Some places are sunny and dry, some dull and wet. A few manage to be sunny and wet, and a strange handful are dull and dry. This is the world's sunshine set against its rainfall, cell by cell, on a map you can re-cut by month and by where you draw the lines. Where does your town sit, and which places share its weather?
Ask anyone in Britain about the weather and you will hear about two things: whether the sun came out, and whether it rained. They are not the same question. Lima, on the coast of Peru, gets less rain in a year than London gets in a fortnight, yet it is one of the dullest big cities on Earth. Surabaya in Java is soaked by the monsoon and still sunnier than Madrid. Put the two measures together on one map and the world falls into four corners and a middle.
I took the Climatic Research Unit's 10-minute climatology for 1961 to 1990 (CRU CL 2.0), which gives every land cell outside Antarctica a monthly average for sunshine and for rainfall. Sunshine comes as a percentage of the possible daylight, which I converted to hours using the length of the day at each latitude in each month. Then I coloured every cell by two things at once: which third of the world's land it falls in for sunshine, and which third for rain. Nine classes, four corners. Drag the lines, pick a month, and the map redraws.
Each cell is coloured by its class. By default the breaks are terciles of land area, so each third of the land falls on each side of each line, and every class shows up. Switch to fixed breaks for round numbers, or drag the handles under the legend. Hover over (or tap) a legend square to see only those cells. Hover over the map to read a cell; click to open it below.
Sunshine: CRU CL 2.0 percentage of possible sunshine, converted to hours with the astronomical day length. Rain: CRU CL 2.0 precipitation. Both are 1961 to 1990 averages on a 10 arc-minute grid of land (no Antarctica). Shares are of land area. Base map tiles © Esri.
Search any of the 34,167 places in GeoNames with 15,000 people or more, or click anywhere on the map. You get its year of sunshine and rain, month by month, where it falls in the world, and the places whose twelve months look most like it. Weather twins can be on the other side of the equator: tick the box to let a southern summer match a northern one.
Loading the place list…
If sunshine and rain had nothing to do with each other, each of the nine tercile classes would hold a ninth of the land, about 11%. They do not. Sunny places tend to be dry and dull places wet (Spearman's rank correlation across cells is −0.24), so the sunny and dry corner holds 21% of the world's land and the sunny and wet corner only 3.0%. Click a city to open it.
The world's big cities by sunshine and rain
573 places in GeoNames with a million people or more, on the classes currently set on the map (whole year). UK cities are ringed. Rain is on a log scale.
The UK's land averages 1,319 hours of sunshine and 1,121 mm of rain a year in this data. Among the 156 countries with at least 20,000 km² of land in the grid, only Iceland, Ireland and Norway are duller. The sunniest UK cell, on the Sussex coast near Bognor Regis, gets 1,774 hours, and 84% of the world's land gets more. Every bit of the UK is in the dullest third of the world's land.
Rain is a different story. At 628 mm a year, London is drier than 415 of the 573 big cities. The UK is wet in the sense that it rains often, lightly, under cloud; it is not wet in total, except in the western hills, where one cell near Loch Lomond reaches 3,300 mm. What really sets Britain apart is the lack of sun: of the people counted in the 34,167 places of this study, 96% live somewhere sunnier than London.
UK cities on the world's sunshine scale
Bars: share of the world's land by annual sunshine hours. Ticks: UK places of 150,000 people or more.
The sunniest big city in the grid is Phoenix (3,817 hours), with the cities of the Sonoran Desert, Arabia, Iraq and Sudan close behind. The dullest is Guayaquil in Ecuador (1,154 hours), then six cities of Sichuan, Chongqing and Guizhou in south-west China, a region known for low cloud. Next comes Birmingham, in the English Midlands, with 1,356 hours. Click a row to open the city.
Big cities: GeoNames places with a million people or more (573). GeoNames lists some districts of large cities separately, so a few names repeat a city's weather. Countries: area-weighted means over cells inside Natural Earth outlines, countries with 20,000 km² of land or more.
Annual totals hide the calendar. In London the sun follows the length of the day, peaking at 192 hours in June. In Mumbai the monsoon turns that upside down: July brings 959 mm of rain and only 90 hours of sun, while May, before the rains, has 304. Across the tropics the sunniest month falls outside the local summer half of the year on 63% of the land; outside the tropics it does so on only 3%. Use the month menu above the map to watch the monsoon move.
Twelve months of sun and rain in six cities
The rain in CRU CL 2.0 rests on many thousands of gauges; the sunshine on far fewer stations, and where sunshine was not recorded the authors converted cloud cover to sunshine (New et al. 2002). In much of Africa, South America, central Asia and the high Arctic, the sunshine surface is interpolated across wide gaps, and the 10-minute cells smooth mountains and coasts. So read single cells as indicative, and patterns as the story.
For rain I could check. WorldClim 2.1 is a separate climatology for 1970 to 2000 built with different methods. Across 566,058 shared cells the two agree closely (Pearson's r = 0.977); the median cell differs by 28 mm a year, about 6%, and the rain tercile is the same on 94% of the land. There is no second sunshine grid on hand to compare, which is the main weakness of the map.
Data. CRU CL 2.0 (New, Lister, Hulme and Makin 2002): mean monthly sunshine as a percentage of the maximum possible, and mean monthly precipitation, 1961 to 1990, on a 10 arc-minute grid over land outside Antarctica (566,262 cells with both). The station network behind them is uneven and sunshine is the thinner of the two.
Sunshine hours. Hours = percentage ÷ 100 × astronomical day length × days in the month. Day length is the FAO-56 formula (Allen et al. 1998) for each day, averaged over the month, at each cell's latitude. It counts the sun from rising to setting, so it is a little longer than the time a sunshine recorder can register; CRU's percentages are relative to day length, so the conversion follows their definition.
Classes. Equal land breaks are weighted quantiles of the cells, weighted by cell area, recomputed for whichever period you choose. Fixed breaks are 2,000 and 3,000 hours and 500 and 1,000 mm a year (3 × 3), scaled by the length of a season or month. The colours come from four corners mixed in the OKLab colour space.
Places. GeoNames cities of 15,000 people or more (34,167 matched), each given the nearest land cell (median 6.7 km away). Coastal and island cities can pick up a cell a few kilometres inland. "People" means the GeoNames populations of these places, not everyone; GeoNames lists some city districts alongside their city, so a few people are counted twice.
Weather twins. For each month, the difference in sunshine hours and in the square root of rain, each divided by its spread across all land and months (80 hours; 4.5 root-mm), then the root mean square over twelve months. A score of 0 is identical; the default limit is 0.5. The flip option also compares each place with the other's months shifted by six.
On the page. The monthly grids are stored as images, sunshine to the nearest 2 hours a month and rain on a square-root scale. The largest error this adds to a cell's yearly total is 9 hours of sun and 25 mm of rain, so the map's class edges can differ from exact values by a whisker. City totals in the cards and tables are exact.
Not covered. 1961 to 1990 is a past climate; sunshine has changed since in places (for example with cleaner air in Europe). There is no temperature, humidity or rain intensity here. Ten-minute cells are about 18 km across at the equator: no valleys, no sea breezes.
Climate: CRU CL 2.0, Climatic Research Unit, University of East Anglia, New, M., Lister, D., Hulme, M. and Makin, I. (2002) A high-resolution data set of surface climate over global land areas. Climate Research 21: 1 to 25. No licence is stated with the files; I publish derived grids and aggregates with this citation. Cross-check: WorldClim 2.1, Fick, S.E. and Hijmans, R.J. (2017) WorldClim 2. International Journal of Climatology 37(12): 4302 to 4315 (only summary statistics are published here). Places: GeoNames cities15000 (CC BY 4.0). Country outlines: Natural Earth 1:10m (public domain). Day length: Allen, R.G., Pereira, L.S., Raes, D. and Smith, M. (1998) Crop evapotranspiration, FAO Irrigation and Drainage Paper 56. Base map, hill shading, shaded relief and satellite tiles © Esri, HERE, Garmin, USGS, NGA, Maxar, Earthstar Geographics, © OpenStreetMap contributors, and the GIS user community.