Shortly after 4 PM today, a thunderstorm moved over the University of Utah, providing a nice downpour of rain with some embedded small hail just to add insult to injury for anyone walking around campus.
Could someone have issued a nowcast with 30 minute lead time that this storm was going to hit campus?
The loop below shows that the storm blew up rapidly in the central Salt Lake Valley and moved northeast over the University of Utah in only 15 min. The first echoes produced by the developing cell were observed at 2149 UTC (3:49 PM MDT), with rain beginning on campus at about 2205 UTC (4:05 PM MDT).
There is no human or computer-based forecast tool available today that can reliably forecast the development of a storm cell from nothing on such short time scales. Very-short-range forecasting, known as nowcasting, is based primarily on the linear extrapolation of the motion of existing radar echoes (there are some exceptions, but for the most part, nowcasting is based on linear thinking). It's extremely difficult to anticipate the intensification or decay of storm cells in such short time periods.
What about computer models? Not at the present time. First you need a really big computer, bigger than anything used for operational forecasting today, because you need a lot of resolution. Then, you need to ability to rapidly assimilate radar and other observations into your model. This is also extremely difficult and we are in the early stages of figuring out how to do this effectively.
Essentially, developing computer models that can provide detailed and reliable forecasts of convective storms at short lead times represents the holy grail of nowcasting. This is an area of active research in the atmospheric sciences and perhaps we will make some inroads in the coming years and decades, if not for storms like the one above, perhaps for severe convective phenomena like derechos and supercells. For a look, click here.