ProbWarn Discussion:

Slider bars give too much control, increments of 5 or 10 should be used (forecasters don’t think in terms of 66 vs 68 %). Group uncertainty together in widget following sliders for motion and speed… peak probability should be labeled ‘trend prob’ or ‘mid prob’–has lead to some confusion. Preview of product is necessary before saving…

Product for low probabilities, allows for communication of threats that forecaster may just wait on otherwise.

pulsing probabilities can lead to confusion to end user… How often do you update grids? and how does the updates get translated to outside user? how can we issue the right product to get the desired reaction?

need to more clearly tie original threat polygon to storm, either clarify naming convention (e.g., alpha, beta, etc–something more obvious than the utc convention currently used) or move polygons along automatically with time using set motion from warning.

Kristin Kuhlman (Gridded Warning Cognizant Scientist)

Operations continue with two teams alternating positions every hr… forecasters seem to appreciate the ease of updating the warnings with this method, though the start-up in its current format is a bit tedious. Can also get confused with multiple warnings on the same storm by highlighting the wrong threat polygon, though hasn’t been a huge issue tonight… Forecasters again would like to see a preview of ProbHail (or other threat) swath before saving it…

Team 1 issuing a lower prob tor warning on storm in northern Chaves Co.; storm in SW Chaves Co has moved out of threat area with lack of attn to storm. New cell has developed to west and the team is debating covering both cells with one warning or beginning new warning on western cell. (Team has joked it would like a “maybe” button on choosing whether a deterministic warning would be issued for this cell, probabilistic warning covers this uncertainty much better).

Team 2 plans on tackling wind threat on storm with HP nature nw of Tucumcari and bringing down probs on tor threat.

Kristin Kuhlman (Gridded Warning Cognizant Scientist)

Possible software guidance for future: Encircle large area and grid points exceeding some threshold value (e.g., MESH > 1.25 in) gives an initial first guess polygon for threat area.

Team 1 is expecting their storm to decrease in intensity as it moves off the terrain and are going to begin issuing warnings on the storm NW of Roswell (Chaves Co). Velocities from KFDX are not good on WDSSII, so will begin with hail threat and then possibly tackle tor next (funnel cloud reported with storm). They have also tackled the hail threat on a separate storm in far SW Chaves Co.

Team 2 is discussing possibilities of probabilities of hail > 2 in or 3 in (sig hail?). Move to allowing others (the public) to decide what is important for themselves (e.g, different thresholds for action)–the NWS issues probabilities of events occurring (not warnings)…

Kristin Kuhlman (Gridded Warning Cognizant Scientist)

We finished a first training exercise using live data in southern NM around 5:30 pm and have moved to issuing warnings on two separate storms in NM. Team 1, Mark and Brad (on Moore), concentrating on the hail threat near Los Alamos. Team 2, Kevin and Eric (on Higgins), are looking at both the hail and tornado threat in southeastern Santa Fe.

Team 2 have noticed that threat areas can be a bit ahead of radar / mesh output as update times 1min for the former, 5 min for the latter. Both teams have discussed how often should update product, see it as a continuous product that you don’t have to wait until the end of the “warning” to update/reissue.

Kristin Kuhlman (Gridded Warning Cognizant Scientist)

Gridded Warning Discussion:

We sectorized warnings in TX–north (Ron and Dan M.) and south (Dan P. and Dave). Both groups focused on all three threats (wind, hail and tornado), targeting 2-4 storms each.

Would like to sync multiple threat types for same storm, such that if storm motion is changed for the hail threat and then update (if chosen) coincidently for wind and tornado for that storm. (Could try on a case such as Greensburg where you have to separate tornado threats with different storm motions.)

Felt like lost concentration as moved in ’round-robin’ from storm to storm. Possibly focused on too many storms.

Sig hail, Sig Wind options–on top of other three threats–add to the amount of data a forecaster must maintain, but seem necessary. 100% pea-sized hail or 100% baseball hail or possibility instead state for exceedance of a level: e.g., 100% prob of nickel size, 40% prob for > golf ball.

Pulsing storms for group 1 caused constant updates in trend, sometimes trends would be changed quickly from one update to the next. *Do the forecasters worry about continuity or accuracy more? * Second group dealt with storms that maintained more continuity and followed conceptional model of HP storms, so confidence in hail and wind threats especially was particularly high.

In terms of software, would like interface such that forecasters could draw trend in line graph (e.g, bell shape or up and then flatten).

Kristin Kuhlman (PAR Cognizant Scientist)

Could note better convergence signature in PAR data with faster updates.

Ron P. has taken over warning contour operations attm. Updating threat area in Okuskee Co. to cover growth of storm. Has higher threat embedded within lower threat.

Kristin Kuhlman (PAR Cognizant Scientist)

ProbWarn Software: Would like function for probabilities such that you can set an initial and a final probability to ramp up to at a certain time and maintain throughout the warning time.

Also, with a high number of warnings in close proximity need to be able to hide threat area contours that are not currently being worked on. Quite frequently will accidentally move or highlight wrong contour while updating a different one.

Kristin Kuhlman (PAR Cognizant Scientist)

Issuing first tornado warning for threat in Pott. Co. panhandle, with evidence of rising motion from helicopter and rotation (vel: 30 out, 15 in) and notch in reflectivity in both PAR and KTLX. Going with low probs now, but expect increase in chance based on environment and beginning evidence of storm motion to the right. Discussion of whether or not deterministic tornado warning would be issued at this point–for this day and in this environment would go with “No” at this time.

Also updating storm motion for hail grids. And updating larger threat area to account for single core that has developed in SE Cleveland Co, smaller area higher probs.

*Software: would like to be able to toggle reflectivity while viewing warning contours as top source

Kristin Kuhlman (PAR Cognizant Scientist)

Testing very low (0%) prob for storms initiating in so Cleveland Co. and moving into Pottawatomie. Using probwarn to communicate expectation of increase in prob (to 40%) in central Pott. Co. in ~45 min. Uncertainty in storm motion and speeds puts warning grid covering almost all of Pott. Co.

Storms continue to back-build along front making it difficult to maintain coverage of threat areas.

Within probwarn software, would like clicking on slider bar able to move # one at a time instead of jumping to a random #.

Kristin Kuhlman (PAR Cognizant Scientist)

Dan. M. would like a “preview” button to see his warning before issuing it. Also has decided to junk previous two warnings and issuing one larger warning with higher uncertainties in storm motion to account for back-building updrafts along the line.

TOA and TOD should have actual times (UTC) in legend instead of 15/30/45 min.

Fast scanning rates from PAR giving rapid updates of individual updrafts, can cause ‘threat’ area to be too small and end up having to move it around too much. New larger warning accounts for this.

Kristin Kuhlman (PAR Cognizant Scientist)