How it works, and what it can't tell you

The question

Many trips have fixed dates: the month is set by work, school or someone else's calendar. The question then is "given this month, what am I walking into, and where should I go?" Weather, visitor season and cost appear separately, with an Overall score for comparing their combined trade-off.

1. Weather

The Holiday Climate Index (HCI) rates a day from 0 to 100 for leisure tourism using four facets: thermal comfort, cloud cover, precipitation and wind. Its rating tables and weights were derived from surveys of tourists' stated preferences, and it is published in peer-reviewed literature: Scott, Rutty, Amelung & Tang (2016), Atmosphere 7:80 for HCI:Urban; Rutty et al. (2020) for HCI:Beach. We use the rating tables as compiled in Demiroglu et al. (2020), Atmosphere 11:911, Tables 1–2, with three changes, described below, so our score is adapted from the HCI rather than the HCI itself.

Our first change to the published index. The published HCI computes the humidex from the day's maximum temperature and its mean relative humidity. Mean humidity is high because it includes the cool, damp night; applied to the afternoon maximum it implies far more moisture than the air holds, so warm days read as muggy even where summers are dry (Lisbon in July: a humidex of 39 where the air supports 34). We use the day's actual vapour pressure instead: the station's mean humidity at the day's mean temperature, or ERA5's mean dewpoint where no station measures humidity. Moisture in the air barely changes through a day, so this pairs correctly with the afternoon maximum. The rating tables and weights are unchanged.

We made the change only after testing it against a benchmark fixed in advance: 37 destinations across eight climate groups, each with the months travel guides commonly recommend for weather and the months they advise against, half used to choose a change and half held back to check it. Agreement on the held-back half rose from 0.67 to 0.78 (1.0 is full agreement at both ends of the year). The biggest gains were Mediterranean cities, which the published version ranked best in winter, and northern cities, which it ranked best in spring rather than summer. The cost: hot interiors such as Madrid now rate July higher than guides do, because even with the correction a dry 36 °C afternoon loses only part of its thermal score while sun and dry air score full marks. Two other changes we tested, a daylight penalty and a harsher rule for wet days, were not adopted: the first bought little on the held-back half at a large cost to hot interiors, the second added almost nothing.

Our second change: humidity at half weight. Even with the right moisture, the humidex marks a humid coastal afternoon of 29–33 °C as harshly as Seville's dry 36 °C, so Athens, Rome and Barcelona rated July and August below January, and beaches rated their high season below May and October. Travel guides say otherwise, so we count the humidity part of the humidex at half weight. We tested fourteen versions of the heat rating, all written down before any result: less weight on humidity, the hot end of the table moved up by 1–4 °C, and an apparent temperature that credits sea breezes. We chose on the 37-destination benchmark, taking among those within 0.02 of the best the one that moved the fewest places' best month; dropping humidity altogether scored slightly higher but is physically wrong and wrecked the hot interiors. Then we checked the choice once on 15 new destinations, signed off before any scoring (Athens, Lake Como, Crete, the Cyclades, Mallorca, Sicily, Malta, Cyprus, Toledo, Évora and five northern controls): agreement rose from 0.84 to 0.88, and on the original 37 from 0.82 to 0.86 (held-back half 0.79 to 0.84). The cost: Sicily and Malta now rate July and August among their best months where guides prefer June and September, and hot interiors such as Seville still rate high summer better than guides do.

Each facet is rated 0–10 from the published bands, weighted and summed. Every page uses one set of words, banded on the weather score out of 10 as shown: 9 or more excellent, 7 to 8.9 very good, 5 to 6.9 good, 3 to 4.9 fair, below 3 poor (roughly 80, 70, 60 and 50 on the index; the displayed /10 is rounded first). Beach regions get HCI:Beach, city and mountain regions get HCI:Urban. The index is a warm-season instrument; it does not know about snow, so mountain pages say plainly in winter that the score measures sightseeing and walking comfort, not ski conditions.

Data. We compute the index for every day from 2011 to 2025 on ERA5, the Copernicus Climate Change Service's reanalysis (0.25°, about 25 km), corrected with open weather-station records from eleven national weather services: Germany, France, Spain, the Netherlands, Ireland, Austria, Switzerland, Norway, Sweden, Finland and Denmark (about 4,700 temperature stations and 6,800 rain stations). ERA5 on its own reads afternoons too cool, most of all on coasts and islands, where a 25 km grid cell is partly sea; it counts too many drizzly days; and its download has no night-time minimum. So three corrections, each learned from the stations and applied by calendar month: afternoon highs get an offset, a share that grows with the sea in the grid cell (more strongly towards the south, where a mostly-sea cell under a summer sun reads island afternoons too cool), and a height term; night lows are the daily mean plus an offset, a sea share, a share of ERA5's own day range and a height term; and rain below a cut-off counts as a dry day. The cut-off depends on the month and the terrain: a 25 km grid spreads mountain showers over dry valleys, so it is higher where the ground is rough and where towns sit below ERA5's smoothed ground (Bolzano, Trento, Innsbruck), and lower over flat country (about 1.2–1.8 mm in winter and 1.6–2.6 mm in summer). Learned from the open stations and tested country by country, this cut rain-day errors in mountain areas from 2.3 to 1.4 days a month and removed their excess of 1.5 wet days; against the official averages in Italy, which it never saw, from 2.3 to 1.7 days. The height term compares where people actually live in each grid cell (the European Commission's GHSL population grid on NOAA's ETOPO terrain) with ERA5's smoothed ground: in the Inn valley ERA5's ground sits near 1,600 m, while Innsbruck is at 580 m. Humidity, cloud and wind are ERA5 as published. Daily scores are averaged by month and across years, then onto the statistical region by where its people live: each grid cell counts in proportion to the region's population inside it, so the score describes towns and resorts, not empty uplands.

How good the corrections are. Each correction was tested country by country: learned from the stations of every other country, then checked against the stations of the one left out. Where a station's height is known, afternoon highs come within 1.0 °C of it on that test; applied as the regions use them, coastal highs improve from 2.4 °C of error to 1.5 °C, night lows from 4.7 °C to 1.7 °C, and rain days from 3.3 to 1.9 days a month, with ERA5's excess of 2.6 wet days a month removed. Against the stations inside each region (135 places with open stations), our highs are typically within 0.5 °C and night lows within 0.6 °C. Weather services in much of Europe (Italy, Greece, Portugal, Poland, Belgium, Czechia, Hungary, Romania, the Balkans, the Baltics and others) publish no open station data, so there the corrections are the ones learned elsewhere. We checked them against the official climate averages those services publish, mostly for 1991–2020, at 163 stations in 23 countries (used only as a test, never to fit anything). Those averages are older than our 2011–2025 years, so we first measured how far they sit from our figures where we know ours are right: at 43 stations in Germany, France, the Netherlands, Austria, Switzerland, Spain and Sweden, the 1991–2020 afternoon highs are 1.1 °C cooler than ours (recent warming) and night lows the same. Allowing for that, our model comes within 1.0 °C of the official averages for highs at the stations themselves, with no lean either way outside the Mediterranean and 0.4 °C on the cool side there; summer afternoons are not read too hot. Night lows come within 1.4 °C and lean mild, most on Mediterranean coasts. Compared the way the pages are read (see How each place was checked), the figures shown are typically within 0.5 °C of the local stations for highs and 0.6 °C for night lows where open stations exist, and 0.7 °C and 0.9 °C of the official averages elsewhere; island afternoons, once 2 °C cool in Malta and the Canaries, now come within about 1.4 °C.

Where night lows are not shown. We compare the night lows on our pages with the stations inside each region (see How each place was checked below). Where they miss by more than 2 °C, we show no night low rather than a wrong one: 5 places against local stations (Tyrol, Mallorca & the Balearics, Melilla, Liechtenstein and Swedish Lapland), 9 against official climate averages (Cyprus, Corfu & the Ionian Islands, Meteora, Pelion & the Sporades, Peloponnese, Abruzzo, Aosta Valley, Friuli Venezia Giulia, Marche and North Macedonia). Cold air pooling in Alpine valleys on clear nights and sea-exposed islands are the usual causes. Highs and the weather score are shown everywhere.

Where rain days are not shown. Even with the terrain-aware cut-off above, a few places still count rain days differently from their stations. Where the stations inside a region count rain days more than 30% differently from ours, we show no rain-day figure: 3 places, too many in Rome & Lazio, too few in Canary Islands and Malta. The weather score there still uses our rain data.

How each place was checked. Every place's page says how its weather was checked, under the weather figures, with the detail under "How we know": against the open weather stations inside the region (135 places; January and July), against a national weather service's official climate averages (126 places; every month, after allowing for recent warming), or not at all (2 places: Central Bohemia and Around Budapest, with no station that publishes monthly averages). The figures on our pages describe where a region's people live, so each station is first brought to the height where the region's people live (at the rate our model learned from the stations) and weighted by how many of the region's people live in its 25 km grid cell; one valley station then stands fairly for a region whose towns sit higher (Bolzano for South Tyrol). A station counts only for regions of its own country.

A trade-off we chose. Accurate temperatures first lowered our weather benchmark (0.82 → 0.79 on the held-back half): with a coastal city's real 30 °C July afternoon, the comfort index marked summer down more than travel guides do. We kept the correct temperatures and fixed the heat rating instead (humidity at half weight, above). The same happened again when we corrected island afternoons (September 2026): against official averages the islands' highs improved from 1.3 to 1.1 °C and lost their 0.5 °C cool lean, but warmer, correct summers on Crete, the Cyclades and Mallorca then rate July and August a little lower than guides do (our fresh check set 0.88 → 0.86). We chose correct temperatures again, then tested how heat is rated where it is breezy.

Our third change: a breeze makes heat easier. On a hot afternoon, wind takes the edge off: above the comfort table's best band, each 10 km/h of wind beyond 10 km/h counts the felt heat 1 °C lower, and never below that band. It helps windy islands and coasts (the Meltemi in the Aegean, Atlantic and Black Sea coasts) and windy interiors such as Zaragoza. We tested eight versions (sea breezes of 1–3 °C by the cell's sea share, a wind term, a more heat-tolerant beach table, and combinations), all written down first, chose on the 82 destinations we had already used, and checked once on 14 new ones signed off before scoring (ten hot coasts and four inland controls): agreement rose from 0.93 to 0.95, with no earlier check set lower (the fresh set back to 0.88).

Why not the station grid. Until September 2026 we used E-OBS, the gridded record from Europe's national weather services, and then NOAA's global station summaries. Both are licensed for non-commercial use only (NOAA's for stations outside the US), so neither feeds this site; every weather number here comes from openly licensed data. 263 of 263 regions have a full weather series.

The Copernicus "climate suitability indicators for tourism" dataset publishes the same indices, but from climate-model simulations with a 1970–2005 historical run. It is a climate-change product, not a description of current weather, which is why we compute the index from observations instead.

What's on. About 30 dated events that fill hotels and move prices for a week or two (Oktoberfest, Carnival in Venice, Cologne and Nice, Holy Week and the April Fair in Seville, Las Fallas, San Fermín, tulip season, King's Day, the Cannes and Venice film festivals, the Monaco Grand Prix, the Palio di Siena, Christmas markets, Ferragosto, Midsummer, Orthodox Easter and a few music and arts festivals). Each is checked against its organiser or official tourism board, and dates come from the event's rule (Easter-based, fixed dates, Midsummer) or the published dates for the year. An event whose next dates are not announced yet is left out rather than guessed. Ranges marked "typically" (Christmas markets, Italy's August holiday weeks) vary by town. Events are shown beside the scores, not folded into them: the crowd figures are monthly and these are weeks.

Nonstop flights from the US. From Eurostat's monthly air passenger statistics by airport pair (avia_par), every route between a US airport and an airport in the region, placed by the airport's coordinates (OurAirports). A month counts when those routes carry at least 1,000 passengers, roughly one flight a week each way. The latest complete year Eurostat publishes is 2024, and airlines add and drop routes every year (United's Newark–Palermo began in May 2025), so we only state service the statistics record, with its year, and never claim that a place has no nonstop flights. Most US travellers connect through a hub anyway.

Sea temperature. For the 113 regions with a coastline of their own, the sea figure is the monthly mean of NOAA's Optimum Interpolation Sea Surface Temperature (OISST v2.1, a quarter-degree blend of satellite, ship and buoy measurements; Huang et al. 2021), 2011–2025, averaged over open water within 30 km of the region's coast. It is not part of the weather score. The words follow common swimming guidance: warm from 24 °C, comfortable 21–23, fresh 18–20 (fine once you are in), cold 15–17 (short dips, or a wetsuit), very cold below 15. Sheltered bays and shallow beaches run a degree or two warmer than open water in summer; exposed Atlantic beaches can run colder when upwelling brings deep water up.

The scores

Region-month guides compare the previous, current and following months using their published regional averages. Overnight-demand percentages compare average stays per day, rather than monthly totals or queues. Short weather tradeoffs use the displayed rounded temperatures and rain days; missing measures stay unavailable. Early/late answers use smoothed historical daily weather, not weekly visitor or room-price forecasts. Packing and swimming answers interpret the existing temperature bands; sea temperature remains separate from Overall.

Every month of every place gets an Overall score out of 10 and a rank among that place's twelve months. It combines three parts, each scored out of 10 so that higher is better for you:

• Weather /10: the Holiday Climate Index below, spread over its useful range: 40 or less scores 1, 85 or more scores 10, so 70 ("very good") scores 7. The word beside it follows the displayed, rounded /10 score: excellent from 9.0, very good from 7.0, good from 5.0, fair from 3.0, poor below. Those are approximately 80, 70, 60 and 50 on the underlying index; rounding can shift the boundary slightly.
• Crowds: this month against the place's own busiest month. Peak is 85% or more of the busiest month's visitor nights, busy 60–85%, moderate 35–60%, low below 35%, and off season below 10%: at that point many resort hotels and restaurants are shut (checked against known closure months: the flag catches 92% of them and wrongly marks 4% of open months). It is a proxy from visitor nights, not a list of closures, so towns and cities in the region usually stay open. Inside the Overall the quietest month scores 10 and the busiest 4, except off-season months, which score 5: empty because the resorts are shut is not a benefit. So that a small revision of the data can't swing a month's score, the crowd score blends gradually from 5 to the normal score between 7% and 13% of the busiest month, rather than jumping at the 10% line. A place whose busiest month has under twice the visitors of its quietest is steady: crowds barely change, so they score 7 every month and don't reorder the months.
• Cost: a comparative proxy for destination price level and season, not a quote for what a visitor will pay. Its index starts with the country's restaurants-and-hotels price level (EU average 100) and adjusts for region and month. Where an official source publishes a regional money level, the hotel part of that level (we take 60%, a judgment call about how visitors split spending between rooms and meals) follows the region's rate that month against the country's typical rate, the plain average over its regions and months. So Madrid in October (€158 average revenue per occupied room-night) reads pricey and Extremadura (€65) cheap. The sources, each compared only with regions of its own country: hotel rate per occupied room in Spain (INE), Norway (Statistics Norway, converted from kroner at the month's average rate) and Finland (Statistics Finland; not Åland, where prices are suppressed); room revenue per guest-night in Portugal (INE) and Slovakia (Statistical Office, quarterly, so the three months of a quarter share one price); yearly accommodation revenue per guest-night in Sweden (Tillväxtverket, 2025) and Hungary (KSH and the Hungarian Tourism Agency, 2024), spread over the months by the national hotel price swing, so there the level is regional but the seasonal pattern is the country's (Swedish Lapland's winter peak, for one, is not captured); in Italy, ISTAT's regional accommodation price index gives each region's monthly swing but no money level. Bulgaria (NSI) and Italy (Banca d'Italia's spending per night) publish regional levels, but their terms don't yet clearly allow reuse on a commercial site, so those levels are withheld pending permission; the site shows the national level there. Everywhere else only the national level and the country's monthly hotel price swing are published, so two regions of the same country don't differ in level. Their seasons do differ: prices follow demand, so a region's monthly swing is the country's swing plus a share of how far the region's visitor numbers that month run above or below the country's pattern (the Alps are dearest in winter, the coasts in summer). The share (about 0.13 points of price per point of demand) is fitted each build on the regions that publish their own monthly prices; tested country by country, it predicts real regional swings better than the national swing alone in five of six countries. Its limit: a city whose summer rise is smaller than its country's countryside rush (Prague, Copenhagen) comes out a little cheaper in July than visitors might expect. We checked every European statistics office: France, Germany, Austria, the Netherlands, Belgium, Switzerland, Ireland, Denmark, Czechia, Poland, Romania, Slovenia and the Baltic states publish nights by region but no regional price we could use; Hungary and Sweden publish monthly regional figures only in interactive dashboards, so we use their yearly tables. Cheap below 80, moderate 80–105, pricey 105–130, expensive above. Inside the Overall, 50 or less scores 10 and 170 or more scores 1.

The Overall on every page is the Balanced mix: weather 50%, crowds 30%, cost 20%, and never more than the weather score plus 2 (plus 1 for cities), so empty streets and cheap hotels can't lift a month with poor weather to the top. Without a weather score there is no Overall. A first version weighted the three equally; crowds and cost then outweighed weather, and Rome's best month came out as January.

How often the Overall agrees with travel guides. We wrote down, before scoring, the months guides recommend for 31 places all things considered, and the months they advise against (Paris, Rome, Venice, the Greek islands, the Algarve, Prague, the fjords and others). In the October 2026 benchmark run, Overall agreement across the 31 places is 0.68 (1.0 is full agreement). The beach and island group scores 0.83 and the northern group 0.81; city breaks score 0.59. Paris and Lisbon still favour summer more than guides do, while Prague and Berlin each have a summer month among their top choices. Our visitor-season measure counts overnight stays across a whole region. It cannot measure queues at sights or day visitors. Changing the weights, cap or crowd scale did not resolve the city gap in our tests, nor did using foreign visitors' nights or regional airport passengers instead. For city breaks, read the visitor-season figure as overnight demand, not a forecast of how crowded a particular sight will feel.

Ranking places against each other. The Overall compares a month with the place's own year, so it answers "when should I go to Rome?" well, but it can't say whether Rome beats the Algarve in May: a quiet, cheap month anywhere scores as well as a great month somewhere famous. So the month lists (where to go in Europe in May) rank differently. First come the 100 most-visited regions, by tourist nights a year, ordered by that month's weather score; the other regions follow as lesser-known places, in the same order. We checked this before using it: for six months we wrote down, before scoring, the places travel guides agree on (the Canaries and Madeira in January, the Amalfi Coast and Crete in May, the fjords and the Dolomites in July, Vienna and Prague in December, and others). Ranking by Overall across all places put 20% of them in our top 20 in the three months held back for the test; the popular-destinations order put 51% there. It still misses the cool north in July and the Christmas-market cities in December, which people choose for reasons the weather can't show. Country-month pages rank a country's regions by weather in the same way. Sorting by Overall, crowds or cost is one tap away on every list.

Beach or sightseeing. Each region gets one weather score, for the trip most of its places are known for: the beach index for mostly coastal regions, the sightseeing index elsewhere. Some regions are known for a place visited the other way: Venice in a mostly coastal Veneto, Seville and Granada in Andalusia, the beach resorts of Bulgaria's coast in a mostly inland region. There, the month page says what the score rates and gives the other rating beside it, so a July visit to Venice reads "good for sightseeing, and hot", not only "very good for the beach". We tested switching those regions' scores to the other index; it helped in some places and hurt in others, so the scores stay as they are for now.

Near here, and how busy against all of Europe. The crowd word on every page is a season (moderate season, peak season): the month against the place's own year. Each month page lists nearby places on the same dates: the two closest regions, plus the three most-visited within about 190 miles, so Rome shows Umbria and Abruzzo next door and also Tuscany and Naples. Distances run between region centres, as the crow flies, so they are a guide, not a travel time. Beside the crowd word, which compares a month with the place's own year, beaches, islands, mountains and countryside also say how busy they are against all of Europe that month (overnight visitors for the place's size): Santorini in June is busy for Santorini and one of Europe's busiest places. We don't say this for cities: a big city's population dilutes the count and day-trippers aren't in it, so Rome in July would read "typical for Europe", which it isn't.

Your own priorities. The right trade-off depends on the traveller, so on month, country and place pages you can change the mix under What matters most?: Best weather (70/20/10), Fewer crowds (30/60/10), On a budget (30/20/50), or Custom sliders. The page then recomputes every Overall and rank in your browser with the same rule; the weather cap applies while weather carries 40% or more. Your choice is remembered in this browser only and never sent anywhere. The three parts always stay visible beside the Overall.

2. Season, for a place against itself

Eurostat publishes nights spent at tourist accommodation (hotels, holiday rentals and campsites) by month and NUTS 2 region, with a per-1,000-inhabitants version. We use each region's latest two complete published years from 2024–2025, falling back to 2023 where a newer complete year is missing, and divide by days in the month to get tourist nights per 1,000 residents per day. Where only one complete reference year is available, the guide names that year.

Season stage = that month's figure ÷ the region's own busiest month. Peak season means 85% or more; busy season 60–85%; moderate season 35–60%; low season below 35%. A region whose busiest month is under twice its quietest is called steady all year, because it has no real season.

This is dimensionless, so it can be compared across places: "you would arrive at Paris's peak but past Santorini's" is a fair statement. It says nothing about how crowded a place feels in absolute terms.

3. Visitors

We start from overnight visitors per resident: tourist nights per 1,000 residents per day, from the same Eurostat table. On a typical February night the Lisbon region has about 43,000 overnight visitors among 2.2 million residents, about 1 visitor for every 50 residents.

A bare figure like that has no frame of reference, so we show where it ranks among all Europe's regions in the same month, the alternatives for anyone whose dates are fixed. In the top tenth we say among the most visited this month; from the 75th to the 90th percentile, more visitors than most; in the middle half, about typical for Europe; in the bottom quarter, fewer visitors than most. Lisbon in February has more visitors per resident than nearly nine in ten regions, although only 2% of the people present are visitors. Across Europe the spread is steep: the typical region in February has about 1 visitor for every 130 residents, while the South Aegean islands in August have about 9 for every 10.

Statistical regions include their suburbs and countryside. Visitors concentrate in centres, so for city regions a historic centre has a higher share of visitors than the regional figure; pages for city regions say so. For countries we rank each country against the others that month.

What crowding cannot mean here

None of this measures how a place feels to stand in. That depends on the sights and hours you choose, including day-trippers and locals. Our overnight-stay data does not count them. Weather, visitor season and cost remain visible beside the combined Overall score so you can see why a month ranks where it does.

4. Prices

Finnish regional prices come from Statistics Finland’s all-accommodation series, including hotels, cottages and camping establishments with at least 20 beds or electrical caravan pitches. An occupied rental unit can be a room, a cottage or an electrical pitch. The average includes VAT and excludes meals and other services. Spain’s hotel ADR excludes VAT. These figures describe the recorded accommodation mix; they are not prices for equivalent rooms across countries. Finnish definitions · Spanish hotel questionnaire.

Eurostat publishes no regional hotel rate. Publishable regional accommodation money levels cover 58 regions in Spain, Portugal, Norway, Finland, Slovakia, Sweden and Hungary. Spain and Norway report average revenue per occupied room; Finland reports it per occupied accommodation rental unit; Portugal, Slovakia, Sweden and Hungary report revenue per guest-night. Only 38 regions have monthly regional rates: Slovakia's figures are quarterly, and Sweden's and Hungary's annual levels are moved through the year using a national monthly index. Monthly regional rate series use 2024–2025 where published; each displayed amount names its actual contributing period. Italy's ISTAT publishes a regional accommodation price index with no euro baseline. All euro amounts here are historical averages with their units and source named; none is a current booking quote. Where direct regional levels are unavailable, Cost starts from the national price level and estimates a regional seasonal adjustment from visitor demand.

National price level is Eurostat's 2024 comparative index for restaurants and hotels (EU average 100). Hotel price season uses the national harmonised consumer price index for hotels in each month against its own average month, averaged over 2024–2025. These are relative historical signals; neither gives a room price in euros or dollars. Eurostat also publishes national indices for package holidays and air transport. Its air transport index is not a fare from a US airport to the destination and is not used in the Cost score.

Selected local prices and hotel openings

Twelve month guides add local historical price comparisons: the Balearics from April to October, and Lapland in January to March, November and December. The local and parent averages use the same month in 2024 and 2025, with the local difference at least 15% in the same direction in both years. These are averages of two monthly observations, not a fixed hotel basket or a current quote. September and October’s Spanish comparisons include provisional 2025 figures. Other months have not been added to this pilot. The local figures guide what to price-check; regional month ranks do not rank individual islands or winter activities.

Balearic March and April also show estimated hotel-room capacity as a share of each year’s peak. This measures the seasonal size of the room supply, not vacant rooms, bookable rooms today, the share of hotels closed, or restaurant and activity openings. A smaller room supply can coexist with the guide’s Low visitor-season label. These comparisons provide context and do not feed any rating.

Price sources: INE tourist-area ADR and regional ADR; Statistics Finland 11ix for Lapland and its wider parent. Finnish parent prices retain the existing weighting by rooms, occupancy and days. Room-capacity sources: INE 2024 and 2025.

Prepared by Bigfoot Wanders from INE and Statistics Finland data, retrieved 6 October 2026. Statistics Finland updated its table on 24 September 2026; the INE capacity exports were created on 21 May and 11 September 2026 for the 2024 and 2025 tables respectively. Reuse follows INE’s terms and Statistics Finland’s terms (CC BY 4.0). Download the observations, statuses and source record. We review these notes yearly; mismatched parent periods or revised parent amounts cause the price note to be withheld pending review.

Snow-trip planning around Rovaniemi and Levi

The Lapland hub and November, December and March guides compare three winter-trip options separately from the warm-weather rating. The recommendation to start with March for snow and daylight is an editorial interpretation of the evidence below, not a new score or a claim that March is the best of all twelve months.

We summarize Finnish Meteorological Institute daily morning snow-depth observations from 2011–2025 at Apukka (about 15 km northeast of Rovaniemi, 106 m elevation) and Kittilä Pakatti (about 16 km south of Levi village, 179 m). These are rural/lowland stations, not town-centre or ski-slope measurements. Nearby railway/airport controls had no qualifying snow records, so there is one usable snow station per area.

A year-month needs at least 90% of its expected snow observations, and a summary needs ten qualifying years. Apukka December 2018 and Pakatti November 2020 fail the daily-coverage rule and are excluded. Missing observations are never zero-filled. FMI code −1 means no snow; 0 means none at the station but some in surrounding open terrain. Both fall below 1 cm.

Both stations have 465 of 465 March observations at or above 10 cm across these fifteen years. This historical record does not guarantee future conditions. Including incomplete years and bounding missing days both ways preserves the March/December versus November distinction. Daylight uses our existing solar calculation on the 15th at each station's latitude; it excludes twilight and does not measure sunshine. Zero direct daylight near Levi in mid-December does not mean continuous darkness.

Winter accommodation comparisons reuse the reviewed Lapland accommodation series, not town-specific rates. They disappear if the current period or parent comparison no longer reconciles; the snow guidance remains available. Operator links are manually checked factual references, reviewed 6 October 2026. Stored-snow facilities can open before the wider natural-snow network; activity dates and current conditions still require the operator's own confirmation.

Contains modified Finnish Meteorological Institute observation data, 2011–2025, summarized by Bigfoot Wanders under CC BY 4.0. Retrieved 6 October 2026. Observation definitions · download station summaries, requests and source hashes. Recheck coverage and locality annually and operator references before each winter; these summaries never enter Weather, Crowds, Cost or Overall.

Other figures

• Overnight visitors per night: the per-1,000 rate × the region's population on 1 January (Eurostat).
• Domestic share: nights by residents of the country ÷ all nights, from the same Eurostat table.
• Platform rentals: Eurostat also publishes nights in short-stay rentals booked through Airbnb, Booking, Expedia and Tripadvisor, collected from the platforms themselves, by month and region. That reaches the small private lets that fall below the national reporting thresholds of the establishment count, but some hosts are in both, so the two cannot be added. We show the platform figure beside the main one, and give the visitor share as a range: the lower end counts establishments only, the upper end treats every platform night as extra. We use only complete platform years that also contribute to the region's accommodation-night baseline, normally 2024–2025 and 2023 where a newer year is unavailable. Each region's guide gives the exact platform period separately.
• School holidays: from the OpenHolidays project (open data, ODbL), which publishes school and public holidays by country, state, region, zone or municipality. We match those to each region; where a region spans several calendars (a Swiss greater region with several cantons, a Dutch province split between holiday zones) we say "in parts of the region". A region's school year is used only when its calendar is complete, tested by the whole region being on summer holiday at some point between 15 July and 15 August; otherwise a missing holiday could read as "schools are open". Single days off are left out. OpenHolidays has no or only partial calendars for Spain, Italy and Greece, so for those we compile the official 2026/27 calendars ourselves: each Spanish community's and each Italian region's published decision or gazette entry, and for Greece the national rules in Presidential Decree 79/2017 (Easter follows the Orthodox date). We record what the text states (first and last day of lessons, and each closure) and join closures to the weekends and public holidays around them. Where a calendar leaves the choice to provinces or schools (Andalusia's provinces, Navarra's schools, the Basque Country's framework) we state only what is true everywhere in the region. A calendar we could not confirm from an official source is left out rather than guessed: Asturias and Basilicata for 2026/27. For each country we also compute the share of its people who live where schools are out on a typical day of the month, weighting regions by population, when the complete regions hold at least 80% of the population. Not yet covered: Ireland, Cyprus, the Nordic and Baltic countries, Asturias, Basilicata, Corsica, and parts of Switzerland; some countries' 2026/27 calendars are not published in OpenHolidays yet.
• "Is it a good time?": the weather wording uses the displayed, rounded Weather score: yes (7.0/10 or higher), mostly (5.0–6.9), mixed (3.0–4.9), not really (below 3.0). For a country, the count likewise uses regions at 7.0/10 or higher. Visitor season and prices appear beside it; Overall combines their scores for a separate comparison.
• What to wear: a fixed rule on the typical afternoon high and night low: 81°F (27°C) and above summer clothes; 70–80°F (21–26°C) T-shirt weather; 59–69°F (15–20°C) light layers; 48–58°F (9–14°C) a sweater and a light jacket; 37–47°F (3–8°C) a warm coat; below 37°F (3°C) a winter coat, hat and gloves. Rain gear once a typical month has 8 or more days with at least 1 mm of rain.
• Night lows: the average daily minimum temperature, from ERA5's daily mean corrected with station minima (see Weather above).
• Daylight: hours between sunrise and sunset on the 15th, as almanacs count them (the sun's upper edge, with standard refraction), computed from the region's centroid latitude.
• Typical afternoon, rain days, heat: from the same ERA5, station-corrected daily data. Typical afternoon is the mean daily maximum; rain days count days with at least 1 mm of rain (median year, and the 90th-percentile wet year); heat is the number of recent years in which the month had at least one day at or above 35°C.
• Region type: from Eurostat's own coastal, mountain and urban typologies of NUTS 3 areas, weighted by NUTS 3 population so a region is typed by where its people live. Beach if at least 90% of people are in coastal areas; city if at least 70% are in predominantly urban areas; mountains if at least half are in mountain-flagged areas and the population-weighted elevation is 500 m or more (the mountain flag is surface-based, so Florence's province carries it); beach if at least half are coastal; countryside otherwise. Coastal metros whose visitors come for the city (Lisbon, Athens, Copenhagen, Barcelona and a few others) are set to city by a short published override list. The basis is stated on each region's page.

Known limitations

• Crowd data is monthly. Nothing finer exists for Europe. Weather is daily underneath, and heat and rain odds use that.
• Regions are statistical regions, not cities. Paris means the whole Île-de-France; central Paris is busier than the regional figure.
• Nights exclude day-trippers and people staying with friends or in their own second homes, so visitor share is understated where those matter. Private lets not on the four big platforms are in neither count.
• Two reference years for crowds. The regional monthly series only starts in 2020.
• Weather is a population-weighted average over the region. That favours the valley towns over the high passes in mountain regions, which is where visitors sleep, but it still describes the region, not one resort.
• The weather index is warm-season oriented. Ski months in mountain regions will score low on weather while being that region's peak.
• Prices are national except where a statistics office publishes regional hotel rates or revenue (Spain, Portugal, Norway, Finland, Slovakia, Sweden, Hungary) or a regional accommodation price index (Italy). Regional money amounts use different units: revenue per occupied room or per guest-night. They are historical averages, not the price of a particular booking.

Sources

• Eurostat tour_occ_nin2m, nights spent at tourist accommodation establishments by month and NUTS 2 region (NUTS 2024).
• Eurostat demo_r_d2jan, population on 1 January by NUTS 2 region.
• Eurostat tour_occ_nin2, tour_cap_nuts2, tour_occ_anor2: accommodation mix, capacity and hotel occupancy, annual.
• Eurostat tour_ce_omn12, short-stay accommodation booked via online platforms, by month and NUTS 2 region.
• Eurostat prc_ppp_ind (price level indices) and prc_hicp_midx (HICP monthly indices).
• INE Spain, Coyuntura Turística Hotelera, table 2059 (ADR by autonomous community). INE Portugal, Inquérito à Permanência de Hóspedes (room revenue and guest-nights by NUTS II). ISTAT, NIC consumer price index, accommodation services (ECOICOP 112) by region.
• ERA5 hourly data on single levels, Copernicus Climate Change Service (C3S) Climate Data Store, DOI 10.24381/cds.adbb2d47, CC BY 4.0, including its land-sea mask and orography. Contains modified Copernicus Climate Change Service information 2026. Neither the European Commission nor ECMWF is responsible for any use that may be made of the Copernicus information or data it contains.
• Weather-station records, aggregated to monthly values by us: Datenbasis: Deutscher Wetterdienst, Einzelwerte gemittelt (CC BY 4.0) · Source : Météo-France, Données climatologiques de base – mensuelles (data.gouv.fr), Licence Ouverte 2.0 · AEMET, Agencia Estatal de Meteorología (AEMET OpenData) · KNMI (Royal Netherlands Meteorological Institute), daily station data, CC BY 4.0 · Copyright Met Éireann; source met.ie; CC BY 4.0; Met Éireann does not accept any liability whatsoever for any error or omission in the data, their availability, or for any loss or damage arising from their use · Datenquelle: GeoSphere Austria – https://data.hub.geosphere.at (CC BY 4.0) · Source: MeteoSwiss (CC BY 4.0) · Data from MET Norway (CC BY 4.0) · Source: SMHI, CC BY 4.0 · Finnish Meteorological Institute open data, CC BY 4.0 · Contains data from the Danish Meteorological Institute (DMI), CC BY 4.0. All modified: daily values averaged or counted by month.
• GHSL GHS-POP R2023A, population grid 2020, European Commission Joint Research Centre, to weight grid cells by where people live.
• NOAA NCEI, ETOPO 2022 Global Relief Model, doi:10.25921/fd45-gt74: terrain height for the region types and the height correction. Public domain.
• NOAA OI SST V2.1 data provided by the NOAA PSL, Boulder, Colorado, USA, from their website at https://psl.noaa.gov (sea temperatures).
• GISCO NUTS 2024 boundaries and territorial typologies. © EuroGeographics for the administrative boundaries.
• Eurostat datasets (CC BY 4.0): figures shown are our own calculations; Eurostat is not responsible for them. Exchange rates: ECB reference rates via Eurostat (source: ECB statistics). Air passenger routes: Eurostat avia_par. Airport locations: OurAirports (public domain).
• Regional hotel prices: Statistics Norway (ssb.no), table 14168, CC BY 4.0 · Statistics Finland, StatFin 11ix, CC BY 4.0 · Elaboración propia con datos extraídos del sitio web del INE: www.ine.es (table 2059) · Instituto Nacional de Estatística – Portugal (Statistics Portugal), modified by us · Istat, elaboration on Istat data (NIC), CC BY 4.0 · HCSO / ksh.hu · Statistical Office of the Slovak Republic (DATAcube cr3806qr), CC BY 4.0 · Tillväxtverket, Inkvarteringsstatistiken.
• Contains information from OpenHolidays API (openholidaysapi.org), which is made available here under the Open Database License (ODbL) 1.0. Our derived holiday data is published separately under the same licence at /data/holidays.json.
• Official school calendars 2026/27 for the Spanish autonomous communities, Ceuta and Melilla, and the Italian regions and autonomous provinces (regional gazettes, decisions and education departments; each region's page links its document); Greece: Presidential Decree 79/2017.
• Scott, D., Rutty, M., Amelung, B., Tang, M. (2016). An inter-comparison of the Holiday Climate Index (HCI) and the Tourism Climate Index (TCI) in Europe. Atmosphere 7(6):80. Rutty, M. et al. (2020). Atmosphere 11(4):412. Demiroglu, O.C. et al. (2020). Atmosphere 11(9):911.

Editorial notes, where they appear, are marked as such. No page depends on them. There are no booking links and no affiliate relationships, because the most useful thing this site can say is "don't go then", and that cannot be said credibly while earning commission on the booking.