Model Verification for Sub-Tropical Rain Event over SoCal
For those unaware, a cut-off low has made the last few days pretty wet across most of Southern California. While this is certainly not an extreme or particularly unusual rain event, it has been significant for how severely the models failed at predicting the rainfall over the region. Models typically have a tough time with cut-off low events, but here we will see just how poorly they performed a 24hr forecast.
Overview of the Event
A low cut off from the westerlies several days ago to the NE of Hawaii. This low slowly made its way eastward over the next few days. On the 23rd (two days ago) the cut off low was loosely picked up again by troughing associated with the main branch of the jetstream (effectively making the low no longer cut-off). The tail of the trough still extended deep into the tropics, however, and additionally, the shortwave to the north cut off from the flow to join up with the base of the trough.
Here's an animation of what was going on at the 500mb level from the 20th to the 25th based off GFS intializations.
Now about that 24hr forecast...
Below I have included the total precipitation forecast through 00z Jan 25th based off model intializations on January 24rd. In other words, this is a 24hr precipitation forecast verification. Models should have a good handle on this, right?
ALL Models Intialized on 00z 1/24. Total Precipitation is shown through 00z 1/25.
CMC showing a 1/4 inch or less across SoCal
GFS showing 1/2 an inch or less across most of SoCal
NOGAPS showing a nice bloch of 1/2 to 1 inch of rain
Last 3 HPC 1/24-1/25 24hr Precipitation Forecast showing little to no precipitation across SoCal
Here's what actually happened
California/Nevada River Forecast Center Precipitation Total (in inches) Ending 00z Jan 25th. 1/2 an inch or more across the Los Angeles basin and Orange county. Coastal areas of Santa Barbara and Ventura received around 1-1.5 inches, with a few areas in the coastal slopes exceeding 2 inches.
So why did the models fail so hard with a 24hr precipitation forecast?
Models have a tough time with cut-off lows (and cut-off highs) because these are very sensitive synoptic situations. To add to forecast challenge, the cut-off low was positioned over the middle of the ocean where observations are few and far in between. The lack of quality observations, combined with the sensitive synoptic set up led to disastrous model performance, considering this was a 24hr forecast.
As far as which model did best, it would appear that the NOGAPS (lol) did the best, followed by the GFS and CMC. Even more significant was that HPC guidance performed the worst. For this forecast, their 2 and 3 day lead forecast did not show a drop of rain over Southern California. Their 1 day lead 24hr forecast did finally show rain over Southern California, but the actual precipitation that verified exceeded their forecast by over 1300% (max precip forecasted: 0.15in, max precip verified: 2+in). For a 24hr forecast from the Hydrometeorological Center, this is unacceptable.
Here's to better forecasting...