I recently joined the climate prediction experiment.

After observing the model output on my computer (still phase 1), I now feel there may severe problems with how the model approximates cloud cover, and especially tropical convection. To the point where I must question the integrity of the model itself.

I understand the probabilistic nature of the experiment, but given some of potential problems I've noted I've got to wonder how accurate the overall endeavour can be considered. Just as the slab ocean model puts serious limitations on the model (hence the need more the more accurate coupled model in the next step), I feel the models treatment of clouds and tropical convection may also put similiar restrictions
on it - despite its statistical nature. I'm not aware of how the model is dealing with these things internally, so I may well be wrong.

I've noticed some very basic problems - clouds and fronts in the extratropics retrogading (easterly) with respect to pressure variations. A seeming lack of any apparent convection in the tropics, or for that matter accurate clouds of any sort. Strange pressure anomalies. Weird snow and ice extents. Again, some of this may be due to the parameters on my particular model... I've put together a page with animations spanning several model days off my computer and omparisons to actual satellite images I've collected over the years:

http://www.fractalfreak.com/ClimatePrediction/index.html

Yeah, it's a rather large page. Kinda blew up on me :(
If you want to skim it, check out the top text and links to the first two .wmv movies. Also check the botton of the page to see some of the strange pressure anomalies I've observed.

I figure modellers always need feedback, which can be hard to come by ... :)

Anyway, let me know if you think there's anything to my concerns.

-Eric B

Interesting stuff. Especially the clouds going westward in the southern hemisphere. I hadn't looked at any time lapses of CPDN output like that before.

I'll let the modelers explain what you are seeing.

But as for Linux users not being able to view .wmv files, not true. Most modern Linux distributions have a media player that will play Windows media files, and if not, VLC or MPlayer can be downloaded and installed to view them.

I'm not sure how you captured images, but using Alt-PrintScrn will capture just the active window to clipboard so the capture is the same size each time. But <a href="http://www.brothersoft.com/Multimedia_Graphics_Screen_Capture_WinGrab_6192.html">WinGrab</a> should be able to capture individual window contents at timed intervals for later generation into an animation. This may be helpful if you want to do more.

Hi,

I read your article, and found it interesting.
As I'm not a climatologist nor a computer geek, so I can't answer rightly but I'm going to give my point of view.
The atmosphere is divided into 19 layers and just 4 of them are shown on the screen-saver.
I don't think seeing 4 layers of cloud is sufficient enought to conclude that the model is unrealistic.
This is why you seem to think that temp and pressure are more or less OK but cloud layers are terrible: you don't see the whole picture.

As for the snow on Texas, you're absolutly right: It's completely bizarre and unrealistic. Perhaps this fact comes from your model.
Others models show cold equator or impossible things like that. I just finished a model last week with a cold spot near South-America with negative temps at the equator.
Reading the Nature article, you'll see that a lot of models are not taken into account by scientists because of strange behaviors.

Second thing: the hadsm3 model is a simplified model, with a slab ocean.
Perhaps the atmosphere is also simplified: Perhaps the globe isn't turning and so no Coriolis Acceleration is created: so you don't see all twirling and convection phenomenas.
Well I don't know: Perhaps someone more qualified than me will answer your questions.
Later in this year a more realistic model (coupled model) will be released and perhaps will take into account more physical parameters.

Last thing: You should use CPview or the Advanced vizualisation package that's going to be released for BOINC to do your work because taking screenshots of the screensaver is not a very accurate tool.

&gt; Perhaps the globe isn't turning and
&gt; so no Coriolis Acceleration is created: so you don't see all twirling and
&gt; convection phenomenas.

I do know the Coriolis effect is included in the model. There was a nice thread in the php board where DaveF gave a list and brief description of the subsystems in the model and the Coriolis effect was one of them. (UK_Nick gave Dave a hard time of the name of the gravity drag wave.)

BTW what is the typical latitude which divides the Polar Cell from the Ferrell Cell in the Southern hemisphere?

---

Visit BOINC WIKI for help

And join BOINC Synergy for all the news in one place.

hi

&gt; Interesting stuff. Especially the clouds going westward in the southern
&gt; hemisphere. I hadn't looked at any time lapses of CPDN output like that
&gt; before.

Yes. In fact, from what I can see, ALL cloud patterns on my computer
move eastwards, including in the northern hemishpere. The clouds
seems to be almost an afterthought compared to the temp and pressure
outputs (the pressure output DOES move westerly in the extratropics,
as it should). Looking at the statisitcal average of the cloud
cover would never have caught the problem, IMO. From what I can
see the cloud model is waaay too facile, and amounts to little more
than eye candy.

&gt;
&gt; I'll let the modelers explain what you are seeing.
&gt;
&gt; But as for Linux users not being able to view .wmv files, not true. Most
&gt; modern Linux distributions have a media player that will play Windows media
&gt; files, and if not, VLC or MPlayer can be downloaded and installed to view
&gt; them.
&gt;

Good to know, I'm glad more than just windows users have
the potential to view the movies - I was worried about that.
If it wasn't for video games I would've have ditched windows
long ago. Though I must admit the latest versions on windows
(XP pro/NT) are quite robust and stable compared to earlier versions.

&gt; I'm not sure how you captured images, but using Alt-PrintScrn will capture
&gt; just the active window to clipboard so the capture is the same size each time.
&gt; But <a> href="http://www.brothersoft.com/Multimedia_Graphics_Screen_Capture_WinGrab_6192.html"&gt;WinGrab</a>
&gt; should be able to capture individual window contents at timed intervals for
&gt; later generation into an animation. This may be helpful if you want to do
&gt; more.
&gt;

Thanks for the link. I see you've got to embed the links in html
style tags to get them clickable. I wondered why my link
didn't come up as clickable... (maybe I can edit my original
post?)

Regards, Eric B

&gt; I do know the Coriolis effect is included in the model. There was a nice
&gt; thread in the php board where DaveF gave a list and brief description of the
&gt; subsystems in the model and the Coriolis effect was one of them. (UK_Nick gave
&gt; Dave a hard time of the name of the gravity drag wave.)

Thanks, good to know that. :o)
I would like to read this thread because I like to understand the science sides of the projects.

&gt; Hi,
&gt;
&gt; I read your article, and found it interesting.
&gt; As I'm not a climatologist nor a computer geek, so I can't answer rightly but
&gt; I'm going to give my point of view.
&gt; The atmosphere is divided into 19 layers and just 4 of them are shown on the
&gt; screen-saver.
&gt; I don't think seeing 4 layers of cloud is sufficient enought to conclude that
&gt; the model is unrealistic.
&gt; This is why you seem to think that temp and pressure are more or less OK but
&gt; cloud layers are terrible: you don't see the whole picture.

True, but I don't think the cloud movement can be explained by
hidden layers. Supposedly the model shows three layers of clouds -
I would hope these would be distributed throughout the modelled layers.
And deep tropical convection should show up on ALL cloud layers.

(...)

&gt; Second thing: the hadsm3 model is a simplified model, with a slab ocean.
&gt; Perhaps the atmosphere is also simplified: Perhaps the globe isn't turning and
&gt; so no Coriolis Acceleration is created: so you don't see all twirling and
&gt; convection phenomenas.

Yes, yes I'm aware of the limits on the slab model. And no, the
coriolis force is there, it's just plain wrong, at least from looking
at the cloud output.

&gt; Well I don't know: Perhaps someone more qualified than me will answer your
&gt; questions.
&gt; Later in this year a more realistic model (coupled model) will be released and
&gt; perhaps will take into account more physical parameters.
&gt;

Yes, a more realistic model is needed. The slab model is cute, but it's
almost an insult to a problem of this magnitude. Years ago I was
interested in writing my own global weather simulator, but I gave
up when I realized the computational complexity - I didn't want to
wait days for each frame to complete. The coupled ocean model being
worked on is a good thing. And I still think the sampling scheme is
too course. I worry they may have crippled the model so it will run
in a reasonable amount of time so people will not loose
interest. I'd be more interested in a much more comprehensive model,
even if it takes a year or two to complete a model run. I'm certain
they'd loose some poeple from the project is thie were the case, but
no doubt many would remain, and the model output would be more realistic.

&gt; Last thing: You should use CPview or the Advanced vizualisation package that's
&gt; going to be released for BOINC to do your work because taking screenshots of
&gt; the screensaver is not a very accurate tool.

Well, the cpdn visualization package (cpdn_viz2, which I downloaded) would
not run on my (windoze) computer. I was told they were working on
getting it running. This would certainly be perferable to using
ctrl-c/ctrl-v :)

-Eric B

---

&gt; Thanks for the link. I see you've got to embed the links in html
&gt; style tags to get them clickable. I wondered why my link
&gt; didn't come up as clickable... (maybe I can edit my original
&gt; post?)
&gt;
A forum post on how to do some stuff in html in these message boards is <a href="http://climateapps2.oucs.ox.ac.uk/cpdnboinc/forum_thread.php?id=140">here</a> if you didn't already know how.

(...)
&gt;
&gt; BTW what is the typical latitude which divides the Polar Cell from the Ferrell
&gt; Cell in the Southern hemisphere?
&gt;

Crandles, I'm not certain. Most of what I know is based on observations
of satellite imagery. I could go paw through my old Met textbooks and
find out. It's not a hard dividing line, obviously.

Point to be made is the prevailing winds off the coast of antarctica
are westerly (and, come June they'll be screaming). Looking at my first
movie, concentrate (radially) on the outer half of the globe, towards the
few continents vis. in the SH. The flow is clearly backwards, at least in
the cloud output.

-Eric

---

&gt; &gt; Interesting stuff. Especially the clouds going westward in the southern
&gt; &gt; hemisphere. I hadn't looked at any time lapses of CPDN output like that
&gt; &gt; before.
&gt;
&gt; Yes. In fact, from what I can see, ALL cloud patterns on my computer
&gt; move eastwards, including in the northern hemishpere. The clouds
&gt; seems to be almost an afterthought compared to the temp and pressure
&gt; outputs (the pressure output DOES move westerly in the extratropics,
&gt; as it should). Looking at the statisitcal average of the cloud
&gt; cover would never have caught the problem, IMO. From what I can
&gt; see the cloud model is waaay too facile, and amounts to little more
&gt; than eye candy.
&gt;
Mostly just meteorological semantics, but so everyone is on the same page, by convention, moving east (west) or eastward (westward) would be a movement from west to east (east to west). However, easterly (westerly) flow would be a air flow from east to west (west to east). It's possible you way you used them above may confuse someone reading that, but then again, I probably just confused the heck out of people reading this as well. ;) So, the "proper" meteorological usage might be: in the tropics there is a general easterly flow in the mid to upper atmosphere, so weather systems generally move westward (toward the west). In the middle latitudes, there is a general westerly flow in the mid to upper atmosphere, so weather systems generally move eastward (toward the east).

&gt;I worry they may have crippled the model so it will run
&gt;in a reasonable amount of time so people will not loose
&gt;interest.

No, don't worry about that.
The sulphur cycle is under Alpha test and it lasts about 2000 hours on my machine (~750 hours for the regular model), so I don't think the model is crippled for time reason.
If devs team needed to do models during 5000 hours or more, I think they would find fanatic users to crunch such models.
Nonetheless, I understand your point about coarse resolution of the model.

&gt; &gt;I worry they may have crippled the model so it will run
&gt; &gt;in a reasonable amount of time so people will not loose
&gt; &gt;interest.
&gt;
&gt; No, don't worry about that.
&gt; The sulphur cycle is under Alpha test and it lasts about 2000 hours on my
&gt; machine (~750 hours for the regular model), so I don't think the model is
&gt; crippled for time reason.
&gt; If devs team needed to do models during 5000 hours or more, I think they would
&gt; find fanatic users to crunch such models.
&gt; Nonetheless, I understand your point about coarse resolution of the model.
&gt;

Oh yes, I'm certain they'd be able to find poeple to run
the expensive models; I'd be one of them. So I hope they
don't hesitate to 'open the valves' on the computational
complexity.

I'm glad you agree on the coarseness of the model. Even they
admit the timestep of 1/2 hour is pushing the limits of stability
on the iterated time-step approach.

When I was younger and more ambitious I used to think 'solving'
(simulating) global weather was possible. More and more I wonder
if the problem is completely intractable on the scale being attempted
here. The atmosphere is very much alive. The energy flows and
comlexities beggar the imagination, especially any attempts at
conceptually glossing the problem over. After watching satellite
imagery for years I've come to this conclusion. Never make the mistake
of confusing any model for reality.

That said I still think this is worthwhile project.

-Eric

&gt; (...)
&gt; &gt;
&gt; &gt; BTW what is the typical latitude which divides the Polar Cell from the
&gt; Ferrell
&gt; &gt; Cell in the Southern hemisphere?
&gt; &gt;
&gt;
&gt; Crandles, I'm not certain. Most of what I know is based on observations
&gt; of satellite imagery. I could go paw through my old Met textbooks and
&gt; find out. It's not a hard dividing line, obviously.
&gt;
&gt; Point to be made is the prevailing winds off the coast of antarctica
&gt; are westerly (and, come June they'll be screaming). Looking at my first
&gt; movie, concentrate (radially) on the outer half of the globe, towards the
&gt; few continents vis. in the SH. The flow is clearly backwards, at least in
&gt; the cloud output.
&gt;

Ug, yes, I meant to say the prevailing winds around antarctica,
and into the mid latitudes usually move eastward. The model doesn't
seem to do a very good job with commonly found blocking patterns
nor extreme events either. The more I look at the pressure output the
more 'sanitized' it appears.

-Eric

---

(...)

I just corrected some stupid mistakes in my hastily
thrown together page, where I said things were moving
East when I meant West and the other way around.

Gist of the <a> href="http://www.fractalfreak.com/ClimatePrediction/index.html"&gt;page</a> is the same though.

-Eric

---

Eric, I think a lot of your questions can be answered in the Hadley Centre <a href="http://www.metoffice.com/research/hadleycentre/pubs/HCTN/index.html">technical notes</a> and <a href="http://www.metoffice.com/research/hadleycentre/pubs/refereed.html">refereed publications</a>. I have not read many of them myself yet, but several from their titles look like they are what you want.

I think you miss quite a few of the things the models get wrong, tropical clouds are difficult, but so is the motion of pack ice (in the HadCM3 model) and some of the more localised climate features (e.g. NAO, I can't remember if the HadSM3/HadCM3 models get this right)

---

____________________________<br>
<a href="http://www.boincforum.info/boinc/">boinc forum</a> and <a href="http://www.uk4cp.co.uk/">United Kindom</a> team, my climate change <a href="http://www.livejournal.com/users/mike_atkinson/">blog</a>.

&gt; Eric, I think a lot of your questions can be answered in the Hadley Centre <a> href="http://www.metoffice.com/research/hadleycentre/pubs/HCTN/index.html"&gt;technical
&gt; notes</a> and <a> href="http://www.metoffice.com/research/hadleycentre/pubs/refereed.html"&gt;refereed
&gt; publications</a>. I have not read many of them myself yet, but several from
&gt; their titles look like they are what you want.
&gt;
&gt; I think you miss quite a few of the things the models get wrong, tropical
&gt; clouds are difficult, but so is the motion of pack ice (in the HadCM3 model)
&gt; and some of the more localised climate features (e.g. NAO, I can't remember if
&gt; the HadSM3/HadCM3 models get this right)
&gt;

Hi,

Thanks for the links, but can you be more specific? There's dozens of
technical briefs on the page. Did you look at my page and the animations
from my model output? (especially the first movie loop showing the
flawed extratropical cloud circulation about the south pole?) What points
did I miss?

I understand tropical convection is very difficult too model, and the
need to simplify the problem to its essence, if possible. It's just
from what I can see it's not being done -- at least from looking at my
model output I see no evidence of it. The cloud model (assuming it's
not just a gltich in the routine to display the clouds from the internal
model representation) is out of whack with the pressure and temp models.
This alone would mess up critical feedback processes.

I was hoping for a quick response and rebutal of my page. I also
sent a note to the climateprediction people, but have yet to hear
back from them. Do you think I should attempt to bother the hadley
people directly?

Thanks, Eric B

---

It's me again. Yes, I'm bogarting my own thread. lol.

Well, I think I found the answer I was looking for, at
least regarding the models treatment of tropical convection.
This excerpted from HCTN-51 (off the hadley site) where
they are discussing the performance of tropical convection
in the coupled models

"
Convectively coupled equatorial waves are a key part of the tropical
climate system. A faithful representation of these wave modes is
needed for predictions on all time-scales. However, at present there
is little understanding of how well they are treated in state-of-the-
art models and knowledge of these waves is very limited
"

There you have it. I would like to add the above should
be amended to say that equatorial waves are ONE of the keys
to understanding the GLOBAL climate system.

Unless some sort of understaning of tropical convection can
be achieved (certainly greater than we have now) you aint
got a working global climate model, in my very honest opinion.

I've developed quite a fetish for tropical convection from
looking at satellite imagery. Awesome stuff. Each wave, each
region, has a certain uniqueness that I find impossible to
quantify.

Regarding the western movement of clouds in the extratropics -
this needs to be fixed, assuming it's not a parameter issue. It
may be something as simple as a display glitch (meaning the
model is correct internally, but the display routine is flawed)
or it may be flawed throughout (in which case it doubly must be
fixed). I hope thay aren't hesitant to admit this (assuming
the later case) given the large number of model runs that may
be of questionable quality because of this.

nuff said

-Eric B

Eric,

On a more capable BB, I'd send this message to you as a PM (Private Message). For obvious reasons. Be that as it may... .

I'm sticking my aged neck out here but, having read your bio and these posts, I have no idea who or what you are. Or your real purpose. ... What sort of "geek"?

Your posts suggest an advanced knowledge of atmospheric circulatiuon. Your language suggests otherwise. (Just an old WX Forecaster's observation.)

Are you perchance associated with a Think Tank? Perhaps one not not favorably disposed to this endeavor/endeavour?

Please forgive an old guy's mis-analysis if that not be the case. Otherwise, a few words about yourself and what's really going on here seem to be in order.

Regards,
Jim

---

"We have met the enemy and he is us." -- Pogo
Greetings from coastal Washington state, the scenic US Pacific Northwest.

&gt; Eric,
&gt;
&gt; On a more capable BB, I'd send this message to you as a PM (Private Message).
&gt; For obvious reasons. Be that as it may... .
&gt;
&gt; I'm sticking my aged neck out here but, having read your bio and these posts,
&gt; I have no idea who or what you are. Or your real purpose. ... What sort of
&gt; "geek"?
&gt;
&gt; Your posts suggest an advanced knowledge of atmospheric circulatiuon. Your
&gt; language suggests otherwise. (Just an old WX Forecaster's observation.)
&gt;

Hi,

I'm not certain how to parse that last paragraph, but I'll take it
as a compliment.

&gt; Are you perchance associated with a Think Tank? Perhaps one not not favorably
&gt; disposed to this endeavor/endeavour?

No. Absolutely not. Did you read my page regarding the model? I
most definitely DO NOT have a degree in meteorology, though I've
always had a love of meteorology. I did take a couple entry level courses
in college, that's it. But I look at satellite imagery and
movie loops on a daily basis. So it's not like I've got no experience,
and I'd count experience more than a degree in many cases.

&gt;
&gt; Please forgive an old guy's mis-analysis if that not be the case. Otherwise,
&gt; a few words about yourself and what's really going on here seem to be in
&gt; order.
&gt;

I'm a 36 year old autistic, with a lot of experience programming and
using computers. Primarily computer graphics. example,
I came to this project not knowing what to expect, but I was very excited to
be able to finally inspect one of these CGM models up close. On the evening news where I live (US) they've been using CGM models to show the local forcast for several years now - a computer generated display of pressure flows, precip,
clouds, etc. But I know that just because something is modelled on a
computer doesn't mean it's correct. I also know (from experience) that
these computer models are not much more accurate than an experienced
forcaster, and beyond 48 hours aren't worth much. They need to be continually
updated with real world observations to maintain any coherency. I
was disappointed there was no real treatment of convection in the SM.

I've grown increasingly concerned with what I think may be serious
flaws with the model. What's wrong with that? You know, when many
people hear that such and such was modelled on a computer, it tends
to give it an air of legitimacy and authority, but as you know it
really doesn't mean shit. A model is no better than the assumptions
that went into making it.

'What's really going on?' Well, first off, I don't have a hidden agenda
if that's what you're implying. Do you have anything germane to add to my original inquiry? Some of the stuff I've noticed is Met. 101 league
problems, like the movement of extratropical clouds. Assuming my
model is not the only one displaying this behaviour I'm very surprised
it's not already been noticed.

If you care to correct or educate me I'm all ears (or eyes as the
case may be).

-Eric B

&gt; Regards,
&gt; Jim
&gt;

(...)

Somehow the example link got omitted to some of my strange work:
www.fractalfreak.com

regards, Eric B

---