The question started badly. I wanted to know when it rains in a city. That is already vague. Does “rain” mean drizzle? Does “when” mean month, hour, or “when I forgot my umbrella”?
So I made it unfairly specific. A rainy hour means at least 0.2 mm of rain. A useful answer must fit into a 1 to 6 hour local-time window. Then I tested all of them.
The surprise: some cities really do have a rain clock. Others are just wet, dry, or annoying.
The rain clock atlas
Each row is a city. Each column is a month. Green means the best umbrella window is useful. Red means time of day barely helps. The big number is Umbrella Edge: the adjusted rain-risk gain from using the marked hours.
- Big number: Umbrella Edge. 45 pp is a rain appointment. 30 pp is a strong habit. 1 pp means the clock is nearly useless.
- Small number: monthly wet-hour rate. This says how often that month needs an umbrella at all.
- Cell color: red means weak timing, white means middling, green means useful timing.
- 24 bars: rain risk by local hour. Amber bars mark the best 1 to 6 hour umbrella window.
Some cities switch personalities. Addis Ababa has an Umbrella Edge range of 29.6 pp across months. Taichung is 27.4 pp. Ho Chi Minh City is 25.1 pp. In these places, asking “which month?” is not pedantry. It changes the answer.
Some cities are consistently unhelpful. Los Angeles ranges from 0.4 to 1.3 pp. Shijiazhuang ranges from 0.2 to 1.2 pp. Cairo ranges from 0.0 to 1.1 pp. Maybe carry an umbrella because the forecast says so. The clock is not doing much work.
I looked for cities that were flat but still strong. That category mostly refused to exist. The closest examples are Chennai, Durban, and Melbourne. Chennai is the best of that flat group, and even it averages only 3.4 pp. So, no, there was not a secret “always rains at 4 pm” city hiding there. Mildly disappointing.
The chart is the evidence. Sort by range to find cities whose rain schedule changes through the year. Sort by average edge to find cities where the clock is usually helpful. Then hover, because the tooltip has the raw inside and outside rain risk. That is where the cheating would show up.
How the number works
Umbrella Edge is the number in each cell. It is in percentage points.
45 pp means the marked hours are much better umbrella hours, and they catch a lot of the month’s rainy hours. 30 pp is still a strong daily pattern. 1 pp means time of day barely helps.
The formula is:
Umbrella Edge = (rain risk inside the window - rain risk outside the window) * sqrt(share of rainy hours captured by the window)
The square root bit is there because I did not want a silly one-hour spike to win if most rain happened elsewhere. This is the part where the metric is trying not to embarrass itself.
Data: Open-Meteo historical hourly precipitation, 2015 to 2024, using GeoNames city coordinates. The master CSV is in data.7z.