The study provides the first revealed-preference estimates of the benefits of routine weather forecasts. The benefits come from how people use the information to reduce mortality from heat or cold. The study shows that more accurate forecasts reduce mortality if and only if mortality risk is convex in forecast errors, which in turn depends on how people use forecasts. Using data on the universe of mortality events and weather forecasts for a twelve-year period in the U.S., the author shows that making forecasts 50% more accurate would save 1,700 lives per year, for gross annual benefits of $16 billion. The effects of forecast errors indicate that adaptation becomes less effective if temperatures are either higher or lower than expected. Forecast-driven adaptation is especially important in extreme heat, which suggests that short-run weather forecasts could be an important tool for managing the effects of climate change.