2011-09-12 01:52:46Climate sensitivity -- some basic questions
John Hartz
John Hartz
john.hartz@hotmail...
98.122.98.161

1. Is climate sensitivity an input to, or an output from, a climate model? (Discussion of this issue was started on Yes Dimwits, Everyone Agrees That Man Causes Climate Change -- Steve Goddard -- Real Climate )

2. Does climate senitivity vary by base year CO2 concentrations? In other words, is the climate sensitivity computed for a base year when atmospheric OC2 was 350, the same as climate sensitivity computed for a base year when atmospheric CO2 was 400? 

2011-09-12 03:06:04
MarkR
Mark Richardson
m.t.richardson2@gmail...
134.225.187.197

Climate sensitivity is an output of the more complicated climate models - the things you or I think of as a 'climate model' that needs supercomputing power and weeks of process time.

 

In simple energy balance models it can be chosen. You might choose a time it takes for Earth to 'react' to changes and also decide which heat capacity is relevant (which typically comes down to: is it 500 metres of the ocean, or 100, or 1,000?), those 2 figures determine a sensitivity.

 

 

Climate sensitivity doesn't very much for <i>small</i> changes in starting conditions. I suspect that for 300 and 400ppm starting points the values will be similar. There's no way we could pick out different values from the uncertainty if we started at 300 or 301 ppm for example, but there is some point where climate sensitivity changes.

'Regime changes' might be, for example, once most Arctic ice and summer NH snow has melted. The 'albedo feedback' contributes to the climate sensitivity, and once all the snow & ice is gone you can't boost warming by melting any more (because it's gone!). That's the most obvious example I could think of...

2011-09-12 04:23:31
Alex C

coultera@umich...
67.194.20.210

[EDIT - please read further responses, esp. starting from time tag 5:02 AM Sept. 12]

1: It depends on what model you are using.  Climate sensitvity must be derived somehow, and is found either through running complex models against known data or deriving the figure statistically from data too.

In order to have any sort of climate prediction, though, you must use the same basic relationship between forcing, temperature, and climate sensivity, but in reverse.  You cannot possibly predict anything without first giving your model a climate sensitivity to work with, in this case it would be an input.  Again, it depends on what you're using the model for - if it's to determine sensitivity then of course it's an output, but if it's to give an idea of what future temperatures will be, then it is an input.

As it pertains to Hansen's 1988 model, climate sensitivity was NOT an output, he states right in the paper that the model used a sensitivity of 4.2˚C per forcing due to doubled CO2, which means it was an input.

 

2: Yes and no - sensitivity estimates are determined by comparisons against proxy/instrumental data, but also through the usage of some sort of relationship between CO2 and forcing.  Myhre et al 1998 give the currently used relationship, which relates CO2 and forcing through ∆F = 5.35ln(C/C_0).  This is not, however, a derivation from first principles nor is it applicable to every CO2 concentration.  It is a best fit for rdiative transfer model output, and works for concentrations ranging from about 200ppmV to several hundreds.  Over the relevant range as it pertains to today's concentrations and feasible 21st century concentrations, in other words.

2011-09-12 04:31:14Alex C
John Hartz
John Hartz
john.hartz@hotmail...
98.122.98.161

Re the 1998 GIS forecasts, how did Hansen et al drive a climate sensitivity of 4.2C? When and how did they realize it wasn't correct?

2011-09-12 04:35:33Mark R and Alex C
John Hartz
John Hartz
john.hartz@hotmail...
98.122.98.161

Which SkS article, in your opinion, best explains what climate sensitivity is and how it is computed?  

2011-09-12 05:02:45Badger
Alex C

coultera@umich...
67.194.20.210

As I read more into Hansen et al 1983 and Hansen et al 1984, I see I will have to modify my statements, but I think this would be a more accurate summary than what I previously posted: the model in question is a multiple-component simulation of grid boxes with a given resolution (8˚x10˚), and I think, before I read into Hansen 1984 which actually describes the sensitivity calculation (Hansen 1983 describes the model), that the sensitivity is derived from running the model against observed initial conditions and seeing how it compares against the observations of several decades; more on this later.  If that is true though, running the model forward past observations would still require, I think, the same basic invserion of its application.

2011-09-12 05:04:49Badger, responding to question...
Alex C

coultera@umich...
67.194.20.210

I am not sure if any particular article comes to mind.  There are various ways to estimate climate sensitivty, some using models simulating physical processes and some using just statistics and raw data - I do not think we have a single article explaining either.

2011-09-12 05:46:45Ok, so if someone would like to double-check my understanding here...
Alex C

coultera@umich...
67.194.20.210

As I gather from Hansen et al 1984, the way that they calculated climate sensitivity was by running a scenario CO2 increase (2xCO2), which in turn affected the processes simulated in each grid box being modeled, and which in turn resulted in some change in surface temperature in each box, temperatures which could be averaged for a global temperature response.

So, the climate sensitivity is a model output here, and wouldn't really be a model input in Hansen et al 1988.  There is no inherent variable for lambda that is actually in the model, it is derived from output of temperature and compared to the CO2.  So, running the model will always result in a climate sensitivity of 4.2˚C, there is no way to modify the model (well, no way that is very simplstic as changing 4.2 to 3.4) to get to a more realistic climate sensitivity and thus more realistic projections.

This indicates an inherent problem with the model, it will always overpredict temperatures.  In order to fix this, either the processes involved in the model climate must be better modeled or there are some factors that are not included in the simulated climate that exist in the real climate.

Another possible factor is the assumed relationship between CO2 and forcing.  These papers were released before Myhre 1998, would that mean that the forcing connection between CO2 and temperature would be different?  The paper (1984) describes a calculation of climate sensitivity using both a 2xCO2 scenario, and also a 2% increase in solar output.  If I recall correctly, such an increase in solar output would be equivalent to an increase in radiative forcing of just over 4W/m^2, which is also the response using the old CO2-->forcing relationship (∆F = 6.3ln(C/C_0).

Their discussion actually supports this some, as you can see from the "Global Mean Heat Balance and Temperature" section in (1984), the CO2 scenario has a forcing of about 4 W/m^2*.  The solar scenario translates into a forcing of 4.8W/m^2 - we know, though, that a doubling of CO2 does not cause that high of a forcing.  So, the temperature response was high because the forcing was high - the model might have the right idea when it comes to sensitivity, just that we would have to rescale it to get a temperature better fitting with a realistic forcing scenario: if the relationship determined from (1984) is 4.2˚C/4.?W/m^2, then using a better forcing estimate for 2xCo2, Myhre's 3.7W/m^2, would get us a value of anywhere between 3.4 and 3.9˚C/2xCO2.

[* This is strange, the paper Hansen cites for the ~4W/m^2 figure shows an instantaneous forcing of about 4.3W/m^2, though his 1984 model simulation gives a much smaller value.  This might have to do with the time resolution in his 1984 model - single years, not months - but I am not sure.  Or, I'm not properly making the distinction between net forcing and forcing at the surface, but wouldn't you want to know the forcing at the surface?]

 

 

I think we need to update our Hansen 1988 discussion, our conclusions are, essentially, wrong.  Either the model is wrong in its functions, which is OK as more modern models are of course better structured, or the model is OK at predicting cimate sensitivity just that the forcing input was too high for what we thought we were giving it (i.e. we call it 2xCO2, but in reality the W/m^2 value we gave is actually due to, just for the sake of getting a number out there, 2.5xCO2 - some value higher than 2).

2011-09-12 05:53:48Here are the papers BTW:
Alex C

coultera@umich...
67.194.20.210

1981, 1983, 1984, and 1988.

2011-09-12 05:56:25Alex C
John Hartz
John Hartz
john.hartz@hotmail...
98.122.98.161

Do your research findings also impact Dana's article, A detailed look at Hansen's 1988 projections?

2011-09-12 05:59:10
Alex C

coultera@umich...
67.194.20.210

Yes.  The conclusion is essentially correct, but for the wrong reason and only if we assume most of the discrepancy is due to the discrepancy in how Hansen handled CO2 v. forcing back in 1988.  What I do know now is that you can't simply say that the model sensitivity ought to have been 3.4˚C, as there is no way to make that adjustment within the model itself.

If the above assumption is NOT correct, then we ought to consider both looking into Hansen 2006 and owning up to the potential that yes, the model had flaws in it that extended beyond the way forcing was handled.

I think looking at Hansen 2006 in more detail either way would be prudent.  Maybe Hansen himself could provide input.

I'd greatly appreciate input from others here who might have more experience with models.

2011-09-12 06:03:06Eureka!
John Hartz
John Hartz
john.hartz@hotmail...
98.122.98.161

I just came across Dana's article, "A detailed look at climate sensitivity" posted Sep 8, 2010.  Now I have something to sink my teeth into.

2011-09-12 06:05:58
Alex C

coultera@umich...
67.194.20.210

Edit: ha, sometimes I get away from myself: I stated before the conclusions are essentially wrong, then said they are essentially correct.  To choose one, I would say our conclusion, that the model is overestimating warming due to a sensitivity issue, is wrong - it is a CO2--> forcing translation issue, which is independent of the model's functionings.  It is just telling us the temperature response to a higher forcing than what we wanted to tell it.  So, we (er, Hansen) misdefined what the sensitivity is even though the model, generally, relates forcing and temperature accurately.

OR, the model really does have problems and I don't have the slightest clue what they are, though I myself am banking that the former explanation is the most important here.

2011-09-12 06:17:44Alex C
John Hartz
John Hartz
john.hartz@hotmail...
98.122.98.161

Thanks for delving into this matter. Most of what you are saying about Hansen's 1998 forecasts is above my pay-grade. I will therefore let you and the other SkS authors with expertise in these matters sort things out.

Wouldn't it make sense to forward A detailed look at Hansen's 1988 projections  to Hansen for review and comment? 

2011-09-12 06:32:33
Alex C

coultera@umich...
67.194.20.210

As I said I think getting Hansen's input would be best, but we ought to sort out a better understanding of where we think we might be wrong, if we are even wrong, and come up with some context to send him our article besides just "here ya go, any pointers?"  We should be able to lead him with some initial discussion.

FWIW I don't have experience with models either, I only recently started college.  My pay-grade is nonexistent right now :-)

2011-09-12 07:01:47
MarkR
Mark Richardson
m.t.richardson2@gmail...
134.225.187.197

The 'climate sensitivity' reported is generally the 'equilibrium' climate sensitivity. One way of telling this is to run the model for a bit, double CO2 and then run it until the global mean temperature stops changing much. IPCC definitions are <a href="http://www.ipcc.ch/pdf/glossary/ar4-wg1.pdf">here</a>, but the above is the most common value reported.

 

 

In modern GCMs (or Earth System Models - ESMs) you put in the physics and parametrisations as best you can, then you run the model. Or you run it lots of times for different values of physical parameters that you don't know exactly but know the possible range...

Then you can calculate the sensitivity either by letting it run for ages or by calculating the transient sensitivity or by studying the output fields to get the effective feedback parameters and therefore sensitivity (or effective time constant/heat capacity I guess).

So in the complex models, it is not an input.

 

 

In simpler models it can be.

2011-09-12 07:29:39
dana1981
Dana Nuccitelli
dana1981@yahoo...
69.230.106.125

The post in question doesn't say - or at least doesn't intend to say - that Hansen's model sensitivity 'should have been' 3.4°C.  It says - or at least means to say - that if the model's sensitivity had been 3.4°C, its projections would have been accurate.  This is the same conclusion Gavin Schmidt arrived at in a post on RealClimate.

"Thus, it seems that the Hansen et al ‘B’ projection is likely running a little warm compared to the real world, but assuming (a little recklessly) that the 26 yr trend scales linearly with the sensitivity and the forcing, we could use this mismatch to estimate a sensitivity for the real world. That would give us 4.2/(0.26*0.9) * 0.19=~ 3.4 ºC."

There are physical reasons why Hansen's model sensitivity was higher.  Probably issues with accurately representing ocean processes, for example.  It's not as simple as just changing the model's sensitivity to 3.4°C.  The point of that part of my post is to say the same thing as Gavin - that we can use the model to estimate a real-world sensitivity of 3.4°C.

2011-09-12 09:13:44Dana
John Hartz
John Hartz
john.hartz@hotmail...
98.122.98.161

Your last post brings me back to my original question. Why would anyone today choose to use the model that Hansen et al used to make their 1998 forecasts when he/she knows that the model has been improved?   

I believe that the emphasis should be on the fact that the 1986 GISS model has been improved during the intervening years. Their 1998 forecasts are basically ancient history. What matters most now is what the GISS model forecasts today. Your article did not cover this ground. That is why I suggested a sequel would be in order.

 

 

2011-09-12 09:17:13
dana1981
Dana Nuccitelli
dana1981@yahoo...
69.230.106.125

Nobody today does use GISS' 1988 model.  The post isn't about models today, it's about the accuracy of the 1988 model, and what it tells us about climate sensitivity.

2011-09-12 10:11:27
Alex C

coultera@umich...
67.194.29.95

My bad then for misunderstanding Dana.

Badger, the original context that you first asked about the 1988 paper was that Goddard brought it up in order to demonstrate that climate sensitivity is not very high - which we know.  Only deniers stick to failed predictions of models plagued by problems we have already solved.  The climate sensitivity problem with Hansen's paper, whether you take a more direct approach to solve for the radiative forcing effect of CO2 or simply scale down the sensitivity the model predicts doesn't matter, the result is a sensitivity around 3˚C, which still very easily leads to a dangerous future if current GHG concentration increases persist as they do.  It's a major blunder on Goddard's part that is obvious to anyone who isn't wearing their ideological blinders.  Oh well though.

2011-09-12 10:13:54Dana
John Hartz
John Hartz
john.hartz@hotmail...
98.122.98.161

The average person reading the article doesn't have a clue about whether the 1984 GISS model has been updated or not. That's my basic concern.    

2011-09-12 15:49:01
dana1981
Dana Nuccitelli
dana1981@yahoo...
69.230.106.125
Well badger, you have to bear in mind that's the advanced version of the rebuttal. The average person should probably stick to the basic or intermediate versions [I think we have all 3 for that particular myth].