THE FORECAST · 4 MIN READ

Why forecasts are wrong

Forecasts are far better than their reputation and far worse than their presentation. Both facts come from the same place.

Modern forecasts are genuinely good. A five-day forecast today is about as accurate as a one-day forecast was in 1980. Tomorrow's high temperature is typically within a few degrees, and hurricane track forecasts have improved so dramatically that the average three-day error is now smaller than the one-day error was a generation ago.

And yet everyone has a story about the ruined picnic. Here's why both things are true.

Chaos is not a metaphor

In 1961 a meteorologist named Edward Lorenz restarted a simulation using rounded numbers, 0.506 instead of 0.506127, and got a completely different weather pattern. That accident founded chaos theory. The atmosphere amplifies tiny differences exponentially, which means a perfect forecast would require perfect knowledge of every molecule.

The practical consequence: there is a hard ceiling on weather predictability of roughly two weeks, no matter how good computers get. Beyond that, you're describing climate, not weather.

ROUGH ACCURACY BY LEAD TIME

Day 1, highly reliable. Day 3, good on the pattern, fuzzy on timing. Day 5, the general idea, not the details. Day 7, a useful signal about temperature trends. Day 10+, treat as a rumor.

The scale problem

Even a good model can't resolve what's smaller than its grid boxes. Summer thunderstorms are the classic victim: the model knows conditions favor storms across a region, but which specific street gets hit is below the resolution of the forecast and, arguably, of physics itself given what we can observe. So the forecast says 40%, and someone gets soaked while someone else waters the lawn. Both are consistent with a correct forecast.

What "wrong" usually means

Most perceived forecast failures are timing errors, not existence errors. Rain arrived at 5 PM instead of 2 PM. The front stalled a county short. The storms fired an hour late. The event happened, the schedule slipped, and the schedule is what you planned around.

How to read one honestly

See it on your own sky
Fat Tail shows these numbers measured by the airport sensor nearest you, and explains each one when you tap it. Free forever, coming soon to iOS.
TRY A LIVE FIELD REPORT →