SPC Day 1 Convective Outlook — Emotional Hazards Issued 0847Z March 4, 2026 Valid: Duration of interview

Coverage probability: HIGH. Conditional Intensity Group: 2, locally 3. Moderate-to-severe personal disclosure likely across central Oklahoma. If candor develops, expect sustained vulnerability exceeding institutional comfort thresholds. All paths conditional.

Norman, Oklahoma. The National Weather Center smells like burnt coffee and static electricity. It is March 4, 2026, thirty-one hours since the Storm Prediction Center's Conditional Intensity system went live. This is the most significant change to severe weather communication since 2014.¹ On the operations floor, visible through interior glass, the new graphics are running on monitors for the first time in public view. Three forecasters are on duty. One of them keeps glancing at the colors.

Reva Windmark has been at the SPC since 2013. She was part of the 2019 Spring Experiment in the Hazardous Weather Testbed that first explored conditional intensity forecasting.² She has been producing these forecasts experimentally since late 2021. Over four years of knowing how to name the worst before anyone outside this building had the vocabulary for it.³ Her title is Warning Coordination Meteorologist, which means her job is the seam between what the models show and what the words say. She exists, as all composite characters do, in the space between what any single person would be permitted to say on the record and what the institution knows collectively.

She has a mug that says PROBABILISTIC THINKING IS MY LOVE LANGUAGE. The mug is chipped. She holds it with both hands like a talisman.

The system went live yesterday. What changed?

Reva: Operationally? One new layer on an existing product. Psychologically? I'll let you know when I stop vibrating. The outlooks still go out. The five-category scale still exists, Marginal through High.⁴ But now, for the first time, we can answer both questions at once. Not just "what are the chances" but "how bad if it happens." Those have always been two separate questions and we've been answering them with one map. Which is... imagine your doctor can tell you the probability of getting sick but can't distinguish between a head cold and lymphoma. Same map. Same color. That's what we were doing. For decades. And everyone knew it, and the process to fix it took seven years, and the process was correct to take seven years, and I need you to hold both of those facts at the same time because I have been holding them and I am tired.

The word "conditional" is doing a lot of work in this system. CIG only applies if storms develop. You're forecasting the intensity of violence that may never arrive.

Reva: People get stuck on that. I understand why. But all forecasting is conditional. Every forecast describes a world that may not show up. What we did is put the conditionality in the name. Most forecasting hides the "if." We're saying: if this happens, here is what it looks like. That honesty is uncomfortable because it means we're asking people to prepare for something we're simultaneously telling them might not occur. But that's the truth. That was always the truth. We just stopped pretending the map was the territory.

You've been producing these forecasts internally since 2021. For over four years, you had this knowledge and the public didn't.

Reva: [Long pause.] That's accurate.

What was that like?

Reva: You watch an event unfold. You've issued the experimental CIG internally, maybe a 2 or a 3. You see it verify. You see the damage surveys come back and they match what you said, in a product that doesn't exist yet, to an audience of nobody. Then you talk to emergency managers and you use the old language, the hatching, the "significant severe" label, which looks the same whether you're expecting a strong EF-2 or a violent EF-5.⁵ You know the difference. You can see it in the data. You have a system that names it. But the system isn't real yet.

[She stops. Looks at the operations floor through the glass.]

Tornadoes get three intensity levels. Hail only gets two.⁶ Someone had to draw those thresholds.

Reva: Someone did.

Was it you?

Reva: It was a team. But yes. I was in the room. The thresholds come from twenty years of mesoanalysis data; they're not arbitrary.⁷ The data tells you where clusters exist. Where the distribution has natural breaks. What the data does not tell you is that this cluster gets called "3" and that one gets called "2." A human being has to say: here is where "dangerous" becomes "historic."

I'll tell you something. The hail question keeps me up more than the tornado question. With tornadoes, the lethality gradient is so steep that the thresholds almost draw themselves. EF-3 versus EF-5 is not a subtle distinction. But hail. We went with two levels instead of three, and I'm still not sure. Two levels means we're saying the difference between significant hail and the worst hail is a binary. Yes or no. I know we couldn't justify a third level statistically. But "couldn't justify statistically" and "doesn't exist" are different things. The space between them is where people's houses are.

The word "tornado" was banned from U.S. weather forecasts until 1938. Even after the ban was lifted, forecasters barely used it for another decade.⁸ You're in a lineage of people who had to make the language bigger.

Reva: I think about Fawbush and Miller constantly. Tinker Air Force Base, 1948.⁹ They predicted a tornado, correctly, and the institution didn't know what to do with that. The language wasn't ready for what the science could see. Seventy-eight years later and the problem is structurally identical. We're not fighting a ban. We're fighting the reasonable institutional caution that keeps you from saying "the worst" until you've verified that your system for identifying "the worst" actually works. And the caution is reasonable. The process worked. We're here.

[She glances at the monitors.]

It's live and the weather today is dead calm. Not even a Marginal risk. God has a very specific sense of comedic timing.

Does naming it change what it is?

Reva: [Very long silence.]

Before yesterday, there was no such thing as a CIG3 day. There were days that, in retrospect, we'd classify as CIG3. April 27, 2011, the Super Outbreak. 321 people killed in a single day.¹⁰ That was CIG3. We know that now because we went back through the data. But on April 27, 2011, the concept didn't exist. The forecasters that day used the tools they had. They issued a High risk with significant severe hatching. They did everything right. And 321 people died.

Now CIG3 exists. The next time a day like that is coming, we can say: this is CIG3. This is the category we built for the worst. I don't know if that saves anyone. I genuinely don't know. But the act of making "the worst" an official category, with a number and a color and a definition...

[She picks up a pen. Writes on a napkin. Slides it across the table.]

That sentence exists now. It didn't exist on Monday. And the next time three to five people on the overnight shift in this building are looking at the models and seeing something that looks like the end of the world, they have a sentence for it. Whether that changes the tornado or just changes the people...

Just changes the people?

Reva: I think that might be the only thing that was ever possible to change.

What does it feel like to be right about CIG3?

[Her phone buzzes. She looks at it. Looks at the operations floor. Looks back.]

Reva: Ask me in tornado season.

She stands. The interview is not over but she is already walking toward the glass. On the monitors, the Day 1 Outlook is updating. Today's conditional intensity: none. The baseline. A calm day in early March, the system running for the second time in its public life, measuring nothing, ready.

Footnotes

1. NWS Service Change Notice 26-11, effective March 3, 2026. https://www.weather.gov/media/notification/pdf_2026/scn26-11_SPC_conditional-intensity.pdf ↩

2. The 2019 Spring Experiment in the NOAA Hazardous Weather Testbed first explored conditional intensity forecasts. https://www.spc.noaa.gov/exper/conditional-intensity-information/ ↩

3. Since late 2021, SPC forecasters have been producing conditional intensity forecasts experimentally. https://www.weather.gov/news/262402-spc ↩

4. The five-level categorical outlook scale (Marginal, Slight, Enhanced, Moderate, High) was introduced in 2014. https://www.spc.noaa.gov/ ↩

5. Under the previous system, hatching indicated significant severe potential but could not differentiate between EF-2 and EF-5 tornado environments. https://www.foxweather.com/learn/noaa-storm-prediction-center-debuts-new-conditional-intensity-groups ↩

6. Conditional intensity uses 3 levels for tornado and wind but only 2 for hail. The asymmetry is unexplained in public SPC documentation. https://www.spc.noaa.gov/exper/conditional-intensity-information/ ↩

7. CIG groups derived from 20 years of SPC mesoanalysis data. https://www.memphisweather.net/blog/2026/03/01/storm-prediction-center-convective-outlooks-to-add-intensity-potential/ ↩

8. Weather Bureau policy prohibited use of "tornado" in forecasts through the early 20th century. The ban was lifted in 1938; forecasters barely used the word for another decade. https://www.spc.noaa.gov/history/early.html ↩

9. The Tinker Air Force Base tornado of March 25, 1948, predicted by Maj. Ernest Fawbush and Capt. Robert Miller, is considered the birth of operational tornado forecasting. https://www.spc.noaa.gov/history/early.html ↩

10. The April 27, 2011 Super Outbreak killed 321 people. Retrospective analysis shows it would have received CIG3 designation. https://www.memphisweather.net/blog/2026/03/01/storm-prediction-center-convective-outlooks-to-add-intensity-potential/ ↩

11. CIG3 definitions per SPC Conditional Intensity documentation. https://www.spc.noaa.gov/exper/conditional-intensity-information/ ↩