I’ve moved to WordPress: http://bobtisdale.wordpress.com/

Thursday, September 11, 2008

GISS Model E Climate Simulations


I have been looking at the GISS Model E Climate Simulations on and off since May 9, 2008. That’s when Steve McIntyre of ClimateAudit posted Steve Mosher’s recipe for retrieving the data.
When I’ve been able to wrench myself away from the Smith and Reynolds SST data, I’d return to it, looking for a simple way to compare the different radiative forcings. The more I researched what the data represented, the more I realized that simple comparisons between the data sets could not be performed. The reasons: GISS has presented ensembles of climate simulations. The efficacies and climate sensitivities GISS used vary between the climate simulations for each radiative forcing. The underlying unknowns are the methods used to create the ensembles and the impacts of the individual runs on the combined data.

Regardless, three of the climate simulations were interesting enough to justify this lengthy post.

Following the introduction and the explanation of the recipe I used to retrieve the data, I present the output of the GISS Model E Climate Simulations based on the data for all radiative forcings and compare it to GISS global surface temperature (GISTEMP) anomaly data. Then I illustrate the Model E outputs individually for stratospheric (volcanic) aerosols and Total Solar Irradiance (TSI). The outputs of the stratospheric aerosols and TSI runs were so unusual they needed to be presented individually.


The GISS Model E is an update of the Goddard Institute of Space Studies (GISS) atmospheric General Circulation Model (GCM). It is one of the coupled GCMs employed by the IPCC in their 2007 report, AR4, on climate change. It is also the basis for many papers by GISS employees and others that predict the catastrophic results of Anthropogenic Global Warming.

The following is not a discussion of code or the capabilities of the model. It is a quick look at three climate simulations, solely in the form of time-series graphs of global temperature anomaly, that have been created by the GISS Model E for the period of 1880 to 2003.


Called hindcasting, GCM modelers compare the outputs of their models to the instrument temperature record to verify the capabilities of GCMs. This practice is somewhat flawed since the output of the GCM is usually represented by surface air temperature while the global temperature record uses surface air temperature only over land. The rest of the global temperature data sets (those other than the more recent satellite-based data) are made up of Sea Surface Temperature, not the surface air temperature above the seas. It’s a minor difference, but it is a difference.

GISS went through this process with their Model E and documented it in the study by Hansen et al, 2007, “Climate simulations for 1880–2003 with GISS modelE”, “Climate Dynam.”, 29, 661-696, doi:10.1007/s00382-007-0255-8.
Supplemental Information:

The radiative forcings used are discussed and documented by GISS here:

GISS also provides a webpage that allows users to create zonal anomaly and global mean temperature anomaly graphs for the various radiative forcings employed by the model, and to download that data.


If you were to open the preceding link, the title of the page would read “Climate Simulations for 1880-2003”. Scrolling down to Table 1, the second column lists forcing agents (radiative forcings) and the eighth column provides links to a data selection page noted as “Lat-Time”. A click on “Lat-Time” on the final row, “All Forcings combined”, would bring you to a new webpage that allows the selection of “Quantity”, “Mean Period”, “Time Interval”, “Base Period”, and “Output”. These are described below the selection field. Under “Quantity”, I will be looking at Surface Air Temperature. If the “Mean Period” is left at 12 months, the output appears smoothed when compared to a graph with a “Mean Period” of 1. I’ll use a mean period of one month, which will allow me to smooth it for a comparison with GISTEMP global temperature anomaly data, which is also available on a monthly basis. All “Time Intervals” used in this post will be for the full range of years available, 1880 to 2003. And I’ll leave the base years as 1951 to 1980, since I will be using the GISS global temperature anomaly data set (GISTEMP) for comparisons and its base years are 1951 to 1980. For “Output Links”, I will use the “Formatted page w/ download links”, because I will use the data to create my own graphs.

After clicking on “Show Plot”, the next webpage displays the “Zonal Anomalies Plot” and the desired time-series anomalies graph for the selections made. The link to the data in text format for the time-series graph is located in the last line on the page.


Figure 1 illustrates the global surface temperature anomalies created by GISS Model E for all radiative forcings. The appearance is that of a noisy exponential curve with the effects of volcanic eruptions added.

Figure 1

In a comparison to GISTEMP Global Temperature Anomaly data, Figure 2, the Model E Global Surface Temperature Anomalies appear to provide a reasonable approximation of global temperature. When both data sets are smoothed with 37-month running average filters, however, some of the apparent correlation disappears. Refer to Figure 3.
Figure 2

Figure 3

To my eyes, the greatest deviation between the computer output and the historical data occurs around 1940. Global temperature anomalies acquired a “hump” at this time due to a significant multiyear El Nino and due to an anomalous rise in Southern Hemisphere SST that is reflected dramatically in SST graphs of the Indian Ocean, Figure 4, and the South Atlantic Ocean, Figure 5.
Figure 4

Figure 5

This is a clear illustration of the known significant weakness in coupled GCMs, including the GISS Model E coupled GCM, when they fail to consider oceanic variables such as the El Nino-Southern Oscillation, the Atlantic Multidecadal Oscillation, and the North Pacific Residual.


Figure 6 (Raw data) and Figure 7 (Data smoothed with 37-Month Filter) illustrate the GISS Model E global surface temperature anomalies caused by stratospheric aerosols ejected from explosive volcanic eruptions. Working back in time, the Mount Pinatubo (1991), El Chichon (1982), and Krakatau (1883) eruptions are clearly visible. Also visible are the periods of little volcanic activity (from the late 1910s to the 1950s) and of increased activity (from the 1960s to the early 1990s).
Figure 6

Figure 7

Figure 8 is a graph of the volcanic aerosol radiative forcing data used by the GISS Model E. It’s based on the Sato Index, created by Makiko Sato of GISS. The data is available through the above radiative forcing links.
Figure 8

The obvious difference between the two data sets illustrated in Figures 7 and 8 is the slope in the GISS Model E surface temperature anomaly, a slope that doesn’t appear in the radiative forcing data set. Refer to Figures 8 and 9. I realize the data being presented by GISS (as they describe in the “Climate Simulations for 1880-2003” webpage), “are obtained using a coupled atmosphere-ocean model with prescribed time varying forcings. Unless otherwise noted, the results are based on ensembles of 5 runs.” That’s all well and good, but how could the underlying baseline change if the only forcing being examined is volcanic aerosols? There’s a logical explanation, somewhere, but I have failed to find it.
Figure 9

On the GISS “Climate Simulations for 1880-2003” webpage, under the Table 1 heading of “Miscellaneous forcings”, the volcanic aerosols link is listed separately. This implies that the data is for volcanic aerosols only. Are other unidentified data mixed in? In “Climate simulations for 1880–2003 with GISS modelE”, under the heading of “5.1.3 Response to individual forcings”, Hansen et al state, “Stratospheric (volcanic) aerosols, despite their brief lifetime (e-folding decay time ~1 year), have a multi-decadal effect on simulated temperature because of clustering of volcanoes near the beginning of the 1880–2003 period and from 1963 to 1991. Thus volcanoes, specifically the minimal activity during 1900–1950 compared with the late nineteenth century and the period beginning 1963, contribute to the relative global warmth at mid-century, as has been noted previously (Tett et al. 1999; Harvey and Kaufmann 2002).”

If the lack of stratospheric aerosols contributes to the warming from 1900 to 1950, why do surface temperatures in their climate simulations decrease?


Figure 10 illustrates the noisy global temperature anomaly response to Solar Irradiance as presented by the Model E climate simulations. Noisy is the operative word in that sentence.
Figure 10

Much is revealed about the Total Solar Irradiance (TSI) data source when the climate simulation output is smoothed with a 37-month filter. Refer to Figure 11.
Figure 11

For those not familiar with the shape of the curve, that’s similar to the curve of the Lean et al (2000) TSI Reconstruction (Data with Background). This is better illustrated by the graph of the GISS solar radiation radiative forcing, Figure 12.
Figure 12

GISS acknowledges the use of the Lean et al data and its problems in their report “Climate simulations for 1880–2003 with GISS modelE”. They state, “Lean et al. (2002) call into question the long-term solar irradiance changes, such as those of Lean (2000), which have been used in many climate model studies including our present simulations. The basis for questioning the previously inferred long-term changes is the realization that secular increases in cosmogenic and geomagnetic proxies of solar activity do not necessarily imply equivalent secular trends of solar irradiance.” Following that, they go on to explain their reasoning for their continued use of the erroneous TSI data set.

Figure 13 is a graph of the Lean et al TSI data compared to a TSI data set based on the current understanding of solar irradiance variations. A data set based on that current consensus is identified as “Svalgaard”. Comparative TSI data is available in spreadsheet form from Leif Svalgaard’s webpage.
Figure 13

The Lean et al (2000) paper was based on the consensus at that time (2000) that a background component of the Sun had increased significantly over the 20th century. That is no longer the consensus and, as far as I can tell, wasn’t the consensus at the time that GISS submitted their paper “Climate simulations for 1880–2003 with GISS modelE” for publication, which was October 2006.

Note: At comment 41 on the ClimateAudit thread, Leif Svalgaard provides a much better explanation of why the solar minimum trend is now considered to be relatively flat.

Why then would GISS continue to use outdated TSI data? An even better question: why would they continue to present the climate simulations for outdated data?

In Figure 14, I’ve scaled and ranged the GISS radiative forcing data for TSI so that it aligns with the climate simulation for TSI. Note how the model output seems to precede the rise in TSI from around 1915 to 1940, making the climate simulation arrive at its new upper range two decades before the TSI radiative forcing data. Since the GISS “Climate Simulations for 1880-2003” webpage only lists solar irradiance as the radiant forcing for that link, I have to assume it’s the only radiant forcing examined by the model run. One remotely possible explanation I came up with for this phenomenon was the lack of volcanic aerosols during that period caused temperatures to increase faster, but that didn’t make sense since volcanic aerosols didn’t kick in again until the 1960s, and, if that was the case, the continued rise in the TSI data from 1940 to 1960 should have been reflected in the climate simulation. So that explanation didn’t work. Also, why wouldn’t that be included in the climate simulation for stratospheric aerosols?
Figure 14

Then I tried shifting the TSI radiative forcing data forward 21 years to see if an error in the model "relocated" the data. Refer to Figure 15. Eyeballing it, I revised the scaling and ranging to get the “best fit” for the trends, but then the minimums and maximums of the solar cycles didn’t always synch.
Figure 15

It almost appears that, in an attempt to make TSI responsible for the 1910-1940 temperature rise that’s present in the instrument temperature record, the GISS Model E somehow “frontloads” the response. But that doesn't make sense. Much of the rise in global temperature at that time (1910 to 1940) is a function of the rebound in SST in many oceans, after the sudden decrease in SST from the late 1800s to 1910. Refer to Figures 16 and 17, which illustrate North Atlantic and North Pacific SST anomalies.
Figure 16
Figure 17


As an afterthought, I revised the comparative graph of solar irradiance forcing and the Model E output of global temperature response to solar irradiance (originally shown as Figure 14) so that the slopes of the increases in both data correlated as best I could from 1880 to 1940. Refer to Figure 18. I used 4th-order poly trends as reference during the adjustments.
Figure 18

That threw the correlation for the years after 1940 off significantly. Now it seems to display the same inconsistency as the volcanic aerosol comparison.
Figure 19


There are unusual, unexpected, and unexplained GISS Model E responses to solar irradiance and to stratospheric aerosols. Do these anomalous responses carry over to other radiative forcings?

It appears as if the sensitivities to volcanic aerosols and TSI decrease with time or an underlying variable alters the slope of the output. If those are in fact model outputs of only those specific radiant forcings, what would cause that? Or why would they illustrate it?

Looking back at Figure 3, the smoothed comparison of the GISS Model E output versus global temperature anomaly, if the “blend” of all radiant forcings correlates with the instrument temperature record without considering the impacts of oceanic variability, then the blend of all radiant forcings has to be incorrect.

1 comment:

Paul Biggs said...

TSI is one measure of solar activity. Leif points out that his work either means the sun has little effect on climate, or it is even more sensitive to small solar changes than we currently think. TSI is the same now as it was 107 years ago.

Looking at yearly averaged sunspot numbers as a measure of solar activity - we can see the Maunder, Sporer and Dalton minimums, plus a big rise to the 1950s, followed by a fall that still remains above all but the 1950s peak - rather like a pan of water on the hob heated until it boils, then turned down to simmer.


Paul Biggs


Tips are now being accepted.

Comment Policy, SST Posts, and Notes

Comments that are political in nature or that have nothing to do with the post will be deleted.
The Smith and Reynolds SST Posts DOES NOT LIST ALL SST POSTS. I stopped using ERSST.v2 data for SST when NOAA deleted it from NOMADS early in 2009.

Please use the search feature in the upper left-hand corner of the page for posts on specific subjects.
NOTE: I’ve discovered that some of the links to older posts provide blank pages. While it’s possible to access that post by scrolling through the history, that’s time consuming. There’s a quick fix for the problem, so if you run into an absent post, please advise me. Thanks.
If you use the graphs, please cite or link to the address of the blog post or this website.