“James Hansen told Congress on Monday that the world has long passed the “dangerous level” for greenhouse gases in the atmosphere and needs to get back to 1988 levels.

He said Earth’s atmosphere can stay this loaded with man-made carbon dioxide for only a couple more decades without changes such as mass extinction, ecosystem collapse and dramatic sea level rises.

“We’re toast if we don’t get on a very different path,” said Hansen, director of the Goddard Institute of Space Sciences who is sometimes called the godfather of global warming science. “This is the last chance.”

My reference to Scenario B in my previous post in the paragraph excerpted below should be Scenario C. Sorry for any confusion this caused.

Also the shortened time period graph for Scenario B shows an almost immediate leveling off effect that seems to go counter to Hansen’s later assertions that stopping GHGs at current levels would have a residual gain of 0.6 degrees C over a century’s time. The extended graph does show a 0.1 to 0.2 degree C blip for Scenario B.

What you say, Lucia about how the different length runs came about in almost reverse order to their stated likelihood makes sense. I would like to throw in my previous conjecture that from the time Hansen et al ran the Scenario A in the BAU mode that BAU changed to be more like Scenario B. Scenario C has a longer run time than B and I would guess might well have been run before Scenario B and perhaps run to show the significant differences that one could obtain from BAU and mitigation. Though Hansen was not as big a shot as now back then I think he was as sensitive to policy issues then as he is now.

Also the shortened time period graph for Scenario B shows an almost immediate leveling off effect that seems to go counter to Hansen’s later assertions that stopping GHGs at current levels would have a residual gain of 0.6 degrees C over a century’s time. The extended graph does show a 0.1 to 0.2 degree C blip for Scenario B.

In the excerpt from this paper, vintage later half of the 1990′s and linked below, I was surprised to learn that as you noted in your post, Lucia, that the longer runs on the Amdahl V/6 would have been no small or easy task for Hansen et al using a 3D model. The linked paper describes the modification that MIT workers made on a GISS 2D model (and the output results) in order to do scenario runs of longer lengths. And the paper is not describing the computer capabilities of the early 1980s when Hansen et al made their scenario runs but the capabilities a decade later.

A variety of scenarios for changes in greenhouse gas (GHG) concentrations also have to be considered. As a result, a significant number of climate simulations, each for 50 – 100 years, are to be carried out. This would be impossible with the use of GCMs, due to their enormous requirements of computer time, even on the most powerful super computers now available. An alternative approach is to use simplified models. The two-dimensional (2-D) statistical/dynamical model developed at the Goddard Institute for Space Studies (GISS) is 23 times faster than the GISS GCM with the same latitudinal and vertical resolutions (Yao and Stone, 1987).

In this same linked paper I was a bit taken aback by what seems a rather arbitrary change in the RH level at which precipitation is allowed (needed to compensate for the low spatial resolution used for the model) and that it can change the cloud feedback from negative to positive.

In the original version of the 2-D model, condensation occurs when relative humidity reaches 100%. As a result, the amount of precipitable water in the atmosphere obtained in the simulations with this version turns out to be larger than the observed value. At the same time, even in some GCMs with low horizontal resolution, condensation is allowed to occur in partly saturated areas, in order to take into account subgrid-scale variations of relative humidity. Such an approach seems to be even more appropriate in a zonally averaged model: moreover, a similar approach is used in the parameterization of moist convection (Yao and Stone, 1987). Therefore, the value of hcon = 90% has been chosen as the criterion for condensation. This small change has a very profound impact on the model’s sensitivity, namely, if hcon = 100%, the model produces a negative cloud feedback; however, when hcon = 90%, the cloud feedback becomes positive.

But yes, it may turn out that one time scale is dominant for certain purposes– and this may be one of them!

But, the planet doesn’t have one time constant for everything.

But it may turn out that one is dominant. As a rough approximation, you could say that there’s an energy flux associated with each time constant, and one (I’m thinking the ocean mixed layer) may turn out to account for the vast majority of the flux, in which case the one lump model may actually be good. The problem is knowing this in advance. Once you start fiddling with multiple knobs, as I’m sure you well know, you find that a lot of very different combinations end up producing very similar and plausible results.

Lucia, since you are working with global climate time constants can you explain what you think Hansen et al mean when they say that, since for the control run they wanted to permit several integrations over several time constants,

I think they have a time constant for the mixed layer, and ran the control run for a bunch of those. My model is simplified and only has one time constant for “everything”. (Actually, I’ll explain how this relates to more complicated problems if my model doesn’t break down. But, the planet doesn’t have one time constant for everything.)

A general question on my reread of Hansen 1988 is why the graph showing the extended curves for Scenarios A, B and C show a track to 2060 for the least likely scenario, A, and to only to 2027 for the most likely scenario, B, and to 2037 for Scenario C?

I’ll speculate, and you can ask Hansen how close I am.

I suspect the true answer has nothing to do with physics, or even what Hansen thought was most likely. Sceneario A ran longest because of the way projects and programs get funded.Let’s begin with the fact, that Hansen’s ran Scenario A first. (It says so in the paper.)

Of course you think of Hansen as a big-wig now, but this was not always so. My guess is Hansen and his group ran A as a numerical experiment, in the background, using discretionary resources which they’d begged others to let him use. ( You’ll read they were using an Amdahl V6, running in the background of other jobs.) These exist at national labs and government agencies, but there is lots of competition because all scientists want to do ‘fun’ stuff.

Hansen also wished to write further proposals. (In fact, saying you hope to write further externally funded proposals is part of the justification for getting access to internal discretionary resources.) To write a proposal that was likely to be funded, there can be some effect is necessary to convince funding agencies for money to do further experiments. Since you don’t know what results you’ll get until you run them, scientiest always pick a “worst- plausible” case scenario. That’s A. To truly convince there might be a problem, they needed to extrapolate out to a ridiculous time frame. So, they ran to 2060.

As these were running, and shortly afterwards, Hansen presented results, discussed things and persuded others to give him snippets of additional funding to run more numerical experiments. Taking guidance based on questions, he ran the other cases for comparisons. Obviously, it made sense to try for some more plausible scenarios.

Still knowing that extrapolation is not all that likely to give good results, they decided not to run these out to 2060 before writing the paper. (Had they done so, the paper would probably be Hansen 1992 or something. )

I’d also guess at some point, the group eventually got real funding and turned their attention and resources to improving the model rather than just running the thought experiments out to infiniti and beyond. Running the thought experiments forever might have been nearly impossible in any case: That old Amdahl computer may have been on its last legs, scheduled to be retired etc. So, at that point, if the computer vanished, and, given change compilers etc. it might have been a ridiculous amount of work to continue on the older code.

Lucia, since you are working with global climate time constants can you explain what you think Hansen et al mean when they say that, since for the control run they wanted to permit several integrations over several time constants, they do not allow heat exchange across the level defined by the annual maximum mixed layer depth? They then say that the isolated mixed layer response time is 10-20 years for a climate sensitivity of 4 degrees C for a doubling of CO2. What would that make the time constant for the control run based on the conditions for GHGs prevailing in 1958.

On looking at the control run results in the 1988 Hansen paper one has to wonder about finding a climate warming signal in there — and evidently with the time constant artificially reduced. I am not sure what this all implies. Lucia what does the control run for your model show?

The Hansen paper talks of another short cut, that would affect the time constant, I assume, for the scenario runs, in dealing with the annual mean temperatures by restricting the heat exchange depth into the oceans below the annual mixed layer depth.

A general question on my reread of Hansen 1988 is why the graph showing the extended curves for Scenarios A, B and C show a track to 2060 for the least likely scenario, A, and to only to 2027 for the most likely scenario, B, and to 2037 for Scenario C?