Hi Sim.
Good morning from North Carolina
Oh, I am the only one here...
Hello, Brian!
Aha! Good morning Sim :o)
Hi from FSU/COAPS.
Good day Henry, this is Brian in North Carolina
Hi Brian. How are you today?
Not bad. It's rather dreary here - 42 and foggy.
I presume the weather is more interesting in your location?
Ha, its about to get REAL interesting :)
That's what it looked like on radar... if we lose you mid-meeting, we will know why
I'm not sure you can hear us, can you?
We are on the phone number.
We also can't see who the attendees are on the chat.
The 646 number?
Nobody else is listed there
Hi Sim! I see your screen...
You are searching... searching for something...
Great! Whos is attendee 5?
Nobody can hear me, though, huh?
Hello Ryan!
Getting all sorts of folks on board now...
Hi Brian, Hi Sim. My other meeting was cancelled.
Great to have you on board
How are things Ryan? All well in the frozen north?
I can't tell who the attendees are?
Sorry Brian, technical challenges here.
Oh?
I think we are OK. Getting audio configured for meeting.
I can see Sim's computer screen, and I can see text messages, but I do not hear yet.
Do you all hear me?
No.
The list says you are 'muted'.
Can you hear me?
I hear nobody
Hmmm...
I do not hear anyone.
Sim can you hear us?
Nothing yet
OK.
We're trying to figure out the sound.
This is yuanfu from boulder. I hear some pings but no other sound.
I am the same - pings
Same here.
On my list, I only see me and Ryan as 'on' with audio.
And Ryan is muted
I see you are muted too, Brian.
Huh... on mine, it says I am NOT muted
I muted myself so that I do not have any back noise.
Ah, OK.
So nobody can hear anyone else?
Ha, apparently.
Are you dialed into the 646 number?
Sorry about the delay
If HRD had a stenographer, they could type in a box on Sim's machine and we could read-along.
On my machine, it says I am "muted by Organizer"
For the people on the phone, can they hear anything?
No.
I see we are trying to 'get help'
Yes.
Yes
I still cannot hear anything.
I am here. I just need to reconfigure to talk.
I muted myself so that we don't have audio problems.
Yes. I can hear you Sim.
I hear you too.
We've lost Henry, somehow
Maybe the line of severe weather got there...
Henry, you can't hear us?
No weather yet.
tech problems.
Henry, are the problems on your end?
I muted myself, so the echo should not be coming from me. I will unmute I need to say something.
I am muted
Fire away Sim!
Henry is drying his hair...
Set up is great
Someone is still 'buzzing'
OK, problem solved.
I can hear you.
All.
It appears that user Liou is not muted.
Are you using the same architecture as NMC?
Hi there Munehiko! Would you be able to mute yourself? A lot of 'background noise' is coming through.
Brian-san: Sorry for that. Now OK?
Matt Eastin says "Hi"
And then heard "bit-wise reproducibility", and left the room
It is bitwise reproducable on the same computer architecture, but not across different computer architectures.
It is also not bitwise reproducable for different compilers
My experience is that the cumulus/radiation schemes are the source of the non-reproducability.
in WRF-ARW
I agree!
I would tend to agree with Ryan,.
I agree Ryan
My understanding is that the *main* difference between NMM and ARW is the dynamic core.
Yes yes yes!
The physics are independent.
So, would not surprise me if we have identical problem.
But, I will do some experiments and try to verify this claim.
The physics are indepenent, but they do not interact the same with ARW as with NMM
Right.
For example, Ferrier mp 'goes' with NMM better than ARW
OK.
But in terms of code, they interact in the same way.
OK.
Ryan - I heard that they did not interact in the exact same way.
It seems this is an engineering issue as opposed to a physics issue.
That buried in the registry, there were subtle differences.
For example: Ferrier microphysics was not well suited for Dudhia shortwave. Clouds (from Ferrier) ended up having the optical thickness of cement
Could be, but they should be using the same subroutines for the dynamical core to talk to the physics
I am not as wise on the innards of WRF as others... just passing along what I have heard
That problem is a dynamical core problem, not a bitwise reproducable problem
Ah yes... agreed Ryan
Heads up to WRF guys, in the ~/WRFV2 subdirectory dyn_nmm exists solve_nmm.F which is the solver for NMM core. In dyn_em exists solve_em.F.
True!
Might be worth spending some time to figure out exactly what the differences are.
Henry... ought to be the same story in WRFV3
At least this was true in V2 of WRF-ARW.
Right. Those routines call the same physics interfaces though
Yes Sim, we have been going amongst ourselves
Ha.
Ah yes... we should be seen and not heard
:o)
Sim did a great job.
Np!
No problem
We will do something with WRF in North Carolina
Brian, what are you doing to initialize the vortex (if anything)?
Probably WRF-ARW... can we even get H-WRF?
I have v2 of HWRF souce.
Initialization... a BIG question.
I found it at NCEP somewhere.
Hmm... this could be good!
Or, perhaps Gopal has the source somewhere.
Can non-NCEP folk get it?
I have it.
Well, I have the tarball.
That's good enough.
I briefly tried to compile.
Was unsuccessful, but hadn't invested much time in trying to debug.
Yeah... we have ARW V3... and may well just stick with it
We are leaning towards an initialization much like yours
EnKF and 3DVar only work if the difference between the first guess and the data is 'small'
For 6-hour intervals, in a TC environment, this maybe isn't the case
Depends if you are cycling the DA system over a long period.
My team is sort of interested in all three: newer-bogus, 3D-Var, and EnKF
Well, I would be happy to communicate what I have learned.
True Ryan - cycling... AND frequent data collection
We can't let the model drift too far from data.
It seems that the first time you use aircraft radar, it might be a big shock to the system
It is still not clear what data is necessary to set the vortex intensity. This is an open research question.
The model system, that is
Agree that different amounts of data can have a big impact on system.
My initial intention was to use the 3d-Var capabilities of ARW. However, for my dissertation question, I needed *more control* over the vortex structure.
So true Ryan, so true!
Which was not permitted with 3dvar.
I agree Henry - we will probably try all 3 methods in NC, and then see what we get.
My guess is that the EnKF will do best, better-bogus #2, and 3DVar will do worse
I'd be very interested to see your results.
I think you might get different answers for different storms. Henry's method could work well for symmetric storms, but what about severely sheared systems?
That's the thing - we need some sort of bogus scheme that can account for that sort of thing
Not sure what that IS yet...
3DVar scares me because your error covariances are derived from 30-day means (PArrish and Derber, 1992) and is closer to geostrophy than anything.
Well... they can be different...
Ryan, yes.
You are absolutely right.
You can do a 30-day average of inner nests of WRF ARW runs, and use them
Sheared systems will perform poorly.
Sheared systems will die off, so nobody cares about them anyway :o)
What about a storm that goes from a sheared environment to a non-sheared environment?
Well... those make a comeback, don't they
However, I tried to be vigilante of this and permit ancillary data sets such as Doppler radar u,v winds into the system.
I like the idea of a 30-day mean vortex covariance.
I am by no means an expert, just sharing my thoughts.
I can't imagine trying somethign else...
I can appreciate the experts insights.
As Henry said, it would be irrelevant
When I am done my dissertation, Ryan, I would like to go back and investigate why my initial experiments with 3DVar performed as they did.
Future work!?! :)
Sounds like someone wants a post-doc in Albany...
Ha.
Future work is always good. Problem #1 is the error covariance length scales and assumptions. Gradient wind balance is necessary.
Yes.
It isn't clear to me how to get that.
Out now, right?
The DA one closes on April 2nd... I think
ZP-4
Excellent!
Which case?
Good day everyone!
Take care, everyone.