By Steven Goddard
I recently had the opportunity to attend a meeting of some top weather modelers. Weather models differ from climate models in that they have to work and are verified every hour of every day around the planet. If a weather model is broken, it becomes obvious immediately. By contrast, climate modelers have the advantage that they will be long since retired when their predictions don’t come to pass.
Weather and climate models are at the core very similar, but climate models also consider additional parameters that vary over time, like atmospheric composition. Climate models iterate over very long time periods, and thus compound error. Weather modelers understand that 72 hours is about the limit which they can claim accuracy. Climate modelers on the other hand are happy to run simulations for decades (because they know that they will be retired and no one will remember what they said) and because it provides an excuse to sink money into really cool HPC (High Performance Computing) clusters.
But enough gossip. I learned a few very interesting things at this meeting.
1. Weather modelers consider the realm of climate calculation to be “months to seasons.” Not the 30 year minimum we hear quoted all the time by AGW groupies. That is why NOAA’s “Climate Prediction Center” generates their seasonal forecasts, rather than the National Weather Service.
2. The two most important boundary conditions (inputs) to seasonal forecasts are sea surface temperatures and soil moisture. No one has shown any skill at modeling either of those, so no surprise that The Met Office Seasonal forecasts were consistently wrong.
For example, just a few months ago the odds of La Niña were considered very low. Compare the December forecast with the May version. How quickly things change!
SST modeling capabilities are very limited, and as a result seasonal weather forecasts (climate) are little more than academic exercises.
Oh and by the way, Colorado will be exactly 8.72 degrees warmer in 100 years. But they can’t tell you what the temperature will be next week.
“If I don’t understand it, it must be simple.”
– Dilbert Principle
In the top picture, which boxer is weather and which one is climate? What do readers think?
Ocean conveyor theory evolution.
http://www.theresilientearth.com/?q=content/ocean-conveyor-belt-dismissed
Ulric Lyons says:
July 1, 2010 at 5:03 am
If you think about it, an accurate climate forecast would be totally dependant on deterministic long range weather forecasts, and that can only be done by predicting changes in the solar signal.
You are assuming only one variable, which is surely wrong.
@Steven McDonald in France
(finally a topic on WUWT I have some expertise in to comment on, heh)
“I think it’s the other way around. Weather is in constant motion, more aggresive, volatile, and packs a punch at short notice. Clearly the heavyweight.”
As a boxer myself my un-peer reviewed “professional” opinion is that the above description much better suits the smaller fighter. That larger fighter is probably only capable of slow haymakers. The smaller boxer will probably take his time, giving the larger fighter a much longer time in the ring that his fighting skills rightly deserve, being carried only through sheer size and momentum (the smaller boxer will want to avoid letting the larger corner him and use his weight against him, though even then the larger fighter’s options are limited). That description of the larger boxer to me accurately represents the nexus of CAGW committed media-political-and activist interests with the smaller boxer representing skilled and sincere scientists and truth-seeking laypeople.
Someone above said the smaller boxer looked like he’s stumbing, he’s not. He has ducked into position to give a left bodyshot and looks like he is about to follow through with a right cross.
Dan says:
July 1, 2010 at 5:56 am
Hi Steven,
Let’s make a bet. I’ll bet you that the average temperature in the United States in July 2100 will be warmer than the average temperature in January 2100. If I’m wrong, I’ll give you $100. If I’m right, you give me $1. (What a deal!)
Do you accept? If you’d like this bet to be a bit more tangible, we can change the year to 2015, or any other year of your choosing. According to your logic, accepting this deal should be a no-brainer, since you’ve claimed that prediction of climate variables and weather variables are mathematically equivalent
These are not climate variables. They are weather variables or seasonal. Apples and oranges. Similar to I bet you will find apples on an apple tree and oranges on an orange tree. You see! not variables.
In the Sea Ice #11 thread “Gavin” wanted to know what people thought of 45F in Siberia.
http://wattsupwiththat.com/2010/06/28/sea-ice-news-11/#comment-420492
45 F in a city in Siberia means global warming is happening. 45F in a city in Siberia is climate and not weather.
😉
For example, just a few months ago the odds of La Niña were considered very low. Compare the December forecast with the May version. How quickly things change!
So La Nina has begun?
http://www7320.nrlssc.navy.mil/GLBhycom1-12/navo/equpacsst_nowcast_anim30d.gif
1. Weather modelers consider the realm of climate calculation to be “months to seasons.” Not the 30 year minimum we hear quoted all the time by AGW groupies. That is why NOAA’s “Climate Prediction Center” generates their seasonal forecasts, rather than the National Weather Service.
This isn’t entirely accurate. The seasonal forecasts put out by the CPC use a standard 30 year climatological norm to determine predictive probabilities. They take into account other factors, such as El Nino and soil moisture, but when you look at a seasonal forecast, like this one:
http://www.cpc.ncep.noaa.gov/products/predictions/long_range/lead01/off01_temp.gif
“Normal” refers to — at present, in the US — a 1971-2000 climatology. So it isn’t accurate to suggest that 30 years isn’t being used to define “climate” in the CPC’s forecasts.
I think what you meant to draw attention to, of course, is that they are not predicting climate out for 30 years. The seasonal forecasts just go out 12 months. That’s as far as the “climate” specialists at the CPC are comfortable projecting using current methods and technology. But to be accurate, what they are projecting is how the climate over the next 12 months will compare to a 30 year climate norm.
Matt
Suppose that a climate model had predicted decreasing sea ice in Antarctica from 1978-2010. The erroneous predicted positive albedo feedback would compound over time, and the model would go further and further off into the weeds. ;^)
There is no mechanism for climate models to correct themselves, which is why they consistently predict higher than observed temperatures.
Climate is weather averaged over time.
Any attempt to be more specific is artificial, and in these politically charged times, likely to be agenda biased.
Climate is a useful concept for weather prediction. For example, I may reliably predict that in July, 2011, in Tucson, Arizona it will be mostly hot, and in Tierra del Fuego, Argentina, mostly cold. The variations in climate (climate change?) are usually sufficiently small that they are measured in fractions of a degree over a period of years and decades. Climate is useful.
However, when the agenda-driven get hold of it, and like so many Cassandra’s, cry “Doom, doom is nigh….doom deserved because we are eco-evil!”, then climate “science” descends into a quasi-religious farce.
This is the whole global warming scam in one article. Unlike real science, climate “science” pretends that it is a science, because it talks and acts like a science and does all the experiments it can … except it can’t because it takes decades for anything to change with the climate.
So … what they try to kid us, is that even despite the fact they have never ever done any actual experiments and almost all their predictions have been wrong, that doesn’t matter because they’ve learn from those and … what they’ve learnt is never to let themselves get pinned down by the necessity of real experimental/scientific procedures.
There is a saying that if you take enough climate “scientists” and allow them to make enough predictions, then sooner or later one of them will predict something that actually happens … which is all they need to prove they were all right all along!
This is an answer given just now by Dr.Roger Pielke Sr. in his website with reference to a question on weather and climate modelling.
http://pielkeclimatesci.wordpress.com/
His view on the subject is quite realistic.
Sorry, correct link is
http://pielkeclimatesci.wordpress.com/2010/07/01/qa-on-gcms-weather-and-climate/
Steven,
Couple points:
1) No models are perfect, and scientists using them will readily admit this. It tends to be the scientific-illiterate media that place an extreme amount of trust in climate model results. The fact is that even simulating evolution of an afternoon thunderstorm is extremely difficult. (Look at cloud-resolving models and single-column models for instance)
2) Models tend to be subjected to rigorous validation. Yes, none are perfect, but there is always one more advanced and accurate than the rest which goes on to beadopted for broad use. If there was a model predicting sea ice decline in Antarctica over the past several decades, and this failed to match the observational record, then any results from this model predicting the next several decades should be regarded with suspicion.
3) The example you give also illustrates the reason that for the majority of serious model results (be it climate, weather, hurricane, or otherwise), an ensemble average is used to determine the most likely outcome. You seed small perturbations in the initial conditions, then use a few different models, and average the results. This also gives you a nice range of possibilities.
4) Just something quick I noticed – in your 2 ENSO model forecasts shown above, several models in the December run did a very good job at predicting the SST anomaly in May and beyond.
carrot eater says:
July 1, 2010 at 6:28 am
…you can still be sure that the Earth will be warmer on average due to the active sun.
And by extension, cooler due to an inactive sun. Wow. “carrot eater” is finally beginning to see the light. Who’d a’ thunk.
Got it, weather is what happens, climate is what we remember….
Matt says: I think an analogy between predicting weather and climate would help here – imagine you have a hot, fresh cup of coffee in front of you. You pour in some cold cream, and stir it up with a spoon. Now, tell me, which is easier, predicting the temperature and cream content at a given spot in the coffee cup in 30 seconds, or predicting the ensemble temperature of the coffee in 20 minutes (given you know the room temperature, etc)? Its clear that it is much easier to predict the coffee temperature in 20 minutes.
Matt, that is an entirely false analogy. A real analogy would be to predict:
1a. the flow of water in a river at one spot (in a rapid/turbulent flow) in a day
1b. compared to the flow in that spot in a years time
2a. compared to the whole flow of the whole river in a day
2b. compared to the whole flow in a year
These comparisons change not only in the spatial extent but also in time. YOU WERE BEING HIGHLY DUBIOUS …. and comparing apples with cars.
If we take local temperature, and compare it like local current to next day or a year. It doesn’t take a genius to work out that rivers change and whilst the state of flow in the river now is a good indicator of the flow tomorrow at one location, it is really a pretty poor indicator of the flow next year.
This is because like climate/weather, local water flow has increasing noise levels with longer periods. In contrast to your bogus example, weather & climate (of one place) actually gets more difficult to predict over longer periods.
And the same is also pretty obvious about bulk flow, just as it is about average global temperature. Today’s flow is a pretty good indicator of the flow tomorrow (even better if you have a weather forecast to hand), but it is a pretty lousy indicator of the flow in a year’s time.
If however, you were to take “speed” of the water rather than “volume/second”, it all gets a lot worse. Because the whole dynamics of the river continually vary, and a section that may have been rapid one year, may be silted the next. And a silted section may have been scoured by a storm to have become fast flowing.
Likewise climatic noise follows a 1/f^n type relationship, in that the longer you look at the climate, the more variation there will be.
In total contrast to your coffee cup bogus humbug example. The truth is that the variation of the climate increases: look it for a decade it is largely than in any one year (on average). Look at it over a century, it will have MORE variation than in any decade. Look at it over a millennium and it will have even more.
AMO data plot added to the Arctic temperature anomaly.
http://www.vukcevic.talktalk.net/NFC1.htm
“There is no mechanism for climate models to correct themselves, which is why they consistently predict higher than observed temperatures.”
Ah but there is and that is the ever fallible “seat to keyboard” interface him/herself.
Climate (if you consider rainfall an element of climate) sometimes varies almost as much as weather. Look at this graph of prehistoric lake levels of Devils Lake in North Dakota.
http://mapservice.swc.state.nd.us/4dlink9/4dcgi/GetContentPhoto/PB-42/640/480
Devils Lake is a lake in the prairie pothole region of the US that has no outlet and thus gives a picture of how changeable precipitation patterns are in this region. The lake levels are inferred from diatom studies like this http://www.jstor.org/pss/3225026 . If it is true that soil moisture is a key difficulty with creating long term climate models, I have to think that the changeable moisture patterns in the western plains suggest that climate models are robustly intractable.
I also remember hearing Bill Gray remark on C-SPAN that momentum fields that weather forecasters use are less troublesome than the energy fields that have to be used to create more faithful longer term models. I took this to mean that momentum being a product (mv) is less prone to error compounding than energy (mv^2) which involves a squaring.
Is it possible that the enterprise of creating statistical climate averages misleads us into thinking something unwarranted? That this notion that we’ve created called climate is really just an average as it seems from the perspective of a lifetime or decades. To me the existence of the Roman Warm Period and the Medieval Warm Period and the Little Ice Age and for that matter the Younger Dryas all suggest that there is no such thing as a stable global climate. In and of itself there is no reason to be surprised or alarmed that we find we are in trend. What would be more surprising would be to find that with a surface as variable and varying as the earth’s that there should be a rock steady average global temperature.
“Weather is chaotic and unpredictable. Climate is weather averaged out over time. Simple analogy: you can’t predict whether a coin will land heads or tails, but you can predict statistically how thousands or millions of coin tosses will go.”
“There are dozens and dozens of good arguments which could be made about the inaccuracy of climate models, how they are misused and how poorly the science is understood. This, however, is not one of those arguments.”
“I award you no points, and may God have mercy on your soul.”
Answer:
“There is no mechanism for climate models to correct themselves, which is why they consistently predict higher than observed temperatures.”
For my chime in, comparing coin tosses to climate models is apples and oranges. There is no mechanism for climate models to correct themselves as said by Steve. The coin toss model by virtue of probabality does have self-correction inherrent in its design.
Although, I might add a caveat here: The problem with climate models is that they are attempting to predict the future. What they should be doing is taking a wide range of input data for their training (say 1800-1950) and see if their model predicts what is known about the climate from 1950-2000. If and only then can you say these models MAY be able to predict the future.
In the coin toss example, we can model 100 tosses as our training data, and come up with good probabalities for what the next 100 will be simply because we are certain of what certain factors are.
HOWEVER, I should add another caveat here. Even though we might be say 95% sure that we will get between say 40 and 60 heads out of 100, there is that 5% chance that we are completely wrong. Its not in the realm of impossibility to get 100 heads out of 100. Even knowing the mechanism of a model does not mean you can predict the future to 100% at any time.
And climate suffers from a very big disadvantage in that even with all the unknowns we have today, there is still the big unknown of “outside events.” I am talking the volcano, Feedback correction loops we don’t know about, mutations in plants that take more CO2 out of the air, asteroids, solar cycles that we have no seen yet…and the list goes on and on.
Coin toss only suffers from the outside event that the person tossing it may be able to “ham” the results by using a weighted coin… Which is probably the only similarity between coin tosses and climate models, the process of “hamming the result set”.
@ur momisugly PDA July 1, 2010 at 5:33 am:
You can’t compare this to coin flips. If we were, then you would be talking about someone claiming to predict a million coin flips to the accuracy of 10 +- 15. Unlikely garbage.
And aside from that, each coin flip is independent. I would not say that each day’s weather is independent, and you certainly cannot say that climate is independent of weather.
Mike,
Your analogy is flawed in several ways – the glaring one is you’re comparing an open system (the river) to a mostly* closed system (the earth). This is why the coffee cup analogy works – its a closed system in everything except energy.
Examining your claim that “Likewise climatic noise follows a 1/f^n type relationship, in that the longer you look at the climate, the more variation there will be,” its true in some sense, but given a broad enough timescale, you can make fairly accurate predictions, because the nature of a long time scale tends to smooth out small-scale effects. Take the Vostok ice core data, for example – http://faculty.gg.uwyo.edu/neil/teaching/Geomorph/lect_images/Vostok-ice-core-petit.png Over a large enough time scale, the temperature follows a fairly regular oscillation, which is driven by the well known Milankovich cycles. Given a sufficient understanding of the oscillation of the past, and the Milankovich cycles, it wouldn’t be unreasonable to give a ballpark prediction for several thousand years from now.
* I say mostly, because the Earth is a closed system in almost all respects except energy, the input/output of which are fairly quantifiable on broad timescales.
vukcevic says:
July 1, 2010 at 3:21 am
I am sure, these, in turn, are closely related to the Sun. Someone, from across the twilight zone told you, in another post, that these field forces are too low to make any change, however it is precisely the contrary, as electric/magnetic fields are 39 orders of magnitude more powerful than Fred Flintstone’s gravity powered axe!
It´s the growing apple tree vs. the apple falling!
I can tell you what the temperature will be like in July 100 years from now. Most likely very similar to the present. Over the last 80 years, summer temperatures in the US have hardly changed at all.
Now you’re not even trying. Hint: as weather is not the same as climate, so mere extrapolation is not the same as modeling.
As I stated in the article, a climate model run is essentially the same as a weather model run, only with more iterations, more input parameters and often a coarser spatial and temporal granularity.
Yes, and a Peterbilt is essentially the same as a bicycle, only with more circular rotational devices, more mechanical motive power and often heated seats. If you can’t haul 160,000 lbs GCW coast to coast with a bicycle, why do you think you can do it with a Peterbilt?