Tornado warnings provide a lead time measured in minutes, and extending it has proven far harder than improving other forecasts. The obstacle is a tradeoff between warning earlier and warning wrongly.
What the warning is based on
A warning is issued when radar shows rotation within a storm at low levels, or when a tornado is reported by a credible observer.
Rotation typically becomes detectable minutes before a tornado forms, and that interval sets the practical ceiling on lead time for a radar-based warning.
Waiting for a ground report would eliminate lead time altogether, so warnings are necessarily issued on a signature that indicates possibility rather than certainty.
Why warning earlier means warning more often
Most storms showing low-level rotation never produce a tornado, and the earlier in a storm's life the decision is made, the higher that proportion becomes.
Issuing warnings at an earlier stage would therefore increase lead time and increase the share of warnings where nothing happens.
The false alarm rate is already high enough to be a recognized problem, which constrains how much earlier warnings can reasonably be issued.
How false alarms degrade the response
People who experience repeated warnings with no consequence become slower to act, and some stop acting entirely.
Response research consistently finds that the decision to take shelter involves seeking confirmation first, and prior experience shapes how long that seeking takes.
A warning system that gained ten minutes of lead time and lost the audience's willingness to respond would be worse than the one it replaced.
What is being pursued instead
Rather than extending the warning itself, forecasters have improved the messaging that precedes it, with outlooks days ahead and increasingly specific discussion as the event approaches.
The intention is that a warning arrives to a population already aware that the day carries risk, so the warning triggers action rather than assessment.
Narrowing the warned area also helps, since a smaller polygon means fewer people warned unnecessarily for each tornado.
What better modeling could change
Very high resolution models run over small areas can simulate individual storms and their rotation, and continually update using incoming radar data.
The aim is a forecast of storm-scale hazards over the next hour, which would shift the warning from detection of what exists to prediction of what will form.
The obstacle remains the same one that limits understanding of tornado formation: the deciding processes occur near the ground at scales that neither observations nor models capture well.