Tornado Probability vs Chase Success

[Skip Talbot]

Many chasers get excited when the SPC day 1 outlook rolls out with higher tornado probabilties (10% hatched, 15%, 30%...) or maybe disappointed when they are low (5%, 2%, <2%), even if they're not really used for chasing decisions. However, I've hypothesized that your odds of documenting a photogenic tornado are only marginally better on days with higher probabilities than days with lower probabilities. My reasoning for this is anecdotal, but I assume many chasers could relate: my best chase days have been slight risk days, my top two career chase days were both slights (5/22/10 and 6/17/10). I've also taken a number of bad busts on high risk days. Then I remember all of the gorgeous tornadoes that occur outside the risk areas or on low risk days. Just this year: the North Dakota oil camp tornado (less than 2%), South Dakota the following day (2%), and a nice landspout in Illinois a couple days ago (less than 2%). Campo is probably the most famous example from any year (2%). Whereas the high risks this year have produced heavily rain wrapped, low contrast, and terrain obstructed tornadoes, that are difficult to see or confirm except by the damage afterwards.

I've had mets challenge this reasoning. That I'm just selectively remembering events, and making sweeping generalizations. This is definitely true, but I was really curious to see how the numbers actually play out. I tallied up all the stats from my chase logs and compared them against the SPC day 1 tornado outlook. The graphic below compares "photogenic" tornado days against the SPC 1 day categorical tornado outlook. It's important to note that I'm considering a chase success as a day that featured a "photogenic" tornado. While I count all tornado intercepts, there isn't as much satisfaction or share worthy videos and photos from events that feature large tornadoes that are heavily rain wrapped or very low contrast, brief dust whirl tornadoes (bird farts), tornadoes obstructed by terrain, too distant, or nocturnal tornadoes without adequate illumination. Several of my chases, including many high risk chases, featured these types of tornadoes, and I discounted them because I have little or nothing to show for them. From a forecasting standpoint, these tornado events may have verified quite well. However, from a storm chasing standpoint, they are less enjoyable or productive events to me. What's photogenic or a successful tornado intercept is subjective, and entirely up to you, however. So here are my results:

I get a photogenic tornado on:
13% of my slight risk chases
19% of my moderate risk chases
23% of my high risk chases

The difference between the categories is higher than I expected. Is it more than the "marginal" claim I made originally? I think so, seeing that I'm twice as likely to get a photogenic tornado on a high risk than I am a slight risk. Is the probability difference proportional though? I don't think so, since the tornado probabilities are 3 to 6 (or higher) on a high risk day. Another important difference is that I've chased a large number of slight risk days for structure, entirely not expecting a tornado. A few other caveats:

* "photogenic" is subjective
* the time of the day 1 outlook referenced varies, and in many cases my chase decision and target was made before it was issued
* I don't always chase for tornadoes, so structure chases will bias these stats toward the moderate/high risks
* I rarely base my decision to chase on the SPC categorical risk level
* I chased less slight risks in my earlier years when I was more likely to bust on any day, including the high risks, skewing the results toward the slight risks

 

[Mike Smith]

Skip, this is a really interesting post. Thanks for writing this up.

 

[JamesCaruso]

Agreed, very interesting contribution! Would be interesting to see the same analysis including photogenic tornados AND/OR photogenic structure!

I agree high risks often disappoint - sometimes too many storms go up all over the place, interfering with each other. Or particularly high moisture results in lots of low clouds and no visibility.

Skip, I am envious of your skills in the visual presentation of quantitative data, wish I could do that in my financial management consulting business!

Jim Caruso

 

[Shane Adams]

In my mind, the only difference between 2%, 5%, 10%, etc is how many storms are expected. For me, it's far easier to forecast (and chase) slight/mod risk days, because I'm trying to find A storm. I look for the area where I think A or THE storm will be, and I go there. On outbreak type setups or even higher end mod days, there's usually several storms (or even target areas) to choose from, and IMO this creates a larger margin for error if 2-3-4 go up at once and you have to choose.

A few years ago somebody did a career-spanning list of their tornado days VS SPC % risk. I did one myself out of curiosity, and found that most of my tornado days were 5%ers.

 

[Brandon Clement]

I'd be willing to bet the location of the chaser has a lot to do with it as well. For us guys in the SE, it is really hard to justify driving 12+ hours to chase a 5% day unless we are already in the plains. No way we are making the drive on a 2% or see text day. In Mississippi a 2% tornado day seems far less likely to produce than an equivalent day in the central to northern plains because landspout type tornadoes, upslope stuff or cold core setups might as well be a unicorn in Mississippi. Plus the terrain here makes spotting a brief tornado more of a stroke of luck than skill. Slow moving tornadoes are also very rare here. In the south 35 mph storm motions and we get excited. Down here, the cap tends to be weaker and getting isolated tornadic supercells during the day is kind of rare except on the big events. We almost never have to question whether the cap is going to break at lunch time because we are already chasing. Which brings me to the next point, many days the parameters justify a much higher category of severe but the cap concerns limit potential and SPC seems to error on the side of predictability in the plains. In the south SPC seems to play up even the slightest events a bit more because storms here are fast moving, higher population densities, hard to spot and they seem to error on the side of caution.

When I'm in the plains for more than a day or two then I really enjoy the slight risk 5% days, particularly when those 5% days are in Ks or the caprock area along the dryline or a boundary. Unless bulk shear is in excess of 50 knots in the south, when a storm blows up on a boundary it tends to go linear almost instantly.

 

[Stan Rose]

Yeah, I definitely think location factors into all of this. You rarely see moderate or higher risks in eastern WY or CO, or W SD, etc, yet you can get plenty of photo ops here with good visibility and pretty storms.
Can't really compare that to chasing in the hills and trees in certain parts of the plains or southeast.

I get Shane's point, but the flip side is that there are some mod-high risk days where the parameters are so huge where all you have to do is jump on the first storm you find and its almost a gimme.

Well, ive chased 7 days so far this month and im 0 for 7, despite being on tor-warned storms 6 of the 7 days. And all days were slights. Your chart has me worried...wish I could be in NE on tuesday

 

[calvinkaskey]

Look up John Calvin tornado on youtube. I was blessed with maybe the most photogenic tornado along the entire east coast this year. I've also wondering if visual cues are just as accurate as radar etc. The storm I chased was one I was watching for a long time, grew very rapidly, looked to rotate, developed big bulging belts of convection ringing the entire storm and then bang, I get a 15 minute 4 mile long tornado. It was highly visible. It would have been better if there was more farmland and larger fields in the area though.

 

[Andy Wehrle]

http://en.wikipedia.org/wiki/Mammatus_cloud

The clouds themselves don't have any impact on the chance of tornadoes, but they are more commonly seen with severe thunderstorms than weaker ones. I've seen them on the back side of linear severe thunderstorms where the tornado threat was very low. However, sometimes very dramatic mammatus have occurred under the anvils of supercells that produced violent tornadoes, thus they have become associated with tornadoes in the popular imagination.

 

[calvinkaskey]

Are they directly associated with large hail? You usually need a strong updraft for a tornado right?

 

[jmperretta]

You also need a twisting component in that updraft to generate anything that could be slightly construed as a tornado. One of the things I look are the mesoscale analysis pages on the SPC website (http://www.spc.noaa.gov/sfctest/). For those who don't know, you can click in the area of interest to zoom into the storm cells in the area. Then, to see if there's any "twist" in the air, click on the radar echo and check out the helicity index. It is my understanding that the higher the helicity index number is, the more twist there is in the air.

http://forecast.weather.gov/glossary.php?word=helicity

Please note that helicity indexes go up quite a bit when an approaching cold front is headed for your area. A few months ago, where I live in Florida (Southwest Florida), we had a cold front come through with a line of showers and thunderstorms and I saw helicity indexes in the 300 range. This did trigger a couple of tornado watches about 2 hours north of me, but once the front moved through, those numbers dropped like a rock.

 

[James Simonds]

First off, well done Skip.

And to answer John's prompt, visual cues can be misleading as mentioned, but I opine that by physically looking at a storm, I tell more about its state than by a radar image, save for a few characteristics like size of a hail core (which is why I use my eyes and radar simultaneously). Hail size and frequency of lightning are good indicators of the strength of a cell, as is the intensity of inflow winds among many others.

As for forecasting, I've come to view it like this: I would rather be pessimistic and be pleasantly surprised, than be optimistic and be dissapointed. I call this the Michigan viewpoint, and I don't recommend others take it on, as Stormtrack would become a pretty boring place

 

[Rob H]

In my mind, the only difference between 2%, 5%, 10%, etc is how many storms are expected. For me, it's far easier to forecast (and chase) slight/mod risk days, because I'm trying to find A storm. I look for the area where I think A or THE storm will be, and I go there. On outbreak type setups or even higher end mod days, there's usually several storms (or even target areas) to choose from, and IMO this creates a larger margin for error if 2-3-4 go up at once and you have to choose.

A few years ago somebody did a career-spanning list of their tornado days VS SPC % risk. I did one myself out of curiosity, and found that most of my tornado days were 5%ers.

It might be implied in there, but for me I definitely do better on the days I need to claw my way to a storm. I was more mentally prepared on the Slight risk days. It's not so much picking the right cell, but paying attention so that I'm in the position for the right cell. Moderate/High risks used to be my worst days because I would just pick a huge target area and didn't nowcast very much - reasoning that tornadoes would just be handed out like candy. After a year or two of tornadoes not being handed out like candy, I realized I needed to work even harder on those days because like Shane said, there would be a lot more targets in play.