I have downloaded the GISS GCM Model II  and have begun to investigate it. Since all the predictions of future AGW depend on the results from such models I decided I really needed to understand just how they work. The model is written mainly in Fortran (pre Fortran 77 !) which is fine by me as I programmed in Fortran for over 20 years. The code is fairly well commented but is not particularly well structured and is full of goto statements. This was NASA’s GCM model that was used up until about 2000 when it was superceded by the latest model IE which uses a finer grid. However model II is the only one you have a realistic chance of running on a desktop PC. The 3-D grid covering the Earth consists of 8×10 degree lat lon grid with 9 atmospheric layers, 2 ground/ocean layers. The solar input energy is modelled using the orbital parameters with a single value for TSI (more on this in the next post) with a step interval of 1 day. Radiative transfer through the atmosphere uses profiles for clouds/water vapor, CO2 etc. and a model for ocean and land albedo is included. Much of this is summarised nicely in a (Masters) thesis I found on the Internet by Melissa Kelly.
To get up and running fast however, I downloaded the front end GUI and analysis system EdGCM released by University of Columbia. This package comes ready to run and a limited number of input parameters that can be tuned with free graphics and analysis software (EVA) included. Without this it would take months to understand how to setup and run the model with the correct input parameters. The code consists of about 15,000 lines of Fortan, which is large but not excessive. The handling of clouds and water feedback is hardwired into the code and cannot be controlled by input parameters in EdGCM
The first model I ran to test the system was one of the pre-packaged models coming with EdGCM called “Doubled CO2”. This model “measures” climate sensitivity to a doubling of CO2. It starts from a 1958 climate and with CO2 concentrations of 314.9 ppm and then introduces an instantaneous doubling of CO2 levels to 629.8 ppm. The model then reacts to the extra radiative forcing running forward to 2100. The total temperature rise is interpreted as the equilibrium climate sensitivity. I have installed the model on a MAC mini running OS10.6.8 with a 2.4Ghz Core2 processor and 2 Gbytes of memory. The full model simulation took 33 hours to complete, generated over 2 Gbytes of output. The resultant overall surface temperature response is shown in figure 1.
The following conclusions can be drawn.
1. The climate sensitivity of Model II is ~4.4 deg.C to a doubling of CO2
2. The inertia of the Earth’s climate system to a sudden change in forcing is predicted to be about 50 years in Model II
In the 2001 IPCC report the range of climate sensitivities of a range of models was between 2.2 and 5.6 deg.C. MGIS model II is therefore at the upper end of this range. By IPCC 2007 the range of sensitivities had reduced to 2.1 to 4.4 deg. C The major uncertainty is the water feedback terms, especially clouds. The cloud model used in Mode II is simplistic :- each grid point is assigned to be cloudy or clear based on whether a random number is greater than the calculated saturation. It is generally accepted that GISS Model II overestimates climate sensitivity. The only model with greater sensitivity is that of the UK Met office. Without feedbacks the model would produce 1.2 deg.C warming to a doubling of CO2.
Of course I would like to run the model with even negative water feedbacks effects, but I don’t see yet how to do that. It may be that the fortran code needs modifying.
1. Hansen, J., G. Russell, D. Rind, P. Stone, A. Lacis, S. Lebedeff, R. Ruedy, and L. Travis, 1983: Efficient three-dimensional global models for climate studies: Models I and II. M. Weather Rev., 111, 609-662, doi:10.1175/1520-0493(1983)111
2. A Survey of Climate Sensitivity and Response Rates in EdGCM, Mellisa Kelly, Dickinson College, Carlisle PA