Forecast Bust: Verification & Lessons Learned
When we looked at the 12Z MOS runs on March 1st, we saw some difference in the forecasted low temperature. The NAM showed a low of 33F, but a lower temperature at 06Z at the end of the period (31F). The GFS model showed a low of 28F, with a temperature of 31F at 06Z at the end of the period.
 |
| 12Z MOS runs from March 1st, 2012. Courtesy of Texas A&M |
Looking at the wind during the early part of the period showed some disagreement between models. However, we learned quickly that the MOS wind forecasts required significant support and were unreliable at face value. Still, both models showed winds building by 18Z and fairly strong diurnal winds in the previous period. So the question was how much they would lay down during the overnight period. Looking back at the synoptic indicators, it was clear with a weak surface gradient that they would likely be weak during this time frame.
Looking at the cloud cover during this period, it seemed to suggest the presence of some scattered - broken high clouds. Recalling the tongue of higher RH values at 700mb, these clouds still concerned me. With the NAM trending warmer than the GFS at synoptic times, I was focused on the cloud cover. At the end of the period, both suggest a clearing trend, which would make sense in the wake of the cold front and high pressure building in behind it.
The forecast soundings below from NAM and GFS showed again some differences.
 |
| 18Z NAM Forecast Sounding from 06Z March 2 - 06Z March 3. Courtesy of Air Resources Lab, NOAA |
The NAM forecast sounding indicates clouds at 650mb at the start of the period. These clear, then return again by 15Z. The GFS forecast sounding (below) does not indicate any cloud cover during the period, with much larger dew point depressions than the NAM indicated. I think I dismissed the clearing interval indicated by the NAM without much thought. Again, I expected high clouds associated with the vort-max to invade the area at the start of the period.
 | |
| 18Z GFS Forecast Sounding from 06Z March 2 - 06Z March 3. Courtesy of Air Resources Lab, NOAA |
The forecast soundings both indicated a slight nocturnal inversion at the start of the period around 12Z. However, I found the forecasted inversion to be weak, with only a degree or two difference from the surface to the top of the inversion. The image below shows the forecast for 12Z, with just a slight inversion at the surface.
 |
| 18Z NAM Forecast Sounding valid at 12Z March 2. Courtesy of Air Resources Lab, NOAA |
As it turns out, the sounding for 12Z at KABQ showed a much stronger inversion, as you would expect with a low of 25F.
 |
| Actual sounding for 12Z at KABQ showing a moderate nocturnal inversion, light surface winds, and few clouds if any. |
To give you the full picture of how March 2nd turned out, here is the meteogram for the entire forecast period.
 |
| Meteogram for KABQ between 06Z March 2 and 06Z March 3. Courtesy of University of Wyoming. |
As the NAM forecast sounding predicted, broken and scattered clouds did occur at 06Z and they quickly cleared. Meanwhile, the winds, which had been strong the day prior, laid down completely with a weak gradient in place over the state. These two factors allowed the surface temperature to plummet, and with a TD below 20F, not much would stop it.
Methodology
I've discussed my methodology a bit thus far, but it's worth reviewing. Having prepared the wind forecast for the class, I did expect the light overnight winds, although I'm not sure I would have forecasted even two hours of calm winds. My primary focus was on cloud cover and how that might modify the temperatures overnight. The NAM suggested more clouds and I assumed, higher overnight temperatures as a result. This seemed to be supported by the 700mb forecast for higher RH values invading the area. So for the start of the period I thought a blend of the MOS forecasts was appropriate, staying on the cold side of the range as the clouds should stay high.
Then there is the question for the end of the period. Both MOS forecasts showed a temperature of 31F at 06Z Saturday. I thought with the clear skies, falling heights, a potentially colder high temperature during the day Friday, and weakening winds, there was a good chance these forecasts were slightly warm. I argued that nudging this temperature down a couple degrees was justified, given those parameters. Remember, though, that 06Z is just 11PM, and only 5 hours beyond dusk.
So, as I saw it, the low temperature at the start of the period would be close, if not the same, at the end of the period. In the end, though, the clouds are zeroed in on, never materialized and my 29F forecast low was 4 degrees too warm. Further, my expectation to see 29F again at 06Z Saturday was dashed as it was 31F, just as both models forecasted.
Lessons Learned
The greatest lesson learned from this instance, is to always connect the forecast to the big picture and to study the details of the big picture. It's easy to gloss over a 500mb prog once you get an idea of the strongest vort-maxes. In this case, the split between the two vort maxes at the bottom of the upper trough required more investigation. Only then, would I have seen the S/W ridge and at least questioned its effect.
The next lesson is to not dismiss inconsistencies. When the NAM forecast sounding showed clouds at 650mb at 06Z and then no clouds, that should have been a clue. Instead, I dismissed this without much thought, relying on the 700 mb forecast progs and higher RH values.
In the future, I'll look more closely at the UA forecasts and the placement of subtle shortwaves. Then question their effect and look for supporting evidence. Overall, the forecast submitted was a good forecast that could have been great if not for the bust with the low temperature.
No comments:
Post a Comment