Jump to content

Sign in to follow this  
Rainshadow

2017-18 Model QPF Comparison Reaches Its Cold Season Conclusion. Snowfall Forecast Evaluation Posted. Surprise 2nd To Last Place Finisher.

Recommended Posts

1 hour ago, Rainshadow said:

Dude,  I messed up.  Could you go back to both runs on the 26th?  Just let me know or post it in.  I have saved your response to this request.  I promise I wont ask you any more favors on this til November.

 

Can't go back any further than 12z 27.. Do you need last nights 0z run or are we stopping it at 12z yesterday?

Share this post


Link to post
Share on other sites
2 minutes ago, tombo82685 said:

Can't go back any further than 12z 27.. Do you need last nights 0z run or are we stopping it at 12z yesterday?

No we can stop at 12z yesterday.  Thanks Tom for all your help with this thread.:)  I will get the results of last night's event in by Monday.

  • Like 1

Share this post


Link to post
Share on other sites

Well we come to the end of the another cool season model qpf evaluation for PHL.  I have to give a big thanks to Tom for putting up with all of my EC/EPS requests throughout the season :clap:. While we always encourage comments/suggestions/questions, this thread has nearly 11,000 views which is fantastic and up about 2,200 views from the last cold season thread.   The new thread will be started on/about November 1st for the next cold season.  Modeling changes come fast and furious and there will be a new GFS & EC for the next cold season at the very least.  So we all have to take that into consideration going into next winter and hope a qpf pattern can be established in the fall that holds once it does start snowing.  The snow part of the spread sheet does always give Tom & I heartburn as the climo 11:1 ratio will underforecast snow when we are in the dgz in January, while overforecast March snows occurring during the day when it is not moderate/heavy.   We may switch to Kutchera for next season.  

I will do some evaluation and post it with the new thread come November.  Anyone else that seriously sees any pattern with this thread sheet, please feel free to pass it along.

So the final (48th) event, the NAM went out a winner and for only the second time this cold season (both in April speaking of patterns), the EPS came in last.  It tied with the OP EC for last.  Regardless of this outcome, the EPS was in a class of its own as the most accurate qpf model/ensemble for PHL with 20 first place finishes.  The NAM which can run either very good (mostly with lower pcpn events) or more often very bad (as in wet) came in last 16 times (one third of all events).  The other note while the SREF and EPS showed more qpf skill than its OP; the same can not be said for the GEFS (affectionately called the WEFS) whose mean showed no improvement over the OP model.

Thank-you everyone for checking in!

 

This is the final  leaderboard of the season:

 

There have been instances of tied for first and tied for fifth, so the numbers in the first group are not going to add up to 48.

Without further ado

model/ # of times it was the most accurate model/ # of times it was the least accurate model (48 events)

GFS.......2/6

GEFS.....4/13

NAM.....7/16

SREF.....6/9

EC.........11/3

EPS.......20/2

 

Leaderboard (1 = event first place finish;  2 = event second place finish;  3 = event third place finish, etc, then scores for each event added, the lower the score the more accurate accumulatively the model)

EPS.......100

EC........144

SREF.....167

GFS.......182

GEFS.....187

NAM......191  (No Kirkland ball for you).

NEW2017-8qpf.xlsx

 

  • Like 4

Share this post


Link to post
Share on other sites

This would of been a hideous bust for all models with this system on precip amounts 

  • Like 1

Share this post


Link to post
Share on other sites
15 hours ago, tombo82685 said:

This would of been a hideous bust for all models with this system on precip amounts 

I went back and looked at the 48 events; 35 of them the models were predominately too wet and 13 they were predominately too dry. This even holds for the EC & EPS, but their strength was they were not "as wet".  The tropical connection with this one remained offshore.  It doesn't take long now too start getting too dry.

 

Share this post


Link to post
Share on other sites

A few maps, including the new core gfs.  BTW I read that this is not the big upgrade for the GFS, that is expected to occur in 2020.

fv3p_apcpn_neus_10.png.efaf2422150fa88a7006603ea7a65f3d.png 

gfs_apcpn_neus_10.png.57350429b11feabd623dfc48ed02b8d5.png

past72.png.2fcfe3c547deaf3179d7803e029a26b1.png

Nothing new with the GFS getting worse closer to the starting time:

gfs_apcpn_neus_8.png.e2f242215ded77c933b5a0197ad237c1.png

fv3p_apcpn_neus_8.png.44b3bce0a4624096973656080e04fcb3.png

gem_apcpn_neus_8.png.92e97a321e713524a00fea2ff0ae09ec.png

namconus_apcpn_neus_16.png.0c5f8bae48a463e5e1fd492d0fc5ea03.png

icon_apcpn_neus_16.png.57f24db5c582d0b9f020b2f306c1fbdd.png

past72.png.2fcfe3c547deaf3179d7803e029a26b1.png

 

fv3p_apcpn24_neus_7.png

Share this post


Link to post
Share on other sites

This is also on the sheet 2 tab.  What jumps out was that the models underforecast snow early, then overforecast snow late (unless in Quakertown where every forecast was an underforecast).  The NAM naturally came in last place (on average) but the OP EC came in second to last place.  It did not have a really great winter forecasting snow and a couple of the beyond 84hr forecasts were really in vapor lock territory.

Capture.JPG.0adc3e7bcf3acc7a2441a1a0cca3a4a6.JPG

NEW2017-8qpf.xlsx

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  

  • Recently Browsing   0 members

    No registered users viewing this page.

×