Wednesday, January 18, 2012

More snow in Seattle? This morning brings a new look...

A lot can change overnight.  Yesterday evening's model runs were increasingly pessimistic about how much snow would fall in the Seattle area.  However, this morning's projections are looking more on the optimistic sie for seeing 4-6 inches of snow in the Seattle area.  So what changed?

In my last blog post I talked about how the location of the surface low moving in off the coast would be critical in determining the amount and duration of the snow event in Seattle. Too far north and we'll rapidly transition to rain.  Too far south and we won't get enough moisture brought north to overcome the large amount of cold dry air at the surface over Seattle.  Yesterday, the models were bringing in the small surface low into the coast at or just south of the mouth of the Columbia River--this was a bit too far south to get the maximum snow in Seattle. We'd see some snow, but not much.

At least, that's what last night's 00Z models were saying.  But then something changed overnight--the surface low got closer to land.  And as the surface low gets closer to land, it enters our observation network--it gets closer to our offshore buoys and weather stations on the shore.  As such, we get a lot more data about the strength and position of the surface low the closer it gets to shore.  We can use that data to improve our model analysis of where the surface low might be.  This is done through a process called data assimilation.  Most of my research revolves around improving this very process.  So what did data assimilation tell us about the low overnight?

Below is a map from our experimental real-time ensemble Kalman filter system (a type of data assimilation system) here at the University of Washington.   This system uses the power of ensemble modeling to really give us a good idea of the amount of uncertainty in a particular forecast, then uses that uncertainty to help determine how observation will impact the model. This is an increment map of sea-level pressure from 6Z last night (10 PM).  You'll see the familiar sea-level pressure contours in black.  This is what the model system thought the sea-level pressure field was before we assimilated data.  The color shadings show the impact of assimilating all the data we have--nearly 10,000 observations of wind, temperature, pressure and moisture from that entire area--into the model system. Anywhere you see red, that means that the observations wanted to increase the sea-level pressure there.  Anywhere you see blue, that means the observations wanted to decrease the pressure there.
The observations wanted the model to decrease the pressure on the eastern end of the pressure trough extending out to the east from the large low center in the central Pacific.  This isn't too much of a surprise--we knew that a surface low was going to form there anyhow.  Now let's go ahead three hours to 9Z last night (1 AM).  Here's the adjustment map for that time:
Notice that by this point the model indeed formed a small surface low in the center of that pressure trough--it responded to those observations.  But now look at the location of this next pressure adjustment.  The observations want to decrease the pressure to the north of that developing surface low!  This is to say, the observations want to pull that developing low further north.  This kind of pattern continued all morning, with the observations wanting to pull the surface low slightly further north.

Now, our operational deterministic models that I show here don't include the data assimilation analyses from our experimental EnKF system.  However, the initial and boundary conditions for their 12Z runs this morning do come from a larger model (the GFS) that does have a lot of data assimilation included.  So, when looking at the 6-hour forecast of where our 4km WRF wants the surface low to come ashore this afternoon, we see this:
It has brought the surface low back up north a little bit, now coming ashore just north of the Columbia River.  Even though the change in the position of the low is subtle, it has huge ramifications.  This brings the warmer, moister air closer to Seattle and increases the amount of snow that could fall.  Here's this morning's model 24-hour snowfall projections ending at 12Z tomorrow morning:

Remember last night the models had backed off to maybe 1-2 inches in the Seattle area.  This morning's forecast shows 4-6 inches (with an interesting convergence-like band over the Kitsap Peninsula and Seattle...).  Quite the change overnight.

And, since this morning's 12Z run (which is based on data from 4 AM), our ensemble Kalman filter (EnKF) system has continued to run.  Here's the 15Z (7 AM) adjustment of sea-level pressure due to observations:
It shows the low analyzed off the mouth of the Columbia River (still offshore) but there is still a strong adjustment signal that wants to lower the pressure to the north of the low (and raise it a little to the south).  This points to the low continuing to be pulled northward.  Hopefully it doesn't get pulled too far northward--then we'll end up with a lot of rain tonight.

So, the forecast is still developing, but I have pretty high confidence that we'll see several inches in Seattle today.  Not only are these short-range forecasts, but now that the low is close to the coast our models can take advantage of all of our observations here and really nail down where the low is going to be.  I also already have two inches or so at my place, so yes--I have confidence.


  1. Very clearly described, thanks.

  2. Nice post!!! I need to become more familiar with all the non-operational models at the UW... very interesting!