Thursday
Aug072014
by 
Bishop Hill
Bishop Hill Forecast = projection = scenario
Aug 7, 2014 Climate: Statistics Matt Briggs has a post addressing the argument that climatologists are not making forecasts, they are doing projections. This argument, which has always struck me as a master piece of obfuscation, seems not to hold up under scrutiny.
Perhaps afficionados of "projections" can explain why he is wrong.
Reader Comments (60)
Even spin-meister-in-chief William Connolley acknowledged some time ago that he was hopelessly confused about the distinction between climate projection and prediction.
I don’t know about scientifically or statistically, but there is some linguistic justification for using scenario or projection as opposed to prediction or forecasts.
“Based on knowledge of existing conditions and normal weather patterns I predict/forecast that tomorrow it will rain.”
“If I project today’s weather into tomorrow then it should be increasingly wet.”
“Scenario A If the wind speed continues then tomorrow it will rain but scenario B if the wind drops it won’t rain until tomorrow night.”
Option 1 suggests more knowledge and certainty than 2 and 3. Option 3 suggests that wind speed is the only unknown. Option 2 is easily ignored because it pretends no understanding of how weather works only knowledge of what has passed and a belief in persistence. Climate catastrophists are far too fond of projections.
In the classic Hansen temperature chart, he gave different possibilities for an unknown – green house gas emissions, which was fair enough. However the scenarios failed for reasons other than the CO2 and methane. Even had the emissions been higher or lower than his scenarios then they would have had value if the resultant temperature had been consistent with the chart (ie more emissions, more temperature).
Of course when the projection, prediction, forecast or scenario is dependant on many more variables than you are able to accurately include what you should call your wiggly line pretending to be future is a ‘guess’.
Gavin, who will survive the climate science disaster better than others, explained some time ago that he can do projections but not forecasts, because of known and unknown unknowns. There is no model, not even in theory, that can forecast when a volcano will erupt or what the solar activity will be, in 2050. So the models, peace be upon them, are meant to tell us how temperatures will change assuming nothing else much happens.
Projections are conditional.
"If you stand out in the rain, you'll get wet." is a projection.
"You'll get wet by standing out in the rain." is a prediction.
A projection becomes a prediction when its conditions turn out true.
Projections can be wrong, climatologist cannot be wrong, therefore there are no projections/predictions/forecasts/(etc.) in climatology.
Is it not possible that he may be right?
I thought this was all fairly obvious. Climate models, for example, can't make absolute predictions because we don't even know what we will choose (or be forced) to do. For example, our future emissions pathway will depend on many factors and so all that climate models can do is use some kind of possible emissions pathway and then use that to determine what may happen if we follow that pathway. Therefore, it's a projection, or a conditional prediction. If our actual emissions pathway differs from what was used in the model, what actually happens will probably differ from what the models predict. Therefore if what happens in the future does not match what the model suggests, that could simply be because the assumptions about some of the future conditions didn't match what actually happened (it could be that the model is wrong, but you can really only know this if all the conditions were met and the model result still didn't match reality).
Similarly, as Omnologos says, there are other factors that we can't predict accurately in advance (Solar variability, volcanoes). Hence, as Nullius says, a projection becomes a prediction if all the conditions turn out to be true (or, I guess, if we come to know that all the conditions will be true). If you can't know this, then you would normally consider many possible conditions so as to cover the range of possible results, each of which will depend on what conditions may actually happen.
I may change Nullius's example
"If you stand outside, and it rains, you will get wet" - projection
The distinction is moot when policy decisions are being demanded on the basis of said projections/predictions.
"Is it not possible that he may be right?"
Briggs is not quite right in saying 'predictions = projections', but he is right in saying that a scientific theory is judged by the accuracy of its predictions, and if it makes no testable predictions then it's not scientific. Furthermore, that if it makes no predictions, it's not much good for policy planning - contrary to the expectations generated by all those charts they publish of future temperatures screaming upwards.
It's the old argument about validating models again. Climate science has to make testable predictions different from all the alternatives, which have to come true, before we can use it for policy. Lots of graphs are presented, which the unwary take to be predictions, and argue for various policies on that basis. But when those charts don't come true, climate scientists say those weren't actually predictions, they were just projections, and their failure doesn't falsify the theory. (Of course, if the charts do come true, that's widely taken to be confirmation of the theory. That's the 'confirming the consequent' fallacy, of course). Briggs is saying you can't have it both ways. Either you admit you can't predict the future, or the failure of the predictions you make falsifies the theory.
It's a bit of a brutal simplification of a more complicated situation, but he does have a point.
Nullius,
Maybe I misunderstood what Briggs was saying then. Of course, if we're talking about climate models, then they're not really theories, they're complex models. They're based on basic theories and do have some parametrisations (which are typically constrained by physics and are typically used to model sub-grid processes) but if a climate model fails to represent reality, what theory has failed? I would say possibly nothing. As you say, it's more about validating models than about theories being falsified. A complex model can clearly be wrong. A wrong complex model, however, doesn't imply that some fundamental theory has been falsified (unless by theory, you mean complex model, and by falsified you mean wrong).
That's in your opinion of course. In an ideal world, this may be the way to go. However, you have to hope that when something does come true, or not, that we don't conclude that we should have done something sooner. Of course, relying only on climate models would appear to be ignoring much other information that tells us something of how we will warm if we continue to increase our emissions.
Of course he has a point, but it is a brutal simplification. This also isn't meant to be some kind of competition in which one side says "told you so". There would seem to be two reasons for doing climate science. The scientific interest in understanding our climate, and the possibility that we might want to provide information to our policy makers to inform policy decisions. Climate scientists can only provide the best information that they have available. What we (our policy makers) choose to do with that is up to them. If they decide that the information is insufficient to be used for policy decisions, then that's well within their rights. We just have to hope that they're not wrong.
One issue I have with this whole line of reasoning is that it appears to be setting an impossibly high bar. In some people's views climate models would need to be verified/validated to such a high accuracy that we would never do anything and that - in my opinion - is not the way in which one would use evidence to make policy decisions. That's pretending to use evidence to avoid making a policy decision.
Splitting hairs by any measure.
A.T.T.P.
What you say is true. The maddening part to skeptics is that the climate community does not inform policy makers of all the relevant information, mainly the uncertainties of the projections, and the limitations of the data. Just the reverse, too many activist scientists obfuscate the limitations to induce a certain policy outcome.
Re Andthentheresphysics, isn't a model an algebraic representation of a number of causal hypotheses? Theories are hypotheses which have empirical support, ie are not contradicted by observed data, and are consistent with other theories which do. So what does it mean to say that climate models are not theories, they are complex models? Nothing in my lexicon - it reads to me as a tautologous statement, devoid of meaning.
Guy,
I'll see if I can explain (assuming that you're actually interested). A climate model is based on the Navier-stokes equations which have been extensively used in fluid dynamics and are essentially representation of well known conservation laws. There's also radiative physics that is again well-tested and almost certainly not wrong. There's methods for introducing convection that may be parametrised but, again, is quite well understood. There are, however, aspects that are less clear. Diffusion into the deep ocean, for example, can be included but there's quite a large range of values (I think) for the diffusion coefficient. There are also other aspects that are parametrised and that although the parameters are constrained in some way, may have a reasonably large range of possible values.
So, let's imagine that we run a climate model, that all of the conditions that we assume (emissions pathways, solar fluxes, volcanoes) behave as we expect, but that the model is clearly wrong (and defining wrong is also non-trivial, but let's imagine that the result is far enough away from reality that we all agree that it's wrong). What would we conclude? Could one of the fundamental conservation laws of physics be wrong? No, that would be unlikely. Could some fundamental aspect of radiative physics be wrong? Again, no, not really. Could convection behave differently to how we think? Again, no. What's much more likely is that some physical process has not been included or has not been included properly. Maybe one of the parameters that represents some of the sub grid processes is wrong. So, a climate model being wrong doesn't really mean that something has been falsified, unless by falsified you mean that the climate model is wrong. There's also issues related to resolution - there are just some processes that these model can't properly represent.
So, to be clear, I'm not suggesting that climate models can't be wrong. I'm suggesting that using the term "falsified" doesn't really make sense since a complex model being wrong doesn't immediately falsify some fundamental hypothesis. For example, what I have seen are those who will essentially say AGW = Climate models, therefore if climate models are wrong, AGW=falsified. This doesn't really make sense since AGW isn't climate models. Climate models are simply one way in which we can investigate the impact of anthropogenic emissions.
I'll add one other point which is that, scientifically, a complex model being wrong (or, more correctly, not properly representing reality) can be quite an interesting outcome. It tells you that your model is missing something and that there are aspect to the system that you're studying that you don't yet understand. It might not be a great outcome from a policy perspective, but it can be interesting scientifically.
ATTP,
" Of course, if we're talking about climate models, then they're not really theories, they're complex models."
I agree. They're an attempt to integrate everything we currently know about the climate system to better understand the implications of our ideas. They're based on a mixture of well-validated physics, speculation, approximation, and (error-prone) measurements. As a tool for exploring possible mechanisms and improving our understanding, they're excellent. As a tool for making predictions, they're unreliable and inaccurate.
"but if a climate model fails to represent reality, what theory has failed? I would say possibly nothing."
It depends what you have previously claimed. If you've taken the perspective I described above - climate models are only tools for exploring ideas - then I agree. We don't expect them to be able to predict the future and they don't. No big surprise.
But a lot of people of the "The Science Is Settled" school consider them to be the embodiment of scientific truth. Temperatures are going to rise 3.5 C or more by 2100 because the climate models say so, and anyone who doubts it is 'denying' science. It's this sort of polemicist that much of the sceptic argument is against. Yes, there are aspects of the mainstream science we don't agree with either, but on points like this, sceptics and climate scientists ought to be in agreement. Climate models are not intended for prediction, nor are they yet capable of doing so.
"A wrong complex model, however, doesn't imply that some fundamental theory has been falsified (unless by theory, you mean complex model, and by falsified you mean wrong)."
That depends on whether you regard the output of the climate model as the embodiment of the theory's predictions. A theory has to make falsifiable predictions, or it's not science. So if climate models are not what the theory predicts, then what is?
The failure to predict implies that our current understanding as embodied by the climate models is at least partially incorrect, or incomplete. That doesn't mean they're totally wrong, but it does mean they're not totally right, either. And the distinction matters, making a material difference to what they say should happen in the short term. The models as they stand now are falsified (as complete representations of the climate). That doesn't mean that models based on similar principles couldn't be built in future that passed the tests.
"A climate model is based on the Navier-stokes equations which have been extensively used in fluid dynamics and are essentially representation of well known conservation laws."
Navier-Stokes is probably pretty accurate, but discretised Navier-Stokes at a resolution of 100x100 km blocks of air is a pretty poor approximation - especially given the rather hairy mathematical properties of NS!
Don't go claiming 'it's based on the unarguable laws of physics' when you're fudging the mathematics like that. Same goes for the other parameterisations.
"That's in your opinion of course. In an ideal world, this may be the way to go."
In my opinion, this is a bedrock scientific principle. Anyone can build a model that makes wrong predictions. There are an infinite number of possibilities, each of them giving completely different answers. What distinguishes science is that it requires that models are tested before use. A model has to be shown to be capable of making accurate predictions before you can treat its predictions as anything other than speculation.
"However, you have to hope that when something does come true, or not, that we don't conclude that we should have done something sooner."
True. And we must also hope we don't conclude we should have waited, or done something else.
Back in the 1960s, overpopulation was the crisis of the day. Influential scientists told us it was already too late to avoid catastrophe, and that we needed to start instituting draconian population control measures before the impending crisis made even more unpalatable measures unavoidable. Governments listened, and a few such schemes were even started, before more level-headed economists in the West pointed out the flaws in the thesis. China's one-child policy came out of that scare. There are stories of mass sterilisation programmes in India and Africa.
When the governments of the world decided, in the end, to do nothing, the same concerns must have applied. What if they were wrong? What if they blew mankind's last chance at survival? But on the other hand, what if they started on the mass sterilisation and mandatory population control and it all turned out to be for nothing as the sceptics were saying? It's horrible to think about.
With these sorts of issues, you can't just think about the costs and risks on one side of the equation, as the people calling for action to save the world tend to do. You've got to balance both sides, and you absolutely *must* have reliable information, or at least be clearly aware of just how unreliable your information really is.
"This also isn't meant to be some kind of competition in which one side says "told you so"."
Agreed. Although it would have been a lot easier to avoid that if sceptics hadn't been denigrated, ridiculed, and dismissed the way they were. But we are where we are now.
"There would seem to be two reasons for doing climate science. The scientific interest in understanding our climate, and the possibility that we might want to provide information to our policy makers to inform policy decisions."
Agreed. And I think both are important.
"Climate scientists can only provide the best information that they have available."
Yes, and they *should*. They need to make clear to the decisionmakers just how reliable or unreliable it is. And they have to take on best practices from engineering to make sure it really *is* the best. And they need to be especially strict in policing the accuracy of the science, checking and auditing results, rigorously weeding out false or unreliable results, and they need to visibly and transparently uphold the absolute highest possible standards of scientific integrity.
This is not your usual academic empire building or back-biting. This is supposed to be the End Of The World we're talking about! They very idea that someone could withhold scientific data on the grounds that it was their own personal 'intellectual property' and they were relying on secrecy to get more grants...!! Are you serious?! And how could the rest of the scientific community let them get away with it?
This is not some competition where you've got to "win" against the sceptics and 'deniers'. Nobody should be refusing to do or allow check on the grounds that it might give ammunition to sceptics. You ought to be relying on sceptics to help you check it, and keep yourselves on the straight path. This is, after all, saving the planet we're talking about.
To the extent that I think the possible impact on climate is a serious concern, I get really *angry* at what certain scientists have done. This is not something to play games with!
"One issue I have with this whole line of reasoning is that it appears to be setting an impossibly high bar."
It's setting a necessary bar. That's what's needed to do the job. If it proves impossible to meet, then it's impossible. Pretending otherwise won't help anyone. Nature can't be fooled.
Science *does* set very high standards, and the higher the stakes, the higher it sets the bar. That's why it works, and why we trust it to provide answers to our problems.
There are, of course, many other ways to draw conclusions and make decisions, and we may indeed have to fall back on those. They're not science, though, and shouldn't claim to be.
ATTP: Do you have shares in BS? 'Cos trying to short 'em here is not going to make you a fortune. That said, I've enjoyed reading your stuff (not for the reasons you think), which, in any other medium, would be considered a great send up of science (science as you know it, that is).
Nullius,
It seems we broadly agree. I'll just make one final comment. I think you've misunderstood what I meant when I said
and responded with
I wasn't implying that because they're using equations that are essentially representations of conservation laws that climate models can't be wrong. I was suggesting that if they're wrong it doesn't imply that there's a problem with the Navier Stokes equations. Of course an issue with climate models (and with any numerical model really) is how you discretise these equations on a grid (assuming one uses a grid). A model performing poorly, therefore, doesn't imply something about the fundamental equations. It's more likely that they've been implemented poorly.
Harry,
You're of course free to take enjoyment in what a write in any way you like.
Then there is calling these projections "evidence" and "data" and feeding them into other models to produce further "projections".
Surely this is sciency coprophilia of the lowest order?
ATTP:
Well, ATTP, you are a gift, as they say, that keeps on giving. However, I can assure you that the 'enjoyment' I get is well tempered by the fact that it is people like you, with your quite barmy theories, who are trying to decide how I should live my life - and bloody well pay for it. And all based on your crap
sciencepredictions (OK, projections, if you like).Whatever: you and your ilk want to change my world to your view of things, a view that is not so much scientific as political. And you come on here with your faux science and double-talk and expect us to take you seriously. In all the stuff you have written here I see so many circumlocutions as you seek to change your beliefs as a sailor tacks to the wind.
On another thread, some time ago, I asked you to falsify the Null Hypothesis. I'm still waiting.
Harry,
Just to be clear, I have absolutely no interest in how you live your life and have absolutely no interest in influencing how you live your life.
Oh, don't be silly. Not even I'm naive enough to think you'd take me seriously. At least give me that much credit.
Yes, I do remember that. You seem to think that there is only one null hypothesis. I'm not an expert at statistics but am pretty sure that this isn't correct. I'm also not particularly good at riddles, so you'll excuse me if I don't try and solve yours.
A Projection is that, this evening, I will go to my local where I will find Sandra Bullock researching British pubs for her upcoming film. She will talk to me all evening and then ask me to take her back to her hotel because she is not sure she can make it after so many drinks.
A Prediction is that, this evening, I will go to my local, meet the usual gang, get rat-*rs*d and stagger home vowing to never do it again.
Since Hansen and others make predictions and claim they are using climate models as their source, would it be accurate to say that ATTP is calling them liars? Or at least incompetent scientists?
The disconnection between prediction and projection is that in paradigm cases, one is an assertion about a future state of affairs, while the other is an assertion about a graph when extrapolated. The connection (no paradigm required!) is that one may (and arguably in alarmism is intended by the language users to) function as a synecdoche for the other. (Or if you up for niceties: a synecdoche if and only if the projection is a representation of physical processes, otherwise a metonymy if it is an abstraction).
Shakespeare would have understood all this - as would have his audiences.
ATTP, you seem to be making an argument much like Bill Clinton and his discussion of the word "is". You're equating climate model with Navier-Stokes equation and making the simplistic argument that if a climate model projection is wrong, it does not mean that the N-S equation is wrong. By doing this you're sidestepping all of the critical issues. No critic of climate modelling that I've ever heard of suggests that the N-S equation is wrong. However, for example, the N-S equation is continuous and every model represents this by some form of discrete approximation (there are a number of choices for how to do this). However, I have yet to see a time stepping version that will behave well using physically realistic dissipation terms. To behave well, you apparently have to use physically unrealistic parameters. I've heard that one of the original modellers came up with a reason for this (related to discretization), but have never traced the source.
So, in summary, a climate model includes many assumptions and parameters. Falsifying a given climate model simply means that one or more assumptions and/or parameterizations is wrong. Which means that that particular model is not useful for stating what will happen in the future. A complete list of the assumptions, approximations, and parameterizations in any given model is huge and this should not be forgotten. For example, in addition to discretization issues, models do not generally allow for both a low sensitivity to CO2 and a low sensitivity to aerosols. Yet this appears to be the situation in the real world. Hindcasting within reasonable limits simply involves having enough adjustable parameters - it does not mean that the choice of parameters has any validity.
MikeP,
Actually, what I was saying appears to be pretty much what you've just said. I wasn't trying to sidestep all the critical issues, I was simply pointing out - as you've just done - that a complex model being wrong doesn't really mean anything other than the model is wrong.
ATTP:
Wrong! What you and your scaremongering policies and NGO friends do is decidedly to define how I live my life, and many of the population of this country. Wind farms, anyone? FiTs? A society that's gone from wood-chip wallpaper to wood-chip power stations - and all because you think you're so bloody right. How will you feel when you are proved wrong? I bet you won't even admit it.
It's still not clear to me what a projection is in climatology. It must be more than just a contingent prediction, since otherwise the last 10+ years constitute a failed prediction by GCMs. Emissions scenarios have clearly attained levels on which earlier projections were based, without the predicted, contingent, outcomes. Climatology is clearly not prepared to accept that, so the definition of projection must also involve the time horizon for the prediction. Projections are then long-term (contingent) predictions (to allow for the effects of random variation)? The problem here is, of course, at what timescale does a projection have to be judged as a prediction? Has Santer, and his many helpers, provided an answer in the form of a time horizon of 17-20 years?
This assumes there is an authentic distinction, not just evasion.
Harry,
I seriously really don't care how you live your life (really couldn't care less, to be honest). You also seem to know an awful lot of things about what I want and what I'm trying to define, some things that I appear to not even know myself.
basicstats,
There is probably a basic way to infer that the models have failed. The range that is presented is (I think) the 95% range and typically represent the range that we would expect the observed temperatures to fall within 95% of the time. Simply falling outside this range would, therefore, not immediately imply that the models had failed or were wrong. However, if it remained outside this range for sufficiently long, then one would conclude that even of the models were wrong (or missing something fairly crucial). Off the top of my head I can't give you an exact number, but 5-10 years would - I think - be enough to conclude that something was happening that had really not been predicted by the models.
Andrew, that climate models are not initialized to match any state of the past climate. This fact came to light back 2007 in Kevin Trenberth’s blog post Predictions of Climate. Trenberth wrote:
We now know that the climate models cannot simulate ENSO or the AMO or the PDO.
Now let’s discuss the subtle differences between a forecast and a projection.
Weather models are initialized from current conditions--that is, they are started in-phase with reality--in an effort to forecast short-term weather. Because in-phase weather models are short-term and because they are continuously being tested, verified and falsified, the models can be improved.
On the other hand, climate models are not initialized from current or past conditions, and because they cannot simulate coupled ocean-atmosphere processes that can contribute to or suppress long-term warming (ENSO and the AMO), they are not intended to represent climate on Earth. They are intended to represent an Earth where CO2 and other anthropogenic forcings are the only means through which climate and global temperatures can change. So modelers use the term projection instead of forecast. Because climate models are not initialized from reality, climate model hindcasts cannot be tested, verified or falsified against past data because they were never intended to simulate climate in-phase with the real world. And as a result, they aren’t being improved.
Now, with that in mind, here’s the second paragraph of Trenberth’s blog post:
Cheers
One might argue that a projection makes assumptions that cannot be known at the time, i.e., an unpredictable volcanic eruption might cool the planet for a time.
But you can re-run your "projection" with the actual numbers once you have them, and see how well your "projection" went as an actual "forecast".
So this distinction is, ultimately, moot.
ATTP: "but if a climate model fails to represent reality, what theory has failed? I would say possibly nothing."
Perhaps the theory that all relevant variables have been properly accounted for.
Aug 7, 2014 at 10:18 PM Bob Tisdale
===============
Thank you an excellent and informative comment.
According to Trenberth the models are "story lines", "what ifs", and other possible fantasies. Maybe it's time to watch reruns of Star Trek for the next story line, hell it might be a true story line if the producer hit on a right idea in their wildest imagination.. I will not bet the farm on a "what if" or "story line" someone cooked up. The most important issue now is "what if not".
Beam me up Scotty, quickly please.
Old grumpy,
Yes, that's kind of the point of trying to make.
Will,
Sure, they can of course be rerun (of course, they'd then have to put up with people complaining about hindcasting etc) but I would still argue that it's important to understand the difference between a projection and a prediction.
ATTP
"Climate scientists can only provide the best information that they have available. What we (our policy makers) choose to do with that is up to them."
With great respect, the distinction you seek to make here is nonsense. For example, would you characterise the President of the Royal Society as other than a "policy maker"? I wouldn't. Obviously it would be lovely if there were an us and a them, with dispassionate scientists advising detached policy makers. But the world has never been like that.
Standing outside in the rain and getting wet is a prediction: It is specific and testable.
Making claims about the future that cannot be falsified and do not shape the underlying thinking about the claims is not a prediction. Those claims are not really projections or forecasts, either. And they are certainly not science. But they are climate science. And they are sufficient for true believers in climate science to demonstrate their profound enlightenment and arrogance. Like true believers in other areas of obsession, the stricken will argue endlessly over trivia in order to avoid the main point.
alan kennedy,
I don't know why that's even relevant, but he's certainly not an elected member of parliament, which is what I meant by policy maker. He doesn't make policy, other than the policy of the Royal Society, maybe.
This is kind of the point. There isn't an us and them. Scientists are simply doing research into whatever field they happen to be studying and presenting that work to other scientists and to the public. What we (our elected officials if you like) choose to do, given that knowledge, is up to us (through our elected officials). In a sense, you've hit the nail on the head. Expecting scientists to be entirely dispassionate is unrealistic and probably impossible to achieve. Therefore, ignoring what they present because they happen to not be as dispassionate as some might like is essentially an argument for ignoring all science.
hunter,
Of course it can be falsified. It's whether or not we're willing to wait that long to find out.
In truth I agree with this. However, I suspect you and I would disagree on who the term "stricken" refers to.
ATTP said
I may be wrong but aren't the IPCC projections based on emissions scenarios? So these scenarios force some constraints about behaviours.
And then:
So in the end it comes back to the precautionary principle again. Which assumes a cost of action that is less than the cost of the impact. The fallacy here is basing decisions on assumptions which can only be by subjective arguments which pile guess on guess.
To use the prisoner's dilemma as an example, maybe my best option is not to open any door. At least I won't be in a worse position.
Actually I don't know why we are discussing semantics here. Let's go to the dictionary. They are synonyms. The dictionary definitions speak for themselves, one word is used to define the others:
Projection
https://www.google.fr/search?sourceid=navclient&aq=&oq=projection+definition
an estimate or forecast of a future situation based on a study of present trends.
"plans based on projections of slow but positive growth"
synonyms: estimate, forecast, prediction, calculation, prognosis, prognostication, reckoning, expectation; More
forecasting, estimation, computation;
extrapolation
"the company claims it has exceeded its initial sales projection"
Prediction
https://www.google.fr/search?sourceid=navclient&aq=&oq=prediction+definition
a thing predicted; a forecast.
"a prediction that economic growth would resume"
synonyms: forecast, prophecy, divination, prognosis, prognostication, augury; More
bet, projection, conjecture, guess;
rarevaticination, prognostic, auspication
"seven months later, his prediction came true"
•the action of predicting something.
"the prediction of future behaviour"
Forecast
https://www.google.fr/search?sourceid=navclient&aq=&oq=forecast+definition
verb
verb: forecast; 3rd person present: forecasts; past tense: forecast; past participle: forecast; past tense: forecasted; past participle: forecasted; gerund or present participle: forecasting
1.
predict or estimate (a future event or trend).
"rain is forecast for Scotland"
synonyms: predict, prophesy, prognosticate, augur, divine, foretell, foresee, forewarn; More
guess, hazard a guess, conjecture, speculate, estimate, calculate, reckon, expect;
archaicspae;
rarepresage, previse, vaticinate, auspicate
"they forecast that shares in the company will start trading at a profit soon"
noun
noun: forecast; plural noun: forecasts
1.
a calculation or estimate of future events, especially coming weather or a financial trend.
synonyms: prediction, prophecy, forewarning, prognostication, augury, divination, prognosis, projection, calculation;
I haven't got time now to check OED, Chambers or Websters, if you feel they refute the definitions above feel free to post.
Technical fields redefine words all the time but if you are not using in their dictionary definition you need to make it very clear in your writings what definition you are using and not try to argue the meaning ex post</I>.
'When I use a word,' Humpty Dumpty said, in rather a scornful tone, 'it means just what I choose it to mean — neither more nor less.' - Lewis Carroll
clovis,
I'm not sure why you think the emission scenarios force constraints. The different scenarios cover a range of possible future emission scenarios and the model results present information as to what might happen for each of those scenario. The range of scenarios is broad enough (from probably unrealistically high to unrealistically low) to cover the possible range of our actual future emission. The models aren't suggesting that we have to follow one of the emission scenarios. The models are presenting how different emissions scenarios might affect our future climate.
I don't quite know how you got that from what I said. I simply said that if we choose to do nothing that we don't regret that decision in the future. Of course whatever decision we make should - in my view at least - be based on an analysis of the climate risks associated with the different possible emissions scenarios and the risks associated with the various possible policy options. I certainly don't think this is a simple calculation and - in my view at least - the most uncertain aspect of this topic is the policy aspect and not the physical science.
Aug 8, 2014 at 11:34 AM | And Then There's Physics
Model Constraints: The paragraph I quoted seems to say that forecasts can't me made because they would be predicated on unknown future behaviour. I was arguing the future behaviour is taken into account using the emissions scenarios. You may have been referring to other behavioural factors but if they are important, why aren't they taken into account?
Precautionary Principle: To me the paragraph I quoted describes the precautionary principle pretty precisely. Especially the third sentence. Maybe that wasn't your intent, but it certainly reads like that.
clovis,
Sure, but isn't that the same point I was making. We still don't know what emission scenario we will actually follow, therefore climate forecasts are still conditional on us following a particular scenario. I think we're saying the same thing.
The precautionary principle relates to taking action now to avoid a future risk. My point was simply that in the future we will be able to look back at what actually happened. I was simply pointing out that there is a chance that we may regret making the decisions that we chose to make.
Let me try to inject some reality and relevence by using an analogy from my career field of mineral exploration science.
Our geophysical group spent years developing modelling to 'predict' the size, shape and position of discrete bodies rich in magnetite and rarely in gold, copper and bismuth as well.
Data came from surveying the ground surface from a grid pattern, using a magnetometer to measure the distortion of Earth's natural magnetic field by a magnetite body. Model inputs were the measured magnetic field strength and surface position data.
Over time and many refinements, the modellers came to achieve impressively good results, such as 'predicting' that the first drill hole would intersect the modelled body at (say) 185 +/- 10 m down the hole. For those interested, a limit was set because of residual plus induced magnetic edffecs, the former requiring oriented samples of the target rock in order to incorporate its influence. IOW, you had to find the body with the model before you could model it optimally, thus makng it a 'wicked' problem to use words of Judith Curry.
Here is the important part.
The model was deemed to be successful when it had adequately predicted the said size shape and position of the discrete body. This post-drilling evaluation was easily done with math and stats. Over time, the ability of the model to predict was measured; after being found acceptable, it was used in many locations
Note that there was no point to developing the model without a clear set of aims and with measured performance criteria. Unlss these are present and shown to be achievable, there is no meaningful use for the word 'prediction.'
In the context of climate modelling, many of the runs seem to be poking around with various data inputs and assumptions to see what comes out the far end. These are not models by my terminolgy and I do not expect them to predict. An overall failure of more formal climate models is the lack of a set of criteria that include 'FAIL - further inqjuiry not warranted.
You see, in the hard world of mineral exploration, you live or die hy the money generated by your efforts, including your models. There is no prize for being almost right, ot right on average. Ensemble comparisons are irrelevant.
If it predicts acceptably, you can call it a model. If it does not, you can call it a drain on resources that could have been better used.
Somewhere there is a concept called 'accountability,. It does not seem to figure large in climate work.
Aug 8, 2014 at 12:26 PM | And Then There's Physics
So let me ask some direct questions:
Do you believe we have sufficient knowledge balance the cost of inaction and action?
Do you still think there is a need for action now?
Do you think the uncertainties are being adequately communicated to the policy makers by the scientists?
clovis,
Accurately, I would say no. I would argue that we understand the economic implications of climate change and of what will happen (economically) far less well, than we understand how our climate will change under different possible future scenarios.
I think it depends on what you're asking me. If you're asking for my opinion, then it's that we should avoid letting atmospheric carbon concentrations rise much higher than they are today. I don't know how we do that and I don't have particularly strong views. There may even be some technologies (CCS for example) that will allow us to continue using fossil fuels.
Yes.
Aug 8, 2014 at 12:43 PM | And Then There's Physics
I think we agree ;)
But it does imply the policy makers are ignoring uncertainties to further policies that cannot be supported by facts.
When a projection is verified, emission scenario already 'happened', so projection = prediction by then.
A climate model is based on the Navier-stokes equations which have been extensively used in fluid dynamics and are essentially representation of well known conservation laws.
No GCMs use the fundamental formulation of the Navier-Stokes-Fourier equations. None. The fundamental formulations are extensively simplified in order that a tractable problem is obtained. The momentum balance in the vertical direction, for example, is usually reduced to the hydro-static equilibrium case.
More importantly, the matters that critically count are all associated with the discrete approximations to the continuous equations and the numerical solution methods used to solve the approximations. This is, after all, where the numbers come from.
There's no way whatsoever that failure of the GCMs to show high fidelity with respect to the physical domain can be a reflection on the basic formulations of any of the fundamental equations.
There are, however, aspects that are less clear. Diffusion into the deep ocean, for example, can be included but there's quite a large range of values (I think) for the diffusion coefficient.
There are such coefficients in the form of the models of the Navier-Stokes equations used in GCMs. In particular, the momentum diffusion terms in the fundamental equations are modified from the theoretical values, and additional terms not present in the fundamental formulations are introduced. These latter are called momentum dissipation. There is no such thing in the Navier-Stokes equations.
The GCMs contain a multitude of parameterizations for various physical phenomena and processes. These are typically introduced whenever resolution of the phenomena and processes is not feasible and when the underlying phenomena and processes are not completely understood. It is the parameterizations that do all the heavy lifting relative to ensuring fidelity with the physical domain. Generally, the parameterizations are what are tuned.
Representations of all aspects of clouds are all parameterizations, including the motions and vertical locations. Nothing whatsoever is based on any fundamental equation. The radiative-energy-transport interactions with the participating medium ( aerosols ) in the atmosphere are all parameterizations. These are two of the more important aspects of climate process modeling relative to high-fidelity representation of the physical domain. And they are two of the fuzziest of all the magic process numbers.
Several of the parameterizations are direct functions of the size of the discrete spatial increments used in the algebraic approximations of the continuous equations. There is no such thing in the physical domain. This approach ensures that the application order of the numerical solution methods is at most one, and potentially less that one. Likewise, the very coarse discrete spatial grid cannot resolve geographical details at the boundaries of the physical domain. As the gird is refined, the geographical details change and thus again a first-order effect is introduced. So far as I am aware, it has yet to be demonstrated that the numbers produced by a GCM are in fact solutions of even the discrete approximations, much less solutions of the continuous equations. It's called Verification.
The parameterizations are generally based on algebraic representations of measured empirical data. As such they cannot ever be representations without some distribution about some mean states. They are approximations and not absolute representations of the measured data in the physical domain. Importantly, the parameterizations are approximations to the states that the materials of interest have attained and are not properties of the materials. All basic formulations of the fundamental equations contain only material properties and never contain information about previous states that the materials have attained.
The parameterizations generally are not limited by considerations of physical-reality boundaries; extrapolations can if fact produce physically unrealistic states.
All in all, GCMs are based on modeling of the important processes in Earth's climate systems. They are not models of the materials that make up the systems. The distinction is critically important. In this sense, GCMs are process models and are not science models.
clovis,
That may well be true, but - in my view - that is an issue that is distinct from the scientific details themselves.
Edim,
If it's already happened, then it's no longer a prediction, but if you're just suggesting that it becomes a prediction if we know that the chosen scenario will actually happen, then yes, I agree.
Dan,
I think I agree with that you describe, although haven't absorbed it all. I would add though, that
depends on the goal. If the goal is to try and understand our climate system and how it is influenced by various physical processes, then I would argue that they are scientific models. If they match reality, then one might assume that they've correctly included the relevant physical processes. If the results don't, then you assume something is missing or wrong, and you try to work out what. It's a pretty standard way of doing science.
To me, this is one of the problems in this debate. From a scientific perspective, climate models are simply a means of trying to understand how various processes will influence our climate. Scientifically, you may not even care if they end up matching reality or not, because either way it will be interesting. On the other hand, these are also the models that are being used to inform policy makers and, in such a case, we may well want to be much more confident about their projections than we currently are. However, given that they're probably the best we have at the moment, that we only have one planet on which to make comparisons, and we can't go forward and backwards in time, they're probably all we have.
A prediction has some skill but a projection doesn't.
ATTP, you can only verify it after it happenned and by then you know the emission scenario. For the purpose of verification, projection equals prediction.
interesting post & comments from the usuals at BH & ATTP :-)
i would just chuck in this from IPCC ORGANIZATION page -
"Because of its scientific and intergovernmental nature, the IPCC embodies a unique opportunity to provide rigorous and balanced scientific information to decision makers. By endorsing the IPCC reports, governments acknowledge the authority of their scientific content. The work of the organization is therefore policy-relevant and yet policy-neutral, never policy-prescriptive."
and ask ATTP if he thinks the "balanced" part is still adhered to ? if not why not.