Storm initiation times?

[Ben Corby]

So I'm hoping to get a little chasing in tomorrow on CO's eastern plains, and after doing my own little practice forecasting run, I picked Bennett, CO, as the starting point, with a general feeling based on veering, directional shear and CAPE that anything happening in CO tomorrow would happen in a rough line between there and Fort Morgan. You can imagine how pleased I was to check the SPC's website and see that they had assigned a 2% tornado and 5% hail possibility for a stretch of CO that, basically, extends from Bennett and Limon up to Fort Morgan.

My question is how do you figure out the likely initiation times? The SPC is suggesting late afternoon, but how do they come to that conclusion? And, likewise, how can I estimate storm initiation time when deciding what time to head out?

 

[Rob H]

The archives are a great source for information:

http://www.stormtrack.org/forum/archive/index.php/t-23056.html

Additionally, if I think initiation is at 21z, I set a goal to be in the target area by 19z. That gives you time to nowcast and potentially move if you need to.

 

[Patrick Marsh]

Convection Initiation (CI) is a very challenging problem, and one that NSSL is beginning to explicitly address, particularly in the Hazardous Weather Testbed.

There is no standard definition of CI, as everyone we've talked to in the HWT over the last four weeks (during the Spring Forecast Experiment) has a different definition. Does CI begin at the point of the first bubbling cumulus field? Is it when a cumulus cloud reaches the congestus stage? Or what about when a cumulus cloud begins to make a certain dBZ threshold on radar? Or is it when lightning is observed? Does it require an updraft of a certain depth, speed, intensity? Do "dry" updrafts constitute CI or must the updraft contain some sort of moisture? Depending on who you ask, we've received all of these as potential definitions.

At present we have three "working" definitions of CI, one based on reflectivity (REF), one based on updraft moisture content and speed (WQQ), and one based on lightning (LTG). Using these definitions we attempt to determine when various storm scale numerical forecasts predict when CI will occur. We are using the NSSL-WRF and 28-members of the CAPS ensemble. (You might also have seen these products added to the HRRR web displays as they are now using these definitions.) You can view how the numerical guidance is performing regarding CI by examining several web-pages that we've put together on the Spring 2011 Experimental Forecast Program's webpage. Of particular interest are

Observations (Mike Coniglio)
Model Forecasts vs. Obs. vs. Human Forecasts (Me)


Regarding the webpage I've put together, the color-filled underlays are the probabilities of convective initiation within 25 miles of a grid point during the previous hour from the various members (using a smoothing of 20 grid points), ensemble probabilities (smoothings of 05, 10, 15, & 20 grid points), and observations (smoothings of 05, 10, 15, 20 grid points). You can overlay the actual CI points from the various models and observations by selecting the field from the OVERLAY menu. You can also click on the human forecasts by clicking on the "show forecast" link. I should add the disclaimer that for the human forecasts, black = domain forecast, green = 10% chance of CI, red = greater than 10% chance of CI, magenta = greater than red, and blue hatched = region where humans think first CI will occur. Also note that the humans are forecasting for a 3-hour window, which varies by day. I've yet to actually list anywhere what that 3-hour window is, so keep that in mind when examining the human forecasts.

I'd be interested to hear thoughts from everyone regarding CI and perceptions of how the models and/or members have been doing.

EDIT:
You can click on a date at the top of the webpage to view previous day's images. Note that the CAPS ensemble guidance is typically only available Monday - Friday. The NSSL-WRF and Observations are available daily.

 

[Mike Moore]

That's great information Patrick - thanks.

Unscientifically I would say a lot of it involves experience. I know that's not very helpful, but over the years I've been burned, badly, by trying to time my arrive at my target anywhere near initiation. To be fair though I usually chase is Florida which is a fairly predicable pattern on most days. Florida has easily visible indicators like the seabreeze to help to time things. When I do chase on the plains it's an all day event, so I'm usually hanging out truck stop to truck stop from the mornings until I home in on my target area. I've not chased in CO, but hopefully some experienced locals will chime in for you.

 

[Jeff Duda]

Patrick,
I find the issue of CI fascinating. My adviser at ISU attended during the first week and had a lot to tell me about the experiment when he came back. He also showed me the write up you guys made before the experiment began so I had a chance to read through it and see what you were doing this year. I've been following along on Jimmy Correia's blog from the CI desk.

Since my masters work involves model skill of initiation of convection at high resolution, I have been deeply involved in such a task myself. Because I needed a hard, fast, and reliable way to define initiation, I used 3 mm of 1-hourly precip as my definition. The value came from an idealized thought experiment using a 10 mm/hr precipitation rate as the minimum threshold for convection that I have seen cited in the literature. I can then use Stage IV precip data as my observations. Clearly the true accuracy of the Stage IV data is questionable at very high resolutions, but it is better than attempting to measure updraft velocity or rain water mixing ratio in a small area using a rawinsonde or something. However, I think you guys are really on to something when looking at such thresholds. To me, convection initiation occurs when a saturated updraft plume reaches a minimum threshold velocity and maintains itself for at least some minimum period of time (like 30 - 45 minutes so that it is clear the updraft went through, at the very least, the "air mass thunderstorm" evolution cycle). However, along with that, you should also see an increase in cloud water or rain water mixing ratio in that area. As far as the lightning goes, I could argue both ways. I think you can get tropical convection that has no lightning (but I suppose the issue is with continental convection).

Great stuff. I look forward to the future of CI research. Hopefully I can continue to be a part of it.

 

[Bob Hartig]

Ben, I'm no whiz kid at predicting initiation. It's still a bit of a mystery to me. But on the practical side, some of the things I look for are:
* When and where does CINH erode first?
* Given an area of decent moisture and instability, when and where will a source of lift arrive? I always try to get a feel for what the 500 mb jet is going to do as it provides a source of both lift and shear vital to storm organization. Also, what boundaries exist that can provide points of convergence?
* Nowadays you can look at the "cheat sheet"--the HRRR composite reflectivity--and get a pretty good idea of when and where precip breaks out, and what quality it's likely to be.
* If you know how to read a skew-T, you can cross-check likely-looking map locations with soundings to get a feel for when the cap may breach and what dynamics will be present once that happens.

On the day-of, satellite is essential, and the hourly HRRR and SPC mesoscale analysis graphics are very useful.

I hope this is helpful.

 

[Ben Corby]

Thanks to all for the responses. Patrick, you raised an interesting point with the start of CI (which is a term I'll add to my vocabulary!). I had always considered CI as occurring when the towering cumulus stage is reached, but even from my amateur viewpoint, I can see the arguments based on reflectivity and lightning as well.

Mike, I grew up in Jacksonville, so I know all about those afternoon seabreeze pulsers! I think an even simpler way to say it would be, "It's sunny until 2:00 PM (3:00 PM inland), then it rains for 15 minutes."

Bob, I'm learning how to read a skew-T as we speak. I know what everything on it means, and am to the point of interpreting it. Which, of course, is the hardest part! I haven't used CINH or HRRR composite reflectivity yet, so new reports to study and learn!

 

[Bob Hartig]

CINH, or CIN (Convective Inhibition), is the flip side of CAPE. If CAPE is the energy that drives convection, then CINH is the negative energy that has to be overcome in order for CAPE to do its thing. Often you'll encounter it in the form of an abrupt temperature inversion around the 700 mb level called the "cap," a term with which I'm sure you're familiar. The cap serves a purpose in containing energy in the PBL, forcing it to simmer and build in response to daytime heating rather than simply dispersing into the troposphere. But at a certain point, normally sometime in the mid to late afternoon, the cap needs to weaken enough to be breached. Unless that happens, all that lovely CAPE just sits there unrealized.

CINH being the corollary to CAPE, it is expressed as a negative value (ex. -50 J/kg). You can look at -100 J/kg as a threshold; from there and beyond, you've got a formidable amount of CINH. Less than that and you're in business, particularly if forecast hours show a weakening trend. The SPC's mesoanalysis graphics use light shading to demarcate CINH of -25 to -100. Often you'll see storms firing in areas of this lighter shading (and, of course, areas of no shading). You don't need CINH to completely disappear in order for storms to initiate; you just need it to weaken enough for the CAPE to power its way through.