The problem is that we are dealing with a rapidly amplifying trough and its interaction with the topography of the western U.S. It is a situation that is highly sensitive to the evolution of the large-scale flow not only in our region, but also upstream over the Pacific Ocean. As a result, predictability is relatively low and we've seen a wide range of possible outcomes in our forecast ensembles.
Over the past few days, the models have converged on a more amplified solution. For example, the GFS forecast initialized at 0000 UTC 31 Oct and valid 5 PM this afternoon (0000 UTC 3 Nov) called for broad southwesterly flow with frontal precipitation over northern Utah.
In contrast, this morning's GFS has a more amplified pattern, with Utah in the warm southerly pre-frontal flow this afternoon.
This is a solution in which we're looking at warm, windy conditions today and tonight with precipitation holding off until tomorrow.
You might wonder why we could not anticipate such an amplified solution a few days ago. Heck, some of the ensemble members were calling for precipitation to come in later rather than earlier, surely we could have figured out such a solution would verify. Nada. When you roll two dice, you know the odds of rolling a 7 (or any other combination from two to twelve), but you can't say what combination you are going to roll. Casinos make money by banking on the long-term odds. Gamblers lose money by deluding themselves into believing that they can somehow determine the outcome (keep blowing on those dice).
In my profession, we are moving toward a future in which we, like the casino, have a good idea of the odds. We're not there yet. We have ensembles, but they have biases and don't always generate reasonable probabilities of weather outcomes. We are, however, making steady progress, and we really need to make better use of them in our forecasts.