How many cover days does your exterior shoot need?

What the tool answers

Two questions decide an exterior block weeks before anyone looks at a forecast: which month to shoot in, and how many cover days to put in the budget. The tool answers both from ten years of daily weather data for the city — not with a percentage you have to interpret, but with the number of cover days that keeps the schedule standing in 80, 90 or 95 out of 100 versions of the shoot.

The grid plays your shoot out 20 times with the city’s rain frequency, so you can see what “70 %” looks like in practice: most versions lose one day, a few lose two, a few lose none. The strip below it shows the same number of days in every month, with sunshine hours as a dashed line — useful when the month is still open.

How we calculate

A rain day is a day with at least 1 mm of precipitation — less than that rarely stops a camera. For each city we count those days per month from daily ERA5 reanalysis data for 2015–2024 and average them over the ten years. Ten years instead of the meteorological thirty keeps recent summers in the picture.

Each exterior day is then treated as lost with probability p = rain days ÷ days in the month, multiplied by the share of rain days that would actually stop you. The number of lost days follows a binomial distribution (Poisson-binomial when your dates cross a month boundary). The cover days are the smallest number k for which “k or fewer days lost” reaches your chosen certainty.

p = rain days ÷ days in month × share that stops you P(≥ 1 day lost) = 1 − (1 − p)^n expected loss = n × p cover days k = smallest k with P(lost ≤ k) ≥ 80 / 90 / 95 %

Where the tool is wrong

• Rain clusters. The model treats days as independent. Real rain comes in spells: the chance of losing at least one day is somewhat lower than shown, the chance of losing two or three in a row somewhat higher. If your block is long, hold the cover days together, not spread out.
• 1 mm is not your threshold. A drizzle day counts the same as a downpour. Use the share slider: if half your scenes can play in light rain or under a roof, set it to 50 %.
• Climate is not a forecast. These are averages of the last ten years. From about ten days before the shoot the national weather service is the better source; from three days the hourly forecast decides the call sheet.
• One grid point per city. The data describes the city as a whole. Coast, mountains and valleys a few kilometres away can be markedly wetter or drier.

From the planning table

A three-day exterior shoot for a brand film in Hamburg, with the client asking for “summer”. July averages 15 rain days: three exterior days are hit at least once with 86 % probability, and 90 % certainty needs three cover days — as many as shooting days. September has 10 rain days: 70 % risk, two cover days. Same light for the brief, one cover day less in the budget.

Most of the scenes can play in light rain under the harbour roofs, so the producer sets the share slider to 50 %. September then comes to 42 % risk and one cover day. That one cover day — booked as a soft hold with crew and equipment — is the number that goes into the quote.