If You Don't Like People Saying That Climate Science is Absurd, Stop Publishing Absurd Un-Scientific Charts

Kevin Drum can't believe the folks at the National Review are still calling global warming science a "myth".  As is usual for global warming supporters, he wraps himself in the mantle of science while implying that those who don't toe the line on the declared consensus are somehow anti-science.

Readers will know that as a lukewarmer, I have as little patience with outright CO2 warming deniers as I do with those declaring a catastrophe  (for my views read this and this).  But if you are going to simply be thunderstruck that some people don't trust climate scientists, then don't post a chart that is a great example of why people think that a lot of global warming science is garbage.  Here is Drum's chart:

la-sci-climate-warming

 

The problem is that his chart is a splice of multiple data series with very different time resolutions.  The series up to about 1850 has data points taken at best every 50 years and likely at 100-200 year or more intervals.  It is smoothed so that temperature shifts less than 200 years or so in length won't show up and are smoothed out.

In contrast, the data series after 1850 has data sampled every day or even hour.  It has a sampling interval 6 orders of magnitude (over a million times) more frequent.  It by definition is smoothed on a time scale substantially shorter than the rest of the data.

In addition, these two data sets use entirely different measurement techniques.  The modern data comes from thermometers and satellites, measurement approaches that we understand fairly well.  The earlier data comes from some sort of proxy analysis (ice cores, tree rings, sediments, etc.)  While we know these proxies generally change with temperature, there are still a lot of questions as to their accuracy and, perhaps more importantly for us here, whether they vary linearly or have any sort of attenuation of the peaks.  For example, recent warming has not shown up as strongly in tree ring proxies, raising the question of whether they may also be missing rapid temperature changes or peaks in earlier data for which we don't have thermometers to back-check them (this is an oft-discussed problem called proxy divergence).

The problem is not the accuracy of the data for the last 100 years, though we could quibble this it is perhaps exaggerated by a few tenths of a degree.  The problem is with the historic data and using it as a valid comparison to recent data.  Even a 100 year increase of about a degree would, in the data series before 1850, be at most a single data point.  If the sampling is on 200 year intervals, there is a 50-50 chance a 100 year spike would be missed entirely in the historic data.  And even if it were in the data as a single data point, it would be smoothed out at this data scale.

Do you really think that there was never a 100-year period in those last 10,000 years where the temperatures varied by more than 0.1F, as implied by this chart?  This chart has a data set that is smoothed to signals no finer than about 200 years and compares it to recent data with no such filter.  It is like comparing the annualized GDP increase for the last quarter to the average annual GDP increase for the entire 19th century.   It is easy to demonstrate how silly this is.  If you cut the chart off at say 1950, before much anthropogenic effect will have occurred, it would still look like this, with an anomalous spike at the right (just a bit shorter).  If you believe this analysis, you have to believe that there is an unprecedented spike at the end even without anthropogenic effects.

There are several other issues with this chart that makes it laughably bad for someone to use in the context of arguing that he is the true defender of scientific integrity

  • The grey range band is if anything an even bigger scientific absurdity than the main data line.  Are they really trying to argue that there were no years, or decades, or even whole centuries that never deviated from a 0.7F baseline anomaly by more than 0.3F for the entire 4000 year period from 7500 years ago to 3500 years ago?  I will bet just about anything that the error bars on this analysis should be more than 0.3F, much less the range of variability around the mean.  Any natural scientist worth his or her salt would laugh this out of the room.  It is absurd.  But here it is presented as climate science in the exact same article that the author expresses dismay that anyone would distrust climate science.
  • A more minor point, but one that disguises the sampling frequency problem a bit, is that the last dark brown shaded area on the right that is labelled "the last 100 years" is actually at least 300 years wide.  Based on the scale, a hundred years should be about one dot on the x axis.  This means that 100 years is less than the width of the red line, and the last 60 years or the real anthropogenic period is less than half the width of the red line.  We are talking about a temperature change whose duration is half the width of the red line, which hopefully gives you some idea why I say the data sampling and smoothing processes would disguise any past periods similar to the most recent one.

Update:  Kevin Drum posted a defense of this chart on Twitter.  Here it is:  "It was published in Science."   Well folks, there is climate debate in a nutshell.   An 1000-word dissection of what appears to be wrong with a particular analysis retorted by a five-word appeal to authority.

Update #2:  I have explained the issue with a parallel flawed analysis from politics where Drum is more likely to see the flaws.

  • RambleOnDude

    I'm not sure what you're asking for, Guest, but I'm referring to the fact that the paper Marcott et all submitted to Science contained the same graph that was in his thesis, *minus* the late uptick. It (the thesis chapter) uses the same proxy datasets. See:

    http://climateaudit.org/2013/03/14/no-uptick-in-marcott-thesis/

    Simply read the other links I've provided in this comment section and start pulling the thread. Nearly as I can tell, the only valid questions raised by Marcott's Science article are: Was the malfeasance intentional? Was it intentionally timed for inclusion in the upcoming (at the time) IPCC policy guide? How deeply were other senior members of the AGW community involved?

    We also typically don't call this sort of thing "skeptic literature", we like to call it "real science". 🙂

  • RambleOnDude

    oferchrissakes i misspelled et al and can't correct it. No OCD here, no sir.

    It's been interesting, gentlemen, unfortunately I have a day job. Have fun!

  • Stuck In A Deep Blue State

    Yeah, all windmills should be located in DC due to all the hot air being expelled. Also Methane Gas is a much stronger greenhouse gas than CO2, and DC is a very large source of this gas as most of the people who work there are full of .....

  • BillWStl

    As an engineer that was involved in some of the most complex computer modelling in the early '90's, I got very interested in climate science, and especially the computer modelling. As I dissected it, and knowing the limitations of the computer modelling intimately, I concluded that all of the models I reviewed were garbage. They had a LOT of great equations, but you boil them all down to a starting point, and the equation of that would best describe the data series that had been observed to that point, and then the belief that it would continue. Then you mash them altogether using various techniques. Two of the major problems are the starting points and the data series. I have been pointing out that using tree rings is an abomination of science. Calling it pseudo-science does not give pseudo enough credit. And then, they excluded tree rings that don't follow the narrative and only use a very small sampling. As the years have gone by, the models and the data sets have becomem more and more secretive- which to me points to an anti-science, pro-politics movement.

    Your statement "The problem is not the accuracy of the data for the last 100 years,
    though we could quibble this it is perhaps exaggerated by a few tenths
    of a degree" - we are only talking about 1.3 degrees, and the data has had to been "corrected" by over a degree, and the issue of heat islands is still being argued about. And oh, the models have shown to be drastically wrong due to the NO WARMING over the last 15 years or so.

    So call me a denier, but you cannot call me anti-science. I believe my approach is much closer to science than "climate scientists" who have more often than not found to been massaging the data.

  • an80sreaganite

    Hmmm.. I'll start believing this crap when someone can explain a few things to me:
    1) Why was the upper midwest (i.e., Wisconsin & Minnesota) once covered by a glacier and it seems to have disappeared (i.e. melted) quite some time before my birth (1966), my parents birth (1923), and for that matter, my grand parents birth (1897). Last time I checked, there weren't a whole lot of "fossil fuels burned emitting greenhouse gases" way back when the glaciers were melting away in the midwest. Given that premise it would seem the planet Earth seems to have been "warming" for quite a long time.
    2) Given that the atmosphere is 78% nitrogen and 21% oxygen with about 1% of everything else, I assume CO2 falls in the 1% category. Explain how 1% of the atmosphere can affect change on this scale?
    3) What is the global science communities accuracy rating for the prediction of atmospheric CO2 and the resulting effects?
    4) What IS the right temperature for plant Earth?

  • VolSeller

    There does not need to be a cabal for it to be very important to point out flawed analysis. The media will always seek out the sensational and disastrous, it sells "eyeballs" and ads. Since most people don't want to dig into the details of even simple tricks like the one described above, the "man-made climate change activists" have a built-in public relations advantage. To the extent that analyses are made public that misinterpret data and mislead the public, they need to be ROUNDLY denounced, otherwise policy decisions will be made in error.

  • FelineCannonball

    I'm not a modeler but I do drink beer with them. Models are models. They aren't complete simulations. They have assumptions and simplifications. They fail in a variety of ways when tested and you really have to talk to modelers to learn what aspects they trust and which they do not. Every modeler you talk to will tell you there are variables they have not included because of processing power, parameters that are probably miscalibrated, feedbacks which are ignored for short term simulations, and chaotic parts of the earths climate system that are not modeled but may be relevent for decadal precision.

    This is why 20 different models create different projections. In fact what you see in IPCC reports for each model is the result of many different model runs which themselves show a much larger scatter. It's a bit like modeling the position of a tennis ball moving down a stream. No model is going to give you great moment by moment projections and the tennis ball might find an eddy and move up stream for a bit, but all the models are going to get it going downstream in the longer term.

    These models really weren't meant for short term projections and in the longer term they model simplified systems. What the IPCC shows are the spread of model results and the mean. The mean projection and spread says X very smart people have tried to model the system and all arrived at ballpark similar answers favoring the existence of AGW. The error bar is huge, but no model is giving a null or negative response to CO2. The controversy instead is the size of positive feedbacks in the current earth system.

    I'd say the ultimate plateau is probably defined by the paleoclimate record and the size of temporarily available feedbacks. Ice albedo feedbacks only work till the ice is gone. Vegetation feedbacks stop when vegetation stops migrating or disappears. Stored carbon feedbacks stop when soils and sediments come to equilibrium with temperature. The potential plateau is probably represented by something a bit bigger than the PETM thermal maximum and is something that can't play out till slow dynamics come to equilibrium in 1000s of years. In the past, extreme warm events have topped out at 1500 to 5000 ppm, forests on the Arctic, and Antarctic, tropical flora and fauna at high latitudes, stable stratified oceans that may trend anoxic, and global temperatures five to ten degrees C above today's. These are natural limits of the system when available feedbacks die out. The PETM is probably an analogy for the most accessible extreme plateau. It's on the low end of the natural spectrum, didn't result in a massive extinction or oceanic anoxia, and again isn't accessible till slow feedbacks kick in (1000s of years). There is no easy way to get to carbonate feedbacks and generate Venus.

    You asked.

  • melanerpes

    Beat the drum slowly and play the fife lowly.

  • FelineCannonball

    As far as I can tell it comes down to these records of warming not being globally simultaneous -- at least with the precision in the original author's age models. That and the fact that other sites, mainly in the southern hemisphere show records of cooling.

    In the end it may be an internal climate oscillation or it may come down to the problem Warren states above. The paleoclimate record of global temperature involves a lot of averaging and short events may get cancelled out because of the precision and accuracy with which individual records can be studied. Events shorter than a century in the middle of the Holocene really are beyond our resolving power. You might see them in individual records but you can't get them to line up unless you force them. Better proxies, better calibration, and higher resolution records may help tease this out -- but you'll never get a sedimentary record that rivals instrument or satellite data.

  • FelineCannonball

    CO2 can be a feedback (insolation change -- soils warm -- decay increases -- CO2 released) or a primary driver (volcanic release, anthropogenic release). Both appear to exist in the geological record.

  • Joe_Da

    The conflict that I see is that the each of the individual proxies that I mentioned show that it is warmer during the MWP than the climate proxy from the closest location. For example the exposed tree stumps in the chilean andes show that it was warmer during the mwp than the tree ring proxies 100 miles away. Which one would be a better proxy, it would seem that the tree ring proxies are not picking up the warmth exhibited by the trees growing during the mwp where a glacier stands today. there seems to be a failure to reconcile the proxies to other known proxy data

  • Cranios

    You sir, are a Sun-caused global warming denier.

  • FelineCannonball

    🙂

    Solar evolution and orbital effects on insolation are pretty much mandatory to explain earth's climatic history (along with CO2, water feedbacks, weathering, plate tectonics, etc.). And in the next two billion years it is widely believed solar evolution will result in an ocean free planet earth as solar output increases and greenhouse gas negative feedbacks die out.

    If you mean sunspot cycles in the 20th century, then yes I don't see much evidence for short term solar output variation being a dominant controller of today's climate. I have a friend in France who is creating an incredibly detailed record of past sunspot activity to test the hypothesis though.

  • Joekeener

    Considering that the SCIENTIFIC METHOD mandates full disclosure of your data and methods I would like to see Michael Mann's unalterd temprature data.
    Where can I find that?

  • Joekeener

    SHHHHHHH, asking questions is unscientific.

  • Joekeener

    Just because every climate computer model has been wrong that has no bearing on believing that new computer models will be correct. You just have to have faith

  • Joekeener

    Honey, you need to get a real job. Or kill yourself.

  • Cranios

    I knew it!

  • Joekeener

    honey, just stop. It is over. Your lifes work is a complete waste of time. It will go the way of other democrat party debacles like slavery, segregation and eugenics.

  • Joekeener

    Oh so you frauds have learned that short term predictions can bite you in the ass when you are wrong so you are now pushing those out so far none of you will live to face the consequenses of your failures.
    Got it

  • Joekeener

    Where does it say in the Scientific Method that a 'consensus" makes a theory a fact?

  • Austexchili

    Oh! Climate graph. I thought it was a graph following Al Gore's PE incidents.

  • FelineCannonball

    Consensus just means consensus. Nothing more. And GCMs for the most part haven't even tried incorporating short term climate oscillations except as noise. Accurate prediction of the state of next years El Niño is still a difficult reach even if it's your goal. As is next months weather in Phoenix.

    Since they didn't try to incorporate that complexity I fail to see how they're moving the goal posts by not worrying about the model not reflecting it. Some people are trying to tease apart potential model problems from design limitations. To me I'd say its a bit like watching a pot boil and publishing predictions as it progresses. For most of us it's probably better to wait till 2030 than watch the play by play.

  • marque2

    It will be a problem when your pants spontaneously catch fire 😛

  • marque2

    I don't know what we had before ICE but I believe it is part of Homeland security now. ICE's goal is to let undocumented immigrants flow over all our borders at about the maximum rate they can be encouraged to vote for Democrats. Not sure what it has to do with Global Warming though,

  • Global Warming is "garbage." If you think not than Al Gore and the dunces at University of East Anglia aren't alone. Beyond garbage it's a complete fraud and their emails prove it, if not the expansion of polar icecaps and last decade-plus of Global Cooling. (Remember the 70s when that was all the rage?)

  • marque2

    Don't forget arbitrarily adjusting temps post 1959 upward and pre 1959 downward in docs. Shutting down stations showing cooling as "cost cutting measures" Shutting down and losting datat from Chinesse weather stations showing cooling. Adjusting satellite data upward. Ignoring data from weather ballons used by third world countries, due to technical incompitence, but not ignoring their weather stations - which are showing extra heat due to lack of maintenance. (The weather balloons were not showing PC data)
    I would guess at least 1/2 the 1.3degC increase is due to arbitrary data manipulation.

  • It's when they manipulate the graphs and cherry pick the data, which of itself is completely questionable, we have a selection of biased scientists designing studies to suit their own political agenda. That is corruption and malpractice. Scientists like Mann and the group of others exposed in those email from UEA should facing an honest inquiry and the truth should be exposed rather than lied about.

  • null_status_of_parameters.sh

  • Whitney

    Perfect. May I use this as my standard rejoinder to conclude any argument with a Leftist?

  • Matthew Slyfield

    I don't think Mann ever looked at unaltered temperature data.

  • dcalfine

    The editors of Science should sue Drum for defamation, but alas they have given him a truth defense.

  • Chemman01

    Thanks,
    I didn't have a lot of time to reply so I chose what is the major issue with modern data. Everything you pointed out are also good points

  • Geno

    I conducted scientific research for 40 years. Your above statement reeks of naivety. Science is about skepticism, not consensus. I also learned that peer review means little anymore because most of what we once knew about scientific ethics was tossed after the progressives began to take over everyone's minds. Too bad for mankind. Regarding the current science on global warming or climate change or whatever is now in vogue, it is nothing more than a steaming pile. The earth's climate is one of the most complex systems in the Universe. Anyone who believes that it can be simulated with such simplistic models and such horrid data is delusional or just plain ignorant.

  • FelineCannonball

    I'm not praising consensus nor saying it creates facts or truth. Consensus simply means what the majority of a group take to be the most likely hypothesis. Consensus has been wrong on large matters in geology, and is frequently wrong in small matters. So my statement that consensus is what it is.

    I'd rather the opinion of the smartest most experienced folks in a field to consensus and I'd rather a scientific consensus to the opinion of a layman. But none of these things are fact. Consensus isn't fact, publication isn't fact, peer review isn't fact, and blog comments don't make fact.

    I've tried to make clear the uncertainty in the models and their limitations. The unknown unknown could shift the "consensus" considerably. But there still isn't a cabal.

  • Also, one cannot even see that despite the dramatic rise in CO2, the increase in temps over the last 17 years is virtually flat. Part of the problem is that solar fluctuation is downplayed to the point of cartoonish absurdity by the warmists; secondly, CO2's effect on warming is not linear, it's logarithmic. I will be unusually fair and not call them outright liars; I'm content to call them stupid and incompetent beyond belief. But non-brainwashed folks who follow these topics closely can take their pick.

  • bbss

    When regulated banks use financial risk and pricing models that are as badly nerfed as most of the climate models I've seen, there are in fact regulatory consequences, and in extreme cases, jail time. Even despite these draconian consequences, banks, and their models, still fail from time to time.

    Not a single climate model has been subject to the rigour or backtesting required of most financial capital models today. Their predictive power is crap, and would attract massive capital penalties at the very least. I dream of the day climate "science" hippies subject themselves and their models to this kind of rigour.

    From a financial mathematician.

  • stephencarter

    There doesn't need to be a 'climate change cabal' etc. The Left has a shared discursive bias for a statist control of science and intellectual inquiry, and for the economy and virtually all sectors of society to be placed under the control of such enlightened 'guardians'. There doesn't need to be a cabal, the Left moves in the same direction with much the same goals. They are doubtless completely independent, and act independently, yet they are taking us to the same horrific brave new world destination.

  • marque2

    Wasn't trying to bag on you - just trying to be of support.

  • AugustineThomas

    If C02 is warming the environment so much, why doesn't it ever actually get warmer? I just read that it was hotter in Roman times.. Did the Romans have coal fired power plants?

    Well, at least you've partially admitted you have a problem..

  • Sharpshooter

    Evasive non-answer.

    Let's try again: Why is the sequence BACKWARDS?

  • FelineCannonball

    That is the answer for CO2 in the geological record. I don't know what else to tell you. Local warming often proceeds CO2 increases and it doesn't explode scientific theory.

    If you are asking about something else, you'll have to clarify.

  • Chemman01

    Sorry,
    I didn't take it as you bagging on me.

  • barryobarma

    If we just went back to the 10 commandments and The Commion Law, it would probably take 2,000 years to evolve to the levels of stupidity currently demonstrated by "settled science", big gummint and over-legislation.

  • Neo

    "... the declared consensus are somehow anti-science"

    First, the declared consensus is anti-science. No somehow is needed. Science isn't about consensus, and it is never settled.
    Besides, the last world-wide governmental consensus was that "Iraq had stockpiles of WMDs" and that didn't work out too well.

  • nickshaw

    I believe the dog ate it. 😉

  • MostlyRight

    Appreciate your responses so far and I'd like your take on something that has bothered me since reading it. I understand the modelers began their modeling with regards to AGW in the 1970s. They modeled "backwards" in time to check their models vs. available historical data (mostly proxies like tree ring, ice core, etc) and forward to give us predictions of the severity of warming. Of the 70-something models predicting forward from the 1970s on, however, ALL of them have predicted high when compared to the 35 years or so of actual temperature data we can compare the models to. As a civil engineer with some background in the sciences, this does not give me a satisfactory level of trust in the ability of the modelers in your field.

  • nickshaw

    It's closer to 0.03% but, look who's counting. 😉

  • nickshaw

    Sorry, I won't play their word games!
    I will call you a skeptic, if you don't mind.
    As am I.

  • FelineCannonball

    Modeling isn't really my field, but I can give you a bit of a different take on this.

    ---------------
    One issue is that pundits and publicity folks on both sides have been providing biased accounts of model data (cherry picking) for public consumption. ALL models and ALL model simulations don't exceed observed global temperature change. Neither are they as alarm-inducing as some of the plots posted by AGW advocates.

    Firstly, old models are historic artifacts. In the 1970's and 1980's they were cartoonish, based on incomplete or inaccurate observational data for starting point and parameterization, and based on single runs without exploration of sensitivity to small differences in starting conditions or errors in observation-based parameters. The heart of these models (and today's) are based on chaotic processes which are very dependent on starting conditions. Running multiple model iterations allows you to tease out the forcing you are trying to study from weather-dependent short-term progression. In longer projections this matters less, but if you are trying to estimate the global air temperature the next year or decade it is extremely sensitive to this months weather. This goes to the point of what models are intended to recreate -- the impact of various forcing mechanisms buried underneath chaotic weather and short-term climate. Back then, and for the most part today, models are not trying to create a precise short term prediction. In the short term they model the impact CO2 forcing has underneath the short term chaos. In the long term, if a significant forcing is imparted by the model it wins out over the noise and the model starts to be testable.

    Considering this 1980's era models did provide ball park results that weren't that bad. Hansen (1981) included most of the major drivers and processes we think about today (using 1D and simplified 3D models) and underestimated current warming by about 30 percent. Hansen (1988) modeled using 3 different scenarios related to uncertainty in feedbacks and future CO2 production. One over-estimated current warming a lot. One over-estimated current warming a bit (but within expected uncertainty). One underestimated current warming by a bit. When you may see figures in blogs which show the first scenario vs. "global temperature" you and I would be lead to say Hansen was full of crap. From a science perspective they are all historic artifacts. Resolution has changed, modeling complexity has changed, understanding of processes has improved, observational inputs have improved. Hansen got a lot right and the physics of these models is still the basis of the more sophisticated models today, but there's absolutely no reason you would fall back on those models for predictions today, any more than a civil engineer would utilize an old model for wind stress on a suspension bridge. It's just a history lesson.

    The first reasonable one dimensional model of CO2 forcing actually dates back to the 1930s. The author actually came up with a reasonable sensitivity per doubling of CO2 and understood for the first time (in publication) the true significance of the atmosphere being opaque or partially opaque in several infrared bands, treating the system as a series of "greenhouse" glass panes rather than a single filter saturated in most CO2 absorption bands (a thought that had tripped up a number of scientists and kept the concept "fringe" much longer than it probably should have). However, he was incapable of deriving it from first principles, and it is likely that he adjusted parameters in the model so that they made some ball park sense for current climate and ice-age climate. I guess you'd call it an empirical model utilizing a largely correct dynamic interpretation of the primary driver. It was more of a "ice ages could be driven by atmospheric carbon cycling" paper than a proof.

    If we go back to "cherry picking" for the moment, it's informative to look at all the model results going into the commonly graphed IPCC mean projections. The full envelope representing a large number of models and model runs includes curves that predicted subtle temperature drops over the last two decades, well below any compiled observational trend. So no, they ALL don't overestimate temperature increases. A useful model archive is here: http://www-pcmdi.llnl.gov/ipcc/about_ipcc.php For any recent 5 year period I'd guess you'll find 25% of the model simulations predicting a temperature decrease. The spread is significant in the short term. What is being modeled however, is the longer term impact of CO2 forcing. Here the models converge toward consistent predictions of warming.

    --------

    In addition to "cherry picking" of model results, you sometimes see advocates republishing model results with shifted Y-axis. I guess the thinking is that where you match up the model and observational data is arbitrary and the authors and reviewers didn't work through that aspect with much thought.

    A different type of cherry picking is done with the instrumental record. I see this with advocates on both sides. As it turns out there is a significant spread in current estimates of global temperature (as it's a derived number no matter how you do it and quite involved to get at). With the instrumental record, one issue is how you weight various sparsely placed whether stations and interpolate between them. I personally think the future is to utilize more sophisticated weather models in the interpolation so that instead of pretending the temperature half way between Svalbard and Pt. Barrow is itself halfway between on the °C scale you utilize concurrent weather models that respect real features of atmospheric circulation. Satellite temperature data is derived from a complex algorithm because looking through the atmosphere in thermal infrared is complex. When we work out the bugs, the best technique will be to integrate calibrated surface and satellite data using concurrent weather models. It's a little harder than looking at a mercury thermometer.

    ---------------

    In terms of short term climate predictions a few more complexities arise. In addition to weather related complexity, we have multi-year climate oscillations related to ocean-atmosphere heat exchange. A simple well-known one is the El Niño cycle which has a large transient effect on global atmospheric temperature. You'll frequently see plots by advocates pointing out that temperatures haven't changed much since 1998. As it turns out, 1998 was the last large sustained El Niño event and a year climate researchers would expect to be a high temp outlier. Basically a lot of energy was transferred to the atmosphere in the equatorial Pacific. This is a process that modelers didn't model at all until recently and model poorly today. Other SST oscillations are probably also important on a decadal scale, and although they are not expected to have significant effects on model predictions for 2100, modelers are trying to incorporate them and better understand the larger problem of the ocean as a heat sink (which will matter in 2100).

    For long term predictions I'd say better modelling of atmospheric moisture feedbacks, more detailed dynamics of the atmosphere and ocean, better observations and radiation impacts of aerosols at different levels in the atmosphere (NASA is doing some cool detailed work on this front) will help constrain future climate trajectories.

    Models give a very large envelope for climate variation for short term and still have a fairly large error bar for long term predictions (but on that scale all predict significant warming). They all see CO2 as a significant long term driver of climate and they use first-principle real physics to get there. The same models with different starting inputs can fairly accurately describe paleoclimatic end-members -- glacier free worlds in the early Cenozoic, Northern Hemisphere ice free in the mid Cenozoic, etc. -- using CO2 estimates from multiple proxies and temperature estimates from multiple proxies. It's cartoonishly accurate and the models don't devolve into anything ridiculously far from observations in the sedimentary record.

    These models will always be somewhat cartoonish, never be precise, and never be that useful at predicting year-over-year global temperature change or ten year trends. They are useful for probing the impact of different potential forcing mechanisms on climate and understanding the dynamics of the system using real-world physics, but they are full of limitations when you try to use them for things they are not designed to reproduce. As I have my hand in paleoclimate and petroleum communities and work with modelers in each, I'll say that paleoclimate models are more sophisticated, more open-source, and more picked over from the ground-up by competing authors. The modelers are circumspect and cautious about over-applying model results.

    It's when you get to the advocate literature that you start having problems. And yes, they are not always distinct. Some scientists advocate in other forums.

    ----------------

    As a paleoclimate researcher, I think about the modelers as playing a minor role in demonstrating the plausibility of causation and feedback mechanisms in the geological record, rates of processes / lag times in response to external forcing, and a glimpse at the dynamics of climate in systems significantly different than today's. The scale of paleoclimate change is obvious, proxies for CO2 (i.e. bubbles in Antarctic ice) and temperature are pretty solid. Without the modelers we'd still have an idea about the instability of earth's climate and what recent past climates with 400ppm, 800ppm, 1600ppm CO2 looked like (spoiler warning: all living things didn't spontaneously combust)

    -----------------

    If you want to learn more about errors in short term climate predictions or see some plots with more than the mean model result and some model simulations that didn't exceed the observed trend you can read this recent post: http://www.realclimate.org/index.php/archives/2013/12/the-global-temperature-jigsaw/#more-16547 I guess it's an advocate forum of sorts, but they have a pretty strong bias toward burying themselves in the science and explaining what isn't constrained or completely understood.