Cameron Nixon

Ever feel like your town gets more nighttime tornadoes than others?

Here's a map of tornadoes by time of day (nighttime tors are plotted on top of daytime tors to stand out). I feel for you Tulsa, Jackson, Birmingham, Nashville...

Lol if this isn't me

That was a gorgeous halo!!

Fantastic example of hail doing what it wants

(Preferentially produced by elevated supercells north of the greatest perceived risk, even though both targets featured supercells)

I was simply beam-ing when I took this last night

📍The Wichita Mountains, OK

Thanks

Best I can do is

Tornado
Hail
Wind
Unified
Mapping
Badaboom
Omg

(THWUMBO)

LOL you guys 🤣🤣🤣

Thanks man!! 😄 I appreciate you

8 months ago I had a vision: what if we had a unified, track-based dataset of all severe events?

Still a lot of improvements I can make to the algorithm itself, but otherwise, looks like we have one now. I can't wait to see what we can do with this 👀

Tornado tracks by rating (left) and duration (right)

While highly rated/long-tracked tornadoes are biased towards the Southeast, long-duration tors are found even up to the High Plains.

Credit to @theasandmael.bsky.social for her incredible dataset!!

Here you go!!

The long tracks here I've found are usually created by some robust, distinctive mesovortex/embedded supercell with an obvious persistent track

So that derecho could be made up of 3 dozen or more smaller swaths

We can chat about this next time we chat!! (sorry, haven't been able to check out the radar stuff yet, but that's definitely on my short-term list). Essentially though, this is a more storm scale report clustering. It won't pick out massive MCS-scale swaths but smaller swaths within it

Oh heck yes this is awesome!! 😄

Convective Mode Classification and Distribution of Contiguous United States Tornado Events from 2003–2023 Abstract Tornadoes and severe thunderstorms represent a significant threat to life and property in the United States annually. Approximately one thousand tornadoes, two-hundred being significant (F/EF-2+), are documented on average every year. Using archived radar and Storm Data storm report information, 21 912 tornado grid hours were manually analyzed for convective mode from the years 2003–2023. This dataset builds upon prior work by Smith et al. (2012) (hereafter S12) by more than doubling the sample size to produce a robust, multidecadal climatology of tornadoes by convective mode. Comparisons were made between tornado samples spanning 2003–2011 and 2012–2023 to assess changes in the frequency and spatial occurrence of tornadoes. Convective mode characterization consisted of a subjective analysis of WSR-88D imagery into three categories: 1) supercell, 2) quasi–linear convective system (QLCS), and 3) disorganized. Spatial climatologies of the different modes were performed and Kernel Density Estimate plots of events per decade were generated as in S12. The highest climatological frequency for tornadic supercells is reaffirmed to extend from KS and OK east-southeastward to MS and AL, while QLCS tornadoes are more frequent from the northern Gulf Coast states into the lower Ohio River Valley. Comparing the new sample to the original, QLCS tornado relative frequency increased by over 100%. Substantial variation in QLCS tornado occurrence may be due to non-meteorological factors such as dual polarization radar and different observing/reporting practices.

I was going to wait until the shutdown was over but the cat is out if the bag. My first journal article is now live!

Convective Mode Classification and Distribution of Contiguous United States Tornado Events from 2003–2023 in: Weather and Forecasting - Ahead of print share.google/M3ty4pVs5Tld...

Next up, wind swaths! 🌬💨

This was created solely using report clustering (no other gridded data). Plenty of events across the entire US, but can see the most intense storms from the Great Plains into the Midwest! 🟦

Yup!!

Ooohh I haven't seen this yet, thanks for sharing this! Yes, 100%, my *only* goal with using MESH is to detect a reasonable portion of observed hail events, such that it may determine if reports were connected. As of this time, no intention of using the actual values except as a threshold!

Yes! I'm hoping that even just the existence of MESH data in these rural areas (without even assessing its "expected size") could help us with the population issue. I'm currently exploring ways to incorporate these late reports, as they're obviously important!

It's possible! I think the biggest issue here though is that the threshold I use for swath-detection (0.75") is simply too high to capture many events (especially in the cool season Oct-Feb). I'm exploring something as simple as using a variable threshold by month to capture these.

I agree wholeheartedly. My near-term goal is to simply get to a point where I have a reasonably high detection rate and MESH swath for each observed hail event, and will *not* be using actual MESH values for any analysis except for fun. Just for object detection!

Ever wondered how hailstorms vary across the year? Here's all radar-estimated hail swaths by month since 2011. Love seeing the northwest flow monsters come alive June-August!

[MRMS MESH > .75", colored by max hail size]

Can confirm it was a late report, late by almost an hour. These are the kinds of things that grind my gears!!

a bit different using this sort of events-based dataset. Also, I actually know a student here at OU who is thinking of looking into modifying MESH based on the background thermodynamic/kinematic environment....

Bingo!! One really interesting thing I'm finding-- MESH bias varies considerably from the cool season to the warm season (esp. August/September), where warm/moist season storms tend to have very high MESH but the smallest hail. I'd imagine our basic climatology of hailstorms could look quite...

Interesting!! I need to dig into this more to figure out the issue; I can confirm I'm not exactly missing it, but it didn't map to the swath correctly. This can happen with late reports. I'll look into this. Thank you so much for pointing this out!!

Actual hail swaths per MRMS MESH >= .75", colored by measured hail size. I'm not sure when the world will be ready for this kind of data but it sure is sweet

I'm actually using Nathan's MESH dataset for this!! I'm essentially just stringing reports together with MESH data. If I write a paper on this I intend to compare the two climatologies.