The best answer is "it depends". The reason why many researchers use ensemble model approaches (incorporating multiple models) is to take the richness of multiple models into account. Some are better at reflecting extremes, and others at reflecting means or information from thoroughly-data-rich parts of the globe. As you will know well, GCMs are complex and based on many (sometimes tricky) assumptions. So I don't think anyone should imply that there is a silver-bullet for all purposes. HADCM3 is indeed widely used by many researchers, and for good reason, but it should not be the only one considered. I'd like to see some responses from climate modeling research units, such as University of Cape Town's Climate Systems Analysis Group.
Title: An appropriate general circulation model (GCM) to investigate climate change impact
Author: Mohammad Reza Farzaneh; Saeid Eslamian; S. Zahra Samadi; Aboulfazl Akbarpour
Journal: Int. J. of Hydrology Science and Technology, 2012 Vol.2, No.1, pp.34 - 47
Abstract: There is a wide agreement in the international scientific society that climate change will modify climatic variables. Atmosphere-Ocean General Circulation Model (AOGCM) is an appropriate model to produce climatic scenarios. The first step to climate change studies is currently using the most appropriate AOGCM output. In this study, AOGCM's for a time period of 1961-1990 in Shahrekord synoptic station are therefore compared and finally the best AOGCM is presented. The employed models in the paper include CCSR, CGCM2, CSIRO and HADCM3 which was mentioned in the Third Assessment Reports of IPCC. The results display that the HADCM3 output has more appropriate correlation than the other models comparing with the observed data of the station. According to the evaluation of climatic scenarios, climate change has the unfavourable impacts on hydrology and water resources of the basin, thus these results are suitable ones of the most achievements to study climate change effects on stream flow changes and variability in Northern Karoon region, Iran.
The best answer is "it depends". The reason why many researchers use ensemble model approaches (incorporating multiple models) is to take the richness of multiple models into account. Some are better at reflecting extremes, and others at reflecting means or information from thoroughly-data-rich parts of the globe. As you will know well, GCMs are complex and based on many (sometimes tricky) assumptions. So I don't think anyone should imply that there is a silver-bullet for all purposes. HADCM3 is indeed widely used by many researchers, and for good reason, but it should not be the only one considered. I'd like to see some responses from climate modeling research units, such as University of Cape Town's Climate Systems Analysis Group.
I agree that the best answer is "it depends," but one has to get starting somewhere ...
If you are new to climate modeling, I would suggest to become acquainted with advanced climate models by learning about two models that are currently being widely used for global earth climate system simulations.
One is the Community Climate System Model, now in version 4. It is hosted by NCAR/UCAR; see their web site: http://www.cesm.ucar.edu/models/ccsm4.0/.
References:
Collins, W. D., and Coauthors, 2006: The Community Climate System Model version 3 (CCSM3). J. Climate, 19, 2122–2143, doi:10.1175/JCLI3761.1
Gent, P.R., Danabasoglu, G., Donner, L.J., Holland, M.M., Hunke, E.C., Jayne, S.R., Lawrence, D.M., Neale, R.B., Rasch, P.J., Vertenstein, M., Worley, P.H., Yang, Z.-L., Zhang, M., 2011. The Community Climate System Model Version 4. Journal of Climate, 24, 4973-4991.
***************
Another advanced suit of climate system models was developed at Geophysical Fluid Dynamics Laboratory (NOAA). The version with the finest oceanic resolution is capable of resolving ocean at a grid with 0.1-degree and with atmosphere grid resolution of roughly 50 km.
Delworth, T.L., A. Rosati, W. Anderson, A.J. Adcroft, V. Balaji, R. Benson, K. Dixon, S.M. Griffies, H-C. Lee, R.C. Pacanowski, G.A. Vecchi, A.T. Wittenberg, F. Zeng, R. Zhang, 2012: Simulated Climate and Climate Change in the GFDL CM2.5 High-Resolution Coupled Climate Model, J. Climate, vol. 25, issue 8, pp. 2755-2781. DOI 10.1175/JCLI-D-11-00316.1
One of the most recent research using GFDL model suit has just been published in Journal of Climate:
Griffies, S.M., Winton, M., Anderson, W.G., Benson, R., Delworth, T.L., Dufour, C.O., Dunne, J.P., Goddard, P., Morrison, A.k., Rosati, A., Wittenberg, A.T., Yin, J., Zhang, R., 2014. Impacts on ocean heat from transient mesoscale eddies in a hierarchy of climate models. Journal of Climate.
You are right about the current skills of all models and that they do not excel in very detailed forecast, mostly because it's so difficult to resolve the cloud physics and thus precipitation locally, even with the finest resolution we never could have dreamed of before. Yet we are talking about simulation of climate system dynamics, and the climate system models do exactly that -- simulate the dynamics or, better, generate climate change scenarios. I am not running global climate models myself and actually don't run any ocean model now, for that matter (I feel that I am still qualified because I used to do quite some ocean modeling and I know how the global models operate). The question was about the best models, so the two I mentioned (and may be a few other models) fit the profile.
I think Phoebe said it all. I would like to add that a model may work well in a region while it may not be as good in another. Additionally, assuming that a "good" model now will remain "good" in the future has been always debatable. The workaround is to consider a multi-model ensemble. The minimum recommended ensemble size is 3, The larger the better but it is also a question of resources including time.
Here comes a question of which 3 (or more as decided) to pick from the available GCMs. Here, the quality of simulating the current climate in the study region becomes relevant. Selecting which criteria to be used to assess this will depend on the purpose of the study. If studying mean climate, then pick performance criteria based on mean, if the focus is on extremes, then the set of performance criteria should focus on those, but without discarding performance related to the mean. The set of models selected for either purpose may be thus different.
Finally, when looking into the future, it is also recommended that selected models to form the ensemble should span as much as possible the uncertainty range.
Phoebe has the best answer. It really depends on what you are doing with those GCMs output and which part of the world or which aspect of climate that you are working on. The best approach to answer the question is however, I believe, to do a comprehensive assessment on all the available GCMs and select a subsets from the GCMs pool. Over the years, there has been a few studies/ideas on how to select a subset of GCMs from the IPCC's PCMDI pool for downstream usage. You can google "GCMs selection" and find extra information on this issue.
I largely agree with what has been said here. If adopting the global perspective, I nevertheless mean that a suite of helpful criteria may be given to which the following three should belong:
1. Climate dynamics in global integrals of motion: does the GCM show realistic atmospheric angular momentum (AAM) dynamics? Note however: a GCM does not take into account all contributors to the Earth's angular momentum; that is, there may be tolerable systematic deviations, certainly in the mean at least (sub-, co-, supperrotational state), and it might not be wise to fit a GCM too closely to the observed mean AAM.
2. Representation of the atmospheric water cycle: A notoriously difficult issue is the qualitatively (nothing to say about quantitatively) correct simulation of the global monsoon systems, with their dynamically interacting major branches. Beyond seasonal mean structures of precipitation and circulation, missing correct intraseasonal monsoon dynamics (notably of the subtropics) might be at the very roots of problems with present-day GCMs. Note: the system might develop chaotic (or otherwise complex, like saddle-node) dynamics at planetary scale that are difficult to cope with in present generation GCMs.
3. Natural variability from intraseasonal to centennial (and beyond): A GCM should show realistic free, not scenario-driven, 'natural' variability. With a view on the episodic nature of the observed global warming, specifically, it should freely develop unforced multidecadal modes of variability.
Altogether, a GCM shoud prove to run in a realistic mode of operation. Certainly there are more helpful criteria to decide whether or not this is the case. In just the same respect, ensemble simulations may even be problematic (to my mind).
the other responders are correct to say that it depends what you are looking at, as no single general circulation model is the best for all applications. Nevertheless, if you have a specific application, then some models are better than others. See for example the assessment of Atlantic blocking by Masato et al, GRL, 2013 which ranks the following four as good models for around the Atlantic basin: MPI-ESM-MR, HadGEM2-CC, MIROC5 and EC-Earth. The HadGEM2-CC model is updated from the HadCM3 model. Data from all are freely available from the CMIP5 database here:
Nowadays CIMP5's models are the best model with new different scenarios (RCP). But in this way you need to investigation uncertainty of different models depend on your case studies with different climate.
First of all you should check the resolutions of models and then depends on study area you can select one or further models with various scenarios. Probably you need a statistical method for downscaling. There are many ways for this aim. you can search on the web and select on of them.
Don't agree with the last sentence. Even a realistic simulation of the past (what is realistic in this context given the partly poorly known forcing?) is no guarantee to get more confidence in simulations of the future.
the sentence reads: "we cannot have much confidence ... unless the model realistically simulates the past" - it does not read: "We can have much confidence ... if the model realistically simulates the past" :-)
Your statements are very general and more political than scientific. I'm not sure if this is the correct website for these discussions. Such a level of argumentation is not very different from discussing wheter you like the colour blue more than purple. So I guess your answer to the initial question "Which is the best General Circulation Model?" is that all models are wrong and useless? I don't understand what you mean with "realistically" unfortunately.
If you don't like that climate change related politics is made based on model projections, talk to politicians. You can count me in on that complaint. For politicians or stakeholders it is convenient to hide behind model results either emphazing the catastropic outcomes (warmists) or the uncertainties and model failures (sceptics). The same is true for observation data. Both sides tend to forget that the theory of global warming due to increased GHG does not evolve from climate models at all. Climate models are used to test certain assumptions how a complex system might react to more CO2 (which sensitivity, what is the timescale and magnitude of change) etc. Scenario simulations are more or less physically plausible experiments for subjective a priori assumptions regarding many factors.
As climate modelers we do our best to improve models. The best we can do is indeed far from enough for many scientific questions. The models cannot be better than our understanding of the Earth system. And that understanding is still quite limited. We even don't know the correct global temperatures today. It is also very difficult to test our models against incomplete, possibly somtimes also not correct observations.
A particular GCM can be suitable for a specific region, and it may not be the suitable one to other areas..... Assessment study must be carried out with a reliable base period data.
The base period may vary region to region. Base period or control period is usually considered before the development start in any region. In literature, it starts from 1961 to 1990. Many researchers have suggested until 2005 depending on the region. We can plot recorded Temperature and can see the trends.....We may get the pattern that temperature is likely to be changed after 1990. So the choice of base period is also necessary for the assessment of climate change impacts. GCM is the global model. However, we use it for local assessment. I understand that base period should not be same for developed, third world or developing the country.
Personally I am not convinced that the GCMs are reliable enough for making decisions based on them. I have a simple question. Can we really simulate evapotranspiration based on the temperature simulations of the GCMs? No matter which GCM is used, all are based on some formulations and assumptions that might not work well. You know, we are telling and claiming that climate change and global warming are basically due to greenhouse gases on another relevant issues. The drought that we are facing in several parts of the world are basically attributed to climate change. However, in 2500 years ago, there are evidences of periodic droughts that seems were severe. So could GCMs prove that in 2500 years ago, a climate changed happened? Lastly, all models are wrong, but some are useful!
Dosio, A., Panitz, HJ., Schubert-Frisius, M. et al. Clim Dyn (2015) 44: 2637. https://doi.org/10.1007/s00382-014-2262-x
Joubert AM, Hewitson BC. 1997. Simulating present and future climates of southern Africa using general circulation models. Progress in Physical Geography. 21(1):51-78.
To understand what the best model is in your area. You should compare the historical period of the model with the historical period of your area. For scenarios, it's better to use different scenarios. For example, RCP2.6 is an optimistic scenario. And the RCP8.5 scenario is pessimistic.
Reliable base period data is crucial to decide the best GCM. You can choose commonly used GCMs by researchers. Do intensive Review and choose the most widely used 3 GCMs. Single GCM is not recommended, however you can identify the best one by analysis.
You can review the uncertainty of GCM models used by many researchers in your area and select a few GCMs done good-fit during control period. Otherwise, you have to study many and select a few GCMs.