Late Feb/March Med/Long Range Discussion

[cae]

I honestly can't remember a storm this year that it was superior over all the other models for several days leading up to the event/non-event. Either the other models are getting better (probably), or the super nino has it all confused. And when you think about it, the Euro went from a blizzard to next to nothing in 2 runs. And the GFS really did best with the blizzard, hands down.

 

The other models are getting better.  I know these don't tell the whole story, but here are the 5-day NH H5 verification scores for the last couple of decades.

 

 

The gap between the Euro and the other models used to be huge, but it has narrowed considerably.  It makes sense - the other models had a lot more room for improvement.  It might be hard for the Euro to squeeze out that last bit of performance. 

 

The gap might widen when you go out to longer forecasts.  Right now, the H5 anomaly correlation gap between the Euro and the GFS peaks at about 7 days.  At that range, we're not paying much attention to the operational runs anyway, so the Euro's advantage might not be as apparent.

[Warm Nose]

Ryan Maue ‏@RyanMaue

ECMWF model upgrade-candidate continues at record-high 5-day skill levels. 0.97 scores or 97% anomaly correlation

[Ian]

Euro weeklies for our area:

 

Feb 29-Mar 7: +1

Mar 7-Mar 14: +10 (minimum)

Mar 14-Mar 21: +5

Mar 21-Mar 28: +2

[EastCoast NPZ]

Euro weeklies for our area:

 

Feb 29-Mar 7: +1

Mar 7-Mar 14: +10 (minimum)

Mar 14-Mar 21: +5

Mar 21-Mar 28: +2

 

Just getting ready for summer.

[showmethesnow]

The other models are getting better.  I know these don't tell the whole story, but here are the 5-day NH H5 verification scores for the last couple of decades.

 

 

The gap between the Euro and the other models used to be huge, but it has narrowed considerably.  It makes sense - the other models had a lot more room for improvement.  It might be hard for the Euro to squeeze out that last bit of performance. 

 

The gap might widen when you go out to longer forecasts.  Right now, the H5 anomaly correlation gap between the Euro and the GFS peaks at about 7 days.  At that range, we're not paying much attention to the operational runs anyway, so the Euro's advantage might not be as apparent.

Notice every time verification scores are thrown up they encompass the whole northern Hemisphere. Do they actually break down verification for different regions? Such as for the East Coast ? If so I would think that they would be a better indicator for our region which during the winter can be a tricky area to nail down

[WinterWxLuvr]

Notice every time verification scores are thrown up they encompass the whole northern Hemisphere. Do they actually break down verification for different regions? Such as for the East Coast ? If so I would think that they would be a better indicator for our region which during the winter can be a tricky area to nail down

They are always at h5 too.  I don't know exactly how, but I think the actual on the ground weather should be part of the verification process.  So what if a model nails the h5 pattern, but predicts 2" of precip and I get 0.3".  That's not exactly "verifying" IMO.

[dtk]

They are always at h5 too.  I don't know exactly how, but I think the actual on the ground weather should be part of the verification process.  So what if a model nails the h5 pattern, but predicts 2" of precip and I get 0.3".  That's not exactly "verifying" IMO.

The centers do much, much more than just look at 500 AC.  For fun, here is Feb. 2016 days 1-3 CONUS precip verification:

Yep, Euro is best (just like H5).  Here is an example of 24 hr RMSE of 850 vector winds over North America:

ECMWF is still best.  I could go on.....

[Ji]

EPS showing that low anomoly in East Day 14-15

[Bob Chill]

EPS showing that low anomoly in East Day 14-15

 

GEFS is kinda doing the same thing. Might be too fast but I have a feeling we're in for an annoying period mid month. There could be some "why couldn't this happen 3 weeks ago" kind of stuff. 

[Ji]

GEFS is kinda doing the same thing. Might be too fast but I have a feeling we're in for an annoying period mid month. There could be some "why couldn't this happen 3 weeks ago" kind of stuff. 

HM was talking on twitter that there could be another winter threat for the interior and the spring  high fiving will stop by mid month and then a true SSW taking place(why didnt this happen 3 weeks ago lol)

[WinterWxLuvr]

The centers do much, much more than just look at 500 AC.  For fun, here is Feb. 2016 days 1-3 CONUS precip verification:

Yep, Euro is best (just like H5).  Here is an example of 24 hr RMSE of 850 vector winds over North America:

ECMWF is still best.  I could go on.....

 

 

Thanks dtk.  The only place I ever see verification is on here and it's always at h5.

[cae]

Notice every time verification scores are thrown up they encompass the whole northern Hemisphere. Do they actually break down verification for different regions? Such as for the East Coast ? If so I would think that they would be a better indicator for our region which during the winter can be a tricky area to nail down

 

They are always at h5 too.  I don't know exactly how, but I think the actual on the ground weather should be part of the verification process.  So what if a model nails the h5 pattern, but predicts 2" of precip and I get 0.3".  That's not exactly "verifying" IMO.

 

You can get a lot of verification scores for different metrics (precip, temp, slp, etc.) at this site.

 

http://www.emc.ncep.noaa.gov/gmb/STATS_vsdb/

 

On those pages, "PNA" is basically North America.  You can also find ensemble verification scores there, but for some reason I don't think they have scores for the mesoscale models. 

 

It's harder to find historical data for North America for all of these statistics. dtk, are those charts you posted publicly available?

[psuhoffman]

Very frustrating for sure that many of the pattern drivers I was paying attention to, the MJO, AO, even NAO to some degree went the way I wanted, but the EPO on the pacific side decided to break down and really limit the potential on this period.  I was kind of just assuming the pacific side would stay in our favor since it mostly was all winter.  It was the Atlantic that was our issue.  So of course just when we finally get the Atlantic to cooperate the Pacific decides to give it to us.   We do get a 5 day window before the Pacific really decides to ruin things for us, but its now looking like both systems within that window want to be pretty pathetic, especially for this time of year.  I hold out some very very low hope that perhaps one of these can become more then just a neussance inch or two but it sure looks like that is our fate.  After that the pacific side really just kills it.  It wont matter what the AO is doing with the EPO flooding the CONUS with the NINO blowtorch.  I still think mid month things are going to be miserable.  The pacific still looks kinda mediocre but with the higher heights all over Canada and the STJ cutting under we will likely have some storms coming across.  But at this point, given what is going to happen the next week, I doubt they will find enough cold to do us any good.  Like was said yesterday it will be a lot of "why didn't this happen a few weeks ago"  THere are still a few rogue runs within both the GEFS and EPS that show something to hold out that lottery type hope but its beyond a long shot right now.  More likely we get some juiced up STJ system coming out of the gulf and up the east coast and we get days of 40 degrees and rain.  I bet the higher elevations above 2500 feet in WV and western MD get something mid month. 

[dtk]

You can get a lot of verification scores for different metrics (precip, temp, slp, etc.) at this site.

 

http://www.emc.ncep.noaa.gov/gmb/STATS_vsdb/

 

On those pages, "PNA" is basically North America.  You can also find ensemble verification scores there, but for some reason I don't think they have scores for the mesoscale models. 

 

It's harder to find historical data for North America for all of these statistics. dtk, are those charts you posted publicly available?

Yeah, there are many resources for verification.

 

NCEP EMC precip:  http://www.emc.ncep.noaa.gov/mmb/ylin/pcpverif/scores/

 

There is additional verification for precipitation, fits-to-obs, etc. from the aforementioned "STATS_vsdb" page at EMC (use the upper left menu).

 

The other plot I showed for 850 wind verification is derived from the WMO verification page here:

http://apps.ecmwf.int/wmolcdnv/

 

You just have to figure out how to navigate things.  The subset of variables is small-ish as these are the agreed upon metrics to share when doing international inter-comparisons.  

 

There is other verification such as object-oriented (MODE) for precipitation from WPC:  http://www.wpc.ncep.noaa.gov/verification/mode/mode.php#page=page-1

 

Lots and lots and lots of stuff out there.....

[MN Transplant]

Maybe the verification stuff should be in the NWP thread so it doesn't get lost? 

[showmethesnow]

You can get a lot of verification scores for different metrics (precip, temp, slp, etc.) at this site.

 

http://www.emc.ncep.noaa.gov/gmb/STATS_vsdb/

 

On those pages, "PNA" is basically North America.  You can also find ensemble verification scores there, but for some reason I don't think they have scores for the mesoscale models. 

 

It's harder to find historical data for North America for all of these statistics. dtk, are those charts you posted publicly available?

Thanks. Shame they don't break it down for specific regions. It would be nice to verify if the tendencies/bias we sometimes attribute to certain models in certain situations was real and not imagined.

[Bob Chill]

GEFS trended colder in the long range and a decent h5 look as well.  PSU may get his elevation paste bomb. 

[Ian]

GEFS trended colder in the long range and a decent h5 look as well.  PSU may get his elevation paste bomb. 

only about +2 by the end of the run!

[Ji]

GEFS trended colder in the long range and a decent h5 look as well.  PSU may get his elevation paste bomb. 

looks like a nice split flow pattern with some blocking on top. March 1958 redux coming. See you in 2 weeks

 

http://www.tropicaltidbits.com/analysis/models/?model=gfs-ens&region=namer&pkg=z500a&runtime=2016030112&fh=384&xpos=0&ypos=333.3333421636513

[Bob Chill]

only about +2 by the end of the run!

 

It's a d17-18 event. Bowling ball style. It's happening!

 

[psuhoffman]

GEFS trended colder in the long range and a decent h5 look as well.  PSU may get his elevation paste bomb. 

There a a LOT of nasty miserable cold rain storm looks on the individual GEFS and EPS runs mid march.  There are also a few that are cold enough for a big wet snow paste job for places NW of the metros and with a little elevation.  I think its unavoidable we get a stormy period but odds favor wet over white given the CONUS will be flooded with pacific air and any storm would have to bomb and track perfect enough to virtually create its own cold, kinda like 58.  I am sure the snow line will set up at 1200 feet.  (I am around 1100)

[cae]

Maybe the verification stuff should be in the NWP thread so it doesn't get lost? 

 

I agree.  That would probably be a more appropriate thread for it anyway.  Maybe one of the mods can move these posts.

 

Thanks. Shame they don't break it down for specific regions. It would be nice to verify if the tendencies/bias we sometimes attribute to certain models in certain situations was real and not imagined.

 

From my observations, the verification scores tend to track local model performance reasonably well, but having more localized data would be helpful.  Perhaps the best would be to provide the underlying data so that people can do their own analysis, although I'm not sure how many weather services would be OK with that. 

[mitchnick]

looks like a nice split flow pattern with some blocking on top. March 1958 redux coming. See you in 2 weeks

 

http://www.tropicaltidbits.com/analysis/models/?model=gfs-ens&region=namer&pkg=z500a&runtime=2016030112&fh=384&xpos=0&ypos=333.3333421636513

temps matter too

[psuhoffman]

looks like a nice split flow pattern with some blocking on top. March 1958 redux coming. See you in 2 weeks

 

http://www.tropicaltidbits.com/analysis/models/?model=gfs-ens&region=namer&pkg=z500a&runtime=2016030112&fh=384&xpos=0&ypos=333.3333421636513

The look is what we would want to see if we did want a march 58 repeat but at the same time that kind of thing is truly like hitting the jackpot.  We could get that same look 100 times and only once would it work out that way in late March.  But at least it keeps hope alive.  I would rather it just be nice out, but since it looks wet and miserable around March 15-20 might as well root for a fluke late season snow. 

[Ji]

temps matter too

it will be cold enough during the storm. Were not getting arctic air in place a week out like we did for the January blizzard. It just has to be cold enough when its storming

[mitchnick]

it will be cold enough during the storm. Were not getting arctic air in place a week out like we did for the January blizzard. It just has to be cold enough when its storming

That way of thinking has not worked so far this season and I suspect the chances of it working in the March 15-20th period are even less.

We need a seriously cold temp anomaly to go along with the right upper level pattern to seriously consider it ever happening/or having a reasonable chance.

[dtk]

I agree.  That would probably be a more appropriate thread for it anyway.  Maybe one of the mods can move these posts.

 

 

From my observations, the verification scores tend to track local model performance reasonably well, but having more localized data would be helpful.  Perhaps the best would be to provide the underlying data so that people can do their own analysis, although I'm not sure how many weather services would be OK with that. 

Other than a small subset of observations for which there are agreements in place, everything that comes out of NCEP is publicly available on some fashion.

[Ji]

That way of thinking has not worked so far this season and I suspect the chances of it working in the March 15-20th period are even less.

We need a seriously cold temp anomaly to go along with the right upper level pattern to seriously consider it ever happening/or having a reasonable chance.

or have a closed ULL track right below us or on top of us. I saw the Carolinas a few years ago get historic snow while we were partly cloudy and mid 40's":(

[mitchnick]

or have a closed ULL track right below us or on top of us. I saw the Carolinas a few years ago get historic snow while we were partly cloudy and mid 40's":(

Well, if you're asking me to consider pipe dreams, OK, but I was talking reasonable hopes/expectations based on ensemble runs.

[Ian]

we're due for a good 44 and rain