BaltimoreWxGuy Posted February 16, 2012 Share Posted February 16, 2012 Didnt the Euro initalize poorly? If so then why are we discussing it and its ensembles? just food for thought maybe? Link to comment Share on other sites More sharing options...
Deck Pic Posted February 16, 2012 Share Posted February 16, 2012 I don't see anything in the logs and I'm not in the building today if they made any announcements. From the status page, it started on time but is definitely running slow. We've been having a lot of machine issues lately..... I'm out to 21hrs, though I am not usually much faster than the NCEP site itself....I've seen them restart the run before, but it is rare and don't remember that in quite a while Link to comment Share on other sites More sharing options...
TalcottWx Posted February 16, 2012 Share Posted February 16, 2012 Didnt the Euro initalize poorly? If so then why are we discussing it and its ensembles? just food for thought maybe? It would be foolish not to study the run poor intialization or not Link to comment Share on other sites More sharing options...
dtk Posted February 16, 2012 Share Posted February 16, 2012 Didnt the Euro initalize poorly? If so then why are we discussing it and its ensembles? just food for thought maybe? First off, the notion that the Euro "initialized poorly" is rubbish. HPC did note that it appeared (subjective) that the GFS lined up better with the BC shortwave when they visualized some of the model fields and water vapor imagery (compared to the Euro and NAM). Further, if you believe that the model "initialized poorly", that would be even greater reason to rely on the ensembles....since part of their purpose is to sample/represent the initial condition uncertainty (analysis error). Link to comment Share on other sites More sharing options...
WinterWxLuvr Posted February 16, 2012 Share Posted February 16, 2012 First off, the notion that the Euro "initialized poorly" is rubbish. HPC did note that it appeared (subjective) that the GFS lined up better with the BC shortwave when they visualized some of the model fields and water vapor imagery (compared to the Euro and NAM). Further, if you believe that the model "initialized poorly", that would be even greater reason to rely on the ensembles....since part of their purpose is to sample/represent the initial condition uncertainty (analysis error). Thanks for posting that. Good point. Is there some issue that causes the models to sometimes run slowly? I noticed over on the psu site, the sref members, which usually update about a panel per minute when they are coming in, are updating at about a panel per 5 minutes. Link to comment Share on other sites More sharing options...
phil882 Posted February 16, 2012 Share Posted February 16, 2012 The UKMet is a fine model. However, the difference in skill between the GFS, UKMet, and CMC models is closer than that figure implies. The only model that is better, using this metric, with 95% statistical significance (at least in recent years) is the ECMWF (this is true for many variables, metrics, and levels). Also keep in mind that it's a global, time averaged metric...so it doesn't always necessarily translate to skill for discrete events. I agree with this... the difference in AC between the UKMet/GFS/CMC are pretty small. I just thought I'd stick up for the CMC and UKMet since they have decent verification scores and others were saying all you need to do is look at the GFS and ECMWF and your set. You aren't getting the full picture of the spread if you do that. Link to comment Share on other sites More sharing options...
yoda Posted February 16, 2012 Share Posted February 16, 2012 I'm out to 21hrs, though I am not usually much faster than the NCEP site itself....I've seen them restart the run before, but it is rare and don't remember that in quite a while Running 15-30 mins late http://www.nco.ncep.noaa.gov/pmb/nwprod/prodstat/ Link to comment Share on other sites More sharing options...
dtk Posted February 16, 2012 Share Posted February 16, 2012 Thanks for posting that. Good point. Is there some issue that causes the models to sometimes run slowly? I noticed over on the psu site, the sref members, which usually update about a panel per minute when they are coming in, are updating at about a panel per 5 minutes. Well, our operational supercomputer and its backup are nearing the end of the contract (and their lifetime). We are running with a pretty high compute load which gives us a lot less wiggle room. [We're scheduled to have our new computers delivered in 2013]. I haven't been paying attention to the details, but I know that both the operational and backup computers have had a myriad of minor issues lately. If a job lands on a sluggish node, or if there are disk issues, things can slow down pretty quickly. The issue with the PSU site could be on their end, as it seems as though the SREF has been running on-time lately...but I'd have to dig around some to say anything useful. Link to comment Share on other sites More sharing options...
usedtobe Posted February 16, 2012 Share Posted February 16, 2012 First off, the notion that the Euro "initialized poorly" is rubbish. HPC did note that it appeared (subjective) that the GFS lined up better with the BC shortwave when they visualized some of the model fields and water vapor imagery (compared to the Euro and NAM). Further, if you believe that the model "initialized poorly", that would be even greater reason to rely on the ensembles....since part of their purpose is to sample/represent the initial condition uncertainty (analysis error). I've thought for awhile that the subjective thoughts on initialization errors are questionable as so much of the data used is not derived from satellite which the subjective evaluation is saying is wrong. I guess I;m saying the objective analysis based on satellite data is getting evaluating strictly based on clouds while the objective analysis uses sensors that measure things. I'm sure there are forecasters who disagree with me. The euro ensemble mean is not that different than the operational so I'm not sure why anyone would think the euro is any worse than any other model in this case. That said, where the srefs is showing the most uncertainty with respect to the mean low position is pretty interesting and say not to grab hold of any solution too quickly as here are some members farther north than the mean. Link to comment Share on other sites More sharing options...
psuhoffman Posted February 16, 2012 Share Posted February 16, 2012 there are always some suppressed, or some way north, or some ots, or some to our west it seems statistically, these systems in a NINA come further north than modeled at 72 hrs more often than not and I don't believe that's an unfair statement this could be one of the few that goes otherwise, but the odds favor us, and that's all we can hope for put another way, anyone in DCA/BWI prefer the Euro or any other model being too far north for hit at this point? You are having a selective memory about a Nina pattern...seems to me the actual trend in a Nina is for storms that would hit us to trend either north or south and miss. THe northern and southern stream tends not to phase well in a Nina and so storms that are northern stream dominant tend to trend north over time and go north of us. Storms that are southern stream dominant tend to stay supressed south of the northern stream and not phase. Either way we get screwed hence why there are not many big snowstorms for our area in this pattern. Link to comment Share on other sites More sharing options...
dtk Posted February 16, 2012 Share Posted February 16, 2012 I agree with this... the difference in AC between the UKMet/GFS/CMC are pretty small. I just thought I'd stick up for the CMC and UKMet since they have decent verification scores and others were saying all you need to do is look at the GFS and ECMWF and your set. You aren't getting the full picture of the spread if you do that. Exactly, especially the UKMet (which I defend all the time). There is a problem if you don't look at these models on a regular basis though, as you have no idea what to look at in terms of their tendencies/usefulness/biases, etc., making interpretation of raw output impossible. It's even worse if you only look at them during high impact events (which have large error growth associated with them). Link to comment Share on other sites More sharing options...
yoda Posted February 16, 2012 Share Posted February 16, 2012 I don't see much difference when comparing 18z h5 at 36 to 12z h5 at 42... perhaps a smidge slower with teh h5 energy in MX on the 18z run Link to comment Share on other sites More sharing options...
dtk Posted February 16, 2012 Share Posted February 16, 2012 I've thought for awhile that the subjective thoughts on initialization errors are questionable as so much of the data used is not derived from satellite which the subjective evaluation is saying is wrong. I guess I;m saying the objective analysis based on satellite data is getting evaluating strictly based on clouds while the objective analysis uses sensors that measure things. I'm sure there are forecasters who disagree with me. Mods: Feel free to move this to the Met 101 section if you want. To Wes: Right...I've been thinking about this for a while and I want to come up with a way to demonstrate this to people. I need to design a way to illustrate the following points Creating an analysis / initial condition is statistical and probabilistic by nature, since all of the information we use has errors associate with it (even observations). Only a small majority (52-58%) of observations have a positive impact on the forecast (or, another way to say this: between 40 and 48% of the observations we assimilate actually DEGRADE the forecast). We have methods to quantify this, and I'll throw some examples in the Met 101 threads. [this does not mean that those observations are actually bad, however] If we blindly force the model to be close to observations, we get worse forecasts (and significantly so). I think I can demonstrate this with a toy model. The best analysis does not necessarily yield the best forecast from a NWP model. Link to comment Share on other sites More sharing options...
stormtracker Posted February 16, 2012 Author Share Posted February 16, 2012 Let me take a moment to plug the conference here...DTK will be at the conference this year with this info and much more re: models/modeling. Just FYI Link to comment Share on other sites More sharing options...
PhineasC Posted February 16, 2012 Share Posted February 16, 2012 Let me take a moment to plug the conference here...DTK will be at the conference this year with this info and much more re: models/modeling. Just FYI Can dtk or another met give a presentation on the computing power side of NCEP too? Would be interesting. Link to comment Share on other sites More sharing options...
tcutter Posted February 16, 2012 Share Posted February 16, 2012 Exactly, especially the UKMet (which I defend all the time). There is a problem if you don't look at these models on a regular basis though, as you have no idea what to look at in terms of their tendencies/usefulness/biases, etc., making interpretation of raw output impossible. It's even worse if you only look at them during high impact events (which have large error growth associated with them). verification graph from nws mt holly.................. Link to comment Share on other sites More sharing options...
mitchnick Posted February 16, 2012 Share Posted February 16, 2012 You are having a selective memory about a Nina pattern...seems to me the actual trend in a Nina is for storms that would hit us to trend either north or south and miss. THe northern and southern stream tends not to phase well in a Nina and so storms that are northern stream dominant tend to trend north over time and go north of us. Storms that are southern stream dominant tend to stay supressed south of the northern stream and not phase. Either way we get screwed hence why there are not many big snowstorms for our area in this pattern. nope, disagree gimme' a system on the models too far south 3 days out any day don't want to clutter the thread so no further comment by me Link to comment Share on other sites More sharing options...
Tropopause_Fold Posted February 16, 2012 Share Posted February 16, 2012 Let me take a moment to plug the conference here...DTK will be at the conference this year with this info and much more re: models/modeling. Just FYI sold. Link to comment Share on other sites More sharing options...
Deck Pic Posted February 16, 2012 Share Posted February 16, 2012 The 0z Euro might be in range before the 18Z NAM Link to comment Share on other sites More sharing options...
Bob Chill Posted February 16, 2012 Share Posted February 16, 2012 vort over mx @ hr39 is definitely a bit weaker than hr 45 @ 12z. Probably insigificant but I have too much time on my hands between panels. Link to comment Share on other sites More sharing options...
yoda Posted February 16, 2012 Share Posted February 16, 2012 The 0z Euro might be in range before the 18Z NAM 18z GFS data prep/dump is already done and analysis is underway Link to comment Share on other sites More sharing options...
RickinBaltimore Posted February 16, 2012 Share Posted February 16, 2012 The storm might be here before the 18Z NAM Fixed it for you Link to comment Share on other sites More sharing options...
yoda Posted February 16, 2012 Share Posted February 16, 2012 vort over mx @ hr39 is definitely a bit weaker than hr 45 @ 12z. Probably insigificant but I have too much time on my hands between panels. Looks about same to me... its just a bit to the SW more. its just open now instead of being closed Link to comment Share on other sites More sharing options...
yoda Posted February 16, 2012 Share Posted February 16, 2012 Small changes at 48... h5 energy that was in W MI on 12z at 54 is in E WI on 18z... h5 energy in TX slightly weaker Link to comment Share on other sites More sharing options...
yoda Posted February 16, 2012 Share Posted February 16, 2012 12z NAM at 60 had a nice 1004 SLP in C LA... 18z NAM at 54 broad 1008 in same place Link to comment Share on other sites More sharing options...
mitchnick Posted February 16, 2012 Share Posted February 16, 2012 12z NAM at 57 had a nice 1004 SLP in W LA... 18z NAM at 51 naso much definitely a tad south than 12Z don;t know if it's just a hair slower or weaker/south than 12Z at this point prefer north considering the options, but we're already deep into the NAM run attm Link to comment Share on other sites More sharing options...
yoda Posted February 16, 2012 Share Posted February 16, 2012 Looks like suppression on this NAM run... 12z at 66 had 0.50-0.75 QPF into parts of TN... 18z has 0.1ish in the same areas Link to comment Share on other sites More sharing options...
usedtobe Posted February 16, 2012 Share Posted February 16, 2012 I'm getting a bad feeling about the run. Hopefully it is nothing more than me being me. Link to comment Share on other sites More sharing options...
TUweathermanDD Posted February 16, 2012 Share Posted February 16, 2012 I'm getting a bad feeling about the run. Hopefully it is nothing more than me being me. No you're right, its going pretty far south, could start to make some worry and have bad feelings about 0z. Link to comment Share on other sites More sharing options...
Huffwx Posted February 16, 2012 Share Posted February 16, 2012 I'm getting a bad feeling about the run. Hopefully it is nothing more than me being me. Vort is about 75 miles SW-- but it's a tad stronger and more consolidated. I think it does me good here, but not up in DC> Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.