• Welcome to TalkWeather!
    We see you lurking around TalkWeather! Take the extra step and join us today to view attachments, see less ads and maybe even join the discussion.
    CLICK TO JOIN TALKWEATHER
  • TalkWeather Tropical Weather Widget
    Active Tropical Systems
    Loading...
    Fetching latest NHC data...

Severe WX Severe Weather Threat 3/14-3/16

Okay, 18z GFS sounding central (tad south of central) MS at 18z on Saturday.

View attachment 35063
So I did a thing using this sounding. My background and work is in Statistics (I have a Master's in it, and I'm a Senior Data Scientist), so the SARS Sounding Analog in the Sharppy-based soundings intrigues me. So I dug into how they are created. It turns out the analogs are ALL deterministic, meaning it takes a few variables from the sounding (like ML CAPE) and goes plus or minus 100 is a match, etc. So it's a very rough and crude technique to find analogs. Instead, I wanted to take a different approach.

So I took the entire SARS supercell tornado database off its Github and ran it through some statistics. Then I took the sounding above as a guinea pig and plugged its parameters into the model. My model uses Mahalanobis distance and k-nearest neighbors to find the 10 closest matches statistically to each sounding's variables.

For anyone curious, this returned the 10 closest soundings ---> 40% had EF2+ tornadoes and 60% were tornadic. Very strong indications of the atmosphere we may face on Saturday.
 
So I did a thing using this sounding. My background and work is in Statistics (I have a Master's in it, and I'm a Senior Data Scientist), so the SARS Sounding Analog in the Sharppy-based soundings intrigues me. So I dug into how they are created. It turns out the analogs are ALL deterministic, meaning it takes a few variables from the sounding (like ML CAPE) and goes plus or minus 100 is a match, etc. So it's a very rough and crude technique to find analogs. Instead, I wanted to take a different approach.

So I took the entire SARS supercell tornado database off its Github and ran it through some statistics. Then I took the sounding above as a guinea pig and plugged its parameters into the model. My model uses Mahalanobis distance and k-nearest neighbors to find the 10 closest matches statistically to each sounding's variables.

For anyone curious, this returned the 10 closest soundings ---> 40% had EF2+ tornadoes and 60% were tornadic. Very strong indications of the atmosphere we may face on Saturday.
wow, really cool and terrifying
 
So I did a thing using this sounding. My background and work is in Statistics (I have a Master's in it, and I'm a Senior Data Scientist), so the SARS Sounding Analog in the Sharppy-based soundings intrigues me. So I dug into how they are created. It turns out the analogs are ALL deterministic, meaning it takes a few variables from the sounding (like ML CAPE) and goes plus or minus 100 is a match, etc. So it's a very rough and crude technique to find analogs. Instead, I wanted to take a different approach.

So I took the entire SARS supercell tornado database off its Github and ran it through some statistics. Then I took the sounding above as a guinea pig and plugged its parameters into the model. My model uses Mahalanobis distance and k-nearest neighbors to find the 10 closest matches statistically to each sounding's variables.

For anyone curious, this returned the 10 closest soundings ---> 40% had EF2+ tornadoes and 60% were tornadic. Very strong indications of the atmosphere we may face on Saturday.
Without a doubt I think we’ll see an EF2+ tornado in terms of intensity. Just hope that doesn’t verify.
 
I can honestly see those being verified as of now with this setup. This thing is on the cusp of looking like a upper-echelon type of outbreak.
Still some questions marks, get about two days out from the event and I think we can start to gauge the caliber of the event better. It definitely doesn't look good though.
 
Back
Top