Global climate models are accurate enough to be relied upon
Resolved: Global climate models are accurate enough to be relied upon
Global climate models: Mathematical models which are used to predict future temperature changes under various scenarios
Relied upon: To trust or place confidence in
Round one will be for acceptance only.
ACCEPTANCE: After that it would be impolite to decline QT's challenge despite the fact I'm very busy this month so I ACCPET the debate.
I would therefore prefer this debate to be concise and focused on the major issues. I hope won't need 8k words per round to express our point.
I shall point out general modelling limitations as well as specific contradictions between models and measurements and quote some resumes from scientific papers dealing with this problem. I assume that climate model's means models that were used or cited by IPCC in last decade as whole (usually more than one is used to get averages) and I do not have to prove every single model wrong in detail in order to fulfill my job as Con.
I don't have English as my first language, but voters can feel free to punish me for big mistakes.
I am looking forward to a good debate.
Models can accurately reproduce past climate changes:
Climate models have successfully simulated many aspects of the climate changes observed during the instrumental period. Most notably, models have reporduced the increase in surface air temperatures remarkably well (1-2). Scientists have also found a high degree of similarity between the simulated and observed evolution of global lower stratospheric temperatures during the past 25 years (3).
Good agreement between model projections and observations has likewise been reported for decreases in Arctic Ocean ice cover. As one researcher concluded, “The simulated decreasing trend in average sea ice extent for 1970–1999 (–2.5% per decade) is very similar to observations" (4). In addition, model projections are consistent with observations of changes in ocean heat content since 1960 (5).
In 1988, Dr. James Hansen predicted future atmospheric temperature changes using several different emissions scenarios. His second scenario most closely resembled the observed pattern of carbon dioxide emissions. Models which employed this scenario predicted that we should have seen .33 degrees Celsius of warming between 1988 and 2005. In actuality, we observed .32 degrees of warming, almost exactly what the models predicted (6).
Climate models predict that atmospheric water vapor will increase as the surface warms. Observations have independently confirmed these predictions. Satellite measurements indicate that the total atmospheric water content, which is dominated by water vapor in the lower troposphere, has increased at a rate consistent with model predictions (7-8). Interestingly, upper tropospheric water vapor has also increased during the past two decades (9).
Climate model simulations indicate that cloud cover changes will most likely amplify greenhouse gas warming. Observations have confirmed that these predictions are also correct. As Dr. Andrew Dessler noted, “The short-term cloud feedback has a magnitude of 0.54 ± 0.74 watts per square meter per kelvin, meaning that it is likely positive...Calculations of short-term cloud feedback in climate models yield a similar feedback” (10).
Observations themselves are not without error. In a few cases, model simulations have been even more accurate than data. For example, climate models in the 1990s could not reproduce the full extent of the Northern Hemispheric cooling in the 1950s as indicated by observational data. However, a careful analysis later revealed that the data had been distorted by a change in the way ocean temperatures were measured after World War II (11).
In another example, satellite measurements in the early 2000s showed essentially no warming in the middle levels of the atmosphere. More direct measurements by balloons and radiosondes likewise showed no warming there. However, a "tropospheric hot spot" had been predicted by all models clear back to the 1970s. This alleged discrepancy was resolved to the satisfaction of most modelers in 2005, when several researchers documented errors in the sets of observations. For example, the observers had not taken proper account of how instruments in the weather balloons heated up when struck by sunlight. Once these errors were accounted for, it was evident that the middle levels of the atmosphere had indeed been warming up (12).
As Dr. Michael Mann remarked, “Current climate models do a remarkably good job of reproducing key features of the actual climate...They also closely reproduce past climate changes. We therefore have good reason to take their predictions of possible future changes in climate seriously” (13).
Thanks Pro for concise and focused argument.
General limitations of Climate modelling
In perfect scenario, we would have simple deterministic mathematical model to simulate deterministic phenomena on which we can run multiple experiments with controlled conditions in order to validate the model. While for example engineering simulations can comply with such scenario to some extent, complex phenomena in economic, social or climate sciences are another story. We don't have enough beforehand knowledge about every important climate feedback so we in fact estimate such feedback's with the model itself based on its output compared to measurements. But unless we know all other forcing that can influence output in similar way, we won't get good estimate of the parameter we look for.
For example Koutsoyiannis or Tennekes (extending on Poper and Lorenz) challenge the notion that complex models could ever be reliable according to their nature. "Theories that are complex may become untestable, even if they happen to be true." etc.
If we look at QT's definition of GCM, it is clear that to fulfill their purpose, those models must have the structure of forcings and feedbacks right. We won't get reliable "scenario predictions", no matter how lucky we are in predicting aggregated mean temperature if the true causes are different then we thought.
What is the reliability of current GCMs in scientists eyes?
"We compare the output of various climate models to temperature and precipitation observations at 55 points around the globe... Besides confirming the findings of a previous assessment study that
model projections at point scale are poor, results show that the spatially integrated projections are also poor."
"We examine tropospheric temperature trends of 67 runs from 22 ‘Climate of the 20th Century' model simulations and try to reconcile them with the best available updated observations (in the tropics during the satellite era). Model results and observed temperature trends are in disagreement in most of the tropical troposphere, being separated by more than twice the uncertainty of the model mean."
"These weaknesses combine to make GCM-based predictions too uncertain to be used as the bases for public policy responses related to future climate changes." 
"So there has been a large activity to bring models and observations into agreement, strangely only by adjusting the measurements instead of adjusting the models. "
Even Kevin Trenberth (from the Mann et al. group) now acknowledges many of the GCM's troubles in his paper "More knowledge, less certainty" publicly.
"The scientific literature is filled with studies documenting the inability of even the most
advanced GCMs to accurately model radiation, clouds, and precipitation."
For those interested, NIPCC report  and their topical updates  provide comprehensive information about studies dealing with GCM reliability.
As you put it, Hansen's predictions look totally perfect and within 100th of degree of Celsius. Lets examine the claim.
"Dr. Jim's 1988 projections weren't looking so good, so he dropped an apple in the middle of his oranges. The red line is land only temperatures, but his projections were for global temperatures." Let us also compare it with satellite  and other  data. Now suppose I made highly oscillating prediction. At some points in time, my prediction would be always spot on as it would cross the real data. Look back at the figures. Where is this precision from 2006 until now, or in early nineties? Also note that curve C assumes "emissions drastically reduced" in 1990 [10,11]. I therefore call this conduct a fallacy of cherry-picking.
On top of that, Hansen is known for not so transparent temperature data manipulation.[12,13]
Dessler's assumptions are refuted by Spencer's satellite observations , supporting rather Lindzen's hypothesis:"Our measured sensitivity of total (SW + LW) cloud radiative forcing to tropospheric temperature is -6.1 W *m^-2 K^-1... This decrease in ice cloud coverage is nominally supportive of Lindzen's 'infrared iris' hypothesis."
"This is exactly opposite of the way all climate models behave," as Spencer put it in his own words in . (technical note: Don't take the whole video as extension of my argument. Its used only as source of the quote in 6:19 and for information purpose as I acknowledge I must make my argument myself on this page within its limitations.)
Now lets see the warming troposphere, because it was predicted it would be significantly warmed by CO2 forcing. See figures 3.4.1 and 3.4.2 for PCM model prediction and 3.4.3 for radiosondes.
Updated study , taking in account errors stated by Pro, supports overall warming trend in troposphere of +0.052 ± 0.07 K per decade (while RSS temperature is somewhat higher then other methods, but much less then models).
That is in good agreement what the figure 3.4.3, but in complete disagreement with figure 3.4.2 (also note the scales). The problem is not whether the troposphere warmed, but how much it warmed (1.2 °C at hotspot vs 0-0.3°C) and how different layers warmed relative to each other. The model predicted much stronger warming in troposphere then on surface, but that is not true. It means that model is wrong about GHG forcing or feedbacks in troposphere.
Models may be useful in furthering our knowledge of the problem, mostly by showing us what our assumptions really mean. If we are humble enough and learn from comparing our assumptions to measurements, we can learn from our mistakes.
But the climate models fail if used as defined by Pro.
 Video: "Why the IPCC Models...."
 http://www.nipccreport.org..., pages 106-108
My opponent's quotes:
Attack on my sources
I find it quite hilarious, that Pro attacks integrity of my sources while she depends on sources like Mann et al. (Climategate, "Hide the decline") or Hansen, who was caught manipulating data [2 and last round] and prevented scientists under him to publish their results that disagreed with him  etc. Or IPCC, that misrepresent what the scientists say , and RealClimate, which is basically web promoting Mann et al. point of view.
It's quite obvious who gets the most money in "climate science" . In fact all those climate scientists funds depend on AGW, so its not skeptics who have the major monetary interests in this phenomenon. I may retract my source of the third quote, but then Pro would have to retract a lot of her sources.
Before Santer tries to accuse others, he should correct the way he works with data . His study was refuted by : "Overall, the conclusion of S08 that 'there is no longer a serious discrepancy between modeled and observed trends in tropical lapse rates must be reconsidered in light of upto-date data. The 'potential inconsistency' between models and observations in the tropical region, as reported by Karl et al (2006), remains an issue."
Santer is also refuted in  (which he tried to delay in peer-review [16b]) and  (see part Tropospheric warming below).
Indeed I was not sure what data did Pro compare, I just looked at those closer to each other (land station data). When I wrote about scenario C, I just wanted to make sure that people don't miss what it actually means, since it is closer to measured data than B.
Lets look again at the figures in [my previous round source 9], that corresponds to Fig 2 in [Pro's first round source 6]. The Land-Ocean (black curve) nowhere near the scenario B prediction (blue curve) since 2000 aside from two single points 2002 and 2005 where L-O is at local maximum and B is at local minimum. My argument stands. Especially later data from 2006 on make it obvious.
First I would like to apologize for confusion around refuting Dessler. That Spencer et al. 2007 paper directly refutes the previous Dessler's paper and Dessler 2010 fail to address the main issues and instead attacks a straw man. I should have clarified with further information:
"To Dessler’s credit, he actually references our paper. But he then immediately discounts our interpretation of the satellite data. Why? Because, as he claims, (1) most of the climate variability during the satellite period of record (2000 to 2010) was due to El Nino and La Nina (which is largely true), and (2) no researcher has ever claimed that El Nino or La Nina are caused by clouds. This simple, blanket claim was then intended to negate all of the evidence we published. But this is not what we were claiming, nor is it a necessary condition for our interpretation to be correct." 
In contrast to Pro's claim, Dessler didn't make the different conclusion because of larger data set, but because of different methodology, using simple regression.
Spencer and Braswell examined the same data set in that same year and came with opposite conclusions . They addressed the time lag in changes by phase space analysis in order to see what is most likely cause and what is effect (which Dessler claims is the other way around). You may get similar regression slope like Dessler derived from the data even for strong negative feedback.
"The bottom line is that, you can not use simple regression to infer cloud feedbacks from data like those seen in Dessler’s data plots." 
To simplify it by analogy, it could be like ice cores "CO2 drives temperature" claim by Al Gore. There was certainly correlation as well, but he ignored the time lag as CO2 follows temperature.
Lindzen acknowledged many parts of the criticism and corrected his hypothesis where it was wrong. However, he still refuted conclusions all the papers Pro mentioned [11,12,13]. As I used my space on Dessler, I won't go into details here.
Conclusions of Sherman and Allen 2008 are refuted .
Not only there is much more uncertainty in indirect method based on wind shear due to temperature gradient, they ignored conclusions of previous research on that matter dealing with bias of stronger 200 mbar winds at higher latitudes .
Haimberger 2008 is refuted by  (addendum to Douglass et al.), because there is bias in homogenization in RAOBCORE 1.4 data as other publications pointed out.
“... papers such as Santer et al. 2008 do no testing, but simply assume that all datasets are equal, such as “new” ERSST or “old” RAOBCORE v1.2, v1.3 and v1.4, and thus ignore the publications which have provided the evidence which document significant errors in the ones they prefer."
Hansen Pinatubo and Energy imbalance
After Hansen was refuted, he made new paper “Earth's Energy Imbalance and Implications” in 2011  and admitted: “We conclude that most climate models mix heat too efficiently into the deep ocean and as a result underestimate the negative forcing by human-made aerosols... Continued failure to quantify the specific origins of this large forcing is untenable, as knowledge of changing aerosol effects is needed to understand future climate change.“
Watts and Tisdale wrote some criticism on some of his new claims .
My findings on general limitations of climate models are unchallenged as well as some of the quotes on reliability of GCM. I offered refutation to the points Pro chose to defend.
Why some of the modelers finds hard to admit the counter-evidence or even cheat? Maybe social study of science by Lahsen 2005  will give us some insight: "...persuasive power of the simulations can affect the very process of creating them: modelers are at times tempted to ‘get caught up in’ their own creations and to ‘start to believe’ them, to the point of losing awareness about potential inaccuracies. Erroneous assumptions and questionable interpretations of model accuracy can, in turn, be sustained by the difficulty of validating the model..."
 Video "Hide the decline"
|Agreed with before the debate:||-||-||0 points|
|Agreed with after the debate:||-||-||0 points|
|Who had better conduct:||-||-||1 point|
|Had better spelling and grammar:||-||-||1 point|
|Made more convincing arguments:||-||-||3 points|
|Used the most reliable sources:||-||-||2 points|
|Total points awarded:||7||0|