Thursday, May 09, 2013

Live Blog

3:52p -  yeah, ok. Data Overload is winning. The participants are restless and the line tweaking is generally reaching OCD proportions. It is safe to say: Uncertainty:1 HWT EFP: 0.  So thats a wrap for today. Feedback is welcome. I think i can only handle 1 live blog a week. Tomorrow I will do a weekly summary to close out this first week. [For those interested we are unlikely to share our experimental forecasts so that there is no confusion with actual public products. The goals of the experiment are tied to both forecasting techniques and forecasting processes. We perform forecasts and verification to generate discussion on all of the above. By sharing the processes and techniques with our partners and participants we hope to share what it is like to forecast operationally.]

3:43p - Minor tweak to the probabilities but an enhancement to the significant severe to try and balance our perception of convective coverage along the retreating dryline between 22-00 UTC. I think our mind changing is a direct result of data overload amidst a relatively messy scenario. The tools we have to deal with this mix of convective modes, messy environment, and rapid updates really cant provide the level guidance needed to be relatively certain. NME forecasts have good signals of precipitation in where we are lining up our probabilities. And storms are going behind the boundary just to the west of CDS.

3:32p -  Many many tour groups are walking thru the NWC and taking our picture. Now I know what the zoo animals feel like.

3:25p -Visible satellite and radar analysis comparing to sfcOA. I think we are trying to talk ourselves into slight modifications to account for the clearing in the east TX panhandle near Childress. This cumulus field that is developing along and behind the boundary suggests storms may still have a chance to fire further west than we previously had allowed for. Also debating how the MCS/line segment is moving more southeast than east and may not make it as is into LA. 

3:12p - Howie Bluestein stopped in to tell us his stories of chasing down in TX yesterday. Dryline storms with odd behavior, said they looked like CO storms.

3:04p - and, we're back! 

2:30p - A bit of trimming and trying to reconcile with current observations. I am running out of steam. Need more Oreos. 

2:19p -  NME forecasts were problem laden. Wont get it in. #codeissues #kentsfault

2:12p - Discussion moving to adjusting the sig severe probs that we put up into OK. The coverage appears low in updated guidance and we cant rule out some big hail there. Word on the street is a tornado (landspout) occurred at 18:34 UTC around Snyder, OK. [Glad we covered that area as our SPC experts had indicated!]

2:00p  -  12 UTC ensemble guidance is loaded and is a bit better with the ongoing convection, showing storms developing now in SW thru central OK in the next few hours. Signals for severe storms are weak in OK and mostly focus on the developing MCS in TX. A couple random members try to produce robust UH tracks out in the TX panhandle anywhere from AMA north. UH tracks are still relatively weak in the panhandle and OK. The MCS  appears to be capturing the focus. Even in its wake where storms are forming now the threat appears to be maximized even though it is small because the mode has been unorganized line segments. Maybe there will be another round on the retreating dry line, but this appears to have low confidence. Perhaps the instability will be reduced in this area since the storms are relatively disorganized.

1:30p - Finishing up map briefing.

1:25p - [thinking aloud] Expressions of vast uncertainty regarding evolution of ongoing convection. When the forecast is dirty and you have to figure out what is actually being represented in model land, reconciling these differences and transforming them into signals you can rely on, is quite challenging. This is especially true when the models dont quite have the details right of the ongoing convection. Suddenly you introduce a whole lot of doubt/uncertainty. Not just uncertainty in the forecast, but doubt in the model forecasts, doubt in our ability to correctly modify the forecast. You try to do this all the while trying not to just throw out the model data (sometimes you have to). But usually the models can be so bad that they end up not too bad. Sounds insane doesnt it? But when the atmosphere and models are equally messy they tend to, en masse, communicate something that is reconcilable with observations. Thus getting it right for the wrong reasons. We take advantage of this with longer time periods of forecasts but with short 3 hour blocks of time this is much much harder to do. The local offices have this problem when further confined in space, so for them this problem can be magnified.

11:55a - Lunch time. Map discussion will commence at 1p. I will be taking a live blog break in order to brief the EWP group with our morning forecasts at some point. Next update will commence at 1:30 and go until 2:30p, then again at 3p.

11:46a - NME SCP forecast initialized at 14 UTC shows  minimal window of SCP > 1 in OK, but maximizes along the dryline in south TX -ish to values of 10. Much less spread in the CAPE fields within the maximum; only spread is located along the dryline indicating spread in dryline position.

11:37a - 21-00 UTC on the west team.  Putting in the probs for the redevelopment to the west including OK, and really delivering the message of sig hail in the 21-03 time periods. Not much has changed in our thinking. About to refine based on the NME, and other 12z operational guidance.

11:30a - Thinking about the gap we left in central OK indicated by the outline of higher probs. Will this area just continue to convect and thus yield an opportunity to produce minimal severe reports or not? Trying to manage the evolution of ongoing convection in the TX area has transitioned into a southeastward moving line segment. 

11:12a - Moving back into forecast mode starting with first guess lines generated from the TD technique that uses our full period human drawn forecast and uses the calibrated SSEO to break it down into three 3 hour periods. Loading data into N-AWIPS from the HRRR and NSSL mesoscale ensemble (enKF data assimilation).

11:06a - Just finished discussing the relative success of the NSSL-WRF and the reduced value of the hot start version for last nights bow shaped MCS across northern OK. Will this bias your opinion of todays forecast?

10:13a - Model evaluations commencing via the web site, but I need a live blog break. Back at 11ish for the production of forecasts for the 3hr periods.Plus I have to fix my code for my graphics. [the coding never stops; why does stuff break after you put it through the ringer the previous week?!]

10:07a - Comparing how our teams updated their forecasts: The primary result for today was that updates changed very little. [Do we really have sufficient guidance to make substantive changes to these 3hr forecasts?]

10:04a - Evaluation for the 00-03 UTC period  showed disparity between going for two seperate clusters and one robust high prob area between the teams. Both do well, but the east teams 30% compared somewhat better to practically perfect.  As far as how we did relative to the automated guidance in breaking down our full period forecast using ensemble guidance, human forecast better for one team and not the other at 18-21 UTC.  Automated in the 21-00 UTC was comparable to the human forecasts though probabilities were higher in human forecasts. This suggests that model timing/evolution still needs significant work. Even in the 00-03 UTC period automated matches human forecasts pretty closely, though automated has more false alarm area in the East Team.

9:53a - Yesterdays forecast evaluation for both teams. Huge SIG area in the practically perfect. Forecasts were relatively good for the full period. Now on to the 3hr periods. Very close but challenging for the 18-21 UTC period in CO, much better with "relative skill" above 1 for both groups due to the spread out yet clustered report distribution.

9:27a - Sig area extended into OK along and ahead of assumed front position by 21 UTC. Discussing implications of the cloud cover over south TX. Full period forecast due in order to generate first guess temporal disaggregation.  The OK extension will be in a short time window before diurnal cooling sets in. Questions remain as to coverage in these redevelopment areas because there could be multiple waves of storms. Will they interfere with each other and become messy quick, or will they remain somewhat well spaced and discrete?

9:20a - Lines being drawn, max probs came in at 30% 35% probability of total severe. Sig severe was drawn primarily in the central to southwest TX centered around SJT when things will be the least messiest in redevelopment. 

9:14a - Ensemble probabilities of UH are really similar with all 3 ensembles at all times. Updraft velocities are showing this hole across southern and southeast OK. And max surface wind speed probabilities are really low. SPC calibrated severe SSEO probabilities pretty much mimics the UH probability distribution from eastern central Texas hooking back into the TX panhandle. Gut feeling is that there will be some stronger storms embedded in a higher coverage area in the south, and perhaps a few stronger storms embedded elsewhere in mixed mode mess. Later on these signals coalesce into an MCS from eastern TX into LA late; whether this system actually is rampant with severe storm reports looks to be rather uncertain from all of this 00 UTC guidance onslaught.

9:07a - CAPS ensemble: first up spaghetti reflectivity showing scattered redevelopment in west TX (and also in NW OK though scattered). SSEO has less clustering in central TX for this morning, then gets convection going in the TX panhandle later today.  UH spaghetti plots have fatter weaker tracks but the strongest, fattest tracks are down in south TX. UH tracks are in the panhandle of TX but tend to be relatively few and far between in western OK, and southwestern OK. The MCS signal moving into LA late has a few UH tracks but its mostly vague UH areas not indicative of supercell structures over the whole area.

8:58a - NAM soundings SJT hodographs start out too wavy then go into long length bu relatively modest SRH. Doc brown discussing how the HRRR assimilates radar data via temperature tendency derived from echo greater than 35 dBZ every 15 minutes in order to maintain fidelity of the assimilation that runs every hour. 1st guess still comes from 13km RAP.  [Too many simultaneous conversations; I'm a Dr not a dictation expert!]

8:50a - Discussing the NAM 06UTC run showing a rapidly moistening area in the area of the TX clusters favoring upscale growth thru afternoon with redevelopment further west later this evening. Shear sufficient for supercells, moisture not really an issue.

8:44a - Just finished map discussion and hand analysis. Now breaking into the satellite observations to check out these subtle short wave troughs. I have already stated that todays storms are going to be early and often. Sure enough storm clusters have started in central TX.

Model Potpourri

 Spent time discussing the forecast issues of the day (and it lasted all day) for 5/8:
1. lack of synoptic forcing on the strong side
2. boundary layer (dry air HCRs initiating storms on the dryline, warm sector HCRs initiating storms east of the dryline in a narrow moist tongue)
3. marginal to sufficient shear for supercells
4. boundary layer oddities (mesoscale drying in the post MCV environment that advected north in OK, moist plume that got sandwiched between the dryline and the dry air pocket)

Tuesday, May 07, 2013

Models: signal or noise?

Todays forecast challenge had to do with storm location and initiation. From the TX panhandle northward thru KS into NE and CO. It seemed fairly evident that storms would form in this deep and relatively dry boundary layer (at least for May). With the PBL depth approaching 3km, models were not shy in breaking out storms to enjoy the roughly 40-50 kts of vertical shear and lovely quarter circle hodographs. It was relatively obvious that the PBL would control where and when storms broke out.

Monday, May 06, 2013

2013 Spring Experiment Week 1 Day 1

The Spring Experiment within the Hazardous Weather Testbed kicked off today. The Experimental Forecast Program has a running list of objectives (there just is too much to explore) so lets hit the high points of producing convective outlooks:
1.  Can we merge human and ensemble forecasts to produce shorter time scale forecasts (in this case create a long period (20 hour) forecast and turn that into multiple 3 hour periods)?

What we are testing involves making the best use of the models and incorporating what the forecasters are good at. The models are terrible in getting severe weather at the right times and in the right places. Challenges in predictability (i.e. small scale errors in observations, errors in how we represent model processes, etc) add up to prevent being truly spot on in the forecasts. But forecasters, after years of experience and training, can be great at pattern recognition and referencing climatology about severe storms and severe storm processes. Merging of the two can thus help make better forecasts as we demonstrated last year. We call this technique Temporal Disaggregation (TD). It just means we take the ensemble forecast and apply it to the area outlined by forecasters. It worked well and compared favorably to what forecasters could draw independently ... in other words: what we drew in fine grained forecasts matched well what the model probabilities would be if we corrected the location!

As far as time, we learned that if the periods were long enough we could account for poor time forecasts. So the next big question is:

2. Can we make good short period forecasts? What resources can we take advantage of to get around the predictability challenge?

So our TD technique will produce a first guess of the forecasts for the 3 shorter periods (18-21, 21-00, 00-03 UTC).  From these first guesses, we will use updated ensemble data to update these forecasts. Here we have more models to use with the latest data (through data assimilation). One of the challenges an operational forecaster would face is: Can this be done timely and accurately while providing good risk information? Can the experimental models be used reliably?  Can we create good proxies or variables that we can extract from the model that relate either directly or indirectly to severe weather?

We have the CAPS 8 member 4-km grid spacing, radar data assimilation ensemble (SSEF), the 7 member SSEO, and 10 member AFWA. We also have the NSSL mesoscale ensemble (NME) run at 18km grid spacing with 36 members but updated hourly with 3 cycles of forecasts out to 03 UTC. This latter ensemble uses the ensemble Kalman filter to assimilate surface data, aircraft and satellite observations.

A big part of the experiment will be spent evaluating the models, the techniques, and the forecasts themselves. We hope to highlight the model capability at this higher temporal resolution. We will also make it a point to identify good metrics that reliably identify good forecasts as compared to what forecasters will also deem as good. In this way we put the metrics to the test and discuss the strengths and weaknesses of them.

Today was mostly about showing all the tools and models at our disposal while making a forecast for thunderstorms with hail in the North Carolina and Kentucky area. The SSEF, SSEO, and AFWA convection allowing models were in moderate agreement about producing a few strong storms. The uncertainty was moderately high (only half of the corresponding members of each ensemble produced these stronger storms), but confidence was high that a few of them would be capable of producing hail. Sure enough we put our forecast for 21-00 UTC for the hail and the first and biggest report occurred at 20:55 UTC, with the rest in NC occurring thereafter.

I think for the next post I can mention some of the high resolution variables that we use as proxies for severe weather. And future posts will discuss some other experimental products that we get to see including the UKMet offices' Unified Model.