Monday, January 28, 2008

Oscillations and Predictions

SEARCH BLOG: WEATHER

I've noticed that frequently the actual weather does not cooperate with weather forecasts. Predicting weather is a tricky business, no doubt about that. You are dealing with jet streams and mixing fronts and traffic conditions on I-75. So it is little wonder that you end up with results like this:



By my calculations, the actual low at about 8 AM was 28° F below the predicted low.
Maybe the low means the low after the high rather than the low before the high. Communicating what is meant is important, too.
Regardless, the point here is that two issue are before us:
  1. Variability [oscillations] make trend determination dependent on starting point
  2. Calculating actual averages and variation from "normal" are dependent on starting point
In this little example, if I choose 7° F as my starting point, the temperature trend for 4 days is clearly upward while the 4-day average is below normal. If I choose 42° F as my starting point, the temperature trend is clearly downward while the 4-day average is above normal.

I've noticed a lot of argument lately about temperature trends...
  • since 1000 BC...
  • since 1200 ...
  • since 1650 ...
  • since 1880 ...
  • since 1930 ...
  • since 1985 ...
  • since 2000 ...
Strangely, they all tell different stories... and they seem to prove that everyone is both right and wrong. We are warming... or not.

I've also noticed that it doesn't matter if predictive models are accurate or represent what really has happened or is happening... we will use them anyway.