Status: Ongoing analysis Material from: Linus, Ivan
1. Impact
In the middle of February 2021 a severe cold spell hit North America and snowy conditions reached as far as Texas. The cold spell started on 9 February and culminated on 15-16 February.
https://www.bbc.co.uk/news/world-us-canada-56095479
2. Description of the event
The time-series below shows the EFI for 2-metre temperature for the first forecast date as a proxy for the analysis. The cold event started on 9 February and lasted to 20 February.
The plots below show analysis of T850 (shade) and z500 (contour) from 12 to 17 February, every 24 hour.
3. Predictability
3.1 Data assimilation
3.2 HRES
The week before the main cold surge (8-14 February) temperatures started to drop. For several days a broad low stratocumulus deck expanded and persisted for several days over a hide area over the Great Plains and east of the Rockies. Several sites reported mist/drizzled/overcast conditions for few days in Texas and, Oklahoma. No indication of snow on the ground for most of Oklahoma, and Texas at the start of the period. Yet temperatures near surface were systematically warmer compared with the few sites checked. Temperature errors could be associated with cloud errors. The plot below show the VIS image at midday local time and the inset figure corresponds to a very short range forecast (H+18) simulated visible image. The cloud area is much reduced and "less thick"- pockets of open cells are visible in the HRES.
Checking the forecast errors of 2m temperatures over Texas it is clear that the onset of the cold was not well handled by the model. Forecasts ranges beyond three days had errors exceeding 15 degrees (model warmer).
3.3 ENS
Prediction of the onset of the event
The plots below show EFI for 3-day mean temperature valid 11 February. The outlined box is centered on Dallas, Texas.
The plot below shows the forecast evolution for 1-day temperature mean averaged on 11 February inside the box outlined in the plots above (1x1 degree around Dallas). The plot includes HRES (red dot), ENS control (purple dot), ENS distribution (blue boxes, median square and mean diamond) and model climate distribution (red boxes). The value from one station inside the box is plotted in green. While the forecast from 4-5 February predicted very cold temperatures, the forecasts from 6-7 February was much warmer.
The plots below show the forecasts from 5 February 00UTC (left) and 6 February (middle) and 8 February (right). The lines represents HRES (red), ENS control (purple), ENS members (blue) and observations (black).
The two panels below show the ENS forecasts expressed by the latitude (x-axis) when 2MT and T850 hPa dropped below freezing (for a fixed longitude near Fort Worth, Texas) as a function of the forecast lead time (y-axis, in days) for a same verifying time, ie, 9th February 2021 12Z (~06 Local time, prior to sunrise). The ENS systematically is under-spread and warmer near surface but not near the top of the boundary layer. Vertical solid represents the outcome latitude (analysis).
Prediction of the peak of the event
The plots below show EFI for 3-day mean temperature valid 14-16 February. The outlined box is centered on Dallas, Texas.
The plot below shows the forecast evolution for 3-day temperature mean averaged inside the box outlined in the plots above (1x1 degree around Dallas). The plot includes HRES (red dot), ENS control (purple dot), ENS distribution (blue boxes, median square and mean diamond) and model climate distribution (red boxes). The value from one station inside the box is plotted in green.
Observations show that cold air arrived in Dallas, TX from north-north-west between 00 and 12 UTC on 9th February. Cold air is estimated to extend in a 100-hPa deep layer from the surface.
Radiosonde measurements at Fort Worth, TX on 09@00 UTC, 09@12 UTC and 10@00 UTC (from left to right).
The model never managed to fully represent the vertical structure of the atmosphere in the cold air including low clouds in Dallas. The forecast shown below is initialised at 00 UTC on 8th February.
Forecast v. observed vertical profiles valid for 09@00 UTC, 09@12 UTC and 10@00 UTC (from left to right). Forecast initialised at 00 UTC on 8 February is displayed.
3.4 Monthly forecasts
The plots below show the weekly 2-metre temperature anomaly for the week starting 15 February from the extended-range forecast system.
3.5 Comparison with other centres
The plot below shows the forecast evolution for 1-day temperature mean averaged on 11 February inside the box outlined in the plots above (1x1 degree around Dallas). The plot includes observation (green),HRES (red dot), ENS control (purple dot), ENS distribution (blue boxes, median square and mean diamond) and model climate distribution (red boxes) plus the ensemble from DWD ICON model in evergreen and NCEP in light green.
The plot below shows the forecast evolution for 3-day temperature mean averaged inside the box outlined in the plots above (1x1 degree around Dallas). The plot includes observation (green), HRES (red dot), ENS control (purple dot), ENS distribution (blue boxes, median square and mean diamond) and model climate distribution (red boxes) plus the ensemble from DWD ICON model in evergreen.
4. Experience from general performance/other cases
5. Good and bad aspects of the forecasts for the event
- Poorly handled forecast for the onset of the cold, also in the short-range
- Good medium-range predictions of the most intense part of the cold wave