Question about NAM Weather Models

[Leominster]

I'm looking for some explanation or maybe a good site to answer my questions about weather models but what I 'm looking for is an explanation of the difference between the 00UTC, 06UTC, 12UTC, and 18UTC NAM weather models. I hope the answer to this question does not require a degree in meteorology...thanks in advance for any help...

[ORH_wxman]

The difference between the those models is the time they initialize. New data goes into the model at each of those times and the model is then run with that new data.

[tombo82685]

usually the 0z and 12z model runs hold the most weight and better data

[baroclinic_instability]

I'm looking for some explanation or maybe a good site to answer my questions about weather models but what I 'm looking for is an explanation of the difference between the 00UTC, 06UTC, 12UTC, and 18UTC NAM weather models. I hope the answer to this question does not require a degree in meteorology...thanks in advance for any help...

No difference at all except the time they run, that is the time they begin ingesting data and observations, perform the data assimilation, create the objective analysis, and begin integrating the model equations to develop the forecast, which, in the case of the operational NAM (North American Mesoscale) WRF-NMM, is 84 hours.

As Tombo82625 said, the 0 and 12Z runs also ingest the 0 and 12Z radiosondes, or balloon soundings, which directly sample the upper atmosphere.

[Leominster]

The difference between the those models is the time they initialize. New data goes into the model at each of those times and the model is then run with that new data.

So basically based on the time of day (taking into consideration the relationship between Eastern Standard Time (where I am) and Greenwich Time, one of the four models would be chosen. So right now I would want to look at the 00 UTC model which updated at 7:00 PM EST? Then the next update will be at 1:00 AM EST and after 1:00 AM I will want to check the 06 UTC? Am I on the right track here?

Thanks so much.

[Leominster]

No difference at all except the time they run, that is the time they begin ingesting data and observations, perform the data assimilation, create the objective analysis, and begin integrating the model equations to develop the forecast, which, in the case of the operational NAM (North American Mesoscale) WRF-NMM, is 84 hours.

As Tombo82625 said, the 0 and 12Z runs also ingest the 0 and 12Z radiosondes, or balloon soundings, which directly sample the upper atmosphere.

Thanks. I'm getting it...I think...slowly but surely...

[baroclinic_instability]

So basically based on the time of day (taking into consideration the relationship between Eastern Standard Time (where I am) and Greenwich Time, one of the four models would be chosen. So right now I would want to look at the 00 UTC model which updated at 7:00 PM EST? Then the next update will be at 1:00 AM EST and after 1:00 AM I will want to check the 06 UTC? Am I on the right track here?

Thanks so much.

You are correct. The NAM is meant to run early (partially why it is not the most reliable model), so you can expect the first panels to start coming in about one hour and 40 minutes after the initialization time. So, in the case of 0Z, start looking around 0140Z and data should be coming in for the first hours of the forecast.

[Leominster]

You are correct. The NAM is meant to run early (partially why it is not the most reliable model), so you can expect the first panels to start coming in about one hour and 40 minutes after the initialization time. So, in the case of 0Z, start looking around 0140Z and data should be coming in for the first hours of the forecast.

Is there a decent website that explains weather models for a weather three-caster like myself? (Not ready to call myself a forecaster)

[baroclinic_instability]

Is there a decent website that explains weather models for a weather three-caster like myself? (Not ready to call myself a forecaster)

Comet modules are great, can be a bit in-depth at times but they explain it well. Highly recommend if you are looking to learn. You need to sign up with an email, but it is quick and easy.

http://www.meted.ucar.edu/nwp/course/modules.php

[CNY Mike]

Southland made an excellent introductory writeup a few years back, I'll try finding it.

[Leominster]

Thanks for All the help everybody...I'll be three-casting the weather soon enough...

[baroclinic_instability]

Thanks for All the help everybody...I'll be three-casting the weather soon enough...

I am a meteorologist and even I don't know what a "three-caster" is. Can you explain?

[Leominster]

I am a meteorologist and even I don't know what a "three-caster" is. Can you explain?

I actually made that up...I'm far from qualified to be forecasting weather so what I do is "three-cast" it...maybe sounds more clever in my head than it actually is.

[baroclinic_instability]

I actually made that up...I'm far from qualified to be forecasting weather so what I do is "three-cast" it...maybe sounds more clever in my head than it actually is.

What kind of forecasting are you looking to do? What is your experience and background? I can help or offer suggestions if you have any.

[JoMo]

oh... 'forecast' = four cast........ three cast.

[am19psu]

usually the 0z and 12z model runs hold the most weight and better data

As Tombo82625 said, the 0 and 12Z runs also ingest the 0 and 12Z radiosondes, or balloon soundings, which directly sample the upper atmosphere.

I don't have the NAM error stats in front of me (anyone have a link to that data?), but the for the GFS, this is an old wives' tale. There is no difference in error scores amongst the run times for the GFS.

[baroclinic_instability]

I don't have the NAM error stats in front of me (anyone have a link to that data?), but the for the GFS, this is an old wives' tale. There is no difference in error scores amongst the run times for the GFS.

If you are right, then I would be pleasantly surprised.

[am19psu]

Over the last month:

Over the last eight years:

Satellite derived data assimilation has improved so much over the last 5-10 years that the radiosonde network is far less important nowadays. If wxbadger made it over here, he could explain it a lot better.

[baroclinic_instability]

Thanks for posting. I will have to analyze before I reply.

[Ed Lizard]

Why does the Canadian version of the NAM end at my house? Whole 'nother question, I suppose.

[Ender]

Over the last month:

Over the last eight years:

Satellite derived data assimilation has improved so much over the last 5-10 years that the radiosonde network is far less important nowadays. If wxbadger made it over here, he could explain it a lot better.

I believe that's basically the crux of the matter, at least at this point. I remember reading a paper of the initialization of the EC, but it referenced the rest of the models as well, satellite derived data vastly, vastly dwarfs the amount of data that's collected via land based and balloon/aircraft observations. I remember reading something on the order of millions upon millions of observations in excess of what could possibly be ingested into a model run. The paper was essentially on the issues surrounding model initialization compute time, which for the EMC back when the paper was written (circa 2007 or so) was nearly up to the amount of time required to actually run the model.

[Ender]

Why does the Canadian version of the NAM end at my house? Whole 'nother question, I suppose.

Because it's the REG and not the NAM...and because its domain is Canada, not North America. Otherwise, no idea.

[baroclinic_instability]

Why does the Canadian version of the NAM end at my house? Whole 'nother question, I suppose.

Yeah, it its domain just cuts out some of the SE and SW states in the U.S. Since it is only 48 hrs, they probably aren't too worried about hurricanes making their way into the model domain and progressing into Canada before that 48 hr run is up

[baroclinic_instability]

Over the last month:

Over the last eight years:

Satellite derived data assimilation has improved so much over the last 5-10 years that the radiosonde network is far less important nowadays. If wxbadger made it over here, he could explain it a lot better.

I believe that's basically the crux of the matter, at least at this point. I remember reading a paper of the initialization of the EC, but it referenced the rest of the models as well, satellite derived data vastly, vastly dwarfs the amount of data that's collected via land based and balloon/aircraft observations. I remember reading something on the order of millions upon millions of observations in excess of what could possibly be ingested into a model run. The paper was essentially on the issues surrounding model initialization compute time, which for the EMC back when the paper was written (circa 2007 or so) was nearly up to the amount of time required to actually run the model.

Thanks for the information. This begs the question then....when will they end soundings?

[Ed Lizard]

Thanks for the information. This begs the question then....when will they end soundings?

The twice a day soundings from a limited number of stations? The SPC will request 18Z soundings when severe weather is expected, there must be some benefit to that. Local NWS office complains sometimes about being between the CRP and LCH soundings in their discos, although they do get some ACARS data.

I get to see actual balloon soundings on the interwebs a couple hours after the fact, I sometimes like to compare the 0 hour NAM and GFS skew-Ts from the NIU make your own sounding site to the actual soundings when playing amateur severe weather prediction dude, to see which model best fits the soundings in nearby states, and ACARS isn't on the internet.