EWP Blog

Notes on Gridded Warning Archive Case

Stream of consciousness notes from Mike Cammarata’s exercise:

Getting used to the software. Mike has some experience with this from earlier this week.

“Calibrated” ourselves with the MESH.

Desire to hide/show products more easily.

Would like to know when the next update is expected in displaced realtime case.

Long list of products (mostly warning output grids) became cumbersome to deal with.

At one point, Mike realized that we were encompassing more than the *current* threat area, made that adjustment.

———————————————————————————————————-

Group discussion: Mike Cammarata and Patrick Marsh, warning participants (Kristin/Kevin M. pw coords.)

A more informed decision could be made with better technology and guidance tools. For example, storm-following loops and cross-sections (even automated).

Took longer to issue due to polygon drawing (hard to get used to different knobology)

How did we feel about issuing probabilities? Probability was very arbitrary. Mike: “At what level of risk are we going to have a tornado?”

Discussion ensued about difference between achieving GPRA goals and current paradigm of warnings.

Discussion about significant call to action. (Are probs the best way? For tornado?)

Every decision maker has their individual cost-loss ratio for each decision made.

Andy feels that the “public” needs to know when to be told to “duck”.

The big issue is how we can objectively calibrate forecasters to the verification and to each other, so that there is a consistent answer for each warning.

Kevin Manross and Greg Stumpf (Gridded Warning Cognizant Scientists)

The following entry highlights initial impressions and perspectives from the forecasters/scientists after day 1.

*workload management

– segment operations by type of threat (allows view of specific products)

– sector by storm; in current system, loss of svr thunder warning during tornado warning.

– users would find separate wind/hail products useful; overlap of products and additional information.

– increases need for interoffice collaboration if different people are issuing separate warning types.

– County Warning Areas (CWA), warnings moving into new CWA-ownership? Discontinuous probabilities at CWA borders.

*Is issuing probabilities for warnings the right direction for NWS?

– 30 min vs 3 days. (tornado vs hurricane). Possible loss of response from public with 30 min/30% tornado warning?

– user cost/loss ratio using warn-on-forecast (NWP) guidance.

– binary threshold, doesn’t always allow for conveyance of information.

– significant wx advisory/warning decision products as pre-warning updates. Same as low probability warning forecast w/long lead-time (1hr+).

– For pubic communication issue low, mod, high probability instead of specific % #’s.

– WAS*IS direction for year 2 of project, working with external groups and how the data may be used.

*First forecast today, uncertainty of development and decay of storms and not communicated well with forecast. Duration of warning lasted much longer than storm threat. Was the warning decision “good?” Are our probabilities for t=end going to be too high? Is the shape of the warning grid an accurate representation of the threat?

*More time spent on software and drawing polygons then analysis of storm and storm environment. Again, automation of initial threat area would be useful in year 2, e.g., hail=mesh & tornado=meso. Movement away from WDSSII so forecaster don’t spend as much time learning the software.

Kristin Kuhlman (Gridded Warning Cognizant Scientist)

–Continued training with each forecaster on the probwarn software–

*Discussion of gridded vs current text products and information conveyed within each:

-Inclusion of checkbox for hailsize and windspeed (standardization)

— Ideas:

3/4 in hail now and golfball expected with linear interpolation.

Probability of particular size hail.

*Software: Suggestion of vector from original warning polygon to current threat at current time. With multiple warnings in close proximity to each other it can be hard to tell which polygon goes with each threat area

Kristin Kuhlman (Gridded Warning Cognizant Scientist)

More comments:

Higher prob initially, fall off faster with greater spread.

Would be nice to be able to trace the trend.

Analogy to GFE. Want to be careful we don’t get consumed by the technology.

Also, manage many warnings and adjusting and tweaking might be too time consuming.

Re-draw threat area automatically each volume scan to have continuously advecting threat area.

Could use hail tracks/rotation tracks to update TA.

Greg Stumpf (EWP Weekly Coordinator, 28 April – 2 May)

We have just switched our warning location to SE VA/NE NC with some small supercells still over land. We’ve kind of transitioned into our IOP. Greg is working the technology, and we are having a group discussion on the concept. Here are some comments.

Change “Peak Probability” to “trend Probability”

What about the users’ perspective when the warning keeps getting re-adjusted and the grid value change?

Are the threat areas too small? Could hurt if there is evolution.

Perhaps we could capture cyclic evolution by increasing the motion uncertainty.

Greg Stumpf (EWP Weekly Coordinator, 28 April – 2 May)