Model Uncertainty: Comparing This Week to the Last Few Storms

[phil882]

Hello All,

I just wanted to pass along an interesting blog post I've made highlighting the differences in the synoptic regime between the last few weeks with split flow vs. this week's single shortwave that will do all the heavy lifting over the midwest in the next few days. This was taking from my blog on my website, but I thought I'd recreate all of the information here so that way we can get a proper discussion going. Should be a fun forecasting week ahead!

This is originally from here http://www.atmos.albany.edu/student/ppapin/local3.html

Comparing Forecast Difficulty: February 24th Storm System vs. This Week's Storm System

I'll focus on the differences between two pattern regimes that have characterized the past couple of systems vs. this weeks upcoming system. There are suble but marked differences that will have a large impact on how well the models will be able to handle the upcoming pattern. Lets dive right in shall we?

Last Weeks Synoptic Setup February 19-25th, 2012

Figure 1. D(prog)/Dt map of 500 hPa heights (black contours), vorticity (shaded), and vertical velocity (blue contours) for the 00z February 24th period.

I'll start off this discussion by showing a d(prog)/dt animation of the GFS for the time verfiying at 00z on February 24th. Most probably remember this period being plagued by model inconsitency with large changes from run to run occuring even well into the short range forecast period. Part of the problem was related to the phasing of a southern stream shortwave (represented by the red circle in the animation above) and the northern stream shortwave (labeled as a red line). Note that the GFS was fully phasing these two pieces of energy as late at the 90 hour period. However for the next 4-5 model cycles, the GFS started seperating each piece of energy. By the time the forecast period was 60 hours out, the southern stream energy now was represented as a cutoff low located near Baja California, while the northern steam shortwave was markedly weaker and less amplified. Such major changes in the modeling within the famed 72 hour period are rare, but happen more frequently when in a split flow regeme where multiple pieces of energy phasing are in play. This was the same synoptic pattern that plagued the United States the week before which lead to great model uncertainty for the Mid-Atlatic snowstorm across Virginia. For the upcoming system this week, we likely won't have to deal with any southern stream shortwave interaction, which will already make this a markedly easier forecast period. The lack of phasing should also make it easier for the models to remain consistent over the next few days leading up to this potential event.

Another complicating factor witnessed last week was the ridge off the west coast of the United States. As shortwaves come off the apex of this ridge they are at their least amplified state. Models are only good at resolving features down to a certain size and when features become too linear it becomes very hard for the model to discriminate between what is suppose to be a weak impulse that has little impact on the forecast versus an impulse that will rapidly amplify on the downstream side of the ridge. However, if the impulse is strong enough to have decent amplitude while crossing the apex of the ridge, it should in theory be a lot better modeled out in time.

Figure 2. Compairison of 500 hPa heights (black contours), vorticity (shaded), and vertical velocity (blue contours) between 12z February 22nd and 12z February 26th Forecast from the GFS

Animated above is a compairison of the last shortwave that drove this past Friday's system that produced heavy snowfall across portions of the Midwest and New England with the next potential system that is just coming ashore across the Pacific Northwest. The point I want to drive home is notice how much more amplified this system already is in compairison to the system that struck last week. Global models always have difficulty handling the intensity of shortwaves rounding the apex of a ridge because the gradient is so tight the model is not able to properly resolve the intensity of the trough. However, we shouldn't see the same uncertaintly play out with this upcoming system because it is far more amplified with a weaker height gradient that should be better captured in the global models. We can see this decreased uncertainty in the GFS individual ensemble members, with a suprising amount of agreement for a 4.5-5 day forecast from the most recent 00z run. We see this agreement also stretch out beyond the NCEP models, with both the CMC (GGEM) and the ECMWF supporting the event for the Northeastern United States. Of course when it comes to specific details, the models naturally still have decent disagreements. The fact is though, it seems that the majority of the model guidence is supporting a specific synoptic regieme, and this stable solution seems to be the bi-product of a easier to forecast pattern. This is worth remembering as we move forward in time and we watch the models continues to shift into next week.

This Week's Synoptic Setup Supporting a Mid-Week Snowstorm

Figure 3. Animation of the 850 hPa heights (white contours), temperatures (shaded), and winds (white vectors) of the ECMWF for this upcoming week starting Wednesday

Animated above is the ECMWF's progression of features at the 850 hPa level. I think it really does a good job highlighting all of the features that will come into play for this upcoming system. First of all, we have a mid-latitude low pressure center that will be occluding across the midwestern part of the United States. It is at this point that the storm will peak in intensity and produce blizzard conditions over a large chunk of the upper mid-west. However, my bias is focusing towards the East Coast, so I'll refrain from going into to much detail here.

What I will dicuss is why the storm starts to move eastward rather than into the Canadian maritimes. In this particular case, there is an upper level ridge blocking the progress of this system at 72 hours. This is the result of an anti-cyclonic wave breaking episode spured by the combination of a digging upper level PV impulse off the west North American coastline and the deeping cyclone over the Midwest. Thus, our storm system is forced eastward across the Great Lakes and eventually into the Northeastern United States. However, the precipitation is likely to start across our region far before then.

Figure 4. 84 hour GFS forecast 250 hPa isotachs (shaded), Potential Vorticity (red contours), Irrotational wind (vectors), 500 hPa upward vertical velocity (blue contours), Precipitable Water (shaded grayscale)

The culprit is a large area of deformation that develops out in front of our synoptic low pressure system. The ECMWF shows this very well at 850 hPa on 12z Wednesday leading the precpitation moving into the northeast. The GFS also supports this general setup, with the Great Lakes region into the Northeast being located in both the right-entrance and left-exit region of two 70+ m/s jet streaks. Thus, its not surprising to see that the intial batch of precipitation moving into the area could be rather intense.

What is more suprising though is that despite strong warm air advection occuring at this time period, the 850 hPa temperatures hardly budge above 0C across the majority of New York State and locations eastward. This is where the damming high really starts to become entrenched. Starting Wednesday afternoon and stretching into Friday, this high will literally not budge and inch as it remains firmly locked in place due to the amplification of a low pressure system off into the Atlantic maritimes. In fact, below zero 850 hPa temperatures will start backbuilding westward as the 850 hPa low pressure over the Great Lakes starts to fall apart in the face of strong confluence. It is only towards the end of the forecast that we see some evidence of a coastal feature show up, but I'd rather keep the details sparce on this system at this time since it is unknown if there will enough energy at 500 hPa to support cyclogenesis offshore that might ramp up precipitation totals from New York City, NY to Boston, MA. Those details should become clearer as we move into the work week in the next couple of days. One thing looks increasingly likely though. This will likely be a long duration event, with several impules that will ride along the boundary established between the strong damming high pressure over Canada and the weakening cyclone to the west.

Figure 5. Initial outlook for this week's winter storm

Given all of this discussion, here is my intial outlook for the upcoming winter storm this week. At this point, the damming ridge should help to halt the progress of warm air advection along the New York/New Jersey border. Some of the cold air should be able to funnel down the Hudson river valley, keeping most of the southern finger of New York in the snow. Areas to the west though won't have as much damming, and thus it wouldn't be suprising to see locations such as Binhamton and portions of northern Pennsilvania switch over to a freezing rain / sleet mix at some point during this event. Again all of the details are very preliminary. In fact, given the strength of the high even at 850 hPa predicted, I wouldn't be surprised to see colder temperatures continue to seep southward as we progress forward in time closer to the event.

[baroclinic_instability]

Nice analysis once again Phil. I am happy to know others in the short term world of forecasting don't rely solely on model-casting.

Also, FWIW, the GFS actually handled the lack of phase w.r.t. the closed/small southern stream anomaly better than the ECMWF which continued to eject the southern anomaly ahead of the longwave trough for a number of forecast cycles after the GFS. Made a difference in the overall evolution of the current strong cyclone tracking through the maritimes. I am tempted to pin the midwest storm here, but I doubt anyone on the forums cares about weather across the northern plains.

[phil882]

Nice analysis once again Phil. I am happy to know others in the short term world of forecasting don't rely solely on model-casting.

Also, FWIW, the GFS actually handled the lack of phase w.r.t. the closed/small southern stream anomaly better than the ECMWF which continued to eject the southern anomaly ahead of the longwave trough for a number of forecast cycles after the GFS. Made a difference in the overall evolution of the current strong cyclone tracking through the maritimes. I am tempted to pin the midwest storm here, but I doubt anyone on the forums cares about weather across the northern plains.

I completely agree with you Jason. The ECWMF had difficulty depicting the cutoff low over the Baja several runs after the GFS locked it in (which you can see in figure 1 happened right around 60 hours out). The run to run shifts up until then were laughable and its incredible that we saw the s/w feature shift from Louisiana to Baja California in a 5 run stretch that took place sub 90 hours out. Its been a while since I've seen the models make such monumental shifts at that range, but it fit the overall pattern this winter that the northern stream is simply moving too fast for these southern stream s/w to interact.

I think the run to run differences in the midwestern low this week are quite a bit more manageable. Of course you have a shift here and there, but there have not really been any crazy swings like we saw with the mid-latitude surface low last week. More evolution rather than revolution.

And for us folks downstream, we better care about what happens in the northern plains! We have yet to have a 6"+ snowstorm here in Albany this winter, and many here are hoping this is the system that finally breaks the drought. Part of the reason why the low stalls and then starts to shift slowly east has a lot to do with the blocking upper level ridge to its north due to anticyclonic wave breaking occurring to the north of our system. Now I think thats the one event that might be more difficult to forecast this week, because its dependent on the amplification of the trough over the Gulf of Alaska... an event which looks really complicated looking at some forecast DT maps and has high uncertainty looking at the GEFS guidence. I wouldn't be surprised if that ends up being a source of a lot of the major model uncertainty beyond 72 hours.

[CoastalWx]

That's a great analysis Phil. Well thought up.

[LakeEffectKing]

Very nice Phil. If I may add, in addition to the split flow, the speed differences of the different jets from which the impulses were/are embedded in, also contribute to the overall degree of error in model ouput, given the same leads. Note the GPH contouring in your two examples. Faster (especially out over the Pac.) vs. slower will (over the long run) show increases in error in all models, wrt s/w feature(s) placement(s)/interaction(s).

Edit:

Actually, I see you touched on the above aspect a bit (after re-reading) by mentioning the tighter gradient. My apologies!

Again, well written.

[phil882]

That's a great analysis Phil. Well thought up.

Very nice Phil. If I may add, in addition to the split flow, the speed differences of the different jets from which the impulses were/are embedded in, also contribute to the overall degree of error in model ouput, given the same leads. Note the GPH contouring in your two examples. Faster (especially out over the Pac.) vs. slower will (over the long run) show increases in error in all models, wrt s/w feature(s) placement(s)/interaction(s).

Edit:

Actually, I see you touched on the above aspect a bit (after re-reading) by mentioning the tighter gradient. My apologies!

Again, well written.

Thanks for the kind works guys! This should be a fun storm to track this week with hopefully less surprises

[Ginx snewx]

Fantastic post, thank you. Great read.

[phil882]

Quick Update: While there have been some minor shifts here and there... I think this has been an excellent case where the forecast uncertainty was significantly lower than normal, especially in comparison to the last few events. There was pretty good agreement 120 hours out that there was going to be a major snowstorm across the midwest (although the track shifted northward somewhat) as well as a good probability for a snowstorm for the Northeast as well.

Again I'll show another d(prog)/dt map to highlight the reduced uncertainty compared the last week.

http://www.atmos.albany.edu/student/kgriffin/maps/dprog/F018/500vort/namer/500vort_namer_dprog.html

From about 138 hours in... its remarkable to see how consistent the GFS was with the orientation and eventual cutoff 500 hPa low that developed across the midwestern United States. Sure the resultant 500 hPa feature was a little bit weaker than expected, but this is still pretty good for a 5.5 day forecast with nice run to run continuity. This is light years better than last week, and I am attributing to both the fact that we didn't have as many pieces of energy phasing in, as well as a better amplitude feature that was easier for the global models to identify without error.