Jump to content
  • Member Statistics

    17,609
    Total Members
    7,904
    Most Online
    Vesuvius
    Newest Member
    Vesuvius
    Joined

Jan 4-5 Mixed Bag Discussion Thread


Bostonseminole
 Share

Recommended Posts

3 minutes ago, Ginx snewx said:

I am shocked Pivotal has these

ecmwf_full_2020010212_066_41.75--71.75.png

They're great, but only really useful if they have those sneaky non-manditory layers between 850 and 700mb that often have the warm nose.  Also, seems like they might be moving towards paid subscription at some point given all the fancy stuff they've added recently. 

Link to comment
Share on other sites

2 minutes ago, OSUmetstud said:

They're great, but only really useful if they have those sneaky non-manditory layers between 850 and 700mb that often have the warm nose.  Also, seems like they might be moving towards paid subscription at some point given all the fancy stuff they've added recently. 

According to the banner at the top of the site it seems like they are trying to keep it free if people can support them through Patreon.

Link to comment
Share on other sites

2 minutes ago, OSUmetstud said:

They're great, but only really useful if they have those sneaky non-manditory layers between 850 and 700mb that often have the warm nose.  Also, seems like they might be moving towards paid subscription at some point given all the fancy stuff they've added recently. 

The temp levels look more sparse than the wind barb ones. We've come a long way with ECMWF.

Link to comment
Share on other sites

40 minutes ago, Ginx snewx said:

 

IMG_20200102_123345.jpg

 

35 minutes ago, dendrite said:

The way it works is the euro will nail every gridpoint under a ridge, but will lose the the 15 points in the heart of the ULL. 

Nobody loves hard data more than me, so I think it’s awesome when people post these to speak to the “model X is better than model Y” weenie-ness.  However, there are at least a couple of points that people never really seem to drive home sufficiently for me:

1)    How does model performance with respect to 500 hPa correlation (which I’m assuming is analyzed globally for the 20° – 80° N region in these data sets?) actually relate to the model’s effectiveness as a guidance tool for sensible weather at the surface?  Beyond that of course, how much utility is there in that analysis for the model’s region-specific performance for a certain part of the globe?

2)    So you’ve already got the above factors in play, and then, how are you going to convince me that a couple of hundredths of a difference in correlation coefficient is even relevant?  The trend in model difference (and actually a bit of correlation improvement) is clear over the course of those years, and with so much data, the stats may be there to support a statistical difference between the anomaly correlation of the two models.  Regardless of that outcome though, it still doesn’t speak to the relevance of a 0.03 difference in correlation coefficient to the actual utility of the models.

Again, I love when people post these types of data, but I’d like to hear more about whether or not they actually speak to a realistic difference in model performance that matters to forecasting.

  • Like 1
Link to comment
Share on other sites

2 minutes ago, J.Spin said:

 

Nobody loves hard data more than me, so I think it’s awesome when people post these to speak to the “model X is better than model Y” weenie-ness.  However, there are at least a couple of points that people never really seem to drive home sufficiently for me:

1)    How does model performance with respect to 500 hPa correlation (which I’m assuming is analyzed globally for the 20° – 80° N region in these data sets?) actually relate to the model’s effectiveness as a guidance tool for sensible weather at the surface?  Beyond that of course, how much utility is there in that analysis for the model’s region-specific performance for a certain part of the globe?

2)    So you’ve already got the above factors in play, and then, how are you going to convince me that a couple of hundredths of a difference in correlation coefficient is even relevant?  The trend in model difference (and actually a bit of correlation improvement) is clear over the course of those years, and with so much data, the stats may be there to support a statistical difference between the anomaly correlation of the two models.  Regardless of that outcome though, it still doesn’t speak to the relevance of a 0.03 difference in correlation coefficient to the actual utility of the models.

Again, I love when people post these types of data, but I’d like to hear more about whether or not they actually speak to a realistic difference in model performance that matters to forecasting.

I can't speak to the specifics but I'm pretty sure the differences are statically significant, despite the AC difference being a small number.

And overall you're right, that 500 heights don't always translate to sensible weather.  QPF is not a good metric for models. It's better than it used to be, but still one of the worst variables to predict. Typically models are better at finding a window QPF will occur, but bad at finding the right amount of QPF.

Think of it this way. Whether you get a wound up coastal with 1" QPF or a fropa with 0.01" in snow showers the model still had the trough for that period. High AC score, low QPF score.

Link to comment
Share on other sites

9 minutes ago, OceanStWx said:

I can't speak to the specifics but I'm pretty sure the differences are statically significant, despite the AC difference being a small number.

And overall you're right, that 500 heights don't always translate to sensible weather.  QPF is not a good metric for models. It's better than it used to be, but still one of the worst variables to predict. Typically models are better at finding a window QPF will occur, but bad at finding the right amount of QPF.

Think of it this way. Whether you get a wound up coastal with 1" QPF or a fropa with 0.01" in snow showers the model still had the trough for that period. High AC score, low QPF score.

Brier scores on the Euro greatly outpace the others. What JSPIN is looking for is sensible wx scores. There is a lot of data online available for review. 

Link to comment
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
 Share

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...