Retrospective of NumPy Weather 11 to 20

After extensive research (took me almost an hour), I gleaned a couple of stylized facts (trivia) about next day temperature prediction by actual meteorological offices around the world:

  • Ninety percent of results falling within a 2 degree error margin for maximum temperatures is considered good. Actually this is measured by applying a 36 month (3 years) rolling average.
  • Seventy percent or more is expected for minimum temperatures. Again with moving average and 2 degree error margin.
  • Generally colder temperatures are harder to predict.

On with the stupid retrospective questions.

What did I do wrong?

I have been told, that it was wrong of me to model the day of year temperature as a quadratic polynomial. I tell you this, because I care. Yes, I do!

What did I do right?

It seems that most models are able to predict the average daily temperature for the next day with an accuracy of around seventy percent (2 degree error margin).

Here is the list of NumPy Weather 11-20 posts:

  1. Examining Autocorrelation of Average Temperature with Pandas
  2. Describing Data with Pandas DataFrames
  3. Correlation between the Weather and Stocks with Pandas
  4. Temperature Autoregressive Model With Lag 1
  5. Autoregressive Temperature Model with Lag 2
  6. Intrayear Daily Average Temperatures
  7. Day of the Year Temperature Model
  8. Modeling Temperature with the SciPy leastsq function
  9. Day of Year Temperature Fit Take 2
  10. Moving Average Temperature Model with Lag 1
By the author of NumPy Beginner's Guide, NumPy Cookbook and Instant Pygame. If you enjoyed this post, please consider leaving a comment or subscribing to the RSS feed to have future articles delivered to your feed reader.
This entry was posted in Uncategorized. Bookmark the permalink.