What does this show me?

The graph below shows a sample "distribution" of how far off temperature predictions have been in the past. Perfect predictions give zero as a difference value. The shaded bars just summarize the graph: the darker the shaded bar, the more predictions fell in that difference zone.


Take-A-Sweater? uses historical weather records to compare currently predicted temperatures with what actually happened under similar conditions in the past. "Similar conditions" are defined in "Settings" to be within "Date Tolerance" days of the present calendar date and "Temperature Tolerance" degrees of the current predictions for each of the next five days.

An Example Imagine that: today is February 9; the high temperature 3 days from now (on February 12) is predicted to be 43°; date tolerance = 6 days; and temperature tolerance = 3 degrees. To create the distribution you’ll see on-screen for 3 days out (February 12), Take-A-Sweater? will search its database, which extends back to 2005, for all past 3-day-out predictions of a highs between 40 and 46 degrees, during a range of calendar dates from February 3 to February 15.

For other time spans (e.g. 1, 2, 4 or 5 days out), matching predictions are similarly used.

Credits & Motivation

This App was created in 2012, for use in the Harvard University General Education course "The Art of Numbers," taught by Prof. Alyssa Goodman. The code was written by Bill Barthelmy of Harvard’s Academic Technology Group. Historical data were kindly provided by ForecastWatch, a product of Intellovations, LLC. Current five-day weather forecast data are provided by NOAA.

Our goal in creating this App was to offer an example of how the "uncertainty" associated with computer modeling and prediction can best be displayed, and how uncertain predictions are used in everyday life. As we continue our work, we hope to add cities beyond Boston, and additional features that will demonstrate geographic and temporal trends. We will also revise our display design using input from students and users like you!