Ensemble Forecasts and Tornado Chasing

[David Williams]

I've been taking some Meted courses recently, and a few have been about ensemble forecasts. I have not really heard any chasers discussing ensemble forecasts in relation to storm chasing. Sure you hear about the GFS, NAM, HRRR, etc. But things never stray too much from there. This is interesting to me since the Meted courses can't get over how great the ensembles are and how much more accurate they are than anything else. They repeatedly mention the ensemble's superiority to other individual models and many of their example ensembles or case studies feature ensemble models from the early 2000s. So... if ensembles are so great, and they've been around so long, why aren't they the preferred source for forecasts? I don't hear storm chasers reference/discuss them? I personally wouldn't even know where to find them since I only use Unisys, COD, and Twisterdata. My personal guess is that ensembles are used for large scale/long-term forecasting; however, I'm not really sure.

David

 

[rdale]

After about 3 days you want to rely highly on ensemble GFS Euro etc. In the 24-72 hour period the SREF is good. I'd love to someday see a HRRR ensemble replace the NAM. The reasons you don't see chasers or amateur forecasts refer to them are multiple. 1) Extreme forecasts like a day 8 outbreak are rarely seen, 2) It requires a better knowledge of meteorology to interpret. 3) It takes more time to interpret. The list can go on The point is that ensembles proved us with the ability to judge uncertainty that is highly necessary in the forecast process.

 

[Stan Rose]

Yeah, what Rob said. Ensembles are highly probabilistic, and storm chasing is by it's nature pretty determinative. Your goal isn't the same as SPCs! You can't put yourself mid-way between two widely spaced targets, you have to make a decision. Decision makers prefer determinism. And of course the closer you are to an event, the deterministic models are generally going do better, so they will get more use. That said, I've been using the ensemble HRRR stuff and I find it useful. Also, when you look at successive runs of any model you're basically using an ensemble.

 

[Mark Ellinwood]

Ensembles are better for putting in a probabilistic forecast as to how the synoptic pattern will evolve in the medium to long range. Within 3-4 days, deterministic models are generally preferred for warm season storm forecasting and for honing in on the more mesoscale features. The lower resolution of most of the ensemble forecasts and the smoothing-out nature of them generally rule out trying to nail down chase locations and subtle features to watch on them.

Now, you can always look at multiple short-range models and kind of create your own "ensemble" forecast in your head based on that, or you can prefer one model over another if you have proper reasoning.

 

[Paul Knightley]

In the coming years, hi-res ensembles at short range will probably be one of the next major advancements in numerical meteorology. As processing power gets to the stage of 5 km resolution global models, you'll get to the stage of being able to have a fairly large set of ensembles at the short-range. Then you can start narrowing down initialisation areas, storm track, etc. I can see the stage where corridors of high risk at the storm-scale may be possible from ensemble output, several hours in advance of initiation. Of course, these will be probabilistic corridors, but all meteorology is probabilistic anyway.

 

[rdale]

Here's a fairly useful short-range ensemble usable for sevwx.

http://www.spc.noaa.gov/exper/sref/...76805474&mLON=-84.79512847290039&mTYP=roadmap

 

[David Williams]

Thanks for the responses everyone and for the link.

Sent from my SAMSUNG-SGH-I337 using Tapatalk