Aurora Forecasts and Apps: How to Chase the Lights in Norway

Here’s the thing that will save you the most disappointment on an aurora forecast in Norway, and almost nothing online says it plainly: if you’re in northern Norway, the space weather forecast is not your problem. The cloud forecast is.

Yr — the Norwegian Meteorological Institute’s public service — publishes a table of the Kp index needed to see aurora at each latitude. For the band from 66° to 74°N, which covers Tromsø, Alta, Bodø, Lofoten and Kirkenes, the figure is Kp 0. Zero. The aurora is essentially always overhead. Every hour you spend refreshing a Kp app in Tromsø is an hour spent watching a number that is almost never the constraint.

This page explains what the forecasts actually measure, why the most-quoted number is a three-hour measurement rather than a prediction, which free Norwegian tools beat the paid apps, and how to build an evening around all of it.

The Kp index: what it is, and its one big flaw

Kp is the planetary three-hour-range index. It’s the mean standardised K-index from 13 geomagnetic observatories sitting between 44° and 60° geomagnetic latitude, on a scale of 0 to 9. The underlying K-index comes from the maximum fluctuation of the horizontal component of Earth’s magnetic field over a three-hour interval. The “K” is from the German Kennziffer, and the whole system was introduced by Julius Bartels in 1938.

Definitive Kp values have been produced since 1997 at the Adolf Schmidt Geomagnetic Observatory in Niemegk, run by GFZ Potsdam. NOAA also publishes a real-time Estimated Kp using just eight magnetometers — Sitka, Meanook, Ottawa, Fredericksburg, Hartland, Wingst, Niemegk and Canberra.

Now the flaw. NOAA’s Kp chart updates every minute, which makes it feel live. But every value still describes a closed or closing three-hour bin. An aurora substorm — the thing you’re actually waiting for, the moment a quiet green arc suddenly breaks into moving curtains — typically lasts 30 to 60 minutes. It can begin and end almost entirely inside one three-hour window and barely move the number.

So Kp is a measurement of what the planet’s magnetic field has been doing, averaged over a period longer than the event you care about. It is useful for answering “is this a stormy week” and close to useless for answering “should I go outside now.”

Note: the criticism above is the settled view among aurora watchers rather than an official NOAA position; NOAA documents the three-hour construction openly, but doesn’t editorialise about it.

Yr’s latitude table — the single most useful thing on this page

The Norwegian Meteorological Institute publishes the Kp threshold needed at each latitude band. It’s buried in a help-centre article and I’ve never seen a travel site reproduce it.

Read the top row again. Everywhere from Bodø north needs no geomagnetic activity at all. That reframes the entire planning exercise: your trip is not a bet on the sun. It’s a bet on the weather. See where to see the northern lights in Norway for how the bases compare on exactly that.

Yr also bands its own aurora index: Kp 1–2 above threshold with a nowcast of 0–30% is “low”; Kp 3–5 or 30–60% is “medium”; 6–7 or 60–80% is “high”; 8–9 or above 80% is “very high.”

OVATION: the actual short-term forecast, and its hidden failure mode

The thing most apps display as “the aurora forecast” is OVATION — Oval Variation, Assessment, Tracking, Intensity, Online Nowcasting — a model developed by Patrick Newell at the Johns Hopkins Applied Physics Laboratory and published in 2009. NOAA runs the current version, OVATION 2020. A 2013 upgrade added data from the TIMED GUVI instrument because the earlier DMSP-only version was only reliable up to Kp 7.

OVATION gives you a genuine 30 to 90 minute forecast, and the reason it can is elegant: satellites at the L1 Lagrange point, 1.6 million kilometres upstream of Earth, measure the solar wind before it arrives. DSCOVR is the current workhorse. Depending on solar wind speed, that buys roughly 15 to 60 minutes of warning.

Here is the part nobody tells you. When the L1 solar wind data are contaminated or unavailable, OVATION silently falls back to being driven by the current Kp index instead — and in that mode it has zero lead time. The map still looks like a forecast. It’s showing you the recent past. There’s no prominent warning on the consumer apps that repackage it.

Practical implication: treat a suddenly static or oddly Kp-shaped OVATION map with suspicion, and never build an evening entirely on one app’s oval graphic.

The solar wind numbers, and how confident to be about them

Below OVATION sit the raw measurements: solar wind speed, density, and the magnetic field’s Bz component. The physics is real and well established — when Bz points south (negative), the interplanetary magnetic field couples efficiently with Earth’s northward-pointing field and energy flows in. Northward Bz largely shuts the door.

You will see specific thresholds quoted everywhere: Bz below −5 nT is interesting, below −10 nT is good, below −20 nT is a serious storm; solar wind at 400 km/s is baseline and 500–700 km/s is elevated. Those numbers are enthusiast-community consensus rather than an official NOAA or MET Norway publication, and I couldn’t find them stated as thresholds by either agency. They’re a reasonable rule of thumb. They are not gospel, and I’d rather tell you that than pass them off as settled.

NOAA’s G-scale, by contrast, is official and precisely defined:

SWPC issues Watches at predicted Kp 5, 6, 7 and 8-or-above, Warnings at 4, 5, 6 and 7+, and Alerts across Kp 4 to 9. One more useful fact from NOAA: a bright aurora can be seen from up to 1,000 km away. It doesn’t need to be overhead — which is why people in Denmark photograph displays physically sitting over the Arctic, and why looking north on any clear night is worth doing.

Norway has its own space weather centre, and it’s free

This gets almost no coverage outside Norway and it’s better than most of what you’d pay for.

The Norwegian Centre for Space Weather (NOSWE) is run by the Tromsø Geophysical Observatory at UiT, and is the official ISES Regional Warning Centre for Norway. It publishes nowcast, one-hour and four-hour aurora forecasts for Svalbard, Tromsø, Trondheim, Bergen and Oslo, using the SvalTrackII model developed by Professor Fred Sigernes at UNIS and the Kjell Henriksen Observatory. There’s a free app for iOS, Android and Windows.

Two more free tools from the same institution that give you ground truth rather than a model:

• The Skibotn all-sky camera, which updates every two minutes at night (it’s off between 05:00 and 16:00 UT). Skibotn is the dry valley behind the Lyngen Alps where the chase tours go. If the camera shows aurora, there is aurora — no modelling involved.
• The real-time magnetometer chain at flux.phys.uit.no, with stations at Ny-Ålesund, Hopen, Nordkapp, Tromsø, Andøya, Dønna, Dombås and Karmøy, computing K indices automatically in real time. A sharp deflection on the Tromsø trace means something is happening right now, hours before it shows up in a three-hourly planetary index.

If you learn to glance at the Tromsø magnetometer trace and the Skibotn camera, you have better information than any consumer app can give you, from the institution that actually operates the instruments.

What no forecast can tell you

Three variables decide your night, and only one of them is forecast at all.

Light pollution isn’t in any model. A display that’s obvious from a dark beach is invisible from a lit street, and no app knows where you’re standing. This is the largest single factor you control and it costs nothing — twenty minutes of walking in Tromsø, or ten minutes of driving in Alta, changes what you see more than any change in Kp would. In Lofoten it means getting off the E10 and onto a beach.

Micro-weather isn’t in any model either. Norwegian cloud forecasts are computed on a grid, and Arctic Norway’s terrain is fjords, ridges and valleys at a much finer scale than the grid. A valley can be clear while the coast five kilometres away is fogged in. Skibotn is the famous case — around 300 mm of annual precipitation against a national average near 1,400, because the Lyngen Alps wring the air out before it arrives. No forecast resolution captures that reliably; local knowledge does, which is a large part of what you’re paying a chase guide for.

Your own patience isn’t modelled at all. The most common way people miss the aurora is going outside, seeing nothing, and going back in after ten minutes. Dark adaptation takes 30 to 45 minutes and a faint arc frequently builds into a substorm over the following hour. The forecast said 40%; the person who stayed out saw it and the person who left didn’t. Same forecast, different outcome.

A note on what the numbers can’t promise

It’s worth being clear-eyed about the limits. Even with perfect information — clear sky, high activity, dark site, dark-adapted eyes — the aurora is a natural phenomenon that does what it does. Some nights it sits as a dim green band on the northern horizon for four hours and never develops. That is a normal outcome, not a failure of planning.

What good forecast-reading buys you is the elimination of avoidable failures: standing under cloud that a two-hour drive would have escaped, staying in on a night that turned out to be the best of the week, or leaving after ten minutes. Those are the losses worth preventing, and they’re most of them.

The rest is what makes it worth going. For the wider context, see our main northern lights guide, what each month is actually like, and — if you plan to shoot it — how to photograph the aurora, where a camera doubles as a detector for displays your eyes can’t yet resolve.

Yr does have an aurora forecast — and most people miss it

Norwegians use Yr for weather, and a large share of visitors download three aurora apps without realising Yr already has one built in.

It’s under the “Nearby” tab. It gives a nowcast from 0 to 100% looking 30 to 90 minutes ahead, plus a three-day outlook, and it will send a push alert when the nowcast exceeds 30%. It’s free, it’s from the Norwegian Meteorological Institute, and — crucially — it sits inside the same app as the cloud forecast you actually need.

Yr also makes a point most apps skip: high thin cloud is often transparent to aurora, while low thick cloud is the real blocker. A forecast showing 60% cloud cover can mean a perfectly usable night if that cloud is cirrus at 8,000 metres, or a dead one if it’s stratus at 500. Look at the cloud type and altitude, not just the percentage. This is the single most valuable interpretive skill in aurora chasing and it takes about ten minutes to learn.

The apps, and what they cost in 2026

My honest recommendation: Yr plus the NOSWE app, both free, plus a browser bookmark for the Skibotn all-sky camera. That combination beats any paid subscription for the specific job of deciding whether to leave the hotel tonight in Norway. SpaceWeatherLive is genuinely good if you enjoy the underlying physics; it will not tell you anything actionable that Yr won’t.

One more free resource that’s more useful than it sounds: the Facebook group Tromsø Northern Lights Q&A, where people in the city post live sightings. On a marginal night, someone standing at Telegrafbukta saying “it’s started” is worth more than any model. Our self-guided Tromsø guide covers where those people are standing.

The 27-day trick, and why it matters more in 2026

NOAA publishes a 27-day outlook — F10.7 solar flux, planetary A index and largest daily K — issued every Monday by 15:00 UTC, plus a 45-day Ap and F10.7 forecast. Almost nobody planning a trip uses it, and right now it’s unusually valuable.

The logic rests on the Sun rotating roughly every 27 to 28 days. A coronal hole — a long-lived gap in the corona that fires high-speed solar wind at Earth — comes back around into a geoeffective position on that same cycle. Aurora researchers describe them as “solar wind lighthouses.”

Why this matters now: Solar Cycle 25 peaked in October 2024 (sunspot number 216 in August 2024, down to 78.1 by July 2026). In the declining phase, coronal holes replace coronal mass ejections as the dominant driver. That makes the aurora more recurrent and therefore more plannable than it was at maximum.

How to use it: if there was a good geomagnetic run four weeks ago, check NOAA’s 27-day outlook for a repeat around the same date. It’s not a guarantee and it won’t tell you about the weather, but for anyone with flexibility on dates it’s the closest thing to a long-range aurora forecast that exists. More on what the solar cycle means for the next few winters in our guide to the best time to see the northern lights.

What I’d ignore

Ignore the Kp index if you’re north of Bodø. You’re at Kp 0 threshold. It is not the variable. This is the most important sentence on this page and the one most likely to change your behaviour.

Ignore three-day aurora predictions. Solar wind measurements come from L1, which is 30 to 90 minutes upstream. Anything beyond that is either the 27-day recurrence statistic or a guess. A confident “aurora forecast for next Tuesday” is not a forecast.

Ignore paid aurora apps. Yr and NOSWE are free, official and better sourced. The paid tiers mostly sell alerts, and an alert that fires when it’s cloudy is worth nothing.

Ignore cloud cover percentages in isolation. Sixty per cent high cirrus is a usable night. Thirty per cent low stratus sitting exactly over your valley is not. Type and altitude beat percentage.

Ignore “aurora guaranteed” claims built on forecast data. Operators quoting 92–99% success rates are self-reporting with no published methodology. The 79% December cloud figure for Tromsø comes from meteorological records; the success rates don’t. Our tour buyer’s guide unpacks what those promises actually oblige.

Ignore the moon phase. It has no effect on the aurora and only raises sky background. Photographers often prefer moonlight for the foreground.

How to actually run an evening

1. Afternoon: open Yr. Check the cloud forecast for your base and two or three places within two hours’ drive. They diverge constantly — that divergence is the whole opportunity.
2. Look at cloud type and altitude, not just the percentage.
3. Around 18:00: decide how far you’re travelling. If everything within 200 km is under thick low cloud, accept the night and go to dinner. Chasing hopeless cloud is how people spend four hours in a layby for nothing.
4. Check the Skibotn all-sky camera and the Tromsø magnetometer trace before you commit.
5. Get in position by 19:00–20:00. Activity peaks 20:00–02:00 around magnetic midnight, but displays start earlier than the folklore suggests.
6. Give your eyes 30–45 minutes. Rods need that long for roughly 80% dark adaptation, and a phone screen undoes much of it. Red torch mode only.
7. Use your phone camera as a detector. A three-second exposure sees a faint arc before your eyes do.

Frequently asked questions

What Kp index do you need to see the northern lights in Tromsø?

Kp 0. Yr’s published thresholds put the entire 66–74°N band — Tromsø, Alta, Bodø, Lofoten, Kirkenes — at Kp 0, meaning geomagnetic activity is essentially never the limiting factor there. Cloud is.

Is the Kp index a forecast or a measurement?

A measurement. Kp is the planetary three-hour-range index, averaged from magnetometers between 44° and 60° geomagnetic latitude, and every value describes a three-hour window that has closed or is closing. Since an aurora substorm lasts 30–60 minutes, a strong display can come and go without moving the number much.

How accurate is the aurora forecast?

Reliable at 30–90 minutes, because that’s the solar wind’s travel time from the L1 satellites 1.6 million km upstream. Beyond that it degrades quickly. NOAA’s 27-day outlook is statistical, based on the Sun’s rotation bringing coronal holes back around, rather than a true forecast.

Does Yr have a northern lights forecast?

Yes, and most visitors miss it. It’s under the “Nearby” tab: a 0–100% nowcast looking 30–90 minutes ahead plus a three-day outlook, with a push alert above 30%. It’s free, from the Norwegian Meteorological Institute, and sits alongside the cloud forecast you actually need.

What does Bz mean in the aurora forecast?

Bz is the north–south component of the interplanetary magnetic field. When it points south (negative), it couples efficiently with Earth’s field and energy flows into the magnetosphere. The commonly quoted thresholds — −5, −10, −20 nT — are community rules of thumb rather than published agency standards.

What is the best free aurora app for Norway?

Yr, followed by the NOSWE app from the Tromsø Geophysical Observatory. Both are free and official. Add a bookmark for the Skibotn all-sky camera, which shows you actual sky rather than a model.

Can you see the aurora through clouds?

Through high thin cloud, often yes — Yr states plainly that cirrus is frequently transparent to aurora. Through low thick cloud, no. This is why cloud type matters more than cloud percentage.

How far in advance can the northern lights be predicted?

Reliably, 30–90 minutes. Loosely, 27 days, using the recurrence of coronal holes as the Sun rotates — which is unusually useful in 2026 because the declining phase of Solar Cycle 25 makes coronal holes the dominant driver.

Photo credits

• A quiet arc like this needs a long exposure; when it starts moving, everything changes — Photo: Georg Botz / CC BY-SA 3.0 via Wikimedia Commons.
• Forecasts are reliable at 30 to 90 minutes — Photo: Georg Botz / CC BY-SA 3.0 via Wikimedia Commons.

Checked in August 2026 against NOAA SWPC, GFZ Potsdam, the Tromsø Geophysical Observatory and Yr’s own documentation. Bz and solar wind speed thresholds circulating online are community conventions rather than official agency standards and are flagged as such above. App pricing reflects US listings and may vary by region.