Talk:Global climate model

From Wikipedia, the free encyclopedia

This climate change-related article is part of WikiProject Climate Change, an attempt to build a comprehensive and detailed guide to articles on Wikipedia related to climate change and global warming. You can help! Visit the project page or discuss an article at its talk page. We are focusing on Global climate model.

Contents 1 Improvement 2 Driver decisions 3 To AOCGM or not to AOGCM 4 Not with my approval 5 rm flat planet 6 rm history of conv param 7 rm van Dorland 8 Parameters 9 Error Found in Models 10 Para to talk 11 Techncal template 12 Predictive value? 13 Relation to weather forecasting 14 S+H 2003 15 grids

[edit] Improvement

WMC, "come on" is not informative as to what is wrong with the following. [1] You've seen the source quotations. (SEWilco 14:53, 20 May 2005 (UTC))

• "Thousands of climate researchers around the world use climate models to understand the climate system. There are thousands of papers published about model-based studies in peer reviewed journals - and a part of this research is work improving the models. Improvement has been difficult, and progress has sometimes led to discovering new uncertainties."

(William M. Connolley 18:54, 20 May 2005 (UTC)) Rv without understanding wasn't very helpful either. In this case, I reverted without reading your edit, on the assumption that "rv" meant revert. The first change is no great matter one way or another. Adding Improvement has been difficult, and progress has sometimes led to discovering new uncertainties. is more questionable. If you insist on putting it back in, I won't like it, but won't re-remove it for now. Much of the model accuracy section needs a rebalance sometime anyway.

Oh yes, please do improve the models so they are more accurate. Although we won't know until AR4 how accurate the IPCC claims they are for their purposes. (SEWilco 02:30, 21 May 2005 (UTC))

[edit] Driver decisions

This is not true? How did models explain these things? (SEWilco 03:04, 21 May 2005 (UTC))

"The computer models cannot decide among the variable drivers, like solar versus lunar change, or chaos versus ocean circulation versus greenhouse gas increases. Unless they can explain these things, the models cannot be taken seriously as a basis for public policy."

[edit] To AOCGM or not to AOGCM

"Coupled ocean-atmosphere GCMs are used to project/predict future temperature changes under various scenarios."

But the IPCC did not use that? (SEWilco 04:17, 21 May 2005 (UTC))

"AOGCMs can only be integrated for a limited number of scenarios due to computational expense. Therefore, a simple climate model is used here for the projections of climate change for the next century. The simple model is tuned to simulate the response found in several of the AOGCMs used here."[2]

(William M. Connolley 22:30, 21 May 2005 (UTC)) That was Silverbacks mistake, too. Read on. Its obvious that AOGCMs were used. See if you can work out the subtle distinction...

Working from memory, I assume you're referring to the following mention of SRES. I looked at that but didn't spot phrasing which indicated that SRES was the "simple climate model". (SEWilco 04:05, 22 May 2005 (UTC))

(William M. Connolley 13:11, 27 May 2005 (UTC)) I'm baffled. What are you talking about? SRES are scenarios.

LOL Citing only the hottest A2 scenario is a bit biased isn't it, and calling the 1% compound CO2 concentration business as usual is also stretching NPOV. Hans Erren 20:57, 24 December 2005 (UTC)

You need to laugh less and study more... 1% is clearly labelled "idealised"; A2 isn't the hottest, is92a is used as well... William M. Connolley 21:25, 28 December 2005 (UTC).

I just did, see below,

'The A2 scenario is characterized by a politically and socially diverse world that exhibits sustained economic growth but does not address the inequities between rich and poor nations, and takes no special actions to combat global warming or environmental change issues. This world in 2100 is characterized by large population (15 billion), high total energy use, and moderate levels of fossil fuel dependency (mostly coal). The A2 scenario is the most well-studied of the SRES scenarios that assume no attempt to address global warming.'

A near tripling of world population? Ever heard of Lutz? ref: Wolfgang Lutz, Warren Sanderson & Sergei Scherbov, 2001, The end of world population growth, Nature 412, 543-545, 2 aug 2001 [3]

I prefer, for climate model intercomparison, the pure mathematical 1% compound CO2 concentration rise over any SRES storyline. BTW the graph [4] is called Global_Warming_Predictions.png. I suggest to change it (and the graph title) to "Climate model intercomparison". Hans Erren 22:20, 4 January 2006 (UTC)

Good grief, I may agree with you here... firstly, that graph has on itself the correct word "projections" and the graph itself should probably be moved to G_W_projections rather than intercomparison. Put a note on the talk page there to be polite. I'm really note sure quite how important the variouis SRES stuff is... its better to just rescale from a time axis to a CO2 axis, which to first order works. William M. Connolley 23:37, 4 January 2006 (UTC).

[edit] Not with my approval

(William M. Connolley 13:16, 27 May 2005 (UTC)) SEW added some stuff, apparently because he thought I recommended it. I'm not sure exactly what he meant, but I regard most of what he added as badly phrased and pretty much POV pushing (the POV being anti-model). So I've removed it.

[edit] rm flat planet

SEW wrote: many models are based on a flat planet. This is so bizarre that I don't know what he meant, but its wrong. William M. Connolley 09:23, 12 Jun 2005 (UTC).

Many models have been based on a flat planet. An example above, "1.25 degrees in latitude and longitude", uses a grid with the characteristic that grid areas become much smaller and closer together closer to the poles as lines of longitude converge. Conversely, there is less resolution around the climatically significant areas near the Equator. The first model to use a geodesic grid was announced in 2001. [5]

• I mentioned one example of a grid with a spacing of "1.25 degrees in latitude and longitude". In what way do its lines of longitude not get closer together toward the poles? (SEWilco 15:22, 12 Jun 2005 (UTC))

(William M. Connolley 19:21, 12 Jun 2005 (UTC)) Of course they get closer towards the poles. What has this got to do with a flat planet?

[edit] rm history of conv param

I've removed:

Convection parametrization was found to be necessary in the 1950s, as climate models otherwise became unstable after simulating a few weeks. Akio Arakawa's widely used solution was to ignore loss of kinetic energy and enstrophy. Although ignoring some basic physical behavior, the technique has been widely used in order to get results which resemble what is expected. [6] [7] [8]

Most of this is history of GCMs, which doesn't belong there.

Of the bits that *aren't* history but are current, its impossible to determine. Current models do (I think) conserve KE, so leaving it in is misleading. And... the refs in the AMIP table... don't mean that the exact Arakawa 1972 scheme is being used. People tend to ref the paper when doing a brief description and not ref local improvemnts.

Also, the AIP page is a bit... mystical in that section. I slightly wonder if Weart is a bit out of his field there.

• A little background seems appropriate rather than just saying "Arakawa's computational trick is in use". (SEWilco 14:51, 12 Jun 2005 (UTC))
• Any sources that current models do conserve KE? (SEWilco 14:51, 12 Jun 2005 (UTC))

(William M. Connolley 19:26, 12 Jun 2005 (UTC)) You have abstracted, with no real understanding, a small fragment from the AIP history. This *isn't* the history of climate models page, please don't try to make it so. Start a page on that if thats what you want. The only reason you've dumped this in here (with absolutely no idea whether or not any of this is still current) is because it looks slightly dodgy. Please try to give up the instinctive septicism. Is A's "trick" still in use? You don't know. What is the trick? HYou don't know. Do models conserve KE? Models go to great lengths to conserve quantities that should be, so the default is, yes they do, unless you've got good evidence otherwise.

Getting into this one a bit late. The AIP page is hopeless, a mish-mosh of allusions to aliasing (not named as such) and quadratic-conservative schemes. In the real world KE cascades from larger to smaller scales and eventually is dissipated at the scale of molecular viscosity, which of course is too small to represent in a GCM. The unresolvable part of the cascade is instead mimicked by viscosity-like terms that dissipate KE in the smallest resolvable scales. Raymond Arritt 04:51, 7 July 2006 (UTC)

[edit] rm van Dorland

I removed:

Rob van Dorland, lead author of the IPCC report scheduled for release in 2007, stated "Truthfully, we are far from knowing with certainty how natural climate factors, such as volcanic eruptions and solar activity, affect the earth's climate" [9].

From the context, its clear that this statement isn't really about the *accuracy* of the models - its about the Hockey stick stuff. Also, the source is distinctly dubious: essentially propaganda.

[edit] Parameters

There is a link to parametrization which redirects to Coordinate System which does not seem very relevant. I think linking Climate parameters, forcings and feedbacks is more relevant. However I am not happy with the quality of the page. Perhaps William Connolley could use his expertise to improve it then change the link.

Hmm yes. I've moved (and heavily edited) Clim param... to Parametrization (climate) and will change the link to that. William M. Connolley 18:57:55, 2005-07-31 (UTC).

[edit] Error Found in Models

I'm not sure how recent this China Daily article's information is, but it portends badly. Into Global Warming I inserted the paragraph "In December, 2005 it was revealed the reflectivity effect of airborne pollutants was about double that previously expected, and that therefore some global warming was being masked. This effect implies that pre-2006 models will underpredict future global warming. [10]" However, I feel that more detail is merited in this article. Does somebody get Nature? Simesa 20:56, 28 December 2005 (UTC)

I've seen this. I would always prefer to let things settle in the literature for a while, for all but the most definite results. This one has the potential to be less exciting than it seems, but its hard to be sure at this stage. William M. Connolley 21:39, 28 December 2005 (UTC).

Follow-up: I contacted a researcher (Jerry Meehl) at NCAR: He replied that, as WMC suggested, the new estimates are within the uncertainty bands for models being used ror the next IPCC report. (I wasn't aware that China had supercomputer-model capability.) Specifically, the researcher said:

The exact magnitude of various forcings is uncertain. The new estimates you refer to for aerosols are larger than what some models use (the magnitude of what models compute, for example, for sulfate aerosols varies depending on the nature of their sulfur cycle models or types of sulfate aerosol concentrations they use) but not out of the range of uncertainty for aerosol forcing used across all of the more than 20 models currently being assessed in the IPCC AR4. This accounts for some of the range of model responses to the simulation of 20th century climate. Even with this uncertainty in aerosols, the GHGs are still the largest forcing by far, and are the big driver for late 20th century warming and estimates of 21st century warming. The latest simulations will be assessed in the IPCC AR4, but many modeling groups are publishing their latest findings in the peer reviewed literature now (for example, from our group see: Meehl et al., 2005: How much more warming and sea level rise? Science, 307, 1769—1772).

Simesa 18:59, 29 December 2005 (UTC)

[edit] Para to talk

I've rm'd:

Accurate predictions of climate effects due to solar forcing and its associated feedbacks are important to prediction of effects due to anthropogenic greenhouse gas emissions, because models which fit observed temperature data but underestimate effects due to solar forcing will overestimate effects due to anthropogenic greenhouse gases, and vice versa.

This para is somewhere between wrong and misleading. If models underestimate solar forcing, then they are likely to underestimate other forcing as well; so underestimating solar is just as likely to lead to underpredicting future change. It also (I think) rather misrepresents the D+A stuff. See-also http://www.realclimate.org/index.php?p=229 (and the comments) William M. Connolley 21:39, 28 December 2005 (UTC).

[edit] Techncal template

I've added the Technical template to this article largely because of the part under "A rather more complete discussion of climate models is provided by the IPCC TAR chapter 8, Model Evaluation (2001)." It really doesn't make much sense out of context, and needs cleaned up. Calion | Talk 06:41, 21 February 2006 (UTC)

[edit] Predictive value?

Have there been examinations of the predictive value of climate models? I'm pretty sure that early models were around in 1990, say, so how well did their predictions accord with the change in climate over the past 15 years or so? "Postdiction" of facts known to the people who made the model isn't a terribly impressive achievement, I would say, and it surely isn't a good measure of how well the model will predict the future. —Simetrical (talk • contribs) 20:45, 12 March 2006 (UTC)

Not a lot, because the question isn't as sensible as it appears. People run coupled GCMs starting in about 1860 ish (for a variety of reasons) so thats not a good test of the 1990-now timeframe, because they aren't constrained at 1990. There is some stuff by Hansen in 1990 predicting to now; I think that came out OK though can't remember the ref. Also climate models aren't supposed to predict year-to-year variability William M. Connolley 20:52, 12 March 2006 (UTC)

[edit] Relation to weather forecasting

Cleaned this part up a bit. Also removed the tangential comments on analog forecasting and such. That material could be appropriate for an article on history of weather forecasting but has no relevance to GCMs. Raymond Arritt 03:42, 8 May 2006 (UTC)

[edit] S+H 2003

I've cut this to talk:

As recently as 2003, Sun and Hansen were calculating and using flux corrections and parameterization of cloud models to achieve cold start of a simulation based on "the assumption that the planet was in radiative balance in 1951".

because I dont know what it means. There is some mixing of flux corr and clouds? William M. Connolley 19:24, 17 June 2006 (UTC)

you should read the article. parameterization of the cloud model was used to achieve the flux correction.--Poodleboy 15:07, 19 June 2006 (UTC)

Um. And you think that will make sense to the reader of the article? The text already says *most* recent models don't use FC; so listing *one* that does (and not even telling us which model it is) doesn't make much sense. I haven't read the article; it shouldn't be necessary, to understand what you are on about. I suggest that if you really want the text in (and I can't see what it gains you) you need to make it comprehensible. So... they are using FC; but they weren't using the normal method, they were including the clouds? (were calculating and using flux corrections and parameterization of cloud models doesn't mean what you're now saying; if you mean param of clouds (clouds are always param!?!) was used to generate the FC... then I'm still baffled) In what way? Do tell William M. Connolley 15:36, 19 June 2006 (UTC)

OK, reading the paper: Ocean A is imposed SSTs, so no FC. We apply no flux adjustments with ocean E because such nonphysical adjustments can have a significant effect on the simulated climate (Neelin and Dijkstra 1995; Tziperman 2000). And ocean B has imposed horiz fluxes, which is FC. So this is a sensitivity study on different models... your text is misleading for that alone. And I don't see your point about the clouds... please quote chapter and verse William M. Connolley 15:36, 19 June 2006 (UTC)

[edit] grids

I took out:

The grid can either have a fixed resolution, or can be specified as a gaussian grid, where the latitude resolution is not constant.

Its confusing as written (many people will interpret "fixed" and "constant" as meaning in time). Its also wrong, in implying that this exhausts the possibilities: as mentioned later, its perfectly possible to have a grid with any variation in latitude. What I think it means is that FD models tend to use constant-spacing and Spectral ones use gaussian... William M. Connolley 08:18, 21 September 2006 (UTC)

Okay, I sort-of see why you have removed this twice. This section of the article is about grids, and I feel that it is worthwhile to mention some common grid types, such as gaussian grids (e.g., NCEP-DOE reanalysis 2, CCCma, ECMWF). I'm not a climate modeler, but I was very confused when I first encountered NCEP's fixed and gaussian grids about 1/2 a year ago. So I read up on them and created the gaussian grid article to help me, and others, understand what they are. To me, it would be more confusing to not mention a few common grid types, as they are encountered by non-climate modelers. Certainly, there are many types of grids, such as polar, etc, and these do not need to be described here, as they are very climate modeler niche material — better suited in technical books and journal literature rather than Wikipedia. Much of the content in this article is very technical by nature, so I don't feel that adding a mention of a few common types of grids would contribute to much more confusion. Also, I was sure to specify "latitude" when using "constant", thereby reducing the confusion with other dimensions (such as time). +mwtoews 20:44, 21 September 2006 (UTC)

Maybe I was too harsh. Sorry. How about the current version? BTW: re NCEPs grids: as I understand it (and I'm sure of this for ECMWF) there is the "native" grid (which is gaussian; and indeed reduced-gaussian to boot) which is actually used by the physics. Then there are various other grids - 2.5 degrees, say - onto which the output can be interpolated William M. Connolley 21:41, 21 September 2006 (UTC)

Perfect, I think that is all that is needed, thanks. I thought that the "native" grid would have been the gaussian grid, but I was never really sure until now. +mwtoews 21:56, 21 September 2006 (UTC)