About Where's Hot When
Curated by Hugo. Updated July 2026.
What this is
A climate reference for people choosing when and where to travel. Pick a month, read the colour, open a place. The map answers two questions plainly: where it is warm in a given month, and where it is cool enough to escape a heatwave in the same month. Both are usually treated as taste. They are data.
It is not a forecast. It will tell you that Marrakech in March averages 23° with four rain days across the last ten years. It will not tell you whether it rains on the Tuesday you land. That distinction is the whole point.
The interactive map is the primary surface; the month guides are the long-form companion.
How the data works
Every colour on the map is a long-run monthly mean of daily maximum temperature, computed per point from ten years of dailies (2016 to 2025, 10 full years). The underlying source is the Visual Crossing global historical archive, which blends physical station observations from the NOAA Integrated Surface Database and national met services with reanalysis fill where stations are sparse. Millions of daily rows are aggregated into twelve monthly means per point.
The map carries 3,091 measured points. Editorial destinations sit as named pins on top of a continuous grid of unnamed cells, so every land square in view has a colour, not just the famous places. Cells are snapped to the nearest station or reanalysis pixel and inherit that record.
The signature stat is swing: the p10 to p90 range of daily maxima across all ten years for the chosen month. A place with a July mean of 28° and a swing of 24 to 32 is a different holiday from one with the same mean and a swing of 19 to 37. Averages hide the shape of a month; swing shows it.
The 2025 anomaly lens flips the same map to last year's monthly means only, so a hot 2025 in southern Europe reads visibly hotter than the ten-year default. The ten-year lens is the climate reference; the 2025 lens is the recency check. Both come from the same daily archive. Map colours between measured points use interpolation; the 2025 comparison describes an observed year, not a projection.
Year-to-year reliability. The destination summary counts how many covered years had a monthly mean daytime high within 2°C of that month's historical average. The Reliable filter requires this in at least 80% of covered years, with a minimum of eight years. Each year-month must have observations for at least 90% of its calendar days. A count such as 9 of 10 describes the 2016 to 2025 record. It is not a 90% forecast probability for a future holiday.
Warm and mostly dry. The expanded explanation also counts years when at least 70% of observed days had a daytime high of 24 to 30°C and less than 1 mm of rain. Both temperature and rainfall must be present for at least 90% of the month. This is a fixed, explained example of holiday conditions, not a prediction or a personalised score.
Daily variability. The daily high spread is the width between the 10th and 90th percentiles. It describes variation between days, separately from year-to-year consistency. A width is not a plus/minus range. Older daily-variability grades use widths below 6°C, below 9°C, below 13°C and 13°C or more, respectively.
Beach mode restricts the map to points whose active reading sits in the warm bands (19 to 25, 26 to 32, 33+). Under the ten-year and 2025 lenses the monthly mean decides; under the Now lens the current reading does. Cold coasts are muted rather than hidden, since geography alone is not the answer.
Who makes it
Editor's note
I built this because the "best places to visit in March" lists were all the same twelve cities in a different order, and none of them told me what the weather would actually feel like on the ground. So I made the tool I wanted, in the spirit of seat61.com: one person, an obsessive dataset, plain prose, no upsell.
The work is unglamorous. Ten years of dailies, cleaned per point, aggregated twelve ways, coloured so a continent reads at a glance. Then the writing, one month at a time, one honest paragraph per place. Where I have an opinion I say so and sign it. Where the data is thin I say that too.
If a page here saves you from a rainy week you did not want, it has done its job.
Hugo
Sources
Weather. Daily observations and reanalysis via the Visual Crossing global historical archive, itself sourced from the NOAA Integrated Surface Database, national meteorological services, and ERA5 reanalysis where stations are sparse.
Coffee and beer estimates. Country estimates cover 193 countries and scale US anchor prices using country price levels. A set of 56 editorial destination estimates accounts for selected city and resort differences. These are not surveyed local prices. Beer means 500 ml draught; figures and exchange rates are a July 2026 snapshot. The card labels all prices approximate. The scaling reference is the World Bank price-level indicator.
Cost of living. Coffee and beer anchor prices scaled by the World Bank price-level indicator (nominal GDP per capita divided by PPP GDP per capita, US = 1.0).
Flight footprint. Return economy kg CO2e from London Heathrow using DEFRA 2024 conversion factors with radiative forcing (0.15298 kg per pkm below 3,700 km, 0.19085 kg above). Great-circle distance underestimates real routing by 8 to 12%, so treat the number as a floor.
Travel advisories. Curated snapshots of the UK Foreign, Commonwealth and Development Office country pages. Each chip links to the live FCDO page; check the latest advice before booking.
Tap water. Curated from the CDC Yellow Book and WHO guidance, reduced to safe, mixed, unsafe.
Place descriptions. Editorial where written; otherwise the Wikipedia REST summary resolved by geosearch, attributed inline under CC BY-SA.
Live Now. Current temperatures are shown separately from monthly daytime highs. A refresh job is scheduled every three hours and processes a bounded batch; coverage and fetch times vary by place. Readings over six hours old, missing readings and paused feeds are unavailable. Grey dots in Now mode have no fresh reading. A current temperature is not directly comparable with a monthly average daytime high, so we do not call that difference a weather anomaly.
Raw data files live under public/data/. The prebuild runs scripts/audit_invariant11.py, which fails the build if any chip cannot trace to its source.
Data terms
The computed statistics on this site (monthly means, swing percentiles, rain days, steadiness bands, the 2025 lens) are viewable and citable with attribution to wheres-hot-when.com. Quote a figure, screenshot a map, link back. That is the ask.
The underlying daily observations are licensed from Visual Crossing and are not redistributed here. If you need the raw dailies for your own work, buy them from the source. Nothing on this page grants an open licence to the underlying archive.