Two things decide this and neither is the one people expect. The auroral oval sits at a geomagnetic latitude, not a geographic one, so Yellowknife at 62.5 °N beats Longyearbyen at 78.2 °N. And the equinoxes beat midwinter: December has the most dark hours and the worst odds per hour.
Right now · September
66° geomagnetic with 52% median winter cloud. One of the better bets on this list.
In the rain shadow of the Norwegian mountains: the clearest winter skies in Scandinavia, which matters more than latitude.
ABISKO · SMHI · 1991–2020 · Checked 2026-08-12
65.5° geomagnetic with 68% median winter cloud. One of the better bets on this list.
Two degrees further north than Rovaniemi, and that buys both a real polar night and a much better aurora latitude.
INARI IVALO LENTOASEMA · FMI · 1991–2020 · Checked 2026-08-12
Dark enough at 63.8° geomagnetic, but 72% median cloud cover is what will decide your trip.
On the Arctic Circle: thirty days of midnight sun in summer, deep reliable snow in winter, and no true polar night in between.
The sky here is overcast more often than not in winter. Latitude gets you the aurora; cloud is what stops you seeing it.
ROVANIEMI APUKKA · FMI · 1991–2020 · Checked 2026-08-12
Dark enough at 67° geomagnetic, but 75% median cloud cover is what will decide your trip.
The most accessible aurora destination there is, and one of the cloudiest — the Gulf Stream keeps it mild and grey.
The sky here is overcast more often than not in winter. Latitude gets you the aurora; cloud is what stops you seeing it.
TROMSO · MET Norway · 1991–2020 · Checked 2026-08-12
Dark enough at 64.5° geomagnetic, but 80% median cloud cover is what will decide your trip.
A world-class aurora and landscape destination, and a poor snow one — January averages 3 °C and the city often has none.
The sky here is overcast more often than not in winter. Latitude gets you the aurora; cloud is what stops you seeing it.
REYKJAVIK · Icelandic Met Office · 1991–2020 · Checked 2026-08-12
75° geomagnetic is often poleward of the oval — fewer auroras than places further south on a quiet night.
So far north it is often poleward of the oval. Its real distinction is aurora in the middle of the day, during the polar night.
Far enough north to sit inside the polar cap on quiet nights, which means fewer auroras than places further south. Good on active nights, worse on average.
SVALBARD AIRPORT · MET Norway · 1991–2020 · Checked 2026-08-12
We show at most two places per country so one country cannot fill the list.
A ranking that will not show its criteria is an opinion. These are ours.
Warm days mean a mean daily maximum of 25–33 °C, scoring nothing below 20 °C or above 36 °C. The upper bound is not a formality: Dubai in July is 41 °C, and a list that treats hotter as better puts it first.
Swimmable means a sea of at least 22 °C. There is no meteorological standard for this — it is the convention northern-European travel publishers use, and we are telling you so rather than implying otherwise.
The criteria multiply rather than add, so a perfect air temperature cannot compensate for a cold sea. That is the whole reason the warm list and the swimming list are two separate lists.
Snow reliability is classified against the 100-day rule (Abegg 1996, after Föhn 1990): snow-reliable means at least 30 cm of cover on at least 100 days between 1 December and 15 April, in seven winters out of ten, with the line at about 1,200 m in the Swiss Alps. We classify against that rule and cite it; we do not compute it, because no free dataset carries per-resort snow depth.
Aurora is ranked on corrected geomagnetic latitude, not geographic latitude, and on median winter cloud. The oval sits at roughly 65–70° geomagnetic and moves about 2° equatorward per unit of Kp.
Tropical-cyclone seasons are applied separately from rainfall. A peak month is excluded outright; a shoulder month is capped no matter how dry it is.
Temperature and rainfall are 1991–2020 monthly normals computed from NOAA's Global Historical Climatology Network station records — the station used for each destination is named on its card, along with how far away it is. Sea-surface temperature comes from NOAA OISST v2.1 over the same period. Rain days count days with at least 2.5 mm.
These are thirty-year averages, not a forecast. They tell you what a month is usually like, never what next Tuesday will be.
Almost always the sea. The Canaries in January are 21 °C in the air and 19 °C in the water — a genuine answer to "somewhere warm" and a bad answer to "somewhere I can swim". Rather than pick one meaning and be wrong for half of you, we run two lists.
Because for about half of these places the town temperature is the wrong number. Chamonix's valley sits at 1,035 m and is often green in January while its lifts reach 3,842 m. The elevation pair tells you what you actually need to know; a valley temperature would just mislead you confidently.
December has the most hours of darkness and the worst odds per hour. Geomagnetic activity peaks around the equinoxes — the Russell–McPherron effect — so March and September–October are roughly twice as likely to deliver. If you can only travel in December you will still see them; you will just need more nights.
No. Every number here is a 1991–2020 average from a named weather station. They describe what a month is usually like. For the next two weeks, use a forecast.
Usually because there is no weather station near it with a long enough record, and we would rather leave a place out than publish a number measured 400 km away. Palawan is out for exactly that reason.
Otherwise January's warm list is eight Spanish islands. The cap is applied after ranking, so it never changes which place is better — only how many of one country's you see at once.
Take the destination and the month straight into the packing calculator — it will use the real forecast when you get close enough for one.