NAMKF and MesoNAM

[Bob Hartig]

What the heck is the NAMKF? Seems to me that this term has only recently cropped up in forecast discussions. I don't recall seeing it last year. How does the NAMKF differ from the standard NAM?

Ditto the MesoNAM, another term I've encountered just lately.

Besides explaining these terms, would someone in the know mind providing a link or two to some good sources.

Gracias!

 

[Scott Weberpal]

Its the NAM Kain Fritsch. Google Kain Fritsch and you will get a wealth of information about the model. It has been around for many years, I remember using it as far back as 2001 or 2002.

http://www.mmm.ucar.edu/mm5/Kain_02.pdf

 

[Jeff Snyder]

Back in the day, the Eta model was run 4 times per day, with the 3z and 15/18z runs not going out as far in time as the 00z and 12z runs (e.g. only out to 33 hrs instead of 48-60-72 hrs). The off-sounding time runs (i.e. not 00z and 12z) were later modified to run at 6z and 18z out to 48 hours sometime around the year 2000. Regardless, the WRF-NMM, which replaced the Eta model in the timeslot that we now refer to as the "NAM", currently runs 4 times per day using the same basic model parameters (i.e. out to 84 hours, etc). As such, the distinction between "regular" NAM and meso-NAM probably doesn't need to be made, unless I'm missing something that the 18z and 6z runs don't use/have that the 00z and 12z runs use/have. You'll more commonly just see terms like "18z NAM" instead...

The NAM-KF just refers to the version of the NAM (again, the NAM slot currently is filled by the WRF-NMM model, which replaced the Eta model a few yrs ago) that uses the Kain-Fritsch convective parameter schemes (both for shallow and deep, moist convection). NSSL used to run an Eta-KF model, if I recall correctly, but I'm not sure where you can access any current NAM-KF output. I'm sure someone in the country runs such a model, as I think I've read SPC discussions referring to it. The current NAM graphics that you see online from NCEP or sites that obtain their data from the "officially-run" NAM model use the BMJ convective parameterization scheme. You can learn more about different convective parameterization schemes used by current models (the RUC, GFS, and NAM all use different c.p. schemes) by going to the COMET page on convective weather -- http://www.meted.ucar.edu/topics_convective.php -- and clicking on the powerpoint presentation link under "Convective Weather Refresher" on the left-hand side of the page. I try to make it a point to go through that PPT every year to refresh myself on the schemes, since they each have very important pros and cons that forecasters should keep in mind when examining model data as they pertain to deep moist convection (i.e. thunderstorms).

On a somewhat related note, this is just a reminder that the Eta model is not run anymore. Some folks have used the terms "Eta" or "MesoEta" in forecast discussions on here even this yaer; I think they are actually referring to the NAM (i.e. WRF-NMM) model. The WRF-NMM replaced the Eta a few yrs back, though NCEP still ran the Eta model to provide Eta-MOS output up until recently (though no forecast graphics were produced). Even that has ended now, however. Just mentioning a minor pet peeve.

Edit: I'm not 100% sure on the historical (pre-2000) Eta model information, so feel free to correct me!

 

[Bob Hartig]

Thanks, guys, for the explanation, the links, and for helping to demystify these different versions of NAM that have seemed to be drifting around. I've been at a point where whenever someone referred to the NAM, I'd think, "Which NAM?" But if I understand this correctly, there is in fact only one NAM extant, the WRF-NMM. The things that make the difference are the CPS (i.e. KF vs BMJ) and the resolution, true? (BTW, I've learned a new term and gained a new acronym: convective parameterization scheme (CPS). A man can't have too many acronyms in his arsenal.)

Jeff, thanks in particular for going into some detail in your response. The Powerpoint on CPSs sounds like it contains valuable information that may clear up a few more questions for me.

As always, I appreciate the knowledge that is shared so freely on this forum.

 

[Jeff Snyder]

Bob,

AFAIK, the only time that "NAM" is used is to refer to the 12km WRF-NMM (with BMJ convective parameterization) that's run at NCEP at 00z, 6z, 12z, and 18z. Note that there are many other versions of the WRF that are run by other entities, and you could even download the WRF code yourself and run it on your own computer! For example, NCAR runs a 3 km, explicit-convection (i.e. no convective parameterization) WRF (-ARW), NCEP runs a 4 km, explicit-convection WRF-NMM, and NSSL runs a 4 km WRF-NMM. In addition, for VORTEX 2, CAPS was running a WRF variant at 1 km, as well as a 20-member ensemble of 10 WRF-ARW, 8 WRF-NMM, and 2 ARPS variants.

 

[Bob Hartig]

Holy crow. Well, you see, right there life just got a whole lot simpler.

I am 53 years old. I have been chasing storms, or at least doing my best to earn the merit badge, for thirteen years now. But I continue to be reminded on a regular basis what an innocent I still am.

I think you've just handed me my next area of inquiry, Jeff. But that's one of the things that make this pursuit so rewarding. It keeps my brain from fossilizing. Thanks to you and others like you on Stormtrack, and with due respect for my own research as well, I've got a far better, working knowledge of forecasting today than I did even six months ago. But man, there's still so much to learn! It can seem daunting at times. I still don't really understand the WRF, beyond the fact that it's a code which processes the data for different forecast models, the NAM and RUC being prime examples. In fact, when I see the term "WRF" pop up by itself in a discussion, I automatically assume it's NAM-based (i.e. WRF-NMM). But I really don't understand why the WRF is used, or why the NAM can't just be the NAM.

Until recently, I've used forecast models without giving much thought to their variations. The GFS has been the GFS, the RUC has been the RUC, and so forth. But I've noticed enough nuances just within the NAM that I'm probably at a point where I need to learn more so I can use the different models more effectively.

"The journey of learning begins with an awareness of where one is ignorant, and a determination to do something about it."

I'll start with the Powerpoint you've recommended, Jeff, then check out the other links. Again, thanks!

 

[Jeff Snyder]

Bob,

No problem! This is always a learning experience, and, like many other areas of life, the more you learn, the more questions you have. This field can be very broad, and I am very far from an expert in NWP (certainly there are others on this board that are better versed in NWP than I). Just to clarify, however -- the WRF (Weather Research and Forecasting) is a computer model itself, a massive set of programming code. IIRC, there are two primary physics cores that can be used in the WRF - the NMM core and the ARW core. When someone talks about the WRF, however, they are talking about the actual program, and, typically, the output from the model. This is similar to the RUC and GFS (and MM5, RAMS, ARPS, etc.), which are other computer models with some commonalities and many differences. The Eta model was popular back in the 90s and early 2000s, but it was replaced by the WRF-NMM at NCEP.

The "NAM" (North American Mesoscale model) isn't really a model itself; rather, it's a generic name for the model that occupies that timeslot on the NCEP computing schedule and serves a particular purpose (grid centered on the CONUS, high enough resolution to capture to mesoscale details, and so forth). The Eta model USED to be used in the NAM timeslot, but now the 12km WRF-NMM is used as the NAM. This naming scheme can cause some confusion. For example, in 2006 and 2007, you could have talked about the NAM model and been just fine, even though the actual computer model changed from the Eta to the WRF-NMM. I guess it's kind of similar to the title of "President" -- you can always refer to the "President", but actual actual, physical person to whom you are referring may change from time to time (i.e. at election time). In this case, the title "President" is similar to "NAM", and the actual person who is the President (e.g. Bush, Obama, etc) is the actual model (e.g. Eta or WRF-NMM). It just so happens that, at this time, the WRF-NMM is the NAM. Graphics for the NAM that you see on RAP/NCAR, CoD, TwisterData, HOOT, and other model sites are from the 12 km WRF-NMM (i.e. NAM) run by NCEP. When talking to other chasers or mets, it's easiest to refer to this just as the "NAM" since we'd all know exactly what model you are talking about. For instance, if you say "model output from this morning's WRF run...", people may be confused by which exact WRF you are talking about -- is it the 12 km NCEP WRF (i.e. the NAM), the 3 km NCAR WRF, the 4 km EMC WRF, or some other locally-run WRF? So, we just say "NAM" since it avoid some confusion as a result of the fact that there is only one NAM.

 

[Bob Hartig]

That is an excellent explanation, Jeff! Now I get it. At least, I finally get the NAM, and that is saying a lot. I still have plenty of questions re the WRF, but you've just cleared up a significant bit of confusion for me. A bit of an epiphany, in fact. I love it when a new light switches on!

 

[Andrew Revering]

I love modeling... I run my own WRF-ARW here at home over my small domain of Minnesota when I'm considering chasing.... a few notes that I'll add to Jeff's excellent posts..

- The term WRF isn't nearly as important as its core... ARW or NMM. "NAM" (as Jeff explained) is using WRF-NMM... however its important to note that the ARW core is MUCH more accurate, but it is also MUCH slower to run. This is why operationally they use the NMM core because its easier to make the domain large and get it done quickly. However, you're better off to look at an ARW variant if you want to increase your odds at a correct forecast.

- The WRF EMS can be packaged with a lot of different variables, and almost nobody identifies the combinations of microphysics, longwave radiation, shortwave radiation, cumulus scheme (Kain Fritsch, none or other), land surface model, turbulence parameterization, type of advection of moisture, how many vertical levels, the time step, domain size and even the resolution that the model is set up to use. This is what makes the model... and sometimes even the core is not identified. Without knowing this information, people might try to compare two WRF runs and have two completely different results and not know why... well the simple answer is they are two different models with the same name!

- The lower the resolution, the better the model can 'naturally' handle convection, so you can actually 'turn off' convective parameterization. If the model is down under 10km (some would argue much less than that), your model will handle thunderstorms better without convective parameterization because CP is essentially stunting the growth of convection and sizing it to fit within the large grid spacing of say a 30km model for example. A storm is not that large, so you need to try to mathematically grow the storm on a more natural scale. If your grid spacing is closer to that of a real storm (say 2-5km) the storm will grow naturally within the grid spacing so you can shut off CP. With that said, if you have to use CP; Kain Fritsch I believe is pretty well accepted as one of the best to use.

- Last note... in my experimenting I'm finding that if you make your grid spacing too small and the domain has to suffer because of it, the model will be crappy. So high resolution isn't always everything. Look at GFS. GFS is a good model, despite having huge grid spacing (0.5 degree, or about 35-40km). It does so well (besides its physics) because of its large domain (whole world). So if you find a 1km WRF out there, but its running a domain the size of a state or smaller, don't think it's going to do well.

There's also nesting of the domain but this is probably way more than what you want to talk about.

 

[Andrew Revering]

Oh I also wanted to add that although WRF-NMM is what we call 'NAM', they still produce several flavors at different resolutions.

I believe you can use 12km, 30km and even 40km for graphics on a web site. This information is also not shown on web sites, so you may have two NAMs at different resolutions. The physics combinations (I believe) are identical so the results typically are not varied greatly other than small details of extrema, but regardless its good to know that within each model you can find resolution variations.... that nobody explicitly tell you what you're looking at other than "NAM"

 

[Geoff Manikin]

just to clarify, NCEP is not producing different versions
of the NAM at 12, 30, and 40 km. There is one 12 km run, and that data gets interpolated onto the many output grids we generate. So, if you see a 12 km NAM and a 40 km NAM, these are not from separate model runs. Andrew, perhaps this is what you meant, but I wanted to make it clear since there is always a lot of confusion about this issue. Heck, even some NWS personnel think we run a separate 40 km NAM and separate 20 and 40 km RUC.

Anything under 12 km, such as the Hi-Res window runs or the 4 km convection run is a separate system. You can see the hi-res window under 'HRW' at
http://www.nco.ncep.noaa.gov/pmb/nwprod/analysis/
and the 4 km output is at http://www.emc.ncep.noaa.gov/mmb/mpyle/cent4km/conus/00/ (change the 00 to 12 for the 12z run)

The only exception to my "under 12" rule is that we generate 'downscaled' 5 km NAM output for CONUS (and 6 for Alaska and 2.5 for HI/PR), and that starts with the operational 12 km run:
http://www.emc.ncep.noaa.gov/mmb/gmanikin/smartinit/

Oh I also wanted to add that although WRF-NMM is what we call 'NAM', they still produce several flavors at different resolutions.

I believe you can use 12km, 30km and even 40km for graphics on a web site. This information is also not shown on web sites, so you may have two NAMs at different resolutions. The physics combinations (I believe) are identical so the results typically are not varied greatly other than small details of extrema, but regardless its good to know that within each model you can find resolution variations.... that nobody explicitly tell you what you're looking at other than "NAM"

 

[Bob Hartig]

Let me see how well I can piece this together as a layman. Here is what Jeff wrote a couple posts back:

Bob,

AFAIK, the only time that "NAM" is used is to refer to the 12km WRF-NMM (with BMJ convective parameterization) that's run at NCEP at 00z, 6z, 12z, and 18z.

If I understand this correctly, there's technically no such thing as NAM-KF, since the NAM specifically uses the BMJ parameterization. The NAM-KF would correctly be called the WRF-NMM-KF. Do I have that much right?

Andy and Geoff, thanks for sharing your insights. So then, if I've got this right, the 30km and 40km NAMs are just coarser resolutions of a single run, the 12km WRF-NMM that is the "true" NAM--correct? But then there's also a 5km NAM, which is another derivation of the 12km? And everything else below 12km is not really the NAM, but some other form of WRF-NMM?

Gosh, that all seems simple enough. I can't imagine how anyone would get confused.

 

[Jeff Snyder]

Bob,

AFAIK, you have it right! I don't think there is a NAM-KF since the NAM = NCEP's/EMC's 12 km WRF-NMM with BMJ convective parameterization. As such, any reference to the KF conv. param. would be more correctly stated as the WRF-KF (-NMM or -ARW). It's my understanding that the NAM's "lower resolution" grids (i.e. 20 km, 40 km, 80 km, etc) are all based on the 12 km NAM run, and this was the gist of Geoff's post. I think Geoff works for NCEP and had a hand in the development of the WRF, so I'll bow to Geoff to make any corrections if I've misunderstood anything

 

[Geoff Manikin]

Bob, I think you have it. Just to clarify, I probably
shouldn't have mentioned it, but the 5 km NAM output
is referred to as downscaled or smartinit NAM. Yes,
these 5 km grids are generated from the operational 12
km NAM, as are the 20, 32, and 40 km grids. There is
a nice summary of all of the various output grids
from the 12 km NAM at
http://www.emc.ncep.noaa.gov/mmb/namgrids/
You can see that we take the full grid and provide
various subsections at various resolutions to assist
users who can't download full-domain, 12 km grids
due to bandwidth issues.

Everything under 12 km (other than the smartinit
grids) is indeed a separate model run.

If I understand this correctly, there's technically no such thing as NAM-KF, since the NAM specifically uses the BMJ parameterization. The NAM-KF would correctly be called the WRF-NMM-KF. Do I have that much right?

Andy and Geoff, thanks for sharing your insights. So then, if I've got this right, the 30km and 40km NAMs are just coarser resolutions of a single run, the 12km WRF-NMM that is the "true" NAM--correct? But then there's also a 5km NAM, which is another derivation of the 12km? And everything else below 12km is not really the NAM, but some other form of WRF-NMM?

Gosh, that all seems simple enough. I can't imagine how anyone would get confused.

 

[Geoff Manikin]

Jeff, I'm not sure where the NAM-KF terminology
comes from. Perhaps SPC is running a local version
or is getting one from NCAR. It's also possible
that they're looking at the control Kain-Fritsch Eta
member of the SREF. We're not running a true
"NAM-KF" at NCEP.

Bob,

AFAIK, you have it right! I don't think there is a NAM-KF since the NAM = NCEP's/EMC's 12 km WRF-NMM with BMJ convective parameterization. As such, any reference to the KF conv. param. would be more correctly stated as the WRF-KF (-NMM or -ARW). It's my understanding that the NAM's "lower resolution" grids (i.e. 20 km, 40 km, 80 km, etc) are all based on the 12 km NAM run, and this was the gist of Geoff's post. I think Geoff works for NCEP and had a hand in the development of the WRF, so I'll bow to Geoff to make any corrections if I've misunderstood anything