Bishop Hill Conveying truth 4
Apr 3, 2012 Climate: Models As my thoughts have turned increasingly to climate models in recent months, I thought I might take a look to see how Julia Slingo described these mathematical behemoths in the briefing she sent up to central government in the wake of Climategate - I have criticised this document on a number of occasions in the past (1, 2, 3).
Considering just how central climate models are to the case for DAGW* I was taken aback by how little Slingo had to say on the subject. And what she said was, to say the least, surprising. Here it is:
Models have been used that take into account all the factors that influence climate. These models have been able to simulate the historic changes in global and regional temperatures and have shown that most the warming over the past half century has been caused by the rise in greenhouse gas concentrations.
There are a couple of passim mentions elsewhere in the document, but in essence that is what the Met Office feels that people in government need to know about climate models.
It's hard to know where to begin. I wonder whether anyone will seek to defend Slingo's paper as a fair representation of the reliability and importance of climate models.
Readers should feel free to critique Slingo's words, but please avoid venting. I'm more interested in what policymakers should be told about climate models. I think we should allow ourselves slightly more space than Slingo - shall we say eight sentences?
*Dangerous anthropogenic global warming - "dangerous" being perhaps a less emotive term than "catastrophic".
Reader Comments (111)
bill
“keeps the show on the road, the day of reckoning at bay...”
Funnily enough, I was trying to make exactly that point to Mrs P last night, but she seemed unconvinced. Nice to know it’s not just me...
If Dr (I presume) Slingo is so enamoured of computer models, perhaps she would like a trip in a freshly-designed aircraft before all the practical pre-flight testing is undertaken?
Roger Pielke, Sr. has long pointed out the deficiencies in climate models.
http://pielkeclimatesci.wordpress.com/2012/03/05/two-further-examples-of-the-misuse-of-climate-models-for-multi-decadal-regional-climate-impact-assessments/
"This paper illustrates how model predictions, which have shown no skill at predicting changes in climate statistics, are inappropriately used to claim forecast skill in the coming decades."
Come on Richard, support your boss here.
Experts are about as good at predicting the future as a monkey throwing darts at a dart board to replace the monkey with a computer will still give you the same dart board guesses.
Less emotive - and watering down the message of doom that the Met Office was peddling.
Having looked at Slingo's lengthy publication record, it is interesting that a lot of it is cloud physics.
That area of the models is wrong.because Sagan's two stream approximation is wrong.
So, it's not her fault the climate models are wrong.
Computer modelling of complex systems is notoriously difficult to do, and does not have a good track record. We know that the complexity of the climate system is such that predictive models of the kind we use for weather forecasting support are inevitably incompetent when it comes to forecasting the climate.
John
I like what you write but this is not quite correct. The UK Met Off claim that their climate/Weather model is the same thing and that they can predict upto 10 yrs. When they announced this in 2009? they were planning to sell the 10yr forecasts commercially.
Dear central government,
What answer would you like?
Yours, JS
*Dangerous anthropogenic global warming - "dangerous" being perhaps a less emotive term than "catastrophic".
Sorry but I cannot see the difference. Both are very emotive adjectives for which there is not reason.
[Venting]
Apr 3, 2012 at 12:15 PM | stephen richards
I must say, BISH, I didn't see it as venting !!! When I vent it is normally much more vitrolic and much less factual than this particular comment.
Beneficial anthropogenic global warming and plant fertilizing. Seven words. Now nine. Now eleven. Now thirteen. Now fifteen. Now seventeen. Now eight phrases; at this point sentences are beyond me.
=======================
Call it CAGW or DAGW or whatever you like -- even though it is dying, its proponents are carrying out a scorched earth policy (i.e. entrenched legislation) that will take decades to unravel.
The Australian Government's Carbon Tax comes with guarantees to companies that will be prohibitively expensive to overturn, leaving the people poorer, as is always the way with the policies created by Leftist zealots.
"they were planning to sell the 10yr forecasts"
Ten-day ones would be pretty useful!
Given the option of having to experience one or other of the following, which one would you choose not to undergo?
ca·tas·tro·phe n. 1. A great, often sudden calamity.
dan·ger n. 1. Exposure or vulnerability to harm or risk.
The Met Office were peddling the certainty of catastrophe - not merely warning of possible dangers.
stephen (4:00PM). Many thanks. But may I take you up on the bit you didn't like? I am not quite sure I see why you don't like it all...
Anyone selling 10-year climate forecasts commercially would either need very good professional liability insurance, or very carefully worded small print to make it clear that if the purchaser chooses to make expensive decisions based on the forecasts then that is entirely their responsibility. In which case, they, and you and I, could sell n-year ahead forecasts for anything you like, from weather, to the FTSE, to the winner of the Grand National, provided we can find the buyers of course.
Can you dig out some link with more info on this claim to be able to do 10-year forecasts? I know they publish forecast for global mean temperatures (and they have had a +ve bias of late) but these are of no obvious practical use by themselves since no-one lives or works in this global mean climate - the global mean temperature is a computed and not an observable quantity, and in general it is the timing and extent of excursions from average values which are of most practical concern.
Is it conceivable that the Met Office may use their practical success with their GCMs in weather forecasting to suggest that these same models must therefore be good for climate forecasting, or perhaps have taken all important influences 'into account' in each case? Something like that was done (as I recall it) in that notorious documentary featuring Paul Nurse as an itinerant savant. My point is that the practical value of these models in weather forecasting does not support their being useful for climate forecasting.
“Call it CAGW or DAGW or whatever you like”
I’d settle for E(exaggerated)AGW, but once the W has stopped for long enough it will become rather academic.
The doom-mongers have already spotted this, of course, and are using climate change/disruption instead, although it hardly matters what you call it if it doesn’t exist.
@James P
10 Day forecasts? No problem - Norway Met Office at your service, for a settlement of your choice and in English. Hour by hour, ten day, detailed, and more.
Infintely better than SlingoBilge™
http://www.yr.no/
Bish
Haven't really followed this thread I'm afraid - busy day - but I'm not sure exactly why you are surprised about Julia's statement:
A figure which illustrates her points is here.
NB when I tested the above link, it didn't seem to work so here is the URL:
http://ipcc.ch/publications_and_data/ar4/wg1/en/spmsspm-understanding-and.html
Two things to note:
1. To a climate modeller, "regional" means continental scale (or several countries) - eg: a "regional climate model" general covers Europe, or the Indian sub-continent, or some region of comparable size. We don't mean regions as small as, say, South West England. So the charts on the figure I link to are what Julia means by "reproducing regional temperatures".
2. The quote is specifically about temperature. As we have discussed before, precipitation is much more difficult!
Cheers
Richard
Apr 3, 2012 at 4:54 PM | SayNoToFearmongers
The Norwegian Meteorological Institute use the UK Met Office model..... :-)
Cheers
Richard
What would impress me is if the people who build climate models that look 100 years into the future could build a computer model that accurately depicts and forecasts the behaviour and fortunes of my 6-year-old daughter.
It seems pretty clear that a 6-year-old child is less complicated than a global climate model. After all, there are roughly 43 million of them in the world so there's no shortage of data points. Lots is known about them too.
So if there are people who can model the climate in 100 years' time then it should surely be a cinch to build a model that makes accurate predictions for my daughter over the same period. How much will she eat, and where will she be living in 2016 for example? How many children will she have between 2025 and 2050? What will she earn? How will she heat her home when she is 35 years old? How many cars will she own?
All of these are things you'd need to know to predict the weather in 100 years' time. It seems to me it ought to be equally easy to apply the same genius to modelling people. Really, what's the difference?
Richard Betts,
Slingo:
All the factors?
I asked at Lucia's in a thread that seemed competently technical if there was a running list of factors. It isn't hard to imagine that the factors which are taken into account involve 90% of the energy of known forcings and interactions, but is that actually the case?
Are all the larger forcings recognized such that the ones ignored are almost certainly de minimus?
I don't mean this to be a trick question, but it is a question which we asked ourselves from time to time during my career when we were setting out to design something in an environment we thought we didn't fully understand. We would make a list of all the things which could participate in, or have an effect, on our project, design for the ones we understood, "ball park" the magnitudes of the effects we knew were there but were uncertain of, worry about others, and ignore the little ones.
This didn't always work, but when it didn't, we could at least share the audit trail which would help the next guys not run off the track at the same place we did.
listing and quantifying the uncertainties?
SNTF - ah yes I look at the Norwegian forecast also. As you say it is often quite different from the UK MO forecast, and not long after I mentioned this on a BH thread Richard Betts then pointed out that they use the same model. I expect that the difference is due to the MO having a faster computer and/or they do the model runs more frequently. The output is similar for the next 24-72 hours but after that they both are pretty inaccurate as to what day the cold front will arrive, or how long the high pressure will hang about. For example, when I looked at both forecasts at the end of last week, the MO did not mention the 10-15cm of snow we have had here today in Scotland, as it was beyond the 5 day period; the Norwegians did mention the coming cold air and snow, but forecast it to come not today but in three days time. So beyond 3 days and even the 'best' model is very hit and miss. Of course we can be sure that this computer model is much more accurate for 50-100 year
forecasts predictionsprojections. /sarcIndeed didn't Slingo say in her evidence to the parliamentary inquiry that they were very confident in the long term climate model as it was the same one they used for the 5 day forecast and as such is verified on a day to day basis?
In talking with a government official about climate models, I might ask them what degree of confidence they have in:
- models that predict the economy (and which ones predicted the collapse of the mortgage market)
- current weather forecasts beyond a few days
In other words, point to something they know doesn't work that well and ask them what makes them think the same thing will suddenly work when applied to climate predictions.
Richard good to see you have popped in. I now see that you have made the point about the Norwegians using the Met Office model again.
Just out of interest, am I right about the reasons for why there are discrepancies in the results of the two forecasts? (MO does more frequent model runs and therefore can update forecasts more often)?
Richard Betts,
"These models ... have shown that most of the warming over the past half century has been caused by the rise in greenhouse gas concentrations."
Do you really not see what nonsense that is? I just mean in terms of logic - it is nonsense. How could the models possibly show what causes the rise?
The error on our measurement of the top of atmosphere energy balance is three times the signal we hope to find for the increased co2 effect.
We have no reliable long term records for cloud cover variability, sea surface temperature, ocean heat content or rainfall.
We don't understand most of the carbon cycle or how we can parameterize it.
The biggest factor influencing oceanic heat loss is wind, and we can't measure it reliably.
Our models can't tell us what will happen next week, let alone in 100 years.
Apart from that we're doing great, send more money and a bigger computer.
Nice one Tallbloke. The bigger computer's in the post.
Thanks Richard, I'll build a proper model based on orbital parameters, solar prediction and upper atmosphere ozone changes based on UV measurement.
Results ready in a few years.
From Julia's 'ask the expert' document:
"At the Met Office we use the same model to make weather forecasts as we do to make our climate predictions, so every day we are testing the model and saying, ‘how well did we do with the weather forecast?"
When the Met Office predict a barbecue summer, or fail to predict that Heathrow will be snowbound, these weather forecasts are created by the same models used in climate predictions.
Do politicians need to know much more?
See:
http://www.metoffice.gov.uk/corporate/pressoffice/in-depth/ask/julia-slingo.pdf
James: ""These models ... have shown that most of the warming over the past half century has been caused by the rise in greenhouse gas concentrations."
Do you really not see what nonsense that is? I just mean in terms of logic - it is nonsense. How could the models possibly show what causes the rise?"
If you believe you know everything there is ti know about climate forcings and you've put it all into the models as Dr. Slingo appears to have told the politicians:
"Models have been used that take into account all the factors that influence climate."
And you the take out the known forcings one by one and the temperature decreases bit by bit until you're left with only one forcing and that is CO2 and you have identified, say, 0.3C of the 0.8C as being taken out by all the other forcings then the model has told you that 0.5C was caused by CO2.
Perhaps Richard Betts will correct me if I'm wrong, but I understand it's something like that.
In engineering this would be seen as an extremely hazardous way to draw a conclusion given the assumption that you couldn't possibly be sure you know everything you need to know to model a chaotic and extraordinarily complex system. However, it was decided a priori that CO2 was the villain of the piece and with the best will in the world it's difficult to understand how you could build a model that didn't point the finger at it if you already believe it caused the warming.
Maybe RB can explain it to us.
Models have been developed to simulate climate because co2 economics - above all else - was identfied, over 20 years ago, as not only a primary driver of wealth creation for rich and powerful oligarchies, but also as a facilitator of the bureaucratisation those same powerful interests have sought in order to fulfil their agenda.
The outcome of any climate model is much less a viable snapshot of a possible future weather simulation, but more a tool upon which vested interests and government(s) can rely as a template to plan and shape society to their ends, decades into the future, for their own security.
@Richard Betts
Thanks for clarifying the model issue for yr.no - given that this thread is about conveying truth, would it be fair to ask how that Norway, a country with a population of well under 10% of that of the UK, can provide far clearer, more useful and in my experience, accurate information for UK citizens from your model than the Met Office choose to communicate? If it is the same model, then either different inputs appear to be being fed to it or different interpretations drawn from its outputs.
climategate emails - Doubts about the models
0850.txt: Tim Barnett: " right now we have some famous models that all agree surprisely well with 20th obs, but whose forcing is really different. clearly, some tuning or very good luck involved. I doubt the modeling world will be able to get away with this much longer. "
4443.txt: Phil Jones: " Basic problem is that all models are wrong - not got enough middle and low level clouds. Problem will be with us for years, according to Richard Jones. "
0419.txt: Mike Hulme: " I am increasingly unconvinced by the majority of climate impact studies - including some of those I am involved in "
4933.txt: Rik Leemans to Mike Hulme: " In the new IMAGE 2.2 release, we discovered very strange behavior of the climate module. At higher C02 concentrations the climate sensitivity tents to be much less than expected. This is probably a result of our new parameterization of the oceanic carbon uptake, which also directly influences the heat uptake. Unfortunately, no-one in the IMAGE team is familiar with the code, the changes that are made over the last years in order to implement the regional aerosol effect. "
thanks to:
http://sites.google.com/site/globalwarmingquestions/climategate-2
plus, just to throw it in..
3066.txt: Peter Thorne again, commenting on a draft of the IPCC report:
" I also think the science is being manipulated to put a political spin on it which for all our sakes might not be too clever in the long run. "
Climate science needs its Alan Turing moment: 'On computable numbers, with an application to the Klimaproblem'.
Models have been used that take into account all the factors that influence climate.
This is arrant nonsense. If it were true climate predictions would be available in a book entitled 'Global Climate 2012 - 2100' which would never need to be updated.
Perhaps the Met Office is about to publish it?
As an aside, a complex model that does work is to be found in 'Planetary and Lunar Coordinates 2001-2020' ISBN 978-0118873123 published by The Stationery Office in August 2000. It will not be updated, nor will it need to be updated. The model takes into account all the factors that affect the positions of the moon and planets (and yes, it is a long list).
Computer models that can neither be verified nor validated can not produce ANY useful or original information on a system. This is a fundamental and self-evident truth. Climate models simulate non-linear and chaotic systems, including a large and uncertain number of dependent and independent variables. Such models can never be verified or validated. Any attempt to add meaning to the output of such models by comparing their results is scientific illiteracy of the worst kind.
The Alan Turing moment isn't needed. The climate models were broken from the very start in 1967 .
Apr 3, 2012 at 10:52 PM Roger Longstaff
Yes, doing this is even worse than claiming that models have been validated by checking they reproduce the climate history used to create them.
Years back I read a book called "Systemantics - how systems work and especially how they fail" by John Gall. It's full of aphorisms such as 'When a fail-safe system fails, it fails by failing to fail safe'.
In the chapter on what goes on inside systems, the author notes: 'To those within a system, the outside reality tends to pale and disappear'.
This is exactly what has happened to climate science. Groupthink rules. Their models have become (to them) the reality and they 'validate' their models by comparison with other similar models.
The total absurdity simply escapes them.
Apr 3, 2012 at 5:23 PM | jferguson
The forcings used to drive the 20th and 21st Century simulations by the AR4 generation of climate models are shown in this table.
The two Met Office models (HadCM3 and HadGEM1) are in the bottom two rows.
Cheers
Richard
Apr 3, 2012 at 5:40 PM | James Evans
Hi James,
Fair point - "show" implies proof and of course the models don't do that. Maybe it would have been better to say something like "indicate" or "suggest", although those words seem a little weak compared to how confident we are in this particular issue. I've been trying to think of a word between "suggest" and "show" but can't!
The point of the sentence is to say that the models embody our current understanding of climate processes, and that when we use the models to estimate how the climate system has responded to different combinations of forcings (GHGs, aerosols, land use, solar, volcanoes) and internal variability. When we do this, neither the natural forcings nor the internal variabilty explain the warming over recent decades, but the response to GHGs (which is offset a bit by aerosols) does explain the warming. All forcings (natural plus anthropogenic) give the best match between models and observations.
Cheers
Richard
Apr 3, 2012 at 7:36 PM | geronimo
Hi geronimo,
Attribution studies are done sort of like you suggest, switching on and off different factors, but it's not as simple as how you describe. We don't just remove the forcings one by one until we are left with GHGs alone. We also do it the other way round and remove the human forcings leaving only the natural ones, and also remove all forcings to see how much the internal variability may have contributed.
As you imply, the possible non-linear combinations of the different forcings may mean that just doing it the first way would be unreliable.
We didn't "decide" a priori that CO2 is the bad guy! The simulated atmosphere responds in a similar way to a given level of radiative forcing (eg: 1.5 Watts per metre squared) no matter whether that forcing is due to CO2 or other factors such as a volcanic eruption.
Cheers
Richard
Apr 3, 2012 at 7:22 PM | ZT
The barbecue summer was a problem with communication of the uncertainty in the forecast. The good success rate of forecasts on daily timescales suggests that the basic atmospheric processes are right - the difficulty with seasonal forecasts (eg: BBQ summer) is that the effects of chaos kick in, so the forecast can only be probabilistic - this was not communicated effectively.
Did we fail to forecast snow at Heathrow? I assume you mean December 2010? I thought we did forecast snow, but the response to the forecast was not effective.....!
Bish
To respond to your point on why Julia's document has relatively little focus on models, this is precisely because it is better to focus on evidence from the real world not the models!
I bet if the document focussed mainly on models and had very little in it about observations, you wouldn't have been very impressed...? :-)
Cheers
Richard
Thanks much Richard Betts,
John
Richard Betts
Re: another word, you might consider (if you haven't already) the word "depict"
It's almost synonymous with "show" in some contexts, but does not have the connotations of a proof, more of a representation or illustration, with more or less precision depending upon details and context.
[O/T]
Slingo: "... models have been able to simulate the historic changes in global and regional temperatures and have shown that most [sic] the warming over the past half century has been caused by the rise in greenhouse gas concentrations". But this finding has yet to be confirmed by econometric (LSR) tests. Why do climate scientists like Slingo et al refuse to perform or commission such tests? Could it be because they do not yield the desired answer? That is certainly the case with Hegerl & Zwiers in their Chapter 9 of AR4 WG1 2007. Zwiers has even co-authored a textbook on statistical analysis of climate change, yet in that chapter there is none at all, only "results" of so-called detection and attribution models.
lapogus
Personally, I find it far better to look at the satellite images on-line and just watch what happens. I would say that it is fairly easy to see what will be coming in over the next few days by watching the progression of images over time. Not that you need a computer to do that as I use to go to the flight weather station at the airport and look at the printouts back when I was a private pilot. Computers only make them more convenient today.
Richard Betts
Ever think about running for office? You might think about it. Always good to have an alternative career path. With your Clintonesque style you would do well in that.
[Response to O/T]
It seems that Dr. Slingo may be conflating temperature with climate. But, of course, there is much more to climate than temperature. If she is not conflating the part with the whole, I wonder if she is familiar with what the IPCC’s 4th Assessment Report (AR4) and the US Climate Change Science Program (CCSPO) assessments of models have to say about the ability of current climate models to retrodict (or hindcast if one prefers) climate. If I may quote from the paper, Trapped Between the Falling Sky and the Rising Seas: The Imagined Terrors of the Impacts of Climate Change, [available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1548711; see pp. 12-13]:
To summarize, if climate models cannot hindcast the past—and they can’t according to even the IPCC and the CCSPO—they are unlikely to forecast the future. And, in fact, they have, if anything, substantially overestimated the rate of warming since the late 1990s. That is, they are failing the acid test of reality.