Can you predict the weather?

All of my friends raise an eyebrow when they hear that I’m a member of a competitive weather forecasting team. What is a weather forecasting team? And what possible qualifications could I have as a philosopher allow me to predict the weather?

This is the first year that the University of Toronto has had a forecasting team. Started by graduate students studying atmospheric physics, the team competes in the WxChallenge, a North American collegiate forecasting competition against approximately 60 other North American Universities. The competition involves predicting the high temperature, low temperature, highest sustained wind-speed, and precipitation, for a particular observation station 24 hours in advance. The next day predictions are compared to the observations made at the station; the higher the discrepancy the more points a forecaster earns. The few thousand forecasters that compete are then ranked according to these points and the top 64 will go into a head-to-head forecasting tournament at the end of the season. To add some variety, the weather station changes every two weeks which continually presents forecasters with new challenges.

Now I admit, my qualifications for making such forecasts are weak. However, I am continually surprised at the amount of resources available for amateur forecasters. Forecasting is a model driven activity, and at least in the United States, the data produced by these forecasting models is freely available. Websites like Wunderground.com amalgamate this information, as well as information gathered from satellites and weather stations. My strategy thus far has been to try to determine where these models are weak, and compensate where appropriate. This semester, I hope to learn more about atmospheric dynamics to improve my forecasts.

If you are interested in learning about the weather or participating in the competition, check out www.wxchallenge.com and see if your university has a team. If you want to go it on your own, check out “The Tools” section on the left side. In future posts, I hope to discuss the nature of weather forecasting and competitive science more generally.

2 Comments

  • Mike Thicke
    Mike Thicke Reply

    Hi Greg,

    This is a really interesting event, thanks for writing about it. There is an obvious comparison to prediction markets here—many of the same incentives seem to be at work. Do you know how competitors (in aggregate?) fare vs. traditional weather prediction methods?

    • Greg Lusk
      Greg Lusk Reply

      In aggregate it is really hard to determine, and I don’t see an obvious way of turning it into a kind of prediction market. I can say that the large models are ranked (NAM, GFS, USL) 885, 830, and 110 respectively overall. So human forecasters typically beat model output, which is expected (since the forecaster has the benefit of the model information).

Leave a Reply

Your email address will not be published. Required fields are marked *