Northern Lights Tonight is a local go / maybe / no for tonight. It is not a percentage, not a map tracker, and not affiliated with NOAA. This page is how we decide whether a night is worth a look at one place.
A local tonight reading is not a national Kp graphic. Magnetic latitude, live oval reach, darkness, clouds, and skyglow are evaluated for the headline point on the place page; Kp is context, not a nationwide yes or no.
How we decide, in gate order
For each 30-minute slot that is dark enough (sun at or below −12°):
- If aurora data is missing, we do not guess (UNKNOWN).
- If the oval does not reach you, that slot is NO.
- If clouds are socked in, NO.
- Horizon-only plus city glow, or a bright moon on a horizon oval, cannot be GO.
- Only a clear-enough sky with real reach can be GO. Otherwise MAYBE.
Slots that are still twilight are skipped (not dark yet) and do not vote. If the whole night never gets dark enough, the night is NO (summer high latitude). We do not average a weighted score. We do not print 73%.
How to use a forecast card
Read the status, viewing window, and named obstacle together. GO still needs a dark northern view. MAYBE names an obstacle rather than a coin flip. NO means skip a special trip. UNKNOWN means current data is missing or stale, so we will not reuse an older color.
Quiet and mid-latitude places will show many NOs. That is the product working: a useful trip filter should reject nights when the oval, darkness, clouds, or skyglow do not line up.
If a window is open, give your eyes 30–60 minutes rather than testing the sky for five minutes from a porch. A phone camera may show a faint green smear before your eyes do, but that is a reason to keep looking, not a second forecast.
Near window vs the rest of the night
The next ~90 minutes use NOAA SWPC OVATION grids (at the point and north along the horizon). Later tonight uses the 3-hour Kp forecast plus, on pages that have a dossier, that place’s typical Kp bands. Mid-latitude and sub-oval pages cannot GO on that far window. A /view pin with no dossier cannot GO on the far window either.
The next hour is therefore stricter about the live oval, and more conservative about “later tonight.” A MAYBE in the far window is not a promise the oval will arrive.
State pages use one headline point
A state is not the max of its cities. Colorado’s headline is the Fort Collins area, not Denver. Oregon’s headline is Baker City, not Portland. Other points can be listed; they do not get their own indexed URL.
A state night is “headline point, then contrast,” not “any city in the state said maybe.” If you need a ZIP, use near me rather than assuming the headline speaks for you.
What UNKNOWN is for
UNKNOWN is a product decision. Missing or stale aurora data is not last hour’s GO with a disclaimer. Refusing to guess is part of the method. If every Wave 1 row is UNKNOWN, the homepage card says readings are paused.
Sources
- NOAA Space Weather Prediction Center public products (OVATION and Kp).
- Cloud-cover data adapted from Open-Meteo under CC BY 4.0 (low / mid / high).
- Local sun and moon from the coordinates in the place dossier.
The browser never calls NOAA. Pages read snapshot files. Stale snapshots display UNKNOWN, not last hour’s GO.
The method is open on this page so a GO is auditable: darkness, reach, clouds, then glow and moon. It is not a neural net, and it is not a vibe.
Related: Tonight · How to see · Best time
Confidence on the card
Confidence is high / medium / low on the card. It is not a percent. Conflicting short-term and overnight signals, missing clouds, or a far-window Kp call all push confidence down. Low confidence still never becomes a fake GO.
Moon and glow
Moon is a blocker when it washes a horizon-only oval. It is not a nationwide kill switch. City glow is a blocker when the oval is weak and low. An overhead storm can punch through modest skyglow. The card names the issue instead of mixing moon, glow, and clouds into one mystery score.
What we refuse to build
We do not average slots into 73%. We do not let the sunlit hours vote. We do not let a /view pin without a dossier GO on the far window. We do not take the max of every city in a state. The method is intentionally conservative so a GO means the gates passed, not that a map looked pretty.
If this page and a social post disagree, this page is the product. The social post is not a data source. How we decide stays on this URL so the gates can be checked.
Worked example of the gates
Suppose OVATION says the oval reaches the headline point, clouds are mixed, the sun is below −12°, and the moon is down. That can be MAYBE (clouds) or GO (clear-enough). If the same oval is horizon-only and the city is a light dome, it cannot be GO even if Kp looks large on Twitter.
Suppose OVATION is missing. The night is UNKNOWN, not last night’s leftover GO. Suppose the sun never reaches −12°. The night is NO for darkness, even if the oval is wild.
These examples are the method, not extra data sources.
What the far window is for
The far window exists so you can see whether later hours are even in play. It is not a second GO button. Mid-latitude pages cannot GO there. Pins without a dossier cannot GO there. If you only have time in the far window, read that as “maybe later, not a promise.”
Related
Tonight · How to see · Colorado example · Fairbanks example
Why the gates are ordered that way
Darkness first because a bright sky cannot show a night-sky event. Reach second because clouds over an oval that never arrives are a wasted wait. Clouds third because a clear sky with no oval is still a no. Glow and moon last because they modulate a weak horizon oval more than they cancel an overhead storm.
How we decide follows that order every night so the main issue on the card matches the first gate that failed, not a blended vibe.
Slots and voting
Each 30-minute slot that is dark enough can fail or pass. Twilight slots do not vote. The night label is not an average of slots. One GO slot can make the night worth trying if clouds and reach agree. A night of NO slots stays NO.
How we decide also refuses to let the far window overrule a near-window UNKNOWN. Missing live oval data is not “use Kp instead and hope.”
How we decide GO MAYBE nights still names the obstacle. How we decide the night label is the first failed gate, not a blend. How we decide UNKNOWN nights is missing or stale aurora data, not a coin flip. Keep those three sentences next to the card if a social post disagrees.
Far-window Kp is context. Near-window OVATION is the live oval. Mid-latitude pages never use the far window as a GO. Summer high latitude can be a darkness NO even when the oval is active.
Confidence on the card stays high, medium, or low. It is not a percent and it is not a second status. Low confidence plus a MAYBE still means an obstacle was named. Missing cloud data can drop confidence without turning a reach-NO into a maybe.
Sources stay NOAA SWPC for the oval and Kp, Open-Meteo for clouds, and the dossier coordinates for sun and moon. The browser does not call those APIs. Stale files display UNKNOWN. That is the whole stack.
Gates decide how we label nights. Clouds decide how we label nights. Darkness decide how we label nights. Same five words, three gates, so the card matches the first failure.