where warm air sits over cold ground or water, can distort refraction dramatically. That is the same physics behind mirages and the flattened, layered look of a sun sitting on the ocean.

The practical consequence:

• On an ordinary day, actual refraction differs enough from the standard value to shift the visible sunrise or sunset by several seconds to a minute or so.
• In unusual conditions — strong inversions, extreme cold, hot ground — the shift can reach a few minutes. In polar regions, anomalous refraction has famously made the sun visible days before its calculated return.

This means every published sunrise and sunset time, from every source, carries an irreducible real-world uncertainty of roughly ±1 to 2 minutes. The calculation of where the sun’s center is geometrically can be done to extraordinary precision. What cannot be predicted in advance is exactly how tomorrow’s atmosphere will bend the light. The U.S. Naval Observatory makes the same point about its own data: near the horizon, refraction is variable enough that quoting rise and set times to better than the nearest minute is not physically meaningful.

Elevation: Your Horizon Is Not the Textbook Horizon

The standard calculation assumes you are at sea level looking at an unobstructed sea-level horizon. Change either assumption and the times move.

Standing higher lets you see further around the curve

If you are on a hill, a rooftop, or an aircraft, your horizon dips below the ideal horizontal plane, and you see the sun slightly earlier at sunrise and slightly later at sunset. The effect is modest but real: from a 300-meter vantage point with a clear low horizon, sunrise arrives roughly two to three minutes earlier than the sea-level calculation suggests. Watch a sunset from a beach, then sprint up a tall dune, and you can genuinely see the sun set twice.

High-altitude cities are an interesting case. A place like Denver sits about 1.6 kilometers up — you can see its daily times on our Denver page — but elevation only produces earlier sunrises when the terrain around you is lower than you are. Denver’s problem runs the other way, which brings us to the bigger effect.

Mountains delay what you actually see

Published times assume a flat, zero-elevation horizon. If a mountain range stands between you and the rising sun, the visible sunrise — first direct sunlight reaching you — can lag the calculated time by many minutes, sometimes more than an hour in deep valleys. In Denver, the Rockies sit to the west, so calculated sunset routinely beats the moment the sun actually drops behind the peaks.

No general-purpose calculator models your local skyline. Published times answer the question “when would the sun cross an ideal flat horizon here?” — the only version of the question answerable without a 3D terrain model of your exact viewpoint.

One City, One Coordinate — But a City Is Not a Point

Every sunrise calculation needs a latitude and longitude, and a “city” is not a coordinate. It is a region.

The sun’s terminator — the line between day and night — sweeps westward across the Earth’s surface. At mid-latitudes it moves fast, but not instantly: across a large metro area, sunrise on the eastern edge can lead sunrise on the western edge by two to four minutes. Los Angeles County spans enough longitude that the “correct” sunrise time differs by several minutes depending on which neighborhood you mean.

So when two sources disagree by a minute, sometimes neither algorithm is at fault — they simply anchored the city at different points. One used the city hall, another used the geographic centroid, a third used an airport. All defensible, all slightly different.

If you want the most honest number for where you actually stand, use the coordinate closest to you rather than the biggest nearby city label.

Why Different Websites and Apps Disagree

Put the pieces together and the one-minute discrepancies stop being mysterious. Reputable sources differ because of some combination of:

• Different reference coordinates for the same named place — often the largest single factor.
• Different rounding conventions. One source truncates 6:12:40 to 6:12; another rounds it to 6:13. That alone creates a one-minute “disagreement” from an identical underlying calculation.
• Different algorithms. Some use full high-precision ephemerides (the approach behind the Naval Observatory and NOAA tools); others use compact astronomical approximations like those in Jean Meeus’s Astronomical Algorithms. The compact methods are typically accurate to well within a minute — smaller than the atmospheric uncertainty — but not identical to the second.
• Different refraction models. Almost everyone uses 34 arcminutes, but some tools apply slight temperature or pressure adjustments, or use marginally different constants.
• Elevation handling. Most calculators ignore observer altitude; a few include it.

Notice what is not on this list: error. Two times a minute apart are usually both correct within the precision the physics allows.

The Hard Cases: High Latitudes and Grazing Events

There is one regime where disagreements grow from one minute to many: near the poles, and anywhere the sun’s path crosses the horizon at a shallow angle.

At mid-latitudes, the sun rises steeply, climbing its own diameter in a couple of minutes — so a small angular uncertainty translates to a small time uncertainty. Near the Arctic or Antarctic circles around the solstices, the sun grazes the horizon at an extremely shallow angle. It can spend twenty or thirty minutes crawling through the last half degree. In that geometry:

• A tiny difference in refraction assumptions shifts the computed time by many minutes.
• Whether the sun rises at all on a given day can flip based on the refraction model. Different sources can legitimately disagree about the first sunrise after polar night by a day or more.
• The same shallow-angle sensitivity affects twilight times at high latitudes in summer, when the sun skims just below the horizon all night.

If you have ever compared sun times for Fairbanks, Tromsø, or Utqiaġvik across two sources and seen large gaps, this is why. The event itself is genuinely ill-defined at the minute level there.

What Precision Is Actually Achievable?

A useful way to summarize the error budget:

• Astronomical position of the sun: computable to fractions of a second of time. Effectively exact for this purpose.
• Standard refraction assumption vs. the real atmosphere: typically tens of seconds, occasionally minutes. This is the floor.
• Coordinate choice within a city: commonly 1–3 minutes across a metro area.
• Local horizon (hills, buildings, mountains): anywhere from zero to over an hour for the visible event.
• Rounding to the minute: up to 30 seconds.

The atmospheric term alone means that quoting sunrise as “6:12:47 AM” is false precision. The seconds digit describes the geometry of an idealized atmosphere, not anything you could verify with a stopwatch and your own eyes. Rounding to the nearest minute is not laziness; it is the correct significant-figure discipline for a quantity with roughly ±1–2 minutes of physical uncertainty.

That is also why the practical value of a sun-times site is rarely the last digit. Knowing that sunrise moves two minutes later each day this month, or that solar noon, altitude, and azimuth explain what the light will do between those endpoints, is worth far more than a second-level timestamp.

How todaysuntimes.com Computes Its Times

For transparency, here is how the numbers on this site are produced:

• Times are computed with a well-established implementation of standard astronomical algorithms, using the conventional 90°50′ definition described above.
• Calculations run in your browser, client-side, from the representative city-center latitude and longitude on each city page — so what you see is a live computation for that coordinate and today’s date, not a cached table.
• Every page uses a fixed, documented coordinate per city, so the times are internally consistent across the daily view, month tables, and year tables.
• Times are shown to the minute, because that is the precision the underlying physics supports.

You can read more about the site’s approach on the about page.

The Bottom Line

Published sunrise and sunset times are geometric predictions under standard assumptions: upper limb, ideal flat horizon, 34 arcminutes of refraction, one representative coordinate per city. Within those assumptions, modern algorithms agree to within seconds. Outside them — real air, real terrain, real metro areas — the honest uncertainty is a minute or two, and occasionally more.

So when two sources disagree by a minute, the right reaction is not “which one is wrong?” It is “both are right, to the precision that sunrise actually has.” Treat any sun time as accurate to about a minute, expect your visible sunrise to arrive late if there is anything tall to your east, and be suspicious of anyone quoting seconds.

References

• U.S. Naval Observatory: Rise, Set, and Twilight Definitions
• NOAA Solar Calculator
• Jean Meeus, Astronomical Algorithms, 2nd ed., Willmann-Bell, 1998