Plugging a Century of Climate Data into Eureqa

Posted on 7th December 2009 by Ryan Somma in Enlightenment Warrior,Geeking Out

Friday I posted some links to the Cornell’s free data analysis software Eureqa. The video tutorials take a collection of data points from a swinging pendulum over time and then have the Eureqa software determine the function that best explains its wave using an evolutionary algorithm. I played with the software a bit this weekend, trying out various datasets, when I remembered the wealth of global temperature data on the World Wide Web.

With HadCRUT’s, NASA’s, UAH’s, and RSS’s data available online, I easily downloaded and imported the data into an excel spreadsheet, which made it easy to copy and paste into Eureqa. The HADCRUT data was the most thorough and extensive with over 21,000 points of data extending from 1850 to the present. It was also the most straightforward, which made it the most accessible for a non-statistician like myself. I had Eureqa search the average of the data points a through i against time for functions that best explain the relationship.

(a+b+c+d+e+f+g+h+i+j)/11 = f(t + (t2*.01))

Below are the fit functions Eureqa came up with from most fit to least. The function is at the top of each image with the result drawn over the data points below it:

Solution 01
Solution 01
Solution 02
Solution 02
Solution 03
Solution 03
Solution 04
Solution 04
Solution 05
Solution 05
Solution 06
Solution 06

Each one of these functions depicts a clear warming trend over the last 160 years of measurements that looks a whole lot like Hansen’s instrumental record of global average temperatures, inferred through software lacking any bias or preconceptions. An acceptable criticism of this data is that it only goes back to 1850; however, any climate skeptic who believes this warming trend is only a cyclical phenomenon is welcome to plug the deluge of ice core data into Eureqa and post their findings.


See Also:

New Scientist Magazine: Why there’s no sign of a climate conspiracy in hacked emails

3 Comments

  1. Very cool; nice job! This software looks extremely useful. I can’t wait to start experimenting with it with my own data sets.

    BTW, a great source of publicly available data sets is here:

    http://archive.ics.uci.edu/ml/datasets.html

    It is a very useful archive of data sets for testing various data mining, curve fitting, and machine learning techniques.

    Comment by Dave — December 8, 2009 @ 8:54 am

  2. Sweet! Thanks for the lead. I’ve been playing with the Processing software lately, so this will be good for experimenting with that too:

    http://processing.org/

    Data.gov’s data tends to be harder to import, but it’s turning into a fantastic resource too.

    Comment by ideonexus — December 8, 2009 @ 10:16 pm

  3. [...] play with and interact with, Google Analytics, (My own recent examples: Daqarta Audio analyzer and Eureqa Data inference program) [...]

    Pingback by ideonexus.com » Blog Archive » Science Online 2010: Connections with mathematics and Programming Through Modeling — January 18, 2010 @ 7:47 pm

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.