A Quick Comparison

Getting a handle on the area of interest across southwestern Wyoming and the panhandle of Nebraska, we see a few cells popping up near the WY/NB border. Taking a look at the four-panel image below, you’ll see the CIRA/RAMMB NSSL-WRF simulated satellite imagery on the left hand side (IR on top, WV on bottom) and observed on the right (time: 20Z). So far, the simulated imagery seems to have a decent handle on storm initiation near the WY/NB border in particular, which allows for increased confidence in its forecast, even if the spatial placement isn’t necessarily exactly spot-on.

The simulated satellite imagery seems to be overdoing it just a bit so it’s not as helpful if you’re trying to assess specifics but if you can compensate by interpolating down in your mind’s eye, it is still a good product to increase situational awareness.

Linda Gilbert

GOES-West capturing CI in SE Wyoming

Convection is developing along the front range of the Rockies over SE Wyoming this afternoon, with GOES-West capturing some of the development. Earlier versions of GOES-West had issues with the 4 minute scan of GOES-West where cloud objects consistently ran "hot". This is a sample 4 minute scan from GOES-West, where only one object at the current time (1945 UTC) displayed higher probabilities (>70%) of CI. Additional database training for GOES-West has helped temper 4 minute scans and brought it more inline with the probabilities seen from GOES-East.

1945 UTC GOES-West CI imagery

NOAA/CIMSS ProbSevere model display

It's a quiet start to the 3rd week of the HWT Spring Experiment. As forecasters become acquainted with the various products being evaluated, we've seen some convection fire in southeast Wyoming, but it hasn't amounted to much yet.

The NOAA/CIMSS ProbSevere model is a statistical model that fuses together satellite growth rates derived from GOES cloud products, MRMS Maximum Expected Sized of Hail (MESH), and RAP MUCAPE and effective shear (EBShear) to predict the probability that any given storm will produce severe weather. The idea is to use the 1st-order NWP environment predictors and observed satellite growth rates to give forecasters more confidence and hopefully more lead time on issuing severe weather warnings.

Figure 1 shows the display of ProbSevere in AWIPS-II. It is driven by shapefiles, which are contoured around radar-identified storms. By scrolling over the storm, you can see the ProbSevere readout. The probability of severe (SVR PROB) is first. This storm in southeast Wyoming had a 17% probability of severe. Next we see the environmental MUCAPE (949 J/kg) and EBShear (48.3 kts) computed from the RAP NWP model. Next is the MRMS MESH, equal to 0.26 inches. Lastly, the normalized vertical growth rate, characterized as "strong", and the cloud-top glaciation rate, characterized as "moderate". The strongest satellite growth rates in a storm's lifetime are used in the model. These growth rates occurred 1.25 to 1.75 hours ago. Beyond 2 or 2.5 hours, the satellite growth rates may not be that relevant to incipient storm development. This is something the forecaster should bear in mind when using ProbSevere, as well as other important environmental aspects. Recent research has shown that severe storms tend to exhibit faster satellite growth rates than non-severe storms.

Figure 1: NOAA/CIMSS ProbSevere model display, overlaid MRMS merged composite reflectivity.

-Cintineo

Wyoming and Nebraska CI

Good afternoon all. Grant H. signing on for the first blog post of the first day of the HWT. In learning about the GOES Convective Initiation (CI) tool, I felt this might be a good first place to start. CI seems to be the most useful in hunting down areas where convection will begin over the first 2 hours of and event. I can see this being good from the warning coordinator/mesoscale forecaster perspective in making decisions on where/how to divide up/sectorize radar mets in ops.  The stuff off to the left over eastern Nebraska in blue has a very low probability of initiation around 10 to 20 percent. Meanwhile, the cumulus field over the southeastern Wyoming in the Cheyenne area is reaching up to 50%. This helps to force focus to the locations of interest. I do see some problems here in that the algorithm has troubles IDing areas that transition from stratus to cumulus or from cirrus with cumulus embedded underneath such as in northwestern Nebraska… very little CI color is showing up, other than the sheared area in this image.

Grant H.

EWP Operations Update – Monday 5/19

Forecasters are getting acclimated to AWIPS-2 as we sit and wait for convective initiation ahead of a shortwave trough in SE Wyoming and W Nebraska. We are currently operating from the Cheyenne (CYS) and North Platte (LBF) CWAs.

The downgrade of the “SLGT” risk to a “SEE TEXT” at the 20z SPC Day 1 Outlook and the “WATCH UNLIKELY” Mesoscale Discussion will not dampen our spirits.

2000z SPC Day 1 Outlook for May 19.

Mesoscale Discussion graphic issued at 2012Z.

KCYS is showing the first 40+ dBZ echo of the day in Goshen County, WY and with the airmass over the region continuing to destabilize, we are holding out hope that we will see some isolated convection in both CWAs in the next few hours.

First 40+ dBZ echo in Goshen County, WY.

-Darrel Kingfield

EWP Week 3 Coordinator

Week 2 Summary

This week, the EWP had forecasters from the Louisville, Buffalo, and  Norman WFO’s, as well as a broadcast meteorologist from WUSA (DC CBS affiliate) participate in the Big Spring Experiment. Operations on Monday began in the Davenport and St. Louis CWA’s. Throughout the week, operations slowly shifted eastward as we evaluated the products with severe weather development along an eastbound cold front. These operations included the Detroit, Cleveland, Wilmington, Charleston WV, Pittsburgh, and Sterling CWA’s. One group on Thursday operated in the Shreveport CWA, where marginal severe weather occurred as an upper level disturbance moved through a region characterized by weak low-level moisture but steep lapse rates and only marginal instability. This unique environment posed some interesting forecast challenges, so it was neat to see how the various satellite products and OUN WRF performed.

Participants were able to use all of the demonstration products this week, which included  GOES-R and lightning products, LAPS fields, and the OUN WRF model finally on Thursday. There were many good blog posts written throughout the week highlighting the use of all of these products in various situations across various regions of the US. Below is some end-of-the-week feedback on each product from this weeks participants:

Simulated Satellite Imagery:

• This gave me a heads up on where clouds would move. There isn’t great guidence for sky grids, so I would look at this to see where stratus is moving, etc. if it was verifying well
• I think it is especially effective on the large scale because it picks up on large scale features well.

NearCast System:

• I liked and used it because it is observed thermodynamic data, of which there is very little
• This added value to my forecast process. For example, in West Virginia no boundary was evident at the surface, but there was a boundary in NearCast, and that is where convection fired. That sold me.
• I do not like to rely on NWP data, so this was nice.
• I really liked seeing the gradients, most of the storms developed in theta-e difference minima or moisture maxima or along gradients.
• There were a few cases where you saw decreasing moisture moving in, which was not picked up in the models, and it did have a big effect on storm development.
• In Wilmington, dry air moved in and storms decreased, but they actually did increase a later, so it was kind of inconclusive in this case.

GOES-R Convective Initiation

• Some times it was giving lead time of 30-45 minutes, other times it provided no lead time.
• It was more useful in rapid scan mode.
• I was very impressed with its performance, but sometime the lead timne just wasn’t worth it.
• I thought this product was really great during the daytime, but I do not see it being at all useful at night as it was very inaccurate.
• It was sometimes hard to get a sense of what the probs meant. If I used it, I would get rid of everything under 50%. I just don’t like that much clutter.
• It was particularly erratic around the Appalachian mountains.

Prob Severe Model

• It works awesome in hail situations. I am a fan of it for hail detection and determining which storms will produce hail
• It does have issues with linear storm modes.
• The best part for me was teh moiseover sampling and being able to look at the predictors. It really enhances your situational awareness
• It would be nice to color code the growth rates in the readout
• I noticed a lot of sat growth rates that were older than an hour, that mad eme lose confidence in the signal.
• I think it did increase my confidence in hail events, because I was saw a clear progression in probabilities
• When I saw over 80%, I had great confidence that that storm would become severe
• I do think it could give additional lead time to warnings
• I am fine with including the lower probs because the display is not obtrusive, and I like seeing the progression to higher probs.
• I think what you have now, for hail I would use this product today.
• The survey questions were good
• It gives you a good idea of which storm(s) you should be interrogating
• All participants agreed they would use this in their local WFO.
• Broadcaster: I would use this on the air. If there were a lot of cells, I would point to this storm [with the higher probs] and say that that is the cell to watch. Would not necessarily show probs, but could show colors, etc.

Overshooting Top Detection

• This was not useful for me.
• I see this being most useful when incorporated into another product. This would be a great benefit
• We were unable to use it at night when it is harder to see OT’s, and when many more OT’s are often detected as storms have matured.

PGLM

• I really like the total lightning data
• I’ve never used total lightning, bit I do like it

Lightning Jump Algorithm

• I think I could use this in a warning environment.
• I don’t mind the sigma values as indicators.
• An outline (like prob severe)  might be better then the blob
• It might be good to incorporate the LJ product in the prob severe tool
• I don’t see the zero sigma being necessary
• I told AWIPS-II to blink sigma values that were greater than 2.

Tracking Tool

• There are too many circles on the screen, too much clutter.
• I would prefer to have one circle that you just put on the cell, and it gives you the meteogram.
• Entering the cell id # to track the storm might be a good idea
• I don’t really mind the circles, but I just can’t see myself using this in a warning situation.
• I can see this being used after the fact, looking at a storm, but not in real-time. It is too labor intensive.
• I like the graph itself, but the actual functionality is bad. 
• It is difficult to move the circle-track to align with the track of the storm, especially when many images are loaded. Also, sometimes it does not track at first, so you have to move it around to get it to track. Finally, changing the size of the circles is frustrating, as making some circles bigger makes other smaller. 

GOES-14 SRSOR (1-minute imagery)

• It's great!
• I saw subtle boundaries that I wouldn't otherwise see
• We want quicker satellite updates, it's a no-brainer
• No worry about information overload with this
• I will prefer to view the raw data, but I do see it being useful as input into other products as well

Other:

• I thought the training was good.
• The week was very well organized, well done, and I liked that we stuck to the schedules, it made things very easy.
• I liked the relaxed environment
• Less structure was good, it gave us freedom to see what works well for us.

- Bill Line, SPC/HWT Satellite Liaison and Week 2 EWP Coordinator

Forecaster notes strong ProbSevere in Arkansas

With the Little Rock radar down all afternoon, a forecaster working the warning desk in Shreveport, LA, noted a strong ProbSevere value on a storm in central Arkansas. This storm exhibited strong satellite vertical growth and glaciation rates at 22:45 UTC, jumping the probability of severe from 8% to 36% (see animation below). In a marginal environment, the MESH slowly increased to 0.76", most likely an underestimate, on account of the missing KLZK radar data.

The forecaster said that using the strong satellite growth rates and MESH information from ProbSevere, the available radar tilt data from KSHV and KSRX, and the history of how other storms have evolved in this environment, he would have issued a severe thunderstorm warning for quarter-size hail (he was on the warning desk in Shreveport, so he could not issue an experimental warning).

The ProbSevere probability for this storm first exceeded 40% at 23:00 UTC, which has been very high for the environment today. This was 24 minutes prior to an official NWS warning at 23:24 UTC (figure 2). The ProbSevere eventually exceeded 80%.

Figure 2: NWS warning issued at 23:24 UTC.

-Cintineo

Synthetic Water Vapor Imagery (lower left pane) handles observed “Water Vapor Hurricane” (lower right pane) over Iowa fairly well.

Synthetic Water Vapor Imagery (lower left pane) handles observed “Water Vapor Hurricane” (lower right pane) over Iowa fairly well.

Shawn Smith

Simulated IR imagery vs. convective cells in DC

The simulated IR imagery showed the cold front’s areas of convection on the leading edge of the storm in our target area of Virginia and Maryland on Thursday afternoon.



It matched up spatially with what we were seeing in reality on the rapid scan GOES IR imagery at the same time stamp, 20Z.



The area in blue on the simulated IR indicates cloud tops colder than -60C. This shading doesn’t show up on the real IR image at all, but the cloud tops do have temperatures below -50C in the same convective regions. It looks like the simulated IR is going to overdo the convection, especially for the southern cell over central Virginia, but I thought I’d keep an eye on it to see if a convective cell spawned a severe warning in that area.

The simulated image valid at 21Z shows the strongest convection has shifted further east and is concentrated into one cell.



That cold cloud top maximum in northern Virginia is much smaller and not nearly as cold in the real GOES IR image from 21Z.



During this time period, a line of strong to severe thunderstorms was pushing through the DC Metro area.



Comparing the radar data to the simulated IR at the same time stamp, it appears that the small clusters of convective cells were not well resolved by this product. In fact, the clusters of storms to the south and west of DC were either completely missed by the simulated imagery, or the placement was off by about 50 miles (clouds too far to the southwest to be a match for the convective cells).

This was a day where we had limited tools for severe weather forecasting in the DC Metro area. The threat for hail and tornadoes was very low. ProbSevere, convective initiation, overshooting tops, and PGLM products were rendered useless because of a lack of convection and lightning parameters.

ProbSevere in a low-topped storm environment

In the mid-Mississippi River valley, storms slowly began popping up in an environment characterized by low MUCAPE and moderate to higher effective shear. Cold air aloft has generated enough instability to initiate storms, however. ProbSevere has generally forecasted 5-30% probabilities of severe for storms in this region (see animation below), exhibiting differing degrees of satellite growth and MESH.

Thus far, only 2 severe hail reports have been recorded in Arkansas. ProbSevere's forecasted probabilities have generally been in a neighborhood you would expect in this environment -- marginal. As might be expected with many storms with low probabilities, few reports have been observed (two severe reports thus far). Several penny to nickel size hail reports and sub-severe wind reports have also been recorded.

The storm with the 1" hail report had a 9% probability of severe (it was not warned by the NWS), and the storm that produced golf ball-sized hail had a 38% probability of severe 7 minutes prior to the report (not shown), and coincident with an NWS warning.

The images below show cases where there have been a some "correct nulls" for ProbSevere. In the first case, ProbSevere only had a 4% probability on a storm in Arkansas, when the NWS issued a warning. It never produced severe-level weather. Similarly, the second case below shows a ProbSevere contour near St. Louis, MO, with only 4% probability of severe when it was warned. No reports were recorded from this storm either.

Example 1:

 Example 2:

-Cintineo