Week 1 Feedback (and 05/08 daily debrief)

After a relatively quiet start, week 1 of the 2014 HWT Spring Experiment turned out to be quite active by Wednesday and Thursday. Participants were able to utilize most of the experimental products in the simulated forecast/warning environment, writing numerous blog posts, and leaving valuable feedback to product developers.

Week 1 EWP Participants

Feedback feedback for week 1:

Simulated Satellite Imagery: Unfortunately, due to problems out of our control, this product was only available halfway through the day Thursday.
- The model didn't pick up convection in Texas. That can be very important: if its showing stuff that’s not there, why continue to use it?

NearCast System: 
- It seemed like a very useful product
- It allowed for the visualization of moisture and instability and tracked how it was moving. You can track where the fuel will be. It helped for CI and ongoing convection.
- One of the more consistent tools, it handled gradients well
- 30 minute model update would be fine, more often then that, there wouldn't be much change
- Might be useful to somehow interpolate data to fill gaps, perhaps this could be contoured, and one could turn it on/off
- I actually like the gaps, it tells the forecaster where we don't have data, instead of interpolating, a forecaster can interpolate in his head.
- I haven't seen anything like this in AWIPS; it is a unique tool
- This is the type of tool that you need to look at daily, and the more you use it, you'll be able to figure out how to correctly use it.
- The PW values at the low levels were not very deep last night, this could help a forecaster to update/remove pops
- This was the most consistent product throughout the week. It gave me confidence where the moisture boundaries would be in the near future, and convection consistently developed along those boundaries.

GOES-R Convective Initiation:
- A line of clouds formed in Nebraska. CI showed some higher probs on the northern end of the line, where storms would first initiate several scans later. Probs increased down the line before convection continued its development.
- It is ok. It can go from fantastic to not so good depending on situation
- One of those tools you have to use a lot to learn what regimes it's useful for, and not
- Applying it it non severe situations would be good
- I can see its usefulness in a forecaster being able to update pops in the near term
- it gives a forecaster a 5-10 min lead on starting to interrogate near storm environments where the      convection is firing
- I can see it in a procedure
- It was hard to get a feel for and get a handle on because of changing domains throughout the week (mountains to plains)
- Yesterday, there was a line developing, but it never showed probs
- Out of all of them, might have been the least useful
- Might have other applications, like fire weather out west in the summer.

Prob Severe Model:
- Later on in development, it had very high probs on storms that ci had missed
- It would be useful to change the color of the text when values reach certain thresholds (eg. MESH >1")
- Right now, it is kind of a short term initial storm development tool. I think there are other fields that could go into it that might be useful.
- Prob tornado, wind, hail would be nice to see
- Does have limitations in linear regimes
- It would take more than 4 days to really learn how to use the probs, develop thresholds in my head
- If there was an upward trend in prob severe, I would pay attention to that storm
- It was really good up until the last day in Iowa and Kansas, where many storms had high probs but didn't go severe. These storms had high growth rates but marginal MESH
- In some cases, this product may be enough to make me issue a warning (or not) when I am on the fence
- When I saw probs over 90% and they weren't going severe, I lost confidence
- I would look at these probs, then at other tools to verify it.
- I think, no matter what tools become available in the future, I will always go back and look at the raw data (dual pol, all tilts, etc) in addition
- I think it was useful in flagging areas where I should be looking. I would take a closer look at those storms with higher probs
- How I would use it: I've done my homework on the environment and the first cell is developing (with some sort of prob severe trend) and it is warned on. Any later storms that show the same trends in prob severe, I would likely also warn on. I would want to calibrate to the situation at hand.
- The ability to effectively use this tool would come with experience
- The color scale was fine. I like that the contours got thicker. Everything about the display was fine. I would just change the readout (by adding colors for certain thresholds)

Overshooting Top Detection:
- I like that it makes obvious where the overshooting tops are
- In areas where radar coverage is a little more sparse or when a radar is down, I see more utility, in helping to identify where the strongest updrafts are.
- I didn't get to see if monitoring trends in OT's would be useful (we didn't work late enough shifts), but I think it could have utility in the weakening stages of convection

GOES-14 SRSOR (1-minute Imagery): Was only available the final day of operations (Thursday)
- Is it useful over 5 minute imagery? It is hard to say at this point.  I would need more time to view the imagery and learn how to properly interpret it. But if it is made available to me, of course I am going to view it.
- This imagery is really cool
- It makes boundaries more distinct, and might give you a little extra heads up on cu development
- It was really detailed when convection was starting to go up in Kansas, you could see tops going up and their shadows. It made me think that I should watch over here.
- I looked at this more than the ci product because the movement was so fluid, I could see things as they were happening
- I could definitely discern areas where stronger lift was occurring
- I stick to a lot of the base data when I am interrogating storms, including satellite imagery
- When you depend too much on derived products, you become lazy and start missing important things

PGLM Total Lightning: There was limited use of the Lightning Products due to lack of convective activity within active LMA's
- I concentrated on it in the Cheyenne area. It really picked up on the storm with the density, I had flash count overlaid, and they matched.
- I like being able to interpolate the data in AWIPS. I switched between that and the gridded look

Tracking Tool
- I think this would be useful. I used it to track reflectivity, and you could clearly see the trend
- It could be helpful to show you where in the column the max reflectivity is (look through all tilts for you).
- I would like to see this track features like gate-to-gate and storm top convergence.
- I did not find the interface to be too intensive to use
- The meteogram takes up a lot of real-estate. It would be nice to have the option of bringing up the meteogram in a separate window.

- - Bill Line, SPC/HWT Satellite Liaison

OT Detection Assists on IR

If you refer to the first image, it is obvious that we have some cold cloud tops over Minnesota and Wisconsin. However, it wouldn’t be quite obvious where the highest tops exist and which of those are necessarily overshooting tops. The OTD (Overshooting Top Detection) tool helps! By detecting a cloud top that is less than or equal to -6K than its surrounding anvil, the OTD will mark the cloud top (seen by the dark blue marks in the second image). The anvil must also meet a temperature threshold. The anvil must be less than or equal to 225K and the top must be less than or equal to 215K. If these conditions are not met, no overshooting top will be detected. This prevents the tool from being over sensitive and creating false detections.

Jared Maples

DSM set-up update

22Z…a broken line of storms continues through western IA.  To this point, strong…but not strong enough to warrant warning.

The LAPS model forecast for 22Z is below

Catches the basic idea of the shape/character of convection to the west, but just a little off in location.

Here’s the Nearcast instability forecast looking ahead four hours…highest instability will lift northeast, and this batch of convection will move with it.  Notice lingering instability (relatively speaking)near the NE/IA border.

The CI product shows the potential for more convection to develop along the main cold front behind our initial line.  Expecting this activity, should it take hold, will be less of a concern with lower instability to work with compared to the first line.

db

Storms develop further southwest along dryline

Strong thunderstorms continue to develop along the dryline in Kansas on Thursday afternoon.  The first severe thunderstorm warning came out at 2152 UTC 8 May 2014 for a storm southwest of Topeka, Kansas.  Figure 1 shows the evolution of this storm up to the point of becoming warned.  The ProbSevere model jumped above 50% at 2146 UTC--good for 6 minutes of lead-time.

Figure 1. NOAA/CIMSS ProbSevere and MRMS composite reflectivity valid 2146 -2152 UTC 8 May 2014.

It will be interesting if storms further southwest become severe, as probabilities briefly went above 50%, but have been cycling a bit as the storms try to better organize.

-Sieglaff

Storms over western Iowa

Strong thunderstorms have developed along and ahead of the dryline over far eastern Nebraska and have moved into western Iowa.  Figure 1 shows the evolution of these storms--notice the ProbSevere values fluctuate--occasionally into the higher ranges (pinks > 70%).  These storms are in a favorable environment, have exhibited strong satellite growth rates, and have cycled in radar-estimated intensity.

Figure 1. NOAA/CIMSS ProbSevere and MRMS composite reflectivity valid 2014 - 2112 UTC 8 May 2014.

A question was posed to the pair of forecasters working the DMX WFO at the EWP:

"These storms have exhibited some high probabilities and have fluctuated as the radar intensity ebbs and flows.  How would you use the storms in these in this scenario and what would you suggest as data inputs into ProbSevere to improve the statistical model?"

"One forecaster replied, I like the fact that probabilities are fluctuating with radar intensity, it tells me the storms are cycling and may not be steady-state ready to produce severe weather.  Another forecaster said it would make it more difficult to use--but combined with analysis of radar tilts and the velocity data--it still tells me what storms are going to pose the highest threat.  We've seen storms that get high probabilities and don't cycle--they produce severe weather--but not these.  So that's the million dollar question.  Both added additional pieces of information that might add value would be melting depth (freezing height) (although they noted not in this case) as well as additional reflectivity information at higher levels."

This type of interaction does NOT occur when researchers are in their institutes and forecasters working a shift in their WFO.  It is experiments like these that facilitate the most rapid improvement in new products and the prospect of 3 more weeks of interactions is exciting.


-Sieglaff

Prob Severe of 99%…Three Body Scatter Spike

While monitoring the Soux Fall radar this morning, this beast developed just east of the radar site. The Prob Severe Model jumped to 99%.The hail size was predicted by the tool to be over 2 inches in diameter. A short time later a nice three body scatter spike appeared. We are still awaiting any reports from the area.

Jared Maples

1 Minute Satellite Feed

Today the GOES-14 SRSOR is updating at one minute intervals. This provides an excellent opportunity to see how quickly cloud tops develop, even well before the radar updates. In the first image you can see the void in the circled area. This general area is the same area in which clouds have already developed and are beginning to can upper level winds.

Update: Showers and thunderstorms have initiated in the same area that satellite feed was showing the high tops. The one minute feed gave great lead time in this instance.

Jared Maples

Severe convection firing near EWP WFOs

Strong thunderstorms have developed this afternoon just west of the DMX EWP domain along the dryline associated with low pressure center over the Northern Plains.  The NOAA/CIMSS ProbSevere depicted a storm near Sioux Falls, SD with strong satellite growth rates, a favorable environment and radar estimated hail size increasing with time (Figure 1).  The ProbSevere model generated probabilities in excess of 50% at 2006 UTC.  The NWS Sioux Falls WFO issued a severe thunderstorm warning on this storm at 2025 UTC, when the ProbSevere value was 91%.  As more storms develop and evolve along the dryline further south, more updates will be made.

Figure 1.  NOAA/CIMSS ProbSevere and MRMS composite reflectivity valid 2004 - 2024 UTC 8 May 2014.

-Sieglaff

CI picks up on line of initiating convection

Line of convection working through Omaha’s CWA. Initial CI hits 50-60% at the northern end of a line of cumulus at 1825Z.

At 1830Z CI probs increase to 80-90%. Weak reflectivity is evident at this point.

At 1855Z cumulus further south along the line shows CI at 60-70%. The northern portion of the line has developed into a strong shower/minimal thunderstorm.

At 1915Z the CI probs have verified – the line of developing cumulus is now a line of showers and thunderstorms.

Scott Rudge.

Early Thursday Severe Convection

While we wait within the EWP for severe storms to get going over the currently emulated WFOs (ICT and DMX) we looked north into two regions with severe thunderstorms earlier today--eastern South Dakota/western Minnesota and the upper peninsula of Michigan.  Figure 1 below shows the evolution of two storms near the Sioux Falls, SD WFO between 1644 and 1730 UTC 8 May 2014.  The western storm had ProbSevere values (with strong satellite growth rates) in excess of 50% at 1650 UTC--36 minutes prior to the storm becoming severe warned at 1726 UTC.  The eastern storm had ProbSevere values in excess of 50% at 1704 UTC--18 minutes prior to becoming severe warned at 1722 UTC.  One subtle point in this example, is the western storm had satellite growth rates, but the eastern one did not due to thicker existing cloud cover--users can expect better lead time when the GOES satellite can observe the growth of storms from the cumulus stage to the cumulonimbus stage, as is exemplified in this case.

Figure 1.  NOAA/CIMSS ProbSevere, MRMS composite reflectivity and NWS severe/tornado warnings valid 1644 - 1724 UTC 8 May 2014.

Further to the northeast over the UP of Michigan, severe thunderstorms developed associated with strong warm air advection ahead of the storm system over the Northern Plains.  Figure 2 shows the evolution of the severe storm between 1650 and 1754 UTC 8 May 2014.  The storm of interested achieved ProbSevere > 50% by 1732 UTC--22 minutes prior to the severe thunderstorm warning issued at 1754 UTC.

Figure 2.  NOAA/CIMSS ProbSevere, MRMS composite reflectivity, and NWS severe thunderstorm/tornado warnings valid 1650 - 1754 UTC 8 May 2014.

-Sieglaff/Cintineo

DSM area analysis

Models agree on low pressure lifting NE or NNE form eastern NE into MN through the afternoon.  As of 19Z, all of IA was in the warm sector, with a cold front expected to cross form the west during the afternoon/early evening.  the 12 Z NAM puts the front near DSM at 0Z; the 12Z GFS still has it just to the west.

Earlier convection is exiting stage right, and some clearing is taking place in western IA.  We’re getting rapid scan GOES data today, so we’ll be monitoring the boundary structures that are evident on the imagery.  Below is the CI product superimposed over the visible image…we’re seeing 67% probability of development of 35 dbz returns in the next hour or so along these boundaries.

As this activity develops and moves eastward, Nearcast forecast theta-e difference by 22Z would indicate pockets of instability (in an area of closer to neutral stability) moving in over the next few hours from the west.  Will be interesting to see how this translates into storm coverage, or even storm mode, as the front moves in.

And in the time it’s taken me to work on the above…we’re seeing gradually developing convection marching to the NE/IA border in an area marked by the CI earlier.  Only low to moderate hits on ProbSevere (45% or less by 1930Z).

Do have LAPS data today, but a late realignment of the domain has the model playing catch-up.  Will comment later.

Wx briefing indicated increasing chances of UH tracks in our area, especially in our northeastern counties…watching for supercells with tornadoes and hail…and some severe wind

db

EWP Day 4: Operations Update

The final day of week 1 is already in progress and a complicated forecast it is (due primary to overnight convection).  Our initial thought was to play in the moderate risk associated with the low pressure and warm front, but with the presence of ongoing convection and the desire to utilize many of the GOESR products meant for the pre-CI and CI environment we decided to shake it up. One group of forecasters (Rudge/Bickford) remains in the moderate risk in Des Moines (DMX) CWA while the other pair (Maples / Thoren) is operating in the Wichita (ICT) CWA and looking for possible development ahead of the surface cold front/dryline.  While some convection is already ongiong in the north and eastern DMX CWA, expectation is that supercell storms (and tornadoes) will develop in the area clearing farther west.  Severe activity in the ICT CWA is a bit more uncertain and timing is likely a later this afternoon and evening if it were to occur.

SPC Day 1 Outlook

Current Watches over operational domain

Super-Rapid Scan (1 min imagery) is in it’s first day of operations.  The animated gif (click to show animation) below shows the Cu development along the cold front into the ICT domain and clearing in the western DMX domain.

Super-rapid scan visible satellite on 8 May 2014

-K. Calhoun, Week 1 Coordinator

Daily Debrief (for activities on 05/07)

Forecasters operated in the Norman, Cheyenne and North Platte CWA's. Sim Sat Imagery still not available.

NearCast: 
- Convection started on gradient of highest theta-e and pw values
- A couple times i leaned on this product a little bit
- Might be helpful to overlay surface information with nearcast
- I thought it did pretty well showing where the greatest instability was and showing the moisture advection so you knew the storms would have some fuel
- Forecast worked well as well
- Gaps are a limitation, but I can mentally interpolate the data. This is where the storms are now, there headed towards this area, they could intensify.
- I would actually use this product in operations

GOES-R CI:
- This was the first day I actually could use it in a non-contaminated cu field
- It gave a good indication that further development would occur, even behind the dry line
- It was a confirmation tool maybe at best
- A lot of lower percentage CI and then there was a cluster of higher CI percentages showing up. Convection did occur here.
- It might be helpful to block out the noise (lower probs)
- I sometimes think 35 dBz is too small of a threshold
- When things started going and the sat picture got dirty, the CI probs didn't do as well

Prob Severe:
- It is my favorite tool so far
- In Oklahoma, I warned on storms based on what this was saying, and I looked in the office, and that was where all the current warnings were.
- It is nice to be able to quickly show you what is happening with a storm or group of storms, especially when I was previously paying attention to some other, already active storms
- It would be nice to figure out a way to make it work with organized convection (linear modes, etc.)

Convection develops along NearCast boundary

Severe convection struck much of the central US on 5/7. EWP participants were monitoring the nearcast instability fields (forecasts and analyses), and noticed convection was developing along a sharp instability gradient. Participants noted that this has been a common theme with the nearcast during the week (convection developing along and moving with boundaries), as it does very well with picking up on moisture gradients and their movement. The loop below is hourly NearCast theta-e difference analyses (with 00 UTC forecast at end) with IR underlaid. Notice convection initiates along the boundary in Oklahoma and continues to develop northward along the instability gradient as depicted by NearCast. Negative values (blues to purples to reds) indicate greater instability.

The NearCast system is a Lagrangian model that dynamically projects GOES Sounder temperature and moisture retrieval data forward through space and time at multiple levels of the atmosphere. This provides forecasters with hourly, rapid updating, observation-based, 1-9 hour forecasts and analyses of the vertical moisture structure of the atmosphere. The sounder observes horizontal moisture boundaries particularly well, while the NearCast system preserves said boundaries in its analyses and forecasts. NearCast products can be loaded in AWIPS-II as past analyses leading into the most recent 9 hour forecast cycle, as is done with this loop.

- Bill Line

Storms fire along and near Theta E difference boundary – Henrietta storm has tornadic tendencies

Storms initiated across far western OK along and near a Theta E Difference boundary.

At 20Z

At 21Z

At 22Z

At 23Z

At 2330Z – Henrietta Storm quickly developed a hook with a report of a rapidly rotating wall cloud.

Scott Rudge.

ProbSevere and Mature Convection

Just some comments and thoughts being relayed about the use of ProbSevere and mature convection.  The EWP participants have found the ProbSevere model useful in its stated goal--providing probability a storm will first produce severe weather in the following 60 minutes.  In the ProbSevere training it was stated the model will likely have some utility in the application of continued severe threat and/or storm decay because, even after the satellite growth rates are no longer used when the storm is mature--the model continues to generate probabilities using the environmental conditions and the MRMS MESH.  But the model has not been validated in such a manner--so it was really up to participants to comment on utility in those scenarios.  Thus far a clear distinction appears to be emerging:  the ProbSevere output is useful for discrete mature storms like supercells but not so for upscale growth of storms--convective lines.  An example of mature supercells with continued high ProbSevere values valid at 2330 UTC 7 May 2014 near the Red River is shown in the Figure 1.

Figure 1.  NOAA/CIMSS ProbSevere, MRMS composite reflectivity, and NWS warnings valid 2330 UTC 7 May 2014.  This is an example of continued utility of ProbSevere output for mature supercell thunderstorms.

It will be interesting to see if these observations continue through the experiment.  If so, the next generation of training should reflect this distinction.

-Sieglaff

CYS SVR threat decreasing

The line of storms that has produced a TOR in NE CO and SVRs in our CYS area has shown a slow weakening trend.  Many of these cells are now outside the CO total lightning coverage area so we can not view this.

Of  interest is an area of stronger storms with multiple SVR and TOR form NE CO into the North Platte, NE CWA…

These cells developed in a much more favorable environment…heating into the 60s and 70s today.  Areas under the “old” convection in WY/CO are in the 40s and 50s.

Also of interest:

The deeper Nearcast instability is now mainly centered in NE.   Best threat for further convective development and more strong to severe storms looks like it will be mainly focused just to the east (and to some extent southeast into NE CO) of the CYS CWA for the duration of the afternoon/evening.

db

CI Picking Up in Radar Void Area

I am watching a void area on the radar (bottom image) to see if there is any additional development northeast of Denver. CI values have picked up a bit (top image) in this general vicinity with some 70% probabilities. In that last few minutes probabilities have dropped. Waiting to see the outcome…

Jared Maples

New storms developing over central Texas

While mature supercells ebb and flow over the KOUN and KCYS/KBOU EWP WFOs, new convection was developing along the dryline in the San Angelo, Texas WFO.  Given the goals of the NOAA/CIMSS ProbSevere model of providing the probability a storm will first produce severe in the following 60 minutes (although EWP participants are noting utility of the ProbSevere model for supercell maintenance/severe threat as the model still uses NWP and radar information for those storms), the developing storms further south have become of more interest.

A developing storm exhibited strong vertical growth rates and strong glaciation rates within a favorable environment.  The ProbSevere model had increasing probabilities associated with storm, exceeding 50% at 2242 UTC 7 May 2014 (Figure 1) with further increase up to 97% when the NWS in San Angelo issued a severe thunderstorm warning at 2257 UTC 7 May 2014--15 minutes after the ProbSevere values jumped above 50%.

Figure 1.  NOAA/CIMSS ProbSevere, MRMS composite reflectivity, and NWS severe thunderstorm warnings valid 2242 - 2257 UTC 7 May 2014.

-Sieglaff

Expanded Lightning Jump View: 2234 UTC (CO/WY)

A wider look at the lightning jump grid at 2234 UTC of the NE Boulder CWA… All tracked storms (with greater than 10 flashes per min and within the 150 km LMA domain) are shown.  Three tracked cells show some type of sigma jump at this same time.  Two in green – 1-sigma (or standard deviation) and one smaller cluster, orange, with a 3-sigma jump.

The three-sigma jump storm corresponds with a ProbSevere 92% and MESH of 1.65 in.

-K. Calhoun, Week 1 Coordinator & Lightning Jump PI

CI miss and hit

First isolated cell builds with a CI of 40-50%.

Initial cell continues to grow with a CI of 80-90%.

Initial cell reaches a CI of >90% with a new cell developing to the southwest showing a CI 80-90%.

Initial cell eventially goes “poof”, with secondary cell continuing to grow and show a CI of >90%.

Second cell continues to grow with CI still >90% on the southwest flank.

Second cell continues to grow. A warning was eventially issued for this storm.

Scott Rudge.

persistent OTs

Convection in CO, NE, WY has exhibited persistent (30 min) OTs as of the 22:15 vis sat pic.

  

These have been over persistent strong…and at times severe…storms.  OTs have not always meant the storms were severe…and have seen SVR and even TOR under cells that have not had OTs detected.

db

Storms firing in southern Kansas

While not within the OUN WFO, EWP forecasters noticed a storm in southern Kansas developing as observed by the KTLX radar.  The NOAA/CIMSS ProbSevere model probability jumped to 74% at 2158 UTC with strong vertical growth rates and increasing MRMS MESH values (Figure 1).  The probabilities continued to increase and eventually the NWS Wichita office issued a severe thunderstorm warning at 2211 UTC 7 May 2014 (13 minutes after the probabilities increased to 74% (first time above 50%)) and was at 96% when the warning was issued.

Figure 1.  Animation of NOAA/CIMSS ProbSevere, 0.5 degree KTLX reflectivity and NWS warnings valid 2158 - 2112 UTC 7 May 2014.

-Sieglaff

CI providing lead time to initiation

As seen in the first image, CI is picking up on come cloud cover out ahead of the main convection. Probabilities were 50% and higher in this area at the time. Shortly after, thunderstorms began to develop just to the south of the main line that had already been warned on. These new storms later merged with the northern cell to produce more of a cluster. This example shows how the CI tool has some skill at time that cirrus isn’t inhibiting detection and can also raise awareness on development in a messy area.

More warnings and experimental warnings over southwestern Oklahoma

The storms over southwestern Oklahoma have recently increased in intensity and have warranted NWS and EWP severe thunderstorm warnings, with multiple reports of quarter to golf ball sized hail.  The ProbSevere values for these storms with golf ball sized hail are well in excess of 90%--even some at 100% for times.  The evolution of these storms and the generation of severe thunderstorm warnings is demonstrated in Figure 1.

Figure 1.  NOAA/CIMSS ProbSevere, 0.5 degree KTLX reflectivity, and NWS/EWP warnings valid 2132 - 2140 UTC 7 May 2014. 

A question was posed to the forecasters on the KOUN desk within EWP today, "There have been a number of storms with ProbSevere values in excess of 50% today, some of which seem to be false alarms, is this hampering the use of the product?"   The forecasters replied,

"No I like that the storms with the highest ProbSevere probabilities are in excess of 90% and these are the ones producing golf ball sized hail--these are not marginal hail stones.  The ProbSevere is drawing me to which storms are likely to need severe warnings in the near future.  The false alarm storms, while they have high probabilities today, are in the 70-85% range--which makes them different than these with 90+% probabilities.  I've easily been able to adjust to the probabilities and use them appropriately.  Also I know the ProbSevere model has not been validated as a tool to reissue severe warnings, but I like how the probabilities diminish a bit when the radar intensity decreases and then come back up if the storm re-intensifies.  If you look at this display (Figure 3) it has high probabilities on the right storms, but the ones further west are only (correctly )10-20%"

Figure 3.  NOAA/CIMSS ProbSevere and MRMS composite reflectivity valid 2204 UTC 7 May 2014.  This was the display the NWS forecaster in the EWP was referring to when he was discussing the ProbSevere continuing to identify the severe storms, but had much lower probabilities for the storms further north and west.

-Sieglaff