Status:Finalized Material from: Ervin, Linus, Mohamed
Discussed in the following Daily reports:
1. Impact
On 27 January 2015 a blizzard hit the northern part of the U.S east coast. The worst affected areas were in a band from Long Island towards Boston and further north. The storm was expected to also hit New Jersey and New York City and strong actions were taken before the event (closing motorways etc.). However, this part only got a little snow and NWS was criticised. The model that gave the strongest indication for severe snow over NYC was ECMWF.
The case was discussed in a ECMWF Newsletter article.
2. Description of the event
The snow storm on 27 January was caused by a cyclone that rapidly formed after 26 Jan 12z outside U.S east coast. The cyclone formed on the eastern side of an upper-level trough. The trough had moved eastward and a part of it originated from a trough that almost reached the tropics 2 days before west of Mexico.
The plot below shows NEXRAD 24 hourly precipitation product which is valid for 00-00 yesterday and the NWS snow total map from snow fall reports (produced from a lot of observations in the order of a hundred). The nexrad is in mm and the snow is in inch. The maximum precip was 20-35 mm in the east of the area, around 7-10 mm in NYC. The snow however was up to over 30 inches (around 36 inches as the highest observation) and about 5-10 in NYC. Most of the precip was between 00z-00z, at least in the southern part of the affected areas anyway, so the nexrad map is more or less relevant estimation. The few observations we have in the GTS (for the whole 36 hour period) are in line with this nexrad map.
The below below shows our NEXRAD verification product with the last HRES forecast included (27 Jan 00z).
3. Predictability
3.1 Data assimilation
The plot below shows the MSLP increment for 26 Jan 18z.
The plots below show the z1000 increments for 26 Jan 18z for LWDA (left) and ELDA (right). In this cycle the DA moved the large-scale (within the cyclone developed) further east.
Observation statistics for 26 Jan 00z for radiosonde mid-troposphere temperatures are shown in the plot below. To stations (72215 and 72230) show large positive observation departures (model too cold), both in respect to first guess and analysis. At this time, the trough was located above the two stations.
The plots below show the time-series of the two stations. Both stations show similar departures on 26 January, although the magnitude of the departures are not extreme.
3.2 HRES
The plots below show the 24-hour precipitation from HRES valid 27 January. Before 24 January it was little trace of the system, although there was precipitation further out on the Atlantic. All forecast but the last had too much precipitation over New York city (marked with an hourglass symbol).
HRES 24-hour total precip. 27 Jan
3.3 ENS
The two series of plot below show the cyclone feature plots for valid 27 Jan 12z. The spread decreased significantly around 24 January.
Cyclone feature plots for 27 Jan 12z max snowfall
EFI snowfall 27 Jan
The plot below shows the CDF for 24-hour snowfall for NYC for 27 January. The last (27 Jan 00z, red) and second last (26 Jan 12z, purple) differs considerably.
The plot below shows the CDF for 24-hour snowfall for Boston for 27 January. Here the last forecasts are much more consistent.
The plots below show the position of the cyclone in each ensemble member at 27 Jan 12z (symbols). The line starts 12 hours before and ends 12 hours later. Grey lines are for ensemble members, green for HRES (currently missing in the right plot), and analysis (red). The plots are from 27 Jan 00z (left), 26 Jan 12z (middle) and 26 Jan 00z (right). The westward shift is apparent in the two earlier forecasts.
3.4 Monthly forecasts
The plots below show monthly forecasts for 2-metre temperature anomalies valid for the week 26 Jan - 2 Feb.
2-metre temperature anomalies
3.5 Comparison with other centres
The plots below show comparison between ECMWF, UKMO and NCEP HRES for MSLP and precipitation. It is quite clear that the last few runs, apart from the last (27.00) UKMO and also NCEP were better, especially the MetOffice model. The UKMO (and with a little bit less clarity NCEP) MSLP fields are very similar to the analysis and also very consistent. However, the EC-HRES had the eastward shift already discussed yesterday. The earlier runs (2412 2500 and 2512) have the same feature, UKMO and NCEP had the cyclone more to the east.
The 2300 2312 and 2400 runs are a mixed bag, with I think UKMO being the best. We had a good forecast only from the 2400 run, while the metoffice model was quite good already 24 hours before and also better 12 hours earlier. However, their 2400 run was worse.
4. Experience from general performance/other cases
201402 - Snowstorm - US east-coast
201401 - Snowstorm - North-eastern U.S.
5. Good and bad aspects of the forecasts for the event
- Strong signal first appeared 24 Jan 00z, 3 days before the event
- Precipitation too far west in the short-range forecasts
- Too slow movement of the cyclone - or a secondary low forming in the forecast?
6. Additional material
- News item on the external web