Who Called Hurricane Isaias Best? I Kept Score.

2 days ago

9 min read

Atomic age comic book illustration of a hurricane bearing down on the Gulf Coast as a TV weatherman, a young forecaster holding a smartphone and a retro robot representing AI race to call the landfall, with a referee holding a scorecard and the headline Who Called Hurricane Isaias Best?

Every time a hurricane gets into the Gulf, my phone turns into a slot machine. Spann on one app, Cantore on another, three Mobile TV stations, the Hurricane Center, a couple of guys on X who run their own weather pages, and now Google's AI model getting passed around like it's the new European model. Everybody has a call. Then the storm comes ashore, the power comes back on, and nobody ever goes back to see who actually got it right.

I wanted to know, so I kept score on Isaias. Here's how everybody did, and then I'll show you how I got there.

The scoreboard

1. James Spann, Alabama Weather Network: 82

2. Jim Cantore, The Weather Channel: 76

3. Alabama EMA: 71

4. Spinks Megginson, RedZone Weather: 68 (won the tiebreaker)

4. AccuWeather: 68

4. NBC 15: 68

7. NWS Mobile: 66

7. FOX10: 66

9. Google DeepMind, the AI model: 62

10. WKRG: 60

11. Jesse Vinturella: 57

12. Ed Bloodsworth, WKRG chief meteorologist: 47

13. National Hurricane Center: 46

Atomic age comic book scoreboard showing who called Hurricane Isaias best, with James Spann first at 82 and the National Hurricane Center last at 46

TV vs. the online guys

Put everybody on teams, and the TV stations and big weather companies averaged about 67. The independent online forecasters came in around 63, and the referees (the Hurricane Center, the Weather Service, and the EMA) around 61.

The establishment won, barely, and that's not the result I wanted. I was pulling for the online guys the whole way. But the rules were locked before the first raindrop, the numbers picked the winner, and the numbers said TV. If I'd bent the grades to get the answer I liked, the whole exercise would be pointless.

How I graded it

The rules were set before the storm hit, and they were the same for everybody. Whatever you said by 5 o'clock Friday afternoon counted as your forecast. Anything after that was reporting, which is a different job, and I'll come back to it.

Every forecaster could score up to 100 points. The biggest chunk, 30, was for location, and anywhere within 20 miles of the real landfall counted as a bullseye. Strength at landfall was worth 25 and storm surge 15. The last three categories were worth 10 apiece: timing, calling the major before it happened, and what Mobile would get for wind and rain.

I didn't dock anybody for a call they never made. You got graded on what you actually said, as long as you said something in at least three categories. Surge is on hold for everybody, because there's no tide gauge at Destin and I'm waiting on the Weather Service's official storm report.

I had AI do most of the gathering. Grok watched X around the clock, right up until I burned through my uses at the worst possible moment, and my AI producer pulled the Hurricane Center advisories, the Weather Service statements, and NOAA's tide gauges. I made the calls on the grades.

The answer key

This is what I graded everybody against. Isaias made landfall near Destin around 8:30 Friday night as a Category 2 with 105 mph winds. That morning it had been a major hurricane, and it lost some steam on the way in. Remember that part, because it ended up deciding most of the grades.

The worst of it stayed east of us. Meteorologist Dylan Federico rounded up the peak gusts: Pensacola 116 mph, Destin 105, Gulf Breeze and Milton 102, Navarre Beach 99. Gulf Shores topped out at 65. Here in Mobile we got a gust around 59 and about a quarter inch of rain at the airport, after a week of being told to expect 4 to 8 inches.

Mobile getting off easy doesn't mean Alabama did. Inland, places like Brewton, Canoe, Lenox, and Peterman took a beating, with trees on houses and folks still needing help days later. Around 900,000 customers across the region lost power at the peak, and at least two people lost their lives. If you're still cleaning up in the Panhandle or anywhere up through south Alabama, I'm thinking about you.

A few grades that need explaining

Cantore got his location points from his own post Friday, saying Isaias was "setting up for a final approach around sunset into the Emerald Coast." Where The Weather Channel parked him backs it up. They had him on Pensacola Beach Thursday night, then moved him east to Miramar Beach, right next door to Destin, on Friday morning while the official forecast still said "near Pensacola." Standing somewhere isn't the same as forecasting it, so call that one a judgment call, but his words and his feet pointed the same direction. He lost his points on strength, for calling a major at landfall.

Google's AI is sitting on the thinnest edge on the board. Its landfall dot was 11 to 14 miles from where Isaias came ashore, just outside my window. If the official landfall point shifts a hair, it jumps to 92 and takes first place. That says something about how to read the exact order. The 20-mile line is a rule I picked, not a law of physics, and a few miles can reshuffle the board. Read the rankings loosely. The broader pattern is what holds up.

WKRG and its own chief meteorologist are graded separately, and that's on purpose. Ed Bloodsworth's last call was Thursday afternoon, and he had the strength dead on, Cat 2 at 100 mph. He also had the center running right through Mobile.

The Hurricane Center finishing dead last looks strange, so here's why. Their cone covered Destin the whole time. But their 4 PM advisory, the last one before my deadline, said landfall "near Pensacola" in plain English, and I graded the words.

Matt Devitt, Mike's Weather Page, and MyRadar aren't on the board because none of them named a landfall spot before the deadline. That didn't stop them from doing some of the best live coverage of the storm.

What separated everybody

Strength. On Thursday, most people were calling a Cat 2 around 100 mph. At 7:20 Friday morning, the Hurricane Center upgraded Isaias to a major, and a lot of forecasts followed it up to 115 or 120. By evening, wind shear had torn into it, and it came ashore at 105.

In fairness to the Friday crowd, that upgrade came from real Hurricane Hunter data, and going up with it made sense at the time. Nobody expected it to come apart that fast. But the people who held their Thursday number scored best, and that lesson applies well beyond weather. The loudest, newest number isn't always the right one.

The computer models

The models are what we were all staring at early in the week: the American, the European, the hurricane models, and Google's new AI. They don't fit my main scorecard, since a model doesn't hand you a timing call or a Mobile rain total the way a meteorologist does. What a model does hand you is a track, so I graded them on how far their landfall point missed.

Wednesday morning, about two and a half days out, Google's AI missed by about 35 miles. The American model missed by about 120, and the two main hurricane models by about 165 and 220. That comparison comes from meteorologist Mark Morrissey at The Atmosphere Explained.

Thursday night, James Spann called it a "model war." Google's AI was pointing at Navarre Beach, and the European was pointing at the western tip of Dauphin Island. When Isaias came ashore near Destin, the AI was about 25 miles off and the European about 105.

On strength, though, the AI was outdone by an old-timer. SHIPS, a statistical model built on decades of past storms, said about 105 mph at landfall on Friday morning, and that's exactly what we got. Google's AI said about 97.

Who to trust two or three days out

One of our listeners, Trin F., asked the question I think matters most. Who can I trust to tell me what's coming two or three days early? That's a different question from who covered the storm best, so I answered it separately.

From this storm, the honest answer is Google's AI and the forecasters who were paying attention to it. In every run I could check, the AI sat on the east side of the pack, and the storm kept drifting east toward it. Everybody else spent the week catching up.

Spinks earned a spot here too. Thursday afternoon, more than a day out, he put out a gust map calling for 100 mph or more from Pensacola to Destin. Every reporting station along that stretch came in between 99 and 116.

Once the storm was on top of us, the best boots-on-the-ground coverage came from Spinks and Spann. Matt Devitt showed birds trapped in the eye on radar, which was one of the more remarkable things I saw all week, and Mike's Weather Page went live from Navarre as the eyewall came through.

Spann summed up that side of the job better than I can. On Monday he told the broadcast meteorologists posting smiling selfies during the storm to go spend the week in Lenox, Peterman, and Canoe "cutting trees off homes and feeding people." Then he added, "If you want to be a social media influencer or a model, fine. But stay out of weather." The forecaster at the top of my scoreboard believes the job is about the people in the path, not the person in front of the camera, and I think he's right.

About that "11 days out" post

You may have seen a post going around claiming Google's AI "crushed" Isaias 11 days in advance. I pulled up the charts the author used to back it up. Ten days out, the AI gave the storm maybe a one in three chance of forming at all. Four days out, it had the spot, and three days out, it had the strength.

That's still the best forecast of the week. It doesn't need the extra seven days. Richard Scott, a TV meteorologist in Tuscaloosa, said the AI called the landfall better than the traditional models and he'll trust it more next time. I'm curious to see how it performs next hurricane season.

The one-man weather channel

The part I keep coming back to is Spinks. He runs a website and an app off the same government data and the same AI models the big stations use. He made his first call Tuesday night, before the storm even had a name, and he finished tied with AccuWeather and NBC 15, ahead of a good chunk of the TV board.

I come from music, and this reminds me of punk bands that got tired of waiting on a label to say yes and just pressed their own records. Weather used to be a business you couldn't get into without a station behind you. The big outfits still have real advantages in staff, equipment, and track record. But the tools aren't locked up anymore, and one forecaster out of south Alabama just showed he can compete with them.

The fine print

All of this is preliminary, and the Weather Service's official storm report could still move some numbers, especially once surge gets measured.

The scorecard is the method. Rules set ahead of time, the same rules for everybody, and the data picks the winner. If you think I graded somebody wrong, tell me, and if you're right I'll fix it on the air.

And it's one storm. Trust gets built over a lot of them, so I'm going to do this every time one heads our way. My rule from here on out is simple. Show me your scorecard, and I'll tell you how much I'll listen.

Scott Tindle and I are breaking the whole thing down tonight at 8 on Beyond The Blockchain, FM Talk 106.5. WHO CALLED ISAIAS BEST? SOURCES:

OFFICIAL / THE ANSWER KEY

National Hurricane Center, Isaias advisory archive: https://www.nhc.noaa.gov/archive/2026/ISAIAS.shtml

National Hurricane Center on X: https://x.com/NHC_Atlantic

National Weather Service Mobile/Pensacola: https://www.weather.gov/mob/

NWS Mobile on X: https://x.com/NWSMobile

Alabama Emergency Management Agency: https://ema.alabama.gov

Alabama EMA on X: https://x.com/AlabamaEMA

NOAA Tides and Currents (storm surge gauges): https://tidesandcurrents.noaa.gov

Dauphin Island: https://tidesandcurrents.noaa.gov/stationhome.html?id=8735180

Mobile State Docks: https://tidesandcurrents.noaa.gov/stationhome.html?id=8737048

Pensacola: https://tidesandcurrents.noaa.gov/stationhome.html?id=8729840