Thursday, June 30, 2011

Postmortem Re Yesterday

First, some old history. When I was an operational forecaster (many years ago), the practice of doing postmortems when the forecast went awry was common. A forecaster might even be assigned to do a quasi-formal, case study of  a missed event (I know that I certainly did quite a few of these). Thus, I am a firm believer in the postmortem as a learning tool, that also helps build experience levels. I think that in operational forecasting today the postmortems can often (usually?) be summed up as: "The models screwed up."
While I didn't state it explicitly, I thought that reasonable POPS for the area around the house were about 20-30% yesterday afternoon and evening, because of the very poor steering-level winds for this location, and a marginal looking morning TWC sounding. The WRF model was, however, very bullish on rainfall for low elevations. The NAM was not as aggressive, seeming to keep precipitation more toward higher elevations.
Finally, at 8 pm, as in earlier post, I thought that the chance for rain had passed for metro Tucson. So, this post is about what I think may have gone differently than what I expected.

Shown above are top: WRF-GFS forecasted 3 pm composite radar echoes from the midnight initialization; middle: same thing from run initialized at 1200 UTC; and bottom: the 1200 UTC run's accumulated rainfall at 9 pm yesterday evening. Both forecasts were reasonably good from the Rincons east (i.e., at higher elevations). But, the forecasts were not good for the Tucson metro region, nor for the lower elevations of the Santa Cruz Valley.
Mike Leuthold is away on vacation, and I don't know if anyone is adjusting the PW in the model initialization - or, if he has this procedure automated. But I suspect that the initial low-level moisture fields (for both runs) were not as good as they needed to be. Further, again not mentioned in my posts yesterday, I noticed that the runs had early morning convection in our area - the convection moved rapidly northward and dissipated. But, this early error in the WRF forecasts may have impacted the rest of the forecast runs.


I grabbed the forecasted 1 pm TWC sounding from the 1200 UTC run of the WRF-GFS and it is at top (just above). The model had forecasted PW into the range of 37 to 40 mm and there was substantial CAPE indicated. The model seemed to forecast enough CAPE that it initiated convection at low elevations primarily due to surface heating (note I don't have enough information to be sure of this). The model forecasted BL at 1 pm is about 100 mb too shallow, and considerably too moist - thus, leading to the forecast of low-elevation, strong thunderstorms. As far as I can tell, midafternoon PW was around 30 to 32 mm - about 5 to 8 mm lower than forecasted by the model.
The 0000 UTC sounding taken at TWC yesterday afternoon is shown in the bottom graphic (from the Univ. of Wyoming upper-air page). The sounding analysis there indicates NO CAPE and a PW of 24 mm. So, we know that the WRF forecast sounding was too moist and, from GPS data, we know that the RRS TWC sounding was about 6 to 8 mm of PW too dry. An ugly mess! There was obviously some CAPE present in the BL over the metro area, that lingered through the night within an elevated mixed layer. This CAPE was released when the outflow from Sonora converged into the Tucson region around around 3:30 am.
Thus, my expectations were too high for the afternoon and evening and too low for the late nighttime. Yesterday's situation was confounded by the bad RRS sounding data, which have to be corrected by guess at the forecast office, using a sounding correction program (unfortunately I don't have access to such software and make subjective guesses through the summer to try to infer what the soundings should have shown).

No comments:

Post a Comment