...
3.1 Data assimilation
3.2 HRES
Looking into the "flip-flop" of the predicted precipitation for Aalborg (Denmark) on 1 Dec (e.g in the forecast evolution plot below), we see in the list below that the forecast from 29 Nov 00UTC similar to the shortest forecast, while 30 Nov 00UTC had much lower amount. We have therefore compared the errors in z850 for these two initial times. For the forecasts valid 30 Nov 12UTC (step 36h and 12h respectively), we find larger errors in general in the earlier forecast, but the later forecast had a positive error in the weak trough west of Ireland. This error later affected the cyclone over Denmark on 1 Dec 12UTC, and made the system weaker. It should be noted that the earlier forecast also had a error related to the cyclone but more like a phase shift. The correction of the wave west of Ireland is detectable in the observation statistics from surface pressure observations in the area, where the first guess had a too weak trough on 30 December. The assimilation fitted the observations well, and the observations were beneficial in terms of FSOI (not shown).
Precipitation forecasts for Aalborg valid 1 December (24h):
2021-12-01 00:00:00,32.3028564453125
2021-11-30 12:00:00,31.48651123046875
2021-11-30 00:00:00,16.02935791015625
2021-11-29 12:00:00,15.2740478515625
2021-11-29 00:00:00,33.538818359375
Forecast (black), analysis (red) and error (shade) for z850 valid 30 Nov 12UTC in forecast from 29 Nov 00UTC (left) and 30 Nov 00UTC (right).
Same as above but valid 1 Dec 12UTC.
Observation statistics for surface pressure observations in 50-55N, 25-15W.
The plots below show forecast for snow depth in Aalborg (Denmark) and Lund Sweden) from 1, 3 and 5 December 00UTC, and nearby observations as markers.
In the forecast for Lund from 1 Dec 00UTC, the model accumulated 5 cm of snow (10 mm of water eq.) on 1-2 December and the depth was 4 cm after 5 days into the forecast. In reality, 2 cm was measured on 2 December 06UTC and it melted later that day. This error is probably related to a too slow melting of snow on the ground during the period of rain in the middle of 1 December. As we did not assimilate the observation from Lund, also the forecast from 3 December was affected, and had 7 cm of snow while the ground was snow free at that point. In the forecast from 5 December this was probably corrected by the satellite product.
3.3 ENS
The plots below show EFI and SOT for 1-day snow fall valid for 1 December. On northern Jutland the SOT values reached 5, which is very high. In the forecast from 30 December the snowfall was weaker on the northern side of the band.
...
5. Good and bad aspects of the forecasts for the event
- Reasonable medium-range signal
- Decrease in signal for northern Jutland and parts of Sweden the day before the event, and later flip back
- Overestimation of total precipitation in Aalborg or undet-catch by observation?
- Too high snow density for fresh wet snow?