Known Knowns, Known Unknowns, and Unknown Unknowns

In this post, I will be channeling Donald Rumsfeld, who once famously said.

"There are known knowns. There are things we know that we know. There are known unknowns. That is to say, there are things that we now know we don't know. But there are also unknown unknowns. There are things we do not know we don't know."

Rumsfeld was lampooned for that quote, which many considered to be nonsense, but I understand exactly what he meant.

Known Knowns

Our computer models have gotten pretty good at predicting large-scale airmass changes.  That's why the massive drop in temperature from yesterday (record high of 73ºF and the latest 73ºF on record at the Salt Lake City airport) to tomorrow morning (low 30s) is a known known.  Think of this period as a time machine in which you are transported from early Fall to late Fall in two days.  We are in transit today as cooler air is already moving in aloft and the surface front, marking the leading edge of even cooler air, is scheduled to arrive this afternoon.

Known Unknowns

Difficulty predicting the timing and amount of precipitation, especially snowfall, are known unknowns.   Adding insult to injury is the fact that we are dealing with a front that does not have a well developed precipitation band this morning, as illustrated by the 1400 UTC (7 AM MST) analysis below.

Some models are, however, bullish on precipitation developing along the front.  Others are bullish on precipitation behind the front.  This leads to significant differences in the timing and amount of precipitation tonight and tomorrow morning.

From the operational models, we see water equivalent totals for the Salt Lake City airport ranging from only 0.1" (0000 UTC GFS) to .26" (0600 UTC NAM).  All models produce some frontal precip (step part of the curves, lags the actual surface front by a few hours), but of differing amounts.  Only couple produce a burst of post-frontal precipitation (increases around 1200 UTC give or take).  The National Weather Service forecast (green line) is the wettest of all (.32"), primarily because they call for post-frontal (and presumbaly lake-effect) precipitation to continue late tonight and tomorrow morning.  Most of the models below do not realistically simulate such precipitation.

For experimental models, we can look at the NCAR ensemble, which has the advantage of being much higher in resolution (3 km).  We'll also shift our attention to Alta as I know just a few of you have an interest in that.  Again, we see dramatic differences in precipitation amount produced during the frontal passage (from about 00-06 UTC) and then during the post-frontal period (after 0600 UTC).  Totals range from .3 to 1 inch of water equivalent.

The final known unknown is uncertainty in the snow-to-liquid ratio for the storm.  Drier snow yields larger accumulations per unit of water.  I expect the snow in the mountains from this event to be at or below average water content.  In the valley, accumulations will be limited by snow densification as it melts on the warm ground (accumulations will be greatest on cold surfaces).  However, there is uncertainty in predictions of snow-to-liquid ratio, and this is an additional source of error in forecasts.

Unknown Unknowns

I have spent much of my career studying lake effect so it pains me to say this, but we really can't predict the location and amount of lake effect worth a damn.  We have gotten better predicting the likelihood of lake effect (will it happen or not) and we can look at why lake-effect is likely tonight using the graphs below.  We have, for example, a forecast for a remarkably cold airmass moving over a remarkably warm lake (illustrated by top two left hand panels), a moist low-level airmass (lower left panel), and NW-N flow (top right).  Based on training to past events, this gives us a probability that lake-effect develops (note: this just means that we get some, not that we get a lot) that peaks at about 90% tomorrow morning.

That's all fine and dandy, but we still struggle with these forecasts for a number of reasons.  One is that the models aren't perfect.  They have errors in wind direction and humidity.  Another is that lake effect comes in many flavors.  It can be banded (intense and localized) or non-banded (weaker, but broader in coverage).  What causes those differences in flavor are basically unknown unknowns.  We don't know why some events feature banding and others don't.  Finally, despite the fact that all of the large-scale factors appear to add up to lake effect, we still have events where it doesn't happen (it's not as simple as cold northerly or northwesterly flow moving over the lake as that happens all the time with no lake effect).  Again, this is an additional unknown unknown.

Putting it all together

I think this remains a forecast with a broad range of possibilities.  I'm sticking with my forecast of 1-2 inches of snow for the benches (cold surface accumulation) and 5-10 inches of snow for Alta as I those ranges represent the most likely outcomes.  Could we get skunked and come in below those numbers.  Yes, if the front is less productive than advertised and the post-frontal doesn't come through.  Could we do better?  Yes, if the post-frontal showers late tonight and early tomorrow, possibly with lake effect, get going.

One thing to note about the lake effect is that the forecast flow for tomorrow morning is somewhat sheared, with the wind veering (turning clockwise) with height from northwesterly to near northerly.   As a result, in the NCAR ensemble, the probability of more than 1" of water equivalent is highest in the Oquirrh Mountains, maxing at 60%.  In the central Wasatch odds are generally around 20%.  

Clearly the best option is to keep expectations low and hope for the low probability, heavy precipitation outcome to verify in the central Wasatch.