Pick two weeks. See what the world's weather usually does then — averaged over 30 years.
Data: NASA POWER daily series for 1996–2025, one request per land cell of a 1° grid — air temperature and bias-corrected precipitation from NASA's MERRA-2 reanalysis, radiation from CERES SYN1deg/SRB satellite retrievals. Every value on this map is a 30-year mean for that ISO week: — covered.
Sunshine hours are not measured by satellites directly. They are modelled from the daily clearness index KT — the share of top-of-atmosphere radiation that reaches the ground — as a fraction of the astronomical daylength N: n/N = -0.691 + 4.541·KT − 3.124·KT². The coefficients are fitted against ERA5's own sunshine duration across 22 sites and graded on the half of them the fit never saw: bias -0.21 h/day, RMSE 1.05 h/day. Textbook Ångström–Prescott coefficients score −2.64 / 2.98 on the same test, mostly by drowning the wet tropics.
"All year" counts windows, it does not average them. With the finder set to all year, a place is shown when it meets your limits in at least the chosen share of the year's 53 windows of the length you picked — 53 fortnights, or 53 three-week windows, wrapping past New Year. That is a different and much stricter question than whether its annual mean sits inside the band: a monsoon climate can average 25 °C and 40 mm and still fail every single window. Windows a baseline cannot judge (week 53 in the five-year view) count in neither half, and the panel says how many windows were counted.
Grid or countries. The grid view shows the data as it is: one value per 1° cell. The country view averages those cells inside each border, weighted by cell area, so a wide country is not dominated by its polar edge — useful for comparing places, misleading for large ones, since a single number for Chile or Russia hides more than it shows. Countries too small to contain a cell centre (Singapore, the Maldives, Malta) borrow the nearest three cells; the panel says when it did that. Scroll or double-click to zoom, drag to pan.
Three baselines. The same fortnight averaged over the last 30, 10 or 5 years. Thirty years is the steadier number — about 210 observed days behind each week — and the one climatology usually means. Five years is noisier but closer to what the place has been doing lately; the gap between them is itself worth a look. Week 53 barely exists in a five-year window, so it is mostly blank there.
Read it as climate, not forecast. A fortnight in one particular year can sit far from its 30-year mean. Cells are 1° wide — about 110 km — so a valley and the ridge above it share one number, and small islands inherit the sea around them.