What's the point of the NSSL WRF?

[Hannah.Taylor]

I will not name anyone by name (if they choose to come forward so be it), but I know of several chasers who have used the WRF

http://www.nssl.noaa.gov/wrf/

(there may be another site for the WRF too, just can't seem to recall it).

and have been mostly spot on when big stuff is going to happen and were in position when big storms fired off and produced violent and destructive tornadoes.

Joplin 2011, Guthrie 2011, El Reno 2011, and Shawnee, OK 5-19-2013 and Moore, OK 5-20-2013.


Are there any reliable statistics from past projections with the WRF, and if they're fairly accurate...why are we not using these projections to warn large populated areas of impending devastation well ahead of time before the event???

15 minutes was the warning time yesterday for the Moore, OK tornado...but if we have this type of technology available to project something of this nature....why are we not doing more?

Forgive me for being asinine (and ignorant), but I'm a bit emotional over yesterday's tragic tornado hitting Moore, OK...it's just frustrating to think that 15 minutes is all the better we can do considering all the technology and advancements we have made in the last 100 years.

That's 15 minutes for those in the vicinity of a smartphone, tornado siren, tv or radio or computer. For those caught in between...they might have had 2-5 minutes warning if they were lucky. That's not good enough.

If the WRF stats show these projections not to be that reliable, then I understand, but I haven't heard or seen this discussion come up, so I'm asking to find out.

 

[rdale]

I guess I'm unsure of the premise of what you're saying? The NSSL WRF is not the only reason that people knew central Oklahoma could see big storms. Plenty of models had that in their forecast. As a result, the SPC and many meteorologists spent a lot of time hours (and days) in advance talking about the threat. The NSSL WRF did not predict the Moore tornado 8 hours out. Because of the warning system, the warning was issued 15-30 minutes in advance of the hardest hit areas. That is the limits of what we can do as a science, and for most users should be ample.

 

[Mike Moore]

It doesn't matter at all if the warning time is 60 minutes if people ignore the warnings. I think looking at warning fatigue and eliminating unnecessary warnings is where science has to move next to further improve public cooperation. If you tone the sirens 50 times and there is no tornado what are the odds anyone is listening when 51 hits and it's real? I'm not talking about this incident in particular, more in general terms.

I admit though it does drive me MAD when I see people on tv (as I did last night) saying there was no warning. How can that be when I knew about the possibility of it happening DAYS in advance? I knew about the more immediate threat HOURS in advance. I watched that tornado in REAL TIME from Florida on the internet. It's so hard for me to understand how anyone can be surprised by these things. It's frustrating given the technology we have.

Then again some of the people killed were sheltering in place, so they obviously knew the warnings were issued and they were doing the right thing. It was a be underground to survive type of storm.

Anyway I don't think any one model can predict these things with that level of detail.

 

[Hannah.Taylor]

I guess I'm unsure of the premise of what you're saying? The NSSL WRF is not the only reason that people knew central Oklahoma could see big storms. Plenty of models had that in their forecast. As a result, the SPC and many meteorologists spent a lot of time hours (and days) in advance talking about the threat. The NSSL WRF did not predict the Moore tornado 8 hours out. Because of the warning system, the warning was issued 15-30 minutes in advance of the hardest hit areas. That is the limits of what we can do as a science, and for most users should be ample.

I know the WRF isn't the only model, but it seems (maybe it's just a jaded perception) that every time I hear of a big storm that's supposed to hit a city and then it happens, I hear someone referring to the WRF and saying the WRF predicted it and hit it right on.

Yeah, they're talking about the "threat", the "potential", and people are just ignoring it, or the information is being watered down by the local media as a likelihood of severe weather. I randomly asked a few friends if they knew what the Storm Prediction Center was and they were clueless. Local meteorologists always make their own forecasts and I rarely ever hear of them mentioning SPC/NWS models unless something is currently in progress.

Any reason they can't do what they do with a Hurricane and say "the models are showing a big tornado hitting around location x at a given time? Or would that just induce panic? And how is it different than that of a Hurricane Warning?

I tend to agree with the discussion in the thread about slight/moderate high risk. The terminology might need to be changed.

A local meterologist says potential for severe weather, the average person/citizen thinks, oh a severe thunderstorm, so some hail and wind and rain and lot's of lightning it'll pass and that'll be the end of it.

 

[rdale]

I know the WRF isn't the only model, but it seems (maybe it's just a jaded perception) that every time I hear of a big storm that's supposed to hit a city and then it happens, I hear someone referring to the WRF and saying the WRF predicted it and hit it right on.

That's just perception. Same as "back when I was a kid we always got bigger snowstorms..." You aren't hearing about the many many more events that the WRF said would happen and didn't.

I randomly asked a few friends if they knew what the Storm Prediction Center was and they were clueless.

That's not a big deal. They shouldn't care about how the NWS hierarchy breaks down.

Local meteorologists always make their own forecasts and I rarely ever hear of them mentioning SPC/NWS models unless something is currently in progress.

Well the SPC is composed of meteorologists too, they aren't super-human forces Every meteorologist at least looks at SPC products. But it doesn't mean they should use them verbatim (for example, they have a good risk of severe weather in my area today and tomorrow and I barely see support for storms, so am not mentioning it at all.)

Any reason they can't do what they do with a Hurricane and say "the models are showing a big tornado hitting around location x at a given time?

I've got a great reason!

The models cannot do that. The models can suggest where areas of concern might be, which is why we say "there is a tornado threat in xxx today."

Or would that just induce panic?

If we could predict tornado locations 3 days in advance, that would be great!

And how is it different than that of a Hurricane Warning?

Hurricanes can be generally forecast 24-96 hours out. Tornadoes cannot.

 

[Hannah.Taylor]

fair enough rdale, it was on my mind wanted to ask to get other opinions. thanks.

 

[Jeff Duda]

I know the WRF isn't the only model, but it seems (maybe it's just a jaded perception) that every time I hear of a big storm that's supposed to hit a city and then it happens, I hear someone referring to the WRF and saying the WRF predicted it and hit it right on.

Sounds like you could use some information on what this term "WRF" means.

The Weather Research and Forecasting (WRF) model is a community model developed mostly by researchers who work at or with people at NCAR. The source code is freely available to anyone with the technology and capacity to compile and run it. The WRF is known for being widely applicable to different weather scenarios including global/climate scale modeling, regional/mesoscale modeling and even storm-scale and smaller scale modeling. And indeed people use it for all of those purposes and more. Since anyone can run it, many entities have their own versions of the WRF running, some in a semi-operational mode. The 4 km NSSL WRF to which you are referring is just one such model. Others can be found at this link: http://wrf-model.org/plots/wrfrealtime.php. Even that list doesn't include all of the versions out there that folks are running.

Also you should know that until either 2011 or 2012 (I'm not sure of the specific time the change was implemented), the NAM was actually a WRF-NMM forecast. Thus a lot of older forecast graphics sites like the RAP site (weather.rap.ucar.edu/forecast) and the COD site (weather.cod.edu) continued to label their NAM forecast graphics as WRF-NMM for a long time after NCEP abandoned the WRF as the model used in the NAM. I believe all of these sites have finally updated their links, but if anyone isn't paying attention to the changes on some websites, I can see why some people would still say things as general as "the WRF nailed the forecast".

Now that I remember this, also the Rapid Refresh and the HRRR use the WRF, so the WRF is everywhere.

The WRF is not necessarily the king of models. It's not unarguably the best model. It can be just as wrong as other models. However, its use as a storm scale model (with grid spacing of 4 km) is touted as a big accomplishment. Many people are starting to run the WRF at 4 km for severe storm forecasting. However, severe weather phenomena are still too small to be fully resolved on a 4 km grid, so forecasting tornadoes, hail, wind etc. is not making leaps and bounds in improvement. It's just a steady climb. You should never expect a 4 km model to accurately forecast a tornado...even a big one. The model is simply incapable of doing so. What it can give a hint of is storm mode (i.e., linear vs. cellular) and the potential for rotating storms (via a product called updraft helicity). It can also resolve small scale areas (maybe a county or so in size) where the atmospheric conditions may be incrementally more favorable for a particular type of severe weather.

 

[Hannah.Taylor]

Sounds like you could use some information on what this term "WRF" means.

The Weather Research and Forecasting (WRF) model is a community model developed mostly by researchers who work at or with people at NCAR. The source code is freely available to anyone with the technology and capacity to compile and run it. The WRF is known for being widely applicable to different weather scenarios including global/climate scale modeling, regional/mesoscale modeling and even storm-scale and smaller scale modeling. And indeed people use it for all of those purposes and more. Since anyone can run it, many entities have their own versions of the WRF running, some in a semi-operational mode. The 4 km NSSL WRF to which you are referring is just one such model. Others can be found at this link: http://wrf-model.org/plots/wrfrealtime.php. Even that list doesn't include all of the versions out there that folks are running.

Also you should know that until either 2011 or 2012 (I'm not sure of the specific time the change was implemented), the NAM was actually a WRF-NMM forecast. Thus a lot of older forecast graphics sites like the RAP site (weather.rap.ucar.edu/forecast) and the COD site (weather.cod.edu) continued to label their NAM forecast graphics as WRF-NMM for a long time after NCEP abandoned the WRF as the model used in the NAM. I believe all of these sites have finally updated their links, but if anyone isn't paying attention to the changes on some websites, I can see why some people would still say things as general as "the WRF nailed the forecast".

Now that I remember this, also the Rapid Refresh and the HRRR use the WRF, so the WRF is everywhere.

The WRF is not necessarily the king of models. It's not unarguably the best model. It can be just as wrong as other models. However, its use as a storm scale model (with grid spacing of 4 km) is touted as a big accomplishment. Many people are starting to run the WRF at 4 km for severe storm forecasting. However, severe weather phenomena are still too small to be fully resolved on a 4 km grid, so forecasting tornadoes, hail, wind etc. is not making leaps and bounds in improvement. It's just a steady climb. You should never expect a 4 km model to accurately forecast a tornado...even a big one. The model is simply incapable of doing so. What it can give a hint of is storm mode (i.e., linear vs. cellular) and the potential for rotating storms (via a product called updraft helicity). It can also resolve small scale areas (maybe a county or so in size) where the atmospheric conditions may be incrementally more favorable for a particular type of severe weather.

thanks Jeff. That definitely cleared up some misperceptions I had about the WRF and what it is and how it works. I had some sort of perception it was indeed able to more accurately predict tornadic activity in a given area well in advance. I'm not sure if that was a perception created by hearing about it, or just my brain discombobulating things, but I appreciate your insight.