The Friday Weather Discussion held on Halloween, 31 October 2014, was titled “A Trick or a Treat? A Discussion on the Predictability of the Impending Coastal Cyclone.” At the time of the discussion, a long-wave trough was found over the eastern portion of the United States, bringing colder air towards the Southeast. The forecast by all of the numerical models was for a shortwave trough embedded in the broader trough to move through it and provide the forcing for coastal cyclogenesis off of North Carolina which would strengthen given strong jet dynamics into a strong coastal low while moving north over the Atlantic just east of our region. The discussion was partitioned into a few key topics including (1) the history of the shortwave, (2) how well the models and some model-derived analysis tools performed with the developing system, (3) the sort of impacts that would result from this strong coastal low across the Northeast.
The History of the Shortwave.
An analysis of the potential temperature on the 2 potential vorticity (PV) unit surface from the ECMWF analysis showed the trough across the Eastern U.S. with lower potential temperature values indicating a depression in that PV surface. Embedded within that trough was a small area of even lower potential temperature values which was the shortwave that would round the base of the trough and help develop a strong coastal low following the succession of a couple weaker coastal lows due to the long-wave trough. This trough can be seen in the image below. Because potential vorticity is a conserved quantity, the lower values could be traced back in time and were found to originate as far around the Northern Hemisphere as Russia. Before interacting with the long-wave trough over the Eastern U.S. the shortwave was found over the data-sparse region of the Arctic. It was for this reason that it was considered a wild card in numerical weather prediction forecasts because if the models were initialized without knowing its full extent, how could they integrate forward in time to a correct solution of the strong coastal cyclone that could impact millions of people?
How well the models and some model-derived analysis tools performed with the developing system.
A quick qualitative discussion followed that looked into how well this event had been forecast. As far as the deterministic operational guidance, the GFS had been trending the position of the cyclone closer to the coast in the last few cycles. The NAM and GFS provided similar guidance with respect to the timing, location and intensity of the coastal cyclone in the next 24-48 hours. Beyond the deterministic models, a few products developed and maintained at Stony Brook University were discussed. The first was Dr. Edmund Chang’s wave packet analysis page. Beyond 5 days before the event, a strong wave packet, or envelope of energy linked to cyclogenesis events at the leading edge, was forecast within the GFS. Shown below is a snapshot of an operational product that shows the 4-day (96h) forecast from 0000 UTC 28 October of (upper panel) the upper-level meridional wind anomalies and height field which shows a very amplified pattern and the forecast Eastern U.S. trough and (lower panel) the wave packet envelope reaching the United States.
Another forecasting tool is ensemble sensitivity analysis developed and maintained by ITPA graduate student Minghua Zheng. Using ensemble model data, the mean and spread of a field such as sea level pressure can be calculated. From that it can be understood where the average of the data lie spatially but also where clusters of other possible solutions are. A statistical analysis can be performed that can deduce what the most important correlations or patterns are within the data and then trace back through the forecast times to the initialization time (regarded as true observations) the sensitive regions in a particular field that could ultimately result in changing an important feature of the pattern. For example, looking at the 4-panel plot below of the day 5 or 120 hour forecast from the NCEP ensemble models shows that within the Mean Sea Level Pressure (MSLP) field there is a mean cyclone off the coast (upper left panel) but the shading suggests that a group of members have a cyclone closer to the coast. The strongest correlated pattern (upper right panel) shows a dipole signature that is indicative of a southwest-northeast position difference in the models. The bottom two panels show other statistically significant patterns.
Using the first and most-strongly correlated pattern, EOF1, one can trace back sensitive regions from the Day 5 forecast time to the model initialization time and can see what features and evolution of those features could possibly result in a changed forecast. The plot below shows the model initialization time highlighting some sensitive regions for a Day 5 forecast. So on 28 October a forecaster could potentially look at the evolution with time of some features, like the shortwave originating from the Arctic, and note that the change in amplitude or speed of the shortwave would allow for the coastal low to be closer to the coast. It’s a fun exercise that a still picture doesn’t do justice, so visit this site to loop through in time.
The sort of impacts that would result from this strong coastal low across the Northeast.
After discussion of the predictability of the forecast of the development of a strong coastal cyclone over the weekend, it was then time to discuss what sorts of impacts could be experienced across the region. Any impacts at a location would be a function of the distance to the cyclone center so Mike Erickson’s extratropical cyclone tracking page was brought up which showed a cluster of ensemble members from different model families taking the cyclone up the coast outside of the 40N-70W benchmark. The first question was about the potential for snowfall. While the air mass within the trough was sufficiently cold, snow was unlikely to occur across the NYC metropolitan region mainly because the precipitation would have likely ended by the time it became cold enough. However, across New England and in the higher terrain in particular the likelihood of significant snow was possible, especially into Maine. The SREF spatial plots and plumes were discussed with some locations in the White Mountains projected to get as much as 12-20 inches of snow. The second impacts question was winds. With such a tight pressure gradient between the intensifying coastal storm and the high pressure building in from the west, high winds were expected. The SBU WRF operational model ensemble was used to see what regions would expect to see the highest winds. Looking at a plot of the 48-h forecast of surface winds valid 12 UTC Sunday shows the areas of highest winds extending over Cape Cod with values exceeding 25 m/s (55 mph) as shown below. While there are some differences between the NAM-WRF and the GFS-WRF, wind advisory criterion is still evident along the western flank of the system along the coast.
To summarize the Friday Weather Discussion, this high impact cool season East Coast cyclone event was interesting from a predictability perspective and provided an event with which to utilize various tools developed in-house ranging from those utilizing longer-range ensemble model data to mesoscale models to get the finer details closer to the event time. The shortwave, although spending time in a data-sparse region of the globe, was more of a treat than a trick for those studying the weather who could get easily spooked by model uncertainty. Happy Halloween!