HWT Day 4: Freed from MCS

 Supercells of Midland, TX

As always, the first thing I did when I sat down this afternoon is AWIPS is immediately start fooling with the color tables. Today's chimera was a combination of OCTANE Cloud Top Divergence and Cooling with MRMS Reflectivity at -20° C. MRMS was modified to only show values that exceeded 40DBZ to interrogate the relationship between cloud top divergence and highest reflectivity. At this point in the week, I'm throwing science against the wall to see what sticks. 

My primary radar was MAF with MESH overlay for estimated hail size greater than two inches. Taking a look at MAF sounding this morning, it's giving me big hail vibes and I wanted to see how MESH correlated with the convergence signatures from OCTANE.

Spice Level: Cayenne  🌶🌶🌶

At first glance of the Midland sounding I'm drawn to the fact that we have a substantial amount of CAPE (~3000J/KG). A little bit of a CAP at 700MB should be quickly overcome as there is a pretty good clearing in the clouds over most of Midland's CWA. Lapse rates from 7-500MB is nearly 9C/KM, with a pair of dry layers from 8-500MB and between 5-300MB. Our hodograph, while not perfect, is generally a straight line from left to right.  Effective shear is around 45kts. My conclusion: hail. Possibly big hail, and we may have a chance of some splitting cells at some point today.

One of the cool things we noticed that the OCTANE Divergence/Cooling product picked up a splitting super cell around 15 minutes before the radar did. This was really cool and could be an excellent way of identifying splitting supercells and getting out warnings with an extra 10 minutes of lead time.

The inverted images on the OCTANE DIV/COOL product I think does a good job of highlighting areas in which those storms have entered an environment in which they are ready to split. Studying splitting supercells using sat imagery may give us abetter understanding of when storms are more likely to split. 

Thanks for accepting me into this HWT! I got a lot out of it and am excited for what satellite can do for ops in the future!

-Charmander

Making it how I like it- A look at different ways to view OCTANE

Below I want to provide some (4) examples of how I wanted or needed to adjust the OCTANE products to make the displays work best for me.

The two below graphics are of the octane direction product. Notice a difference between the two? The one on the bottom should look more sharp. The difference was normalizing the reflectivity of the red visible satellite channel (CH 2) so that the visible imagery comes out sharper. While done automatically o nthe CIRA slider product, it is not yet functional in AWIPS. This is a reminder that if you are using this product in AWIPS (currently) you need to keep manually updating the sharpness.

Here is another example from the speed product where the top image has not been normalized for the zenith angle and the bottom one has been.

This is definitely a feature that I would want to see included before release on AWIPS because while it is not an inconvenience to adjust it, it is easily forgotten in the midst of warning operations.

Example two of making OCTANE look how I like it.

While trying to learn how to interpret the direction product from OCTANE, I was offered the barb alternative available on the CIRA slider. I was not a fan. While the direction product has been the most difficult to interpret 

Example three of making OCTANE look how I like it.

Here I changed the bottom left of the octane 4-pannel. This was the Divertgence and cooling product. I was discussing with the developer how my eyes were tiring looking at the cyan field after a shift (bottom image) and that I thought it was washing out the divergence and the cooling fields some. We remedied this by lowering the saturation on the cloud top height field. I thought 45 percent saturation was preferred because I was cautioned that going any lower would not allow a forecaster to differentiate the cooler convective cloud tops and the white Red-visible cloud tops of lower clouds.

Example four of making OCTANE look how I like it.

I was struggling to interpret and use the direction tool, which I believe to be on me and not on the tool. To try and remedy this, I was playing around with the color scheme of the direction parameter. The top display shows the regular color scale from the developers and on the bottom is the one I provided input to create. The reason I went for this color comboe

-Kilometers

 

Using GREMLIN/OCTANE alone to Assess Mesoscale Conditions and Monitor Storm Intensity

 Quick MesoAnalysis Comparison

Using the OCTANE speed sandwich we were able to quickly assess the SPC mesoanalysis of upper level winds entering our CWA. The SPC mesoanalysis indicated upper level winds of around 100 kt spreading through the Big Bend part of TX ahead of an upper level trough. This analysis was quickly confirmed by the OCTANE speed sandwich data. High cirrus was streaming into the region ahead of ongoing convection and was sampled at 105 kt in the OCTANE data. This real time analysis supports the objective analysis by SPC and yields higher confidence in the data. 

Image 1: The OCTANE speed sandwich showing velocities >100 kt across the Big Bend region of TX spreading into the southern parts of the SJT CWA.

 

Image 2: The SPC mesoanalysis of 300 mb winds indicating a 100 kt upper jet spreading into the Big Bend region of TX.

 

 

GREMLIN/OCTANE to Monitor Convection without Radar

 

We were tasked with monitoring convection across the SJT CWA with an ongoing DSS event in San Angelo. While convection remained to our west and east, there were a few thunderstorms that developed across the northern part of our CWA. We did not have radar data, but were able to use OCTANE speed/direction products to assess storm strengthening trends though the afternoon. In combination with the GREMLIN radar emulation, we were confident in our assessment of storm strength even without supporting radar/MRMS data.

 

Image 3: The two storms of concern are to the north of San Angelo and are noticeable in the speed/direction imagery. There were a couple notable periods of cloud top cooling and increased divergence, but the uniformity in the mid/upper level wind speeds and direction increases confidence that these storms are either maintaining intensity or weakening.

 

Image 4: The GREMLIN data showing two distinct cells across the northern part of the CWA. These storms maintained intensity for about 30 minutes without any noticeable change in the OCTANE data. As the OCTANE wind speeds decreased, the GREMLIN data showed that the storms diminished overall.

 

These two products (OCTANE speed/direction and GREMLIN) used together can increase confidence in warning/decision making, particularly in the absence of radar/MRMS data. While there are some latency concerns with GREMLIN, further testing of these data should be continued.
 
 
 

Making Four Panels into one with GLM Data Quality

 When we started the testbed the GLM products were being displayed on a a four panel display (GIF. 1). This display works fine to find areas of poor data quality, like where the white and yellow pixels pop up in the top left of image 1. However, in it's four panel setup, I felt that the display took up too much space. As a group, we worked to merge them into a single panel.

Image 1: GLM data quality suite, with Data Quality in the upper left, Background image i nthe upper right, Day cloud phase in the bottom left overlaid with GLM flash density, and the Ch.2 Red visible in the bottom left. Please note that the background image has an erroneous color curve on this display; it is usally closer to the red visible in brightness.

Below are examples of single panel displays created. In general, folks preferred the display in GIF two, but I thought there were ways for ways to merge the two designs. My desire would be to have the GLM background image provide texture to the data quality product like in GIF three, but to have the data quality product maintain the sharp good/bad color curve shown in GIF two. Even more preferred is a color bar like in GIF four, where the "could be poor" but not nearly or fully saturated values are highlighted in another color (red in this example).

GIF 2: A single panel version of the GLM suite of products. The GLM background image is the background and unchanged. The GLM Flash Density is overlayed unchanged next. The GLM data quality product Is overlayed on top and has been altered such that the nearly saturated and totally sutured bins show up as pink pixels. 

 

GIF 3:  A single panel version of the GLM suite of products. The GLM data quality product is plotted unchanged. Overlayed and the reason for the blue-ish hue is the GLM Background Image which has been made transparent in the black. 

GIF 4: The same as Gif 2, except that the poor-ish data quality from 10 to 40 percent is highlighted in red.

What did you think was the best single panel layout?

 

 

I also want to share something about the data quality scale (Img. 1). In its current format, the scale is not intuitive. Good data quality ranges from roughly 50% to 90% on this scale (blues and greens). Poor data quality is from 50% to 10. Then, the poorest data quality is white and yellow, and represents nearly and full saturation. I was ready this wrongly, such that 50% to 90% was poor because it was closest to the nearly and fully saturated parts of the scale. Instead, it is reversed. Before this product goes to operations, I would want to see the color bar made more intuitive.

 

 Image 1: GLM Data Quality, focusing on the scale in the upper left. Poor data goes from 10%, at worst, up to 90%, then nearly and fully saturated after 90% in the center-right of the scale.

-Kilometers

Color Curves Reign Supreme in OCTANE Products

OCTANE Night Direction and Speed Products

    Today's challenge was to try and see if we could draw out features better in both the OCTANE Night direction and speed products. The night products rely on IR satellite imagery and the lower resolution washes out a lot of the features we can normally see in daytime imagery. The easiest way to try and attain this was to first play around with the color curves in the IR brightness temp (CH-13-10.35um) overlay. Adjusting the colormap starting with black at 40 to white at -45 back to black at -80 then interpolating the 'Alpha'.

Original OCTANE Night Color Curve - Direction

 

 

 

New OCTANE Night Color Curve - Direction

 

Original OCTANE Night Color Curve - Speed

 

New OCTANE Night Color Curve - Speed

MesoAnalysis Support and Lack of Radar

More emphasis has been put on mesoanalysis in severe weather operations over the last several years and utilizing the OCTANE products has demonstrated that it has utility in confirming objective analysis. Scattered thunderstorms were developing across parts of the Carolinas today, but the mesoanalysis suggested that severe thunderstorms would be favored farther south and east of our area of concern through the late afternoon. A quick look at the SPC mesoanalysis of effective shear showed values of 40-50 kt to the south of our area of concern. As thunderstorms started to develop, we were able to use the OCTANE products to confirm that larger areas of shear were present with our southern storms.

Thunderstorms were ongoing across parts of northern SC and southern NC through mid afternoon. The OCTANE speed products clearly show slower wind speeds aloft to the north, while our southern storms show much higher wind speeds aloft and even some divergent signatures to the southwest of our event. A few of these southern storms would eventually become severe. This product will be beneficial to short term operations in confirming local objective analysis.

Our case also required us to look at information without utilizing radar. The GREMLIN product filled the role of radar data, but some notable issues were present. The GREMLIN latency approached 20 minutes at time, which really makes the product unusable for real time warning operations. However, discussion with the developers indicated that some of these latency issues could be overcome. The utility in the product comes from its visual representation. Forecasters are used to looking at radar data, and this product offers something similar. Latency issues must be overcome for it to be usable in operations. The product overlaid with real time lightning data would offer utility in areas with poor radar coverage or during times of significant radar outages. After working a few days with the other products, the GREMLIN utility really became evident in the absence of regular radar data. An example is shown below.

Columbia SC: Trucks with Food

 Protecting the Foodies

    A Severe Thunderstorm watch that covered most of Columbia's CWA was where we concentrated our forecasting efforts today. We found ourselves with no radar and were forced to make decisions on warnings with satellite only. As a result we made some modifications and combined the power of GREMLIN and OCTANE. Here's what that love child looks like:

One of the issues we had in making decisions based on GREMLIN data was the lack of information it provided. GREMLIN provides a radar emulation and given that it's a satellite based product it would be nice to see more information in the sample tool of what's being shown. Values like temperature at highest reflectivity and echotops could be inferred by GREMLIN to help forecasters make better decisions if radar wasn't available. The other issue we had with GREMLIN was the latency. Products were running anywhere from 15-20 minutes behind the rest of the satellite products that we were using. 

A few minutes later we issued our first SVR warning for the eastern edge of CAE CWA for winds over 60MPH and nickle sized hail (sub severe). 

As our storm moved out of the CWA we allowed the SVR to expire and took a look at the PHS Forecast model and compared it to the HRRR to prepare for the next round of thunderstorms. But both models seemed to agree that more TSRA was unlikely:    

Storms in GSP came together and eventually created a good line of thunderstorms from GREMLIN's point of view. GREMLIN was picking up some areas of higher DBZ and a lightning jump through the line was consistent with what we'd expect to see on radar for a SVR. A warning was issued on a line of storms:

The line started to fall apart as soon as it hit the CWA border. We allowed the warning to expire without feeling the need to re-issue downstream. After that, storms no loner had access to some of the peak daytime heating that allowed them to become sub-severe during the afternoon. 

-Charmander

LightningCast Good For A DSS Heads Up...Dashboard Needs Some Work

    Storms off the the west in New Mexico approaching our DSS location in far NW Texas were a great way to test the LightningCast and the associated LightningCast Dashboard. Our DSS location was focused on the "Rita Fire" that was along Hwy 385 between Dalhart, TX and Boise City, OK and partners wanted notice of a 50% probability of lightning within one hour within a 25 nm radius around the fire. Convection was ongoing across NE New Mexico one storm showing deviant motion. This storm was warned on at 2043Z for 1.5" hail and winds of 60 mph based primarily on radar; however, Octane was being used to monitor the storm as it tracked southeastwards towards the CWA boundary. Octane showed consistent divergent signatures in the direction product and good speed decreases on the upshear side of the updraft before being contaminated by an anvil from another storm to the southwest. 

    As the supercell dropped southward from NE New Mexico, the LightningCast product probabilities started to increase into the Rita Fire range ring. It was noticed that the LightningCast Dashboard remained at around 10% despite 75% probabilities along the range ring boundary indicating that the dashboard is focused on a point. This is fine if you are only concerned about one point, but most outdoor events need time to evacuate people or move equipment from the location.

While it was nice that the LightningCast Dashboard allows you to choose different time ranges and an auto refresh rate per a drop-down, I think some other options might be more helpful for support of an event. To help provide a head's up to partners, it would be valuable to have an option to input a specified area that you would like to be alerted for...with the option of inputting those criteria. For example, being able to have a range ring or being able to draw a polygon for the area of interest and then inputting lightning criteria in there to alert you when those values are reached within your range ring or polygon. The criteria provided for the Rita Fire of 50% probability within the range ring would have alerted us on the Dashboard that the threshold had been reached. As far as the LightningCast product overall...it was an excellent resource for monitoring then alerting the Rita Fire partners once their criteria had been reached. A quick screen grab can be sent to partners as well to indicate where the greatest lightning threat probabilities are located.

- Vera Mae

Would you warn?

 

I want to walk you through a thought experiment that I had today during this event in the Panhandle region of TX/OK, as it enters the southeast corner of the Amarillo forecast area. From the Amarillo forecast area, highlighted in white in GIF one, the environment becomes increasingly more unstable. The storm in question is moving through the bottom right corner of the forecast area moving southwest to northeast. 

Here is the question I want you to ponder. Would you warn on this storm with only the environmental information and the two satellite products provided in GIF one? We are assuming we don't have radar data right now. 

GIF one: CH-2, Red Visible GOES-16 1 minute imagery from the MESO-1 Sector. 

Okay, I know, we would normally have the prob-severe tool, and lightning data. But I didn't look at it. So what was my answer? A hard maybe because I would want to make it a group discussion with the other forecasters on shift. My vote would be to warn, but with the lack of information, I would have no problem waiting for a storm report or some other data source.

Now, ask yourself the same question. Would you warn on the cell moving through the bottom right corner of the forecast area? But this time, you have access to the OCTANE suite of products. 

Have an answer? Let me share some further information I collected from the developer before we share out answers. 

I discussed with the developer what correlations existed between the OCTANE speed and divergence products and the severity of a storm. None had been established yet and the rule of thumb I was left with was that the storm top diverge observed in OCTANE speed products were about 30% to 50% less than the radial velocity divergence signals we normally use with data from WSR-88D. So where we measured a divergence speed of 40 kt in the OCTANE Speed product, the radar velocity should be around 50 to 80 kt, which could definitely support severe hail. 

So, I said I would have been comfortable warning with just this additional OCTANE information. That said, a discussion may still be held, but I think this would be the information that gets the warning out more quickly. If I had a table of correlated values and severity? My confidence would sky rocket!

-Kilometers

Update: There was a storm report generated from a mesonet observation that observed a wind speed of 78 mph. 

Monitoring Convective Initiation

 Convective Initiation along Dryline

We were monitoring convective development along the dryline in northern Nebraska with DSS being provided to a chemical spill across the northern part of the CWA. While most of the initial convective activity remained to the north of our DSS site, additional convective development occurred along a dryline to the south along HWY 83. These storms struggled to maintain updraft intensity and the OCTANE speed and cloud top cooling product was used to monitor the strengthening of these storms. Initial radar data showed spotty convective development, but the OCTANE speed product was particularly useful to show that uniform speed/shear information over the developing convection. None of these storms exhibited any significant divergence aloft and remained nearly steady state for about 20 minutes before weakening. Of note, the PHS and HRRR both showed this area to be void of convection with storms to the north and south. This activity diminished as the sun set. The OCTANE products did increase confidence in our thoughts about further development through the late evening.

The Machine Learning Foretold the Future

While monitoring convective initiation in a sea of Altocumulus castellanus clouds as they moved into a more convectively unstable environment, I noticed that the GREMLIN Emulated reflectivity in Briscoe County was higher than the MRMS (GIF 1).  The GREMLIN emulation was producing a DbZ of 50+ while the MRMS was showing a 30 DbZ. To me, this difference was operationally significant as I would pay closer attention to the developing 50 DbZ feature than I would the 30 Dbz feature. So far in our experiment, this was the opposite of what I had experienced and I initially thought the emulated radar was wrong. I jumped over to the OCTANE products to get a sense of how rapidly the cell was building (GIF 2). My assessment of the cell seemed to confirmed my initial opinion that it was not growing as rapidly as I would have expected given that there was not a strong gradient in the speed or direction products. 

Gif One: GREMLIN Emulated Radar on the top and MRMS Composite Reflectivity on the bottom.

Gif Two: OCTANE Products (Speed Upper-left, Direction Upper-Right, Divergence Bottom-Left) and Day cloud Phase RGB (bottom-right)

My opinion changed 10 minutes later.  Seen in image one, the MRMS with the next update showed a similar storm intensity to the GREMLIN emulation. This is impressive when one considers the latency of GREMLIN is greater (about 10 min) than the latency of MRMS (about 2 min). The GREMLIN product actually delivered a more operationally useful product sooner, despite having a greater lag. The GREMLIN product continued to show this ability for two more storms developing further back in the line!

Image One: GREMLIN Emulated Radar on the top and MRMS Composite Reflectivity on the bottom.

 -Kilometers

LightningCast in the Rockies and Appalachians

Forecasters are working in the Amarillo, TX and Columbia, SC county warning areas today. One tool they are evaluating is ProbSevere LightningCast, which helps forecasters by providing objective guidance on lightning potential in the next 60 minutes. 

Afternoon thunderstorms on high terrain are a common occurrence in warm weather. Upslope flow can provide lift for convective initiation. It's difficult to monitor all of the convection developing across a region at once, but LightningCast can help forecasters quickly see which clouds are glaciating and quantify the near-term lightning potential. 

Lead time to the first GLM flash varies widely, but we've found that on average you can expect about 20 minutes of lead time to lightning initiation, measured from the 25% threshold. We are still trying to skillfully extend the average lead time to initiation.

Out east, a short wave trough is forcing convection today in and around the Appalachian Mountains. One new product that forecasters are looking at is the probability of ≥ 10 flashes in the 60 minutes after a GOES ABI scan. This product is meant to give forecasters a heads up on intensifying convection, specifically the initial intensification from benign thunderstorm to one that could develop into a severe threat. The ≥10-flash probability is also an indicator for an increased chance of a cloud-to-ground strike.

In the animation below, two cells achieved flash rates > 10 flashes. On the eastern cell, LightningCast had 35 min (10 min) of lead time to the first 10-flash obs, measured from the 10% (25%) contour. On the western one, there was about 10 min (5 min) of lead time from the 10% (25%) thresholds. Users should note that we have found the P(≥ 10 flashes) generally produces better lead time in "clearer" or "cleaner" scenes, with little convective debris near the developing cells.  

- John Cintineo

Consitency with Lightning Cast

 As a QLCS moved through Tennessee and the threat for severe dropped, we were thinking of a way to message the persistent lightning threat that would still be present. We leaned on the Lightning Cast tool to message this threat. 

 

 Image one: SPC risk categories with colors.

 

Before we made the image though, the idea came up to re-do the contours such that they better aligned with the style guide the NWS, or SPC more specifically uses (img. 1). Image two shows how we added two contours and realigned the colors to add consistency with other operational areas of the NWS. 

 Image two: Lightning Cast with a new 5 and 90 percent contour added and colors of the contours aligned with the risk colors used elsewhere in the NWS. In the background is channel 2- Red Visible satellite.

 

We compared this with the base style from the lightning cast tool and we felt that our updated style better captured our eyes and made it simpler to interpret by us. We also felt that the public would have a better chance of understanding the product if the colors were more consistent.

HWT Day 2: Protecting the Quartet

Protect the Quartet!

Today we were charged with the noble task of watching over the the Quartet Festival located in Lawrenceburg TN. Gaps in the cloud coverage allowed us to utilize some of the satellite products a little more efficiently today. Our first sign of trouble came as convection began to form out in front of the main line moving WNW out of Alabama. LightningCast 60-min prob gave us our first initial heads up that lightning was possible with storms forming out ahead of the main line. A combination of Octane overlaid with GLM data was the primary source in our decision making to issue a notification to the event organizers. Thanks to our quick decision making, everyone is alive to sing again another day.

We combined lightning cast and ENTLN data with the radar to provide ground truth on when lighting was first scene within the 15 mile range ring which allowed us to issue follow up messages regarding the likelihood of ongoing lightning potential. Requesting a LightningCast point too also gave us confidence in issuing notifications to the event organizers.

Discussing the LightningCast Probability data with Kilometers we were discussing ways to get more information out of it. We settled on loading the LightningCast as an image rather than contours. This combined with the sample tool and overlayed with Total Lightning products was more useful when forecasting for a specific DSS point. We also went ahead and limited the data being showed on the lower end of the GLM Flash Density. We didn't want to exclude the Flash Density on the lowest end all together, but we wanted to highlight and compare the GLM Flash Density to the areas with the highest Octane SpeedSandwich. Our end result was this GIF below:

Today was a day to dive into GLM and Lightning probabilities. Once we settled on what we wanted to look at to make DSS related decisions, we realized that it wasn't intuitive to the public. We needed a way to redesign the the Lightning Cast data to be easy to look at to the public. Because the NWS already has a color table for threats utilized by our National Centers, we decided to model our threat level based on the SPC's convective outlook to create new colors for the contours. The following graphic is the end product of that:

As the system moved past, identifying areas that we could issue the all clear on was our next priority. The LightningCast created a nice looking bell curve that lined up with the time that the MCS moved over the event and showed the trends of the storm began to wind down.