Weather or Not

Severe Outflow by R. Edwards

  • Home
  • About
  • Archives

Powered by Genesis

AI in Weather Forecasting (Not the Last)

July 19, 2023 by tornado Leave a Comment

This is the first, likely not last, entry in this here ol’ BLOG that will be devoted to the use of artificial intelligence (AI), including machine-learning output, in forecasting. If you want scientifically vetted, peer-reviewed documentation of some of what I’m talking about, check out:
* Random forests for short-fused applications
* Using random forests to derive day-2 probabilities from CAM ensembles
* Random forests at various severe-day periods

Both public- and private-sector AI usage and claims are presented in this fairly presented Washington Post article.

As someone who uses, and oversees the use of, a set of AI-informed tools as a part of the daily forecast regimen, I’ll state this is not something to be afraid of, nor to lazily and mindlessly use as a crutch.

To address the fear factor—a lot of that is rooted in the black-box aspect Schumacher has discussed. He and other university and lab researchers (such as Loken) have published peer-reviewed papers (see above) discussing how their methodologies work, in reproducible ways. They’re in AMS journals, to which we in NWS have ready access. So don’t be afraid, be informed.

Admittedly it’s harder to be informed on the private-sector black boxes, because they are unscientific in the sense of not being reproducible and peer reviewed. They’re hidden behind a “proprietary” shroud of secrecy—the very antithesis of science! All you see are results—typically only the cherry-picked best, of course, with perhaps a token minor failure if the presenter is allowed to reveal such. But we don’t really know fully where those are consistently weak, due to (again!) lack of peer-reviewed documentation with reproducible methodologies, analyses and results. Caveat emptor!

To the second point (tools, not crutches), this is another in a long line of beneficial advancements in improving the forecasts overall. They add confidence, for example, to SPC day-3+ areas, but the forecaster does not just mindlessly “draw around them”. Meteorological reasoning still must be used, and discussed, by the outlook forecaster. The more uncommon a scenario is, the less the machine learning is trained on situations like them, and the more they’ll miss, even in an ensemble sense. Bulk statistical verification of AI-based forecasts may look great, but smooths over the outliers, which tend to be disproportionately destructive and important. And there’s still the “garbage in garbage out” factor, mainly due to lack of high-density upper-air obs here and even surface obs in much of the world, that renders uncertainty and output spread in ensembles right from the start.

AI/ML-informed models are imperfect, especially for the rare events that most depend on our own physical understanding as meteorologists to tease out. Lose that understanding, and we might as well be automatons regurgitating models. That might be “good enough for government work” most of the time, maybe even accidentally excellent sometimes.

Yet for anyone tempted to steer that direction, in a personal or policy sense, I offer these sobering and disturbing words:

Air France 447.

Filed Under: Weather Tagged With: artificial intelligence, atmospheric science, ensemble forecasting, extreme weather, forecast automation, forecast uncertainty, forecasting, machine learning, peer review, probabilistic forecasting, science, severe storm forecasting, severe weather, storm forecasting, understanding, weather forecast, weather forecaster

Endorsing Kelvin Droegemeier for White House Office and Science and Technology Policy

August 17, 2018 by tornado Leave a Comment

The audience of this BLOG knows that, as a third-party voter for President the last cycle, I think independently, evaluate actions and issues on their own merits and not through major-party lenses, have not hesitated to air grievances regarding the current occupant of the White House, and also, have credited the administration when it has made good choices.

This announcement is its BEST choice yet—a grand-slam home run.

The new White House science and technology advisor nominee, Kelvin Droegemeier, is as smart as they come, a former professor of mine in atmospheric dynamics (think I still smell those clouds of chalk dust flying off that board as he wrote equations at near-warp speed!), a sharp and well-published atmospheric scientist, mentor of many students, skilled communicator, and most importantly, a fundamentally good man who has served state and national science advisory roles, including the National Science Board in both the Bush-43 and Obama administrations.

Kelvin has a powerhouse 55-page CV: broad, deep and well-suited for any national science advisory role. Personally, I’ve known Kelvin for over 30 years and can vouch for his high character and unwavering integrity. He also is a keen dynamicist who can “out-math” some physicists I’ve known. Now, if confirmed, Kelvin gets the ear of the White House as perhaps the most influential meteorologist ever in government.

It’s telling how many atmospheric scientists (dozens, so far), whom I know to be firmly on the right or left sociopolitically, have one thing in common: they have supported this nomination on social media. There is a good reason for that. Kelvin is a true “servant leader” who has the interpersonal skills to build bridges across disparate groups and personalities, and before diving into something, makes sure he has the knowledge base to back up everything he says.

In addition to many of us who know Kelvin, and/or are familiar with his scientific and/or communications work, the American Association of Universities enthusiastically endorses him for the role. I hope this helps to offset some opposition he’ll have on purely partisan grounds from some members of the D side, at least to enough extent to get him resolutely confirmed.

Regardless of which of the two major parties is “in charge”, there always will be hyperpartisan hacks on both sides of the aisle ready to support or oppose a candidate without knowing a flipping thing about that person. That’s the ugly reality. It’s also the only reason Kelvin would not get a unanimous support vote, based on his background and endorsements.

Count this as one more on the endorsement pile. Kelvin Droegemeier has what it takes to skillfully navigate this mine field, and I endorse his nomination to the utmost degree.

The Senate committee will hold a confirmation hearing next Thursday at 9:15 a.m. CDT (10:15 a.m. EDT). Kelvin—go do us all proudly. In this climate (pun intended) it won’t be easy, but if anybody can, it’s you!

Filed Under: Weather AND Not Tagged With: atmospheric science, Congress, Donald Trump, Kelvin Droegemeier, meteorology, science, Senate, White House

Some Lessons from the Sulphur Tornado Intercept

May 15, 2016 by tornado Leave a Comment

Sleeping after an overnight shift, I awakened to Elke’s informing me of a storm erupting a couple counties to the south. I had asked her in advance to do this conditionally–the condition being a storm forming W of I-35 that could cross the Interstate not far away. This was close to the Interstate, likely necessitating a more eastern approach, especially considering all the road work and likely delays around SW Highway 9 and I-35 in Norman.

When I looked through sleepy eyes at a surface map, satellite image and radar loop, those eyeballs damn-near exploded out of my head–this storm was moving off the dryline and into a tremendously moist and well-sheared environment and obviously would stay that way for another 1-2 hours, and I had to get out the door fast. More on the environment later…

As it turns out, and unbeknownst to me at the moment, the supercell already was becoming tornadic before I pulled my vehicle out of the house. I quickly fueled, picked up my intercept partner for the day (Mateusz Taszarek), and booked out of town as quickly and safely as traffic and lights would allow. Five more minutes delay at any point in the process and we might have missed everything, for we would have had to divert the long way around through Ada instead of plunging straight southward on US-177 toward Sulphur to get ahead of the main mesocyclone.

Anticyclonic forward-flank tornado–inside the core!

Every year we learn new things about the atmosphere. Out of hundreds of supercells I’ve intercepted, this is the only one with an anticyclonic tornado inside the forward-flank core! Scarier still, another few minutes and we might have driven into it unknowingly. Then again, another few minutes and I wouldn’t have attempted to go southward through an even meatier part of the core with the risk of huge hail. We made the plunge because I still could see through it to the other side, and because the obviously intense mesocyclone still would be a few miles W of the road when we passed. Below is an annotated screen capture from a RadarScope display including our location at about image time.

This is the first time I’ve even heard of a well-developed anticyclonic tornado buried inside the forward-flank core. In making the transect, we saw no evidence of it in the gray murk of heavy rain to our W, nor any suspicious wind shifts. We had NW winds W of Roff (as it turns out, on the clockwise side of the mesoanticyclone); but you’d also, often experience the same in that part of a “normal” forward-flank core. Winds then shifted to SE as we exited the forward flank. This is nothing surprising either; a violently tornadic circulation lay a few miles to our WSW.

What was surprising was to glance at velocity information a bit later and see that clockwise-spinning couplet racing away from us to the NE as we positioned temporarily in that great photographic location directly in the big tornado’s path.

Why did the anticyclonic tornado happen? We only can offer speculation and conjecture at this point, pending more detailed mobile-radar information (were they scanning northward at all?) and numerical simulations, so I’ll offer my best guess.

Conceptualize a localized anticyclonic vorticity field along the forward-flank gust front (FFGF), perhaps cast there or reinforced by negative BL vorticity from an absorbed split or left-moving updraft…then reinforced by anticyclonic shear (negative vorticity) ambient to that distance from such a big and intensely tornadic mesocyclone. I’ve seen small, brief anticyclonic tornadoes on FFGFs before. Still, even when reckoning that improbable (yet physically plausible) stack of dominoes, I’m at a loss for how that tornado survives for >12 miles while buried in the static stability implicit to 59–64 deg F (based on car thermometer) surface conditions inside that FF core!

Below are a couple 0.5-deg radar loops Steve Miller (TX) made of the process. He saved me the trouble (thanks Steve!).

Will this change intercept strategies in the future when north of a storm? Maybe. Playing probabilities, the odds are extraordinarily tiny of ever seeing another anticyclonic tornado entombed in the forward-flank core of even a violently tornadic supercell. Yet now we know it’s possible, and we must be vigilant of that. This means paying more attention to wind cues and, when available, radar velocity imagery in and near that part of the storm.

Wet wedge, debris-filled RFD

Okay, this isn’t a great surprise after the fact, in a mesoscale-diagnostic sense, given: rich boundary-layer moisture, low cloud bases (low LCL), favorable deep shear, and strong low-level shear and helicity. The supercell interacted with a vorticity-laden outflow boundary from morning storms for a long time by moving eastward at about the rate and direction the boundary itself eroded.

This is not something that can be forecast more than a few hours out, as analytic details become better-defined, because neither the synoptic operational guidance nor the high-resolution convection-allowing models (CAMs) tend to predict such features well. Some CAMs forecast a supercell in about that area (right for the wrong reasons?), but none finished before 12Z had updraft-helicity (UH) tracks characteristic of a violent-tornado-producing supercell. Regardless, the value of careful human diagnostics shines through here, and helped to motivate a well-reasoned pre-watch mesoscale discussion, then tornado watch and follow-up discussion from the SPC.

The tornado itself was a well-developed wedge when we first saw it to our SW (video capture from Mateusz), breaking out of the forward-flank region, and remained over a mile wide as it crossed US-177 north of Sulphur, based on the annotated NWS Norman path map linked here. The next two photos were shot from the first good vantage we could find outside the forward flank, a spot that also happened to be in the tornado’s path (see map linked above). This also was in a beautiful country setting with green grass, trees, a pond and wildflowers-and unfortunately, a beautiful ranch home that was hit (survey photo and damage-assessment tool image courtesy NWS Norman).

The next four images follow the tornado from our second vantage (green dot on the above map), after we got out of the tornado’s way and let it pass to our immediate NW and N.

Violent tornado(es) and disconnect between winds and damage

The storm already had produced one violent tornado before I got there, and another was underway. For the record, I refer to the first (“Katie”) as “violent” given its EF4 damage rating to one house, and the second (“Sulphur” wedge) also as “violent” due to EF5-level mobile-radar winds ~50 feet above ground, even though the highest damage level found was EF3. Official damage ratings are based on the EF scale, which is a damage scale used to estimate winds. It does contain damage indicators that extend 50 ft or more above ground, which also is where the mobile Doppler radar did its sampling in this case. The EF scale, however, does not contain accommodation for actual wind measurements at or near ground at the level of varius DIs, such as by wind instruments or mobile radar.

We can argue about the problems with this until our faces turn blue, but regardless of science or what you and I think, official policy since mid-2013 forces ratings to be based strictly off damage only, regardless of any and all other evidence. This event piled more fuel onto the flames of disconnect between tornado intensity and tornado damage that began with 2013’s El Reno OK and Bennington KS radar-EF5 tornadoes (the latter of which I also observed and photographed) that “only” produced EF3-level structural damage due to their most intense winds missing structures capable of yielding EF5 results. That happened with Sulphur.

The American Society of Civil Engineers has commissioned a group of meteorologists, engineers and a forestry expert to revisit and standardize the EF scale, which ultimately will refine and add damage indicators. It’s a years-long process. A lot of cooks are stirring that stew, including some I respect hugely. [I’m not part of that process due to lots of night-shift work and associated inability to be coherently present for many meetings.] However, the official policy of rating tornadoes based only on damage is a separate matter, and imparts a strong bias toward populated and densely constructed areas. As such, the official tornado rating is essentially a pathetic joke from a scientific perspective.

The EF Scale as it now stands also is about as useful as a $3 bill in rural areas where violent tornadoes occur, but damage indicators capable of EF5 results either aren’t present or aren’t spaced fortuitously with respect to the tiny area of a tornado that has those top-end winds. That’s a good thing from the human-impact perspective! However, it leads to unrepresentatively low ratings for many high-end Great Plains tornadoes.

A well-developed tornado between supercells!

We had a little trouble finding Highway 7 east out of Sulphur, but once we did, we noticed a new supercell well to the southeast (but conceivably reachable). This would become the Wapanucka/Atoka supercell, its first tornado causing a fatality SE of Connerville. Meanwhile the Sulphur storm charged eastward past Hickory, appearing to lose organization as a rampart of convection erupted between it and the Wapanucka storm to its SE.

We had a decision to make: go after what was left of the Sulphur storm, go home, or try to intercept the Wapanucka storm from a typically unsafe NW approach, as that supercell headed toward the steep terrain, road voids and thick forests immediately east of Atoka. Imagine our surprise, then, when this materialized from a small, ragged, otherwise unimpressive updraft several miles to our ESE, NW of Connerville and in between the supercells!

The atmosphere made our decision for us: “None of the above!” Even though the updraft was rotating, it’s a stretch to call the feature a supercell, since its radar presentation was hardly recognizable, and since the parent cloud form was so disorganized and shredded in appearance. The cloud base is uneven and scuddy, what little updraft there is very tilted and small. Imagine this “storm” with no tornado—would you then expect a tornado to form any second? Below is a zoomed-in perspective.

Coincidentally, we viewed this remarkable tube from 7 miles away at a spot along the same road and very near where I observed another tornado (in a different direction), five years minus two weeks before this. The Connerville tornado offered a spectacular, protracted rope-out.

Thus delayed by observing this tornado, and concerned by the highly unfavorable logistics of intercepting the southern supercell around or east of Atoka from the NW, we crept E on the nearest county road behind the old Sulphur storm in case it revealed any tubes from the back side. That not materializing, we headed home in time for late dinners.

The lessons of this intercept are many, but they include:

  1. Beware the possibility of an anticyclonic tornado buried inside the forward-flank core of an intense, cyclonically tornadic supercell.
  2. Exercise due diligence and great caution with any approach of a potentially or actually tornadic storm from the north (its left flank).
  3. In high-vorticity (high-low-level-shear) environments, such as old outflow boundaries. watch for spinups from any persistent updraft, even if it appears disorganized visually and/or on radar.
  4. When sufficiently interesting storm-observing potential occurs amidst a set of night shifts, have someone alert/awake you to nearby development in a potentially ripe situation (such also was the case for me on 3 May 1999).
  5. Official (EF) ratings are not scientific and don’t necessarily represent tornado intensity. I’ve known this for many years. Chuck Doswell has too, and reinforced the point in a recent entry on his BLOG. One can have violent, EF5 tornado winds without a violent EF rating.

To quote the late, great Paul Harvey: “Good day!”

[EDIT: 19 May 16 for added image links]

Filed Under: Photographic Adventures, Weather Tagged With: anticyclonic, anticyclonic tornado, atmospheric science, damage rating, EF Scale, hook echo, Katie tornado, meteorology, mobile radar, photography, rear-flank downdraft, RFD, rope tornado, severe storms, severe weather, storm chasing, storm observing, Sulphur, Sulphur tornado, supercell, supercells, tornado, tornadoes, violent tornado, wedge tornado, Wynnewood

  • 1
  • 2
  • 3
  • Next Page »

Search

Recent Posts

  • Norman “Stormwater” Utility: An Unpublished Letter
  • Better Choices than Woke Cult vs. Trump Cult?
  • Critical Thinking as Applied to an Overseas News Item
  • AI in Weather Forecasting (Not the Last)
  • The Sound of Freedom: An Important Movie

Categories

  • Not weather
  • Photographic Adventures
  • Scattershooting
  • Weather
  • Weather AND Not

Twitter API temporarily busted. Check back later.

Blogroll

  • CanadianTexan
  • Chuck's Chatter
  • Cliff Mass Weather & Climate
  • Digital Photography Review
  • DMN Dallas Cowboys BLOG
  • Dr. Cook's Blog
  • Dr. JimmyC
  • E-journal of Severe Storms Meteorology
  • Eloquent Science
  • Image of the Week
  • Jack's Cam Wall
  • Jim LaDue View
  • Laura Ingraham
  • MADWEATHER
  • Michelle Malkin
  • Photography Attorney
  • Severe Weather Notes
  • SkyPix by Roger Edwards
  • Tornatrix
  • With All My Mind

Meta

  • Log in
  • Entries feed
  • Comments feed
  • WordPress.org