Technique could help climate models sweat the small stuff
PROVIDENCE, R.I. [Brown University] — A team of physicists and mathematicians has come up with a statistical technique that puts the fine details back into computer simulations of large-scale phenomena like air circulation in the atmosphere and currents in the ocean.
Computer models are generally good at capturing the big picture, but they are often forced to ignore things that happen at small scales. For example, models of a planet’s atmosphere capture the large-scale dynamics of jets and airflows, but they don’t include small-scale dynamics created by things like clouds and localized turbulence, despite the fact that those dynamics can often influence the larger scales.
“There are simply too many numbers for the computer to simulate it at a reasonable speed,” said Brad Marston, a Brown University physicist. “It might take years to simulate a day of the atmosphere, which wouldn’t be good.”
The traditional approach to dealing with the problem is to simply lop the small scales off of the simulation. A few ad hoc ways of putting some of that information back in exist, but they tend not to be mathematically rigorous.
“These schemes have always suffered from the criticism that they lack predictive power,” Marston said. “You have to make a lot of decisions that you really shouldn’t have to make but you’re forced to make.”
In a paper published in the journal Physical Review Letters, Marston and his colleagues show a method of averaging out those small-scale dynamics in a way that is computationally tractable, which allows those dynamics to be simulated and their effects to be captured in a rigorous way.
“We’re retaining the degrees of freedom at the small scale, but treating them in a different way,” Marston said. “We don’t have to simulate all the little swirls, so to speak. We treat them by using their averages and the sizes of their fluctuations. It allows us to capture the contributions of these small-scale dynamics that would normally not be included.”
In their paper, the researchers used the technique to model air jets forming on a round surface. They showed that the method produces results similar to brute-force numerical simulations of the same jets.
There have been prior attempts to treat small-scale disturbances statistically, Marston said, but those haven’t fared very well. Prior attempts have treated disturbances as being homogeneous and assumed they were not traveling in any one particular direction.
“But that almost never happens in nature,” Marston said. “Turbulence almost always has some directionality to it. That directionality is what makes these kinds of approximations work. It makes these approximations tenable.”
The researchers hope that the method might make for more accurate simulations of a wide variety of natural phenomena, from how the churning interiors of planets create magnetic fields to how air flows across the surfaces of cars or airplanes.
The method could be particularly useful in modeling Earth’s changing climate because the technique can more rigorously capture the influence of cloud formation.
“Cloud formation is seen as the largest source of uncertainty in climate models right now,” Marston said. “There are famous examples where different climate models that have different ways of dealing with the clouds give you qualitatively different results. In a warming world, one model might produce more clouds and another might produce fewer.”
By averaging those cloud dynamics and then simulating them in the models, it might be possible to reduce some of that uncertainty, Marston said.
The team has already started working to incorporate the method in climate simulations, as well as simulations of ocean currents and problems in astrophysics dealing with the behavior of plasmas.
“There are a whole bunch of problems out there where we feel this could be helpful,” Marston said.
###
Generalized Quasilinear Approximation: Application to Zonal Jets
J. B. Marston, G. P. Chini, and S. M. Tobias
Phys. Rev. Lett. 116, 214501 – Published 27 May 2016
Paper: http://journals.aps.org/prl/abstract/10.1103/PhysRevLett.116.214501
ABSTRACT
Quasilinear theory is often utilized to approximate the dynamics of fluids exhibiting significant interactions between mean flows and eddies. We present a generalization of quasilinear theory to include dynamic mode interactions on the large scales. This generalized quasilinear (GQL) approximation is achieved by separating the state variables into large and small zonal scales via a spectral filter rather than by a decomposition into a formal mean and fluctuations. Nonlinear interactions involving only small zonal scales are then removed. The approximation is conservative and allows for scattering of energy between small-scale modes via the large scale (through nonlocal spectral interactions). We evaluate GQL for the paradigmatic problems of the driving of large-scale jets on a spherical surface and on the beta plane and show that it is accurate even for a small number of large-scale modes. As GQL is formally linear in the small zonal scales, it allows for the closure of the system and can be utilized in direct statistical simulation schemes that have proved an attractive alternative to direct numerical simulation for many geophysical and astrophysical problems
Well sure. It might improve the performance of the GCMs.
You actually need to HAVE a performance, before you can improve it.
GCMs don’t even take notice that the whole earth is circulating every 24 hours. KT has it stationary, no circulation at all.
G
Must have missed the seminal paper explaining how deeply non-linear dynamical systems are now readily computationally accessible. The usual total junk “science” which has become so tedious it’s barely worth commenting on.
Here is my main comment on this new method’s applicability to climate science.
The problem is not that we are unable to model the clouds properly on a computer.
The problem is that we don’t understand the clouds properly in the real world.
As a result, I suspect that this new method of modeling the clouds may indeed be of use … but only once we actually understand the clouds.
w.
Willis Eschenbach June 3, 2016 at 5:17 pm
————————————————–
Touché Willis.
Again, how clouds work is basic physics at the partial differential level. The problem has always been integrating that to a macro description of behaviour.
In the context of climate we aren’t for example interested in the exact shape of every cloud: we want to know how BIG they are and maybe what temperature, and that’s about it.
The problem is that numerical integration leaves a huge amount of detail as to the exact shape that is ultimately ‘too much detail’ You need to know it to predict how the cloud will evolve, but its not part of what you are doing the analysis for.
So this is typical of a branch of maths that gets a good approximation to the value you want to examine, and ignores the data that is present in the real world, that you dont want to know. As all linear systems are. You dont need to know the speed and location of every molecule to know how much pressure is in your car tyres.
Hence the term ‘quasi linear’ which to me means ‘linear enough so we can use linear systems analysis without ending up in the doggy doo’.
So a word of caution: Just because 97% of what ends up reported here as science is grant grabbing gobbledeygook, doesn’t mean it all is.
I have to query what exactly is meant by ‘dont understand’ here. If it means ‘don’t have a usable model for’; then OK, but then this methodology may allow the development of such.
And as I understand it, the problem is exactly the other way around. We do understand how clouds work – at the partial differential level – but we can’t model that on a computer less than the size of the whole planet on any sort of detail in any sort of sane time.
And the various techniques for putting in a fudged value for clouds in GCMs are simply not good enough.
The proof of the pudding is in the eating however, and I can assure you as an engineer, that if this technique works and gets popped into the top level CFD packages and if it works, money will flood into the coffers of these scientists if they play their cards right.
At the moment whole car or whole boat or whole aircraft analysis for turbulent flow – which is a the main component of drag at speed – is impossible to simulate numerically, which is why we still have wind tunnels and boat pools full of water.
If this technique works then its the equivalent maybe of a couple of orders of magnitude increase in computer power.
The jury is still out, but these guys may have just opened the door to a step change in accuracy of fluid dynamic numerical modelling. And that’s a small goldmine and a major breakthrough.
I dont understand the cynicism and negativity. AGW models don’t work, and yet computer models in general often do. This ought to make them better. Better AGW models might actually start to track reality. Surely we aren’t afraid of that?
We aren’t all here because we believe and want to believe that AGW isn’t happening surely. I thought we we here to understand the real situation, whatever it is.
Leo Smith June 4, 2016 at 1:22 am
We aren’t all here because we believe and want to believe that AGW isn’t happening surely. I thought we we here to understand the real situation, whatever it is.
———————————————–
“””whatever it is.”””
Combination of a lot of different factors in the realm of natural variability. Separating the different factors out from the human contribution has become quite challenging.
Do I think that mankind has raped, pillaged, burned and destroyed areas of our planet.
For the sake of, taking care of an ever perpetually growing population base.
Sometimes we just need to think about what mankind has really done.
Sewers, tunnels, ‘electrical wiring, roads paved, structures, cities, depopulation of animal kingdom, over fishing, deforestation, ‘mining iron ore, gold, copper lead, water use, land use, just to name a few.
There is a big growing problem…but don’t forget about the greenhouse gases and pollution, which are only a symptom of the bigger growing problem.
AGW theory seems to revolve around LWIR and CO2 and “mean global lower tropospheric temperature anomaly” ,whatever the hell that is.
Absolutely none of that has anything to do with clouds. (meaning it doesn’t link to clouds).
But solar TSI, clouds and “mean global lower tropospheric Temperature” do seem to be related.
But nobody is making measurements of the latter feedback system; only the first one; which isn’t even a feedback system.
No atmospheric dynamic differential process is anywhere near fast enough to have any effect on the transit of a photon through that atmosphere. As far as solar energy absorption by the earth, the atmosphere is a completely static system. The most violent of tornadoes, or hurricanes does not change the direction of a TSI photon, as a consequence of the differential motions of molecules in that atmosphere. The effect would be the same if every single molecule in the atmosphere stood absolutely still long enough for the photon to pass through to the surface (or not).
G
I would have thought you can reduce a system to statistical output simplicity once you’ve captured all of the data and understand the mechanism/s in their entirety. Then you can show a mean trend.
How they plan to predict the future with most of my above not even being true, is beyond me.
“As a result, I suspect that this new method of modeling the clouds may indeed be of use … but only once we actually understand the clouds.”
It’s clouds’ illusions after all?
We really don’t know clouds at all.
What’s to understand Willis ?? Clouds reduce the solar energy that earth absorbs. How much clouds, affects how much solar energy we don’t get. We don’t have any monitoring system UNDERNEATH the clouds to measure how much solar energy we get, so we don’t know how much we get.
Nothing to understand. More clouds less solar energy; less clouds more solar energy.
G
But you are correct. X-box clouds are of no use to us.
The difference between clouds and no clouds make as much as 100W/m^2 in surface forcing from space (to space), and water vapor carried in the air can make a +20F change in daily high temp, and we are comparing these 2 natural regulating mechanisms to the added forcing of 3.7W/m^2
Sorry for the pass first read and comment well done will I really have no idea how they can publish this stuff and sleep at night .
Cloud formation is too complex, localized and random to model on a global scale without way more observation and analysis of the cloud nucleation mechanisms with present (and future) technology.
You have to be able to predict cloud formation before you can predict it’s impact on climate.
Since the atmosphere is a smallish secondary effect of the ocean, start with that.
Wonderful Climate….
Just as long as they realise that every one of those small scale phenomena is a negative system response to any forcing element that might be trying to move system temperature away from that set by insolation acting on mass suspended off the surface within a gravity field and able to convect freely.
If the new model bumbles along the current trend and looks really boring I will be pleasantly amazed.
If it rockets up towards Thermageddon I will do a Gallic shrug.
If it plummets towards a new glaciation I will move my money in to coal.
Who knows, but the quasilinearity (whatever that means) of the method at small scales seems to ignore a crucial (mathematical) issue in this type of dynamics (whether physically relevant is another matter, it seems). To quote Tao of NS and other fame,
“behaviour of three-dimensional Navier-Stokes equations at fine scales is much more nonlinear (and hence unstable) than at coarse scales”.
On the topic of clouds and cloud formation, I recently found this article.
Focus on high energy particles and atmospheric processes
RGiles Harrison1, Keri Nicoll1, Yukihiro Takahashi2 and Yoav Yair
6 October 2015
Abstract
The letters published in the ‘Focus issue on high energy particles and atmospheric processes’ serve to
broaden the discussion about the influence of high energy particles on the atmosphere beyond their
possible effects on clouds and climate. These letters link climate and meteorological processes with
atmospheric electricity, atmospheric chemistry, high energy physics and aerosol science from the
smallest molecular cluster ions through to liquid droplets. Progress in such a disparate and complex
topic is very likely to benefit from continued interdisciplinary interactions between traditionally
distinct science areas.
3.2. Layer cloud global circuit mechanism
Vertical current flow occurs throughout the atmosphere in fair weather regions through the global atmospheric electric circuit—a conceptual legacy of CTR Wilson—and cosmic ray ionisation. This provides a coupling mechanism between electrically-induced changes, for example from ionisation changes, and low level layer clouds. The ion current’s passage through the cloud-clear air boundary, which also represents a change in electrical conductivity, leads to local charge separation in the droplet formation and evaporation region, in proportion to the current.
Voiculescu et al (2013) found a positive relationship between mid-latitude cloud cover and the interplanetary electric field, which they considered could be occurring through the global circuit mechanism. A further suggestion of a global circuit effect was made by Lam et al (2013), as part of the atmospheric response to the By component of the Interplanetary Magnetic Field. Lam et al (2013) showed differences in the surface pressure patterns between large and small circumstances of By. A defining characteristic of the global circuit is its single maximum diurnal variation, known as the Carnegie curve (Harrison 2013). Harrison and Ambaum (2013) reported an averaged diurnal variation in cloud base properties similar to that of the Carnegie curve, in separate series of data obtained during the polar night in the northern and summer hemisphere. Harrison et al (2015) have shown a sensitivity of cloud droplet distributions to charging of small droplets, such as that typical of layer cloud electrification induced by the global circuit….
http://iopscience.iop.org/article/10.1088/1748-9326/10/10/100201/meta
One of the things I like to tease the modelling community with is how much energy is dissipated by a wee little tropical cyclone – not a Katrina or anything big enough to earn a name, just a little one with an eye perhaps 1km across and a surface peak wind speed of 100km/h. The calculation of the mass flows and energy involved in accelerating the air in that volume to those speeds is actually quite simple. The power is in the 100TW range – and is omitted in every model because the scale is just too small for them to capture. So perhaps they can start to take these little things into account – it might just improve their ability to predict something, which right now they can’t do.
” The power is in the 100TW range”
I estimated Hurricane David carried about 1/4-1/3 the volume of Lake Erie dropping it all the way into Canada.
That’s a lot of power.
Your storm, which is a category 2 hurricane, is put in perspective by the 173,000 terawatts of incoming solar radiation. The power of your storm represents .06% of incoming radiation. This is probably not going to throw off the conclusions by very much. Particularly as we only get tens of these each year, and most of the time there is not one of these going on. Perhaps the modelling community needs to tease you a little more.
No, I described a tropical storm, not a hurricane. National Hurricane Center reports “the 1981-2010 averages are 12 tropical storms, 6 hurricanes, and 3 major hurricanes”The energy of a hurricane is of the order of 10 greater and a major hurricane 10 greater again, so the average annual energy dissipated by all classes of storm is some 370 times higher, i.e. about 37 000TW, or some 20% of the incoming radiation.
I wish that a true expert in computational fluid dynamics would weigh in on this. The global climate models all use the Navier Stokes equations as a starting point. Why they do that is anyone’s guess. The analysis of the flowfield around a subsonic aircraft (whose surfaces are closely modeled, and have few features) with the Navier Stokes solvers available still requires Reynolds averaging, because no computer can handle the number of grid points needed to resolve the finest scale of turbulence. And steady-state solutions are accurate to only a few percent. Still, aircraft CFD models can get down to the millimeter range if necessary, and that is fairly impressive resolution. But an aircraft may be 180 feet long. The earth is 20,900,000 feet radius. And the atmosphere is 80,000 feet deep (for CFD purposes). An aircraft model might have up to 1 billion nodes, and achieve a turbulence resolution of a few millimeters. AN earth model using the same number of nodes, achieves a resolution no better than 100 km horizontal, and a few 10s of meters vertically. But the scale of significant turbulence on earth is on the order of a few feet (a tornado).
No one knows whether Reynolds averaging is valid over such a tremendous scale range. But the fact that aircraft CFD is no more accurate than a few percent for steady state, combined with the fact that there is no such thing as a steady state solution for the earth’s atmosphere, should raise some questions about the validity of models. How can a high-resolution model no better than a few percent accurate for a steady state solution be expanded to a low-resolution model run for a hundred years with no steady-state solution possible even in principal, and be expected to yield meaningful results? I just don’t get it.
None of that fluid dynamics, whether at the aeroplane scale or the global scale, can have ANY possible influence on the trajectory of a solar photon transiting the atmosphere in less than one millisecond. ( I took the liberty of setting the atmosphere at 300 km thick. I have no idea how many Navier Stokes or Reynolds feet that is)
G
Life, the Universe, and Climate Change… As a glider pilot, the core challenge is to locate “lift”, which could be columns, rivers, or pockets of rising air and fly in them to gain altitude. Different terrain produces different types of lift; deserts, forests, lakes, oceans, fields of corn; it’s all different. Different latitudes produce different lift at different times of the year; it’s all seasonal too. Golly, I wish I had thought to check with the experts at Brown University to take some of the guess work out what’s happening with the winds aloft. That’s why I always laugh when I see cartoon-like drawings that are said to depict the function of a multi-variable dynamic system. Watts per meter squared = 42 lol If only I had known THE answer!
mairon62 – June 5, 2016 at 1:06 am
HA, that reminded me of many years ago (late 1960’s) when a good friend and work associate who was an Electrical Engineer ….. and an avid glider pilot …. told me that he was going to design himself a “thermal detector” to attach to the wing tips of his glider.
“Original thinkers” are far n’ few between.
I’ve looked at clouds from both sides now…
So in the end the folks at Brown have created a way to incorporate something huge into models that hasn’t ( by the admission of the experts) been adequately taken into account. Of course, even without this, the models were somehow accepted as being something we could rely on. Now with the inclusion of the Brown methodology, they will include the effects of clouds…or at least they will be advertised as including the effects of clouds.
Maybe in 500 years we will know if this actually works.
YES The past 400 years Humans are causing SOME Global Warming fossil fuels and Deforesting releasing carbon , Something Bigger Is Happening.
Need to think outside the box, the solar system is not alone and not stationary it is on the move , the solar system is reacting to things humans do not understand in the galaxy the Milky Way.
The system and “”all of its planets”” are showing signs of change, even the Oort cloud showing signs of change more deep space Comet diving into the sun, Earth’s north and south auroras have been massive in the past years, this strange activity has to be caused by something.
Here’s my out-of-the-box thinking , Scientists throughout the World know about Dark Energy and Dark Matter Scientists have measured it but still do not know what it is , my thinking is our solar system and all of its planets and the Oort is Passing through a Swirl or an Arm of Dark Energy or Dark Matter which is not affected by the Heliosphere Causing Unusual Friction, Causing These Changes on the Earth and Solar System and its planets.
“”Sun Is Reacting to This Strange Galaxy Matter Energy more solar flares “” causing planet atmospheric change. Warming up planet Earth releasing methane deposits in the permafrost regions.
A nether thought ,, The Earth’s surface is a Natural Heat Disperser or Heat Sink of Solar Energy, also the earth reflects Solar Energy back into space , Solar Panels and Solar Collectors for heating sodium or oils capture the sun’s energy and releases it straight into the atmosphere.
(( Have you touched a Solar Panel in Full Sunlight , you can not do it too hot , it will Severely Burn You )). There are Billions of Black Solar Panels scattered over the Entire Globe.
Hot Black Panels “Releasing Heat Directly into the Atmosphere” heating the atmosphere.