Saturday, October 7, 2017

Relative Time

'Rel Time'

Leap years don't really cause a problem -- except for 1900.


These errors are due to the use of monthly averages where the differing number of days in each month are not accounted for.

Friday, October 6, 2017

ENSO modeling

An ENSO model has been developed by Paul Pukite and Paul has given me a copy of his spreadsheet.  The spreadsheet was designed to find a best model fit to the Southern Oscillation using one of the Southern Oscillation indexes (SOI); Paul was using the NINO34 index.

I decided to expand the choice of index to include all that I could find on the web.  I found 10 in total.  Five that stretch all the way back to 1880 and five from 1950 to the present.  The five that go back to 1880 are the NINO34 index, the 'Cold Tongue Index,' the SOI_signal index, the SOI_noise index, and the SOI_boma (Bureau of Meteorology - Australia).

I trained the model on the period  from 1885 - 1935 using the CTI dataset.  I then calculated the correlation coefficients for the Model output versus the various datasets over various time periods. Here are the results.



This result was not wholly unexpected the different ENSO datasets are highly correlated - except for one outlier in each group.





Sunday, October 1, 2017

Natural Variability in the GMST record

What is the natural variability in the surface temperature record?

One way to explore this is to take the temperature record and subtract all the known non-random forcings and fluctuations.  What's left over is our best guess at natural variation.

I did this with the HADCRUT global land/ocean temperature dataset from 1881 thru 2010. The known forcing that I used were:

  • CO2
  • TSI
  • LOD
  • ENSO (using two different ENSO indexes for comparison purposes)
The result looks like this:


Using the NINO34 ENSO dataset the standard deviation is 0.105503°C
Using the CTI ('Cold Tongue Index') the standard deviation is 0.107541°C

With global warming since pre-industrial at approximately 1°C, natural variation is not the answer.  It's known forcings -- and CO2 is most of that answer.

We should also check the sensitivity of the analysis to the various datasets used.  In the first analysis we looked at two different ENSO indexes.  A second analysis replaces the HADCRUT gmst data with the BEST gmst dataset.  The results are the same to within a few thousandths of a degree:


One other analysis that might be of interest is to gauge how 'The Tidal Model of ENSO' that Paul Pukite has developed fares in this regard.  I have one of his model spreadsheets at hand and so I went back to the first analysis using HADCRUT and substituted the model output for the CTI:


For this application, the model output works as well as the NINO34 dataset to within 4 ten-thousandths of a degree and actually slightly better than the CTI.