And that results in this evolution in the model. It doesn't quite capture the extreme deviant motion of the observed storm or the longevity, but the storm makes UH. (Its motion is also pretty darn close to the environmental Bunkers motion estimate: 280/21 kts.) 3/6

Compare to the previous (22Z) cycle, which warmed and moistened the column a lot less during the DA process. The 22Z is the first cycle where the DA is trying to add precip, so it's basically starting from dry. 4/6

The 22Z run produces this forecast. Cell falls apart in the first hour. 5/6

The practical upshot of this is that in environments where the HRRR is not pre-disposed to produce convection (like in central Texas this afternoon), the DA has to produce the storm, which nearly always takes a couple cycles to catch on. This can appear as a late bias in CI. 6/6