Friday, May 08, 2009

Anatomy of a Well Forecast Bow Echo

Above is an example of one of the forecasts from the Spring Experiment models from Friday. This bow echo moved across southwest Missouri early Friday morning and these images are centered on Joplin, MO (JLN). On the left is the 13h forecast from the WRF-NMM 4km model initialized at 00Z 08-May-2009 and valid at 13Z. On the right is the verifying 1km base reflectivity image with the model fields for winds overlaid on the radar. The barbs in each of the images are the model's instantaneous 10m winds in knots (with the grid skipped to lessen the clutter). The isotachs are plotted from the WRF "history variables" for maximum U,V 10m winds (no grid skip). These are the maximum 10m wind speeds in the model over the past hour ending at 13Z.

Instantaneous 10m winds in the model at 13z, near the rotating bow head, are at least 50 knots. The maximum model 10m winds over the past hour range from 60-70 knots near and north of the weak echo channel and around the comma-head of the bow.

This was only one of several exceptional forecasts of this feature from the models being evaluated in this year's Spring Experiment. To see more output on this case and more, check out the Spring Program website here:


WRF/CAPS Reflectivity Loops on V2 Domain

Plots of WRF/CAPS 1km AGL simulated reflectivity are now available using "V2" as the centerpoint in the URL. The following V2 domain image combinations should work as of today's 00z and 12z model runs:

At 00Z:

At 12Z, you can compare prior runs:
NMM4 / NMM4_12 / AFWA4 / AFWA4_12
CAPS1 / NCAR3 / NCAR3_12 / NMM4_12

Mix and match the models to build the loop you need based on the URLs, above. You can also add &date=YYYYmmdd to review another date.

The 00Z NCAR run goes out to forecast hour 48 while the other 00Z runs go to forecast hour 36. The loop URLs, above have a start hour of 12z go out 24h. This can be adjusted by using the &starthr and &frames variables in the URL.

The 12Z NMM goes to forecast hour 36 while the NCAR and AFWA versions of the WRF only go out 24h. These loops also start at 12Z and go out 24h. Note, some imagery will drop out when looping beyond the available forecasts from that model. Imagery may also be uavailable if model data fails to arrive prior to the execution of the cron job for the image generation script. These loop pages should not to be considered operational and will not always be available.

Thanks to Ryan for building such an adaptable web-based interface that allows us to build these thumbnail image loops!

Thursday, May 07, 2009

Th 7 May 2009 Noteworthy aspects

Centerpoint for tda's activity is KGMJ in extreme ne OK. With the last 2 days activity being farther E, this area was quiet yesterday, so no significant preexisting convection was present to complicate matters. The exception to this was some midlevel convection that at 12Z today was over ern KS into MO.

Early on we identified at the surface 2 weak transition zones (these were too diffuse to call them "boundaries". The first of these was along the Red River, separating shallow cool, moist air over OK (partly the consequence of the persistent cloud cover and pcpn the previous 2d) from more humid air (dew pt > 21C) over n TX. The second was a weak e-w zone of confluence marking a weak sfc front from roughly IL, srn IA across much of NE. The parent sfc low with this feature was over srn MB this mrng, if memory serves correctly.

Aloft, strong zonal flow crosses the W coast and the wrn 2/3 of the CONUS. This is providing a shear environment plenty adequate to support supercells. The temperature contrast up and down the W coast of the CONUS at 12Z this mrng was noteworthy: +13C at KNKX (San Diego) to -11C at KUIL (Quilliute). Only weak disturbances are present in this flow, revealed by 6.7 micron water-vapor imagery primarily.
There aprs to be a weak upper lvl PV perturbation in the flow at 12z this mrng centered nrn NV, srn ID and aligned E-W. Associated with this is enhanced N to S temp gradient across NV at 700mb and evidence (from METARs) of a weak surge of slightly cooler surface air heading ewd along I-80 in srn WY. This surge can be argued to have some relationship to the cooler air that this mrng was over the Nrn high plains behind the aformentioned sfc front.

So, we had in place early this mrng warming conditions over the Southwest, very moist, high CAPE air over the southern plains, and no distinct sfc boundaries or marked upper-air features. This made for another challenging forecast day for today.

Our outlooks for the 2000 - 0000Z and 0000 - 0400Z periods this mrng focused on two areas. The first and most important was KS and MO, where indications were from the 00Z Th 7 May initialized hi-res models (CAPS ensembles, NMM4 from NCEP, NSSL ARW3) that activity would initiate in the general vcty of Great Bend KS in late aftn and move into MO durg the evening. The proximate cause of initiation appeared likely to be [based on 10-m wind field from the CAPS control runs and the other so-called "deterministic" (poor term, but I use it for lack of a better one) models] convergence alg a wind shift derived from the frontal confluent zone noted earlier in NE. Updraft-helicity and CAPE-shear parameters argued that there was a chc of sig severe with this stuff.
During the forecast praparation period it became more apparent that the frontal confluent region at the surface noted earlier was going to be a focus of activity, if there was going to be activity in this area. Complicating the picture somewhat was the mid-level stuff noted earlier as being over ern KS into MO.
We noted mesoscale pressure perturbations and fluctuations with this, suggesting that there was some associated wind perturbation, probably just above the surface in early morning. Falling pressure in ern KS and rising pressure in MO appeared to be associated with acceleration northward of the OK low level moisture into KS during the morning, aprnt on visible satellite imagery as a tongue of the OK Sc advancing into central and e KS. The possibility of outflow cooling at low levels over MO reinforced the decision to go with initiation over central to ern KS.

Another factor was the behavior of both the NCAR 3km (initialized from RUC at 00Z and using GFS LBCs), and the 12Z initialized HRRR. Both these models had shown initiation of radar reflectivity over central KS before 15Z, and that the resulting storms would move eastward. Since there was no evidence for such initiation by then in the observations, and it seemed unlikely that such initiation would occur before, say, 1900 to 2000Z, the west team, at least, largely discounted these fcsts as providing useful guidance.

(As an aside, in the Sc over OK and KS mid-morning through mid aftn there were southward-propagating waves in this low-level Sc. I speculate that these were manifestations of low-level bore-like features, initiated by penetrative downdrafts from the earlier convection, though that these continued obvious into mid-afternoon raises some questions about this.)

Both teams, then, pretty much bought off on the scenario of the 00Z NAM-based intialized hi-res models in their forecasts, tho the teams differed in details. Both teams figured that this was worth a 15% chance of svr ... no higher due to a lot of concern that storms would not even form ... and a 10% conditional probability of sig severe.

The second area of concern was along the Red River in AR and SE OK, where the aforementioned 00Z NAM-based initialized models were indicating moderately intense storms forming in aftn, but that storms this area would become weaker after 00Z. The W team put a 5% chc of svr this area, but the E team did not consider this area worth even that much.

These fcsts were completed by 1600-1620Z.

By 1800Z or so clumping of sfc-rooted Cu was occurring along I-80 WSW of KOMA along a WSW-ENE line. These cells were partly flagged by the GOES-R algorithm for detecting growing clouds along boundaries. (I would have to question whether this algorithm added anything to what an experienced forecaster knowledgeable about the overall convective environment could provide.) This convective development contributed to the final forecasts for the 20 - 24Z period and the 24 (00)Z - 04Z period having the areas of forecast severe shifted Nwd from the preliminary forecasts, touching the nrn border of our forecast domain near the IA-MO border. This was confirmed by the 12z-initialized forecasts, as well as teh 16Z HRRR. The 12Z NMM from EMC in particular had initiation very close to where it actually occurred, and subsequent movement of the activity into MO with rotating updrafts. However, we tended to discount the massive outflow generated by this large cell.

We had less confidence in the liklihood of sig severe in preparing our final forecasts. The temp-dew-point spreads in the air farther N were roughly mid 80s to mid 50s, arguing that strong tornadoes would be very unlikely. So, the main threat for sig severe would have to come from wind gusts or hail. Accordingly the sig severe areas were reduced somewhat and the E team might have eliminated theirs altogether (don't remember attm).

A couple of observations.
The NMM 4km from NCEP initialized at 12Z today also initiated a cell near CDS that put out a circularly spreading outflow that covered much of OK after a few hours.
The HRRR seems to be producing overall too much coverage of radar reflectivity, though in terms of forecast reflectivity I believe it is doing a little better than the CAPS forecasts initialized same time. (NOTE that these forecasts are really not directly comparable since different model configurations (phyics options, lateral boundary conditions, grid spacings, domain sizes, etc.)

Comments and corrections welcome ...
John Brown

Wednesday, May 06, 2009

Another "messy" day.

Another day with ongoing storms to mess-up the afternoon convective environment. Attention was focused over the southeastern and mid-Atlantic states. While the observations led to plenty of forecast uncertainty, today was notable in that it was the first day of the Spring Experiment where the different 0000 UTC models showed some consistency...but of course just because they were consistent doesn't mean they'll produce good forecasts.

One of the more intriguing forecasts was from the 4-km WRF-NMM which developed several discrete, rotating storms along a warm front located in northern North Carolina. The other models also produced storms in this region, but they were smaller, more numerous, and not nearly as organized. The behavior of the NMM can probably be attributed to its surface wind field, which predicted a more easterly wind component north of the warm front (compared to the other models), thus, enhancing shear along the boundary. As of this posting, there are numerous storms in that area, but I do not believe any are supercellular.

Once again, the 1200 UTC model runs were out to lunch and didn't provide any helpful guidance, probably due to ongoing convection influencing initialization. I don't think either forecasting team used the 1200 UTC model runs as guidance for the final forecast products.

It seems like we're just waiting for a day without morning convection. Perhaps then, the 1200 UTC models will have a better handle on the situation and can provide useful guidance.

Craig S

A first for the 2009 Spring Experiment!

It looks like we have full 18-06z loops from all available data from 00Z, May 6, 2009!

This is a first for the 2009 SE.

With the center point change to CLT today we are outside the domain of the 00Z CAPS CNA and C0A models.
GEMPAK just plots a blank image with a title for those frames. Everyone else appears to exist for the
18z-06z period. This includes the following 00Z runs:

CAPS-SSEF-ALL (18 members)
CAPS-CNA - no grids in selected domain and not shown in the loop links, below.
CAPS-C0A - no grids in selected domain and not shown in the loop links, below.

Here are the links to the loops for this 00Z run of the models (the verifying base/composite reflectivity will fill-in tonight):

6-panel 1KM REFL w/ verifying BREF w/ time:

6-panel Composite REFL w/ verifying BREF w/ time:

4-panel Synthetic Severe (no HRLY REFMAX from AFWA4 until next week) w/ verifying hrly reports w/ time:

Where to focus?

Gulf Coast states look pretty worked over...early day reflectivity pattern shows serpentine line of convection over this area..."junked up". May focus on NC-SC where they may get some heating.

Monday, May 04, 2009

I blog

OK, I think I'm blogging. We had a decent first day in the Spring Experiment. Things are starting to take shape and we will soon have lot's of interesting things to blog about. Hope there was no serious storm damage in Norfolk today.

Sunday, May 03, 2009

HRRR Update

Looks like the WRF3-HRRR (high-rez rapid refresh) has been coming in. I had to update datatype.tbl for the $MODEL/wrf3hrrr/ grid file name. Also, there does not appear to be a 1km REFL parameter in the GEMPAK output. There is REFC (composite reflectivity) and also the history parameters that allow the generation of the SSVR plots.

So, as of the 12Z run May 3, 2009 the following thumbnail image names are available:

Composite reflectivity (compare to crefr):

date_runtime/WRF-HRRR3_CREF_YYYYmmdd12_f000.gif through f012.gif


Hourly synthetic severe (compare to lsr/tlsr):


For May 4th: the severe potential across much of the US looks marginal (it figures). However, there is some possibility for high-based or elevated convective initiation across parts of southeast CO into KS/OK/TX areas later in the day and perhaps again overnight. There is also the chance for an isolated LP storm with heating across TX but the chance is low. With that in mind, I decided to set the center point at CDS for Monday.

The weekend saw some remarkable storms! I tried to capture as much as possible. LIT was a good pick for Saturday's actvity and I picked TCL for the long-track bowing MCS/Derecho event on Sunday. The WRFs from NSSL and NMM both showed some amazinglt accurate simulations and were integral to forecast decisions made at SPC. The convective evolution/mode on both days was relatively well handled. This was especially true on Saturday afternoon/evening across the Arklatex:

The 18-24h forecasts form the NSSL and NMM WRFs valid for Sunday afternoon were also really interesting to review. I think the NMM takes the prize in this comparison with its incredibly accurate depcition of the two linear systems over the MS/AL area Sunday morning. However, the forecast was about 3h too slow.

Compare this 17h forecast from the 4km WRF-NMM:

with this radar image about 3 hours earlier:

One last observation:

The 12Z NMM run on Friday and Saturday was poor guidance and seemed to suffer from the cold start. I would shy away from the 12Z NMM for making decisions about storm initiation based on what I saw Friday and again Saturday. There was no 12Z NMM run on Sunday morning. Not sure what happened there.

That's it. One more forecast shift on Monday and then back to some hacking on Spring Experiment graphics!