GISTEMP: Base Period Selection

March 8, 2009
Previous articles gave an overview of GISTEMP and described how the baseline was established and validated.  That baseline is used for comparison purposes during the testing of changes to the GISTEMP software and/or data files.  This article examines the effects of using 1951 - 1980 as a base period on the surface temperature analysis produced by GISTEMP. 

All of the global land and sea temperature anomaly series used by the IPCC are created with reference to a base period of 1961 - 1990 and all of the major temperature series produced by the University of East Anglia Climate Research Unit and England's Met Office, Hadley Centre, such as HadCRUT3, HadSST2, and CRUTEM3 use 1961 - 1990 as a base period.  GISTEMP produces the only major temperature anomaly series that uses 1951 - 1980 as a base period.

Modifying GISTEMP to use 1961 - 1990 as the base period is very easy.  Only two files require changes to compute the anomaly series with a different base period - to.SBBXgrid.f and zonav.f. - and it is only necessary to modify a single PARAMETER statement in each file.  After making those changes, the run_gistemp and run_compare scripts were executed.  The detailed output from run_compare was archived and the summary output is as follows:

		STEP3 Global anomalies - Found 2323 total differences, 2323 >.01C
		Higher than baseline 0 times, lower 2323 times
		Differences >.01C higher 0 times, lower 2323 times
           
		STEP3 NH anomalies - Found 2323 total differences, 2323 >.01C
		Higher than baseline 0 times, lower 2323 times
		Differences >.01C higher 0 times, lower 2323 times
           
		STEP3 SH anomalies - Found 2323 total differences, 2323 >.01C
		Higher than baseline 0 times, lower 2323 times
		Differences >.01C higher 0 times, lower 2323 times
           
		STEP3 Zonal anomalies - Found 1688 total differences, 1654 >.01C
		Higher than baseline 0 times, lower 1688 times
		Differences >.01C higher 0 times, lower 1654 times
           
		STEP4_5 Global anomalies - Found 2323 total differences, 2323 >.01C
		Higher than baseline 0 times, lower 2323 times
		Differences >.01C higher 0 times, lower 2323 times
           
		STEP4_5 NH anomalies - Found 2272 total differences, 2066 >.01C
		Higher than baseline 0 times, lower 2272 times
		Differences >.01C higher 0 times, lower 2066 times
           
		STEP4_5 SH anomalies - Found 2323 total differences, 2323 >.01C
		Higher than baseline 0 times, lower 2323 times
		Differences >.01C higher 0 times, lower 2323 times
           
		STEP4_5 Zonal anomalies - Found 1706 total differences, 1552 >.01C
		Higher than baseline 129 times, lower 1577 times
		Differences >.01C higher 27 times, lower 1525 times

As can be seen from the summary, every anomaly value in five of the output files was significantly affected and nearly all of the values in the other three output files also had large changes.  In total, 17,281 of the 17,504 anomaly values were changed by modifying the base period.  The largest effect was in the Southern Hemisphere data where changing the base period to 1961 - 1990 caused an average .104C cooling in the land only data and .095C in the land and sea data as compared to the base period 1951 - 1980.  The smallest effect was on the Northern Hemisphere land and sea data where there was an average cooling of .042C compared to the baseline.  The effect on the global land only temperature record was an average cooling of .093C as compared to the baseline.

In order to produce a graphical representation of the effects of the different base periods, two additional scripts were written.  The first is named extract_columns and it extracts whatever column is requested from two temperature anomaly files and produces output in tab delimited format suitable to import into a spreadsheet program.  Each line of output contains the year and the values for the specified column from both files.  The second script is named run_extract and it executes extract_columns on each of the anomaly files produced by a GISTEMP test run and our baseline anomaly files to build a new set of files containing the annual and global temperature anomalies in tab delimited format.  The new files were imported into an Excel spreadsheet and the data was used to create charts graphically depicting the effects.  On the charts, the dashed blue and red lines are the actual temperature anomaly values for the specific test case and the baseline respectively, the thick blue and red lines represent the five year moving averages, and the thin yellow line indicates the difference between the test case and the baseline anomaly values.

As shown in the chart below for the Global Land Only Temperature Anomaly series, the slope of the temperature variation is virtually identical no matter which of the two base periods is used, but they are offest by about .1C.  The offset difference is also evident if you look closely at the global land surface temperature graphic on the GISS web site, which uses the 1951 - 1980 base period and peaks at just under .7C in 2005, and then compare it with the GIS temperature line in Figure 3-1 of the IPCC Fourth Assessment Report, which uses the 1961 - 1990 base period and peaks at just under .6C in 2005.  It is our understanding that the IPCC requires the use of a 30 year base period where temperature changes average out to near zero for their assessment reports, which is why 1961 - 1990 was chosen.  Clearly, the base period 1951 - 1980 does not meet that criteria or there would be virtually no discernable differences between the two lines on the chart.

The Global land and sea chart has the same pattern with an offset of about .07C as compared to the baseline.  All of the other files produce very similar looking charts with the Southern Hemisphere having the largest offset of .104C in the land only and .095C in the land and sea data.  All of the charts are available in the Excel spreadsheet.

Conclusion:
We refuse to speculate, at least here, on why the leadership at GISS has made a conscious decision to continue using 1951 - 1980 as a base period instead of conforming to the now standard 1961 - 1990 base period.  We will just state that the net effect is that land temperature series produced by GISTEMP appear to be an average of about .093C warmer and land/sea series appear to be an average of about .068C warmer than they otherwise would and that it is a simple matter of changing one line of code in two Fortran programs to use a different base period.  However, it must be noted that James Hansen, the director at GISS, is a very outspoken global warming alarmist.  He is also certainly aware of the effect that decision has on the data and charts produced using GISTEMP - especially since GISS had to use the 1961 - 1990 base period to produce the results included in the IPCC assessment reports.