Running with SZAs: Lessons to Carry Forward From the Spring 2026 SPG

Today marks the final testing day of the 2026 Satellite Proving Ground. It has been an illuminating week for myself – a pun we’ll get back to in a second. But first, some serious business. The Satellite HWT has proven to be a beneficial experience for a multitude of reasons. Foremost of those is the opportunity to test and help shape development on a whole new generation of satellite-based forecaster tools. But one shouldn’t discount the benefit that comes from the process, the collaboration, and the chance to work with scientists one might not get to meet otherwise. For me at least, this collaboration has been the most satisfying part of the week. A sincere thanks to the organizers, developers, and other forecasters who have made this week such a treat.

Ok, now back to the pun. This has been an illuminating week testing out five different satellite-based tools. The one tool that I haven’t shared imagery from this week is the SZA azimuth-corrected imagery. There’s a simple reason for that: it is designed to increase the visibility of day-cloud satellite products in low light scenarios around daybreak and dusk. The actual experimental time for this project is from early-to-late afternoon across the CONUS. Not exactly an ideal time.

So with that in mind, one of the first things I did today was to check in on SZA imagery from off the coast this morning. This side-by-side comparison of Day Cloud Phase between SZA and non-corrected imagery shows how powerful of a tool this could be.

Figure 1: SZA-corrected Day Cloud Phase (left) and non-corrected imagery (right) over the Atlantic Ocean early this morning

It’s one thing for Day Cloud Phase to gain more definition in the updraft/anvil pinks right at daybreak. That is valuable, but only so much. We already kind of know what’s happening at that level. The forecaster can benefit so much more from increased brightness right by the surface, where dynamic processes and even the texture of the clouds can help us discern so much.

I spent most of today as the “warning operator” at the Topeka simulated WFO. This meant I wasn’t experimenting with satellite products as much as I was testing how they could be used in warning operations to increase confidence in severe impacts from a thunderstorm. Invariably, my real-life operations rely on Day Cloud Phase as just one of the best products to detect vertical motion trends within convection. There are multiple forms of Day Cloud Phase that one can use within this experiment. I am particularly intrigued by MesoAnywhere’s ability to level the playing field, so to speak, when Mesoscale sectors aren’t available for the GOES satellites. Unfortunately, I did not proof the gif pulled off of AWIPS to try and demonstrate that point, and it is not time-matched between different products.

Figure 2: A gif that I did not realize was going to be frame-unmatched showing the three different types of Day Cloud phase available without a meso sector

The products that we tested this week all have the potential to enhance operations in the NWS. I look forward to reviewing them with my colleagues in the weeks to come.

Sabrina Carpenter

Tags: None

Cloud Top Cooling Advanced Warning

Widespread thunderstorms and several supercells were visible moving east-northeast over Kansas this afternoon. The GOES-19 EMESO-1 Cloud Top Cooling showed several large bursts of cooling as visible in two separate supercells in the image below. The northern one showed a tight circulation and a possible brief tornado roughly 9 minutes after the initial burst of cooling. The southern cooling burst strengthened the outflow burst and produced damaging winds that also showed on radar 9 minutes later. A supercell to the west of the region in the image also showed major cooling about 5 minutes before radar indicated a possible tornado and a warning was put out. Based on these few observations, it seems reasonable that severe thunderstorm warnings could be put out long before radar confirmation based on significant cloud top coolings, at least in mature thunderstorms where conditions are favorable for severe weather. In areas where radar is not accessible, this could also be used to help assist in putting out tornado warnings.

Figure 1: On left: GOES-19 Octane Meso Anywhere CH-13-10.35um and CH-02-0.64um with GOES-19 EMESO-1 Cloud Top Cooling overlaid. On right: Ktwx 0.5 velocity (kts).

It is also worth noting the displacement of the cloud top cooling vs the radar signature due to parallax. Figure 2 shows that this area has a parallax of roughly 19 km to the northwest which would be very important to keep in mind when putting out warnings for any area based on satellite imagery.

Cloudius

Tags: None

Last Impressions

 SZA

Love it! Can’t wait for it to be implemented. The amount of detail that can be ascertained both earlier in the morning and later in the evening is unbelievable. Main applications in my area would be for fog detection and snow squalls. While maybe a bit niche, the snow squall phenomenon is always a challenge. It’s a shallow feature that really ramps up in the afternoon and evening hours. If it’s too far out our radar really can’t pick up on it and the day cloud phase is often one of the best tools for tracking.

OCTANE

Continues to impress! The cloud top cooling tool remains a slam dunk for identifying cells that are quickly growing upscale. The example below shows further cooling after a decent anvil had already developed.

Image below is from the same time above 2112Z

8 minutes later the storm looks to be producing hail

Lightning Cast

Another solid product. I didn’t notice too much difference between the two versions but the parallax fix is very welcome. I personally won’t use the dashboards as I like to see the data overlaid with satellite and radar. Just seeing a chart with numbers moving doesn’t work for me. However, if I would recommend adding a sound feature to the dashboard. The only other suggestion I have is using less contours. I prefer the 10, 25, 50, 75 that are already in AWIPS.

Lightning Stoplight

First time ever using this tool but it will definitely become a mainstay in my arsenal for helping with DSS. I really don’t have a lot of feedback to give. The tool is simple and easy to use. I like overlaying radar on top of it in my procedures and I’m neutral to the idea of changing green to blue.

Geoxo products

These were neat to look at over Africa and Europe, but I found myself really struggling to use them stateside. I may need more training to fully understand the benefits of the synthetic satellite data, but I can’t see myself using it. I already have a ton of different CAMs I can look at, so I’m not sure I understand the benefit of looking at the synthetic satellite. I do have high hopes for the WVT tool though. The color table is very difficult for me to tease out what I need from it though. I tried manipulating the ranges a few times and also changed dry to brown and moist to green, but it still seems pretty difficult to pick out features for myself. What I really want is something to help me track boundaries. Jason shared a really neat animation over Africa that highlighted boundaries very neatly, would love to see if that could be implemented somehow. My end goal is I want to see moisture boundaries from lake breezes, decayed thunderstorms, or different moisture fields such as evapotranspiration

IsthataTOR

Tags: None

When A “Bolt From the Blue” Shatters Your Confidence

Today presented one of those moments where you feel pretty confident in your messaging, but then mother nature throws you for a loop. For the Red, White, and Lavender festival, a line of general thunderstorms was moving over the festival right around start time of our forecast period so this gave me an opportunity to test the Stoplight tool and LightningCast in a messaging scenario for lightning cessation.

What seemed like a pretty straightforward case of DSS messaging to festival goers regarding the line moving east and lightning threat ending turned into a stressful situation for me. At roughly 1935Z, LightningCast showed probabilities were trending downward while the Stoplight tool was also trending toward yellow and green to suggest we would be approaching that “all clear” state. See the LightningCast dashboard graphic below with the main decision timeframe boxed in yellow. So, I confidently messaged my Festival Coordinator an estimated all clear time to be the top of the hour.

 

Figure 1 . LightningCast Dashboard over Manhattan Regional Airport, near the Red White and Lavender Festival depicting lightning probabilities trending downwards during my DSS decision timeframe.

But then at ~1955Z, the Stoplight tool turned red again WEST of my site (while the Festival itself was under green) as the FED picked up on what I’m guessing were anvil crawlers. I was surprised because the anvil looked pretty thin to me over that area so I honestly didn’t know what to think. Then, ENTLN lightning plot showed a number of cloud flashes within 10mi east of the site and eventually plotted a negative CG barely 6miles to the east of the festival at 2015Z (see Figure 2). Talk about shattering a DSS forecaster’s confidence!  It seemed to be one of those “bolt out of the blue” strokes, too close for comfort! To LightningCast’s credit, probabilities never went below 20%, with V1 even hovering around 40% the entire decision timeframe.  Mother Nature sure taught me yet another lesson in humility that one cannot be too confident no matter what!

Figure 2. The Stoplight Tool plotted with FED, ENTLN and DSS range rings at 2015Z. The black arrow denotes the location of the negative CG about 6mi east of the site.

I’d like to think that in a real scenario, I would have been a little more conservative in estimating a time of all clear. But with this being an experiment, I felt a little more bold in just “going for it”.

Figure 3. A longer loop of the Stoplight Tool, ENTLN and FED products, including the decision timeframe and point of confidence ruined.

And just for fun, because I’m hooked, below is a loop of the CONUS OCTANE CTC, Speed and Direction products for a rapidly intensifying storm over Cloud and Clay Counties.  Although not as seamless as the mesoscale versions of these tools, there is still a lot of utility in diagnosing storm development and strength with CONUS vs traditional satellite imagery. I’m even becoming more and more a fan of the “blues” colorscale for the CTC product (left hand panel) when plotted on the Ch 13 product. The stark contrast between the dark blues/purples against the rainbow of the IR anvil really catch the eye.

Figure 4 . ECONUS Octane CTC product (left), Octane Speed (middle), and Octane Direction (right).

Astrophage

Tags: None

Expanding my Horizons Beyond OCTANE

Took a look at the WVT data from meteosat 12 this morning on the coast of the Red Sea and Yemen and western Saudi Arabia. Can see thunderstorms initiating off the higher terrain and very high WVT values along the coastline and low values inland. This was reflected very well by the surface dewpoints on the Synoptic viewer.

Figure 1: Meteosat 12 WVT band over western Saudi Arabia and Yemen

 

Figure 2: Synoptic viewer data 0800 hours PDT.

Finally got a chance to use stoplight data at Miles City, MT this afternoon. The product is easy to use for the cessation of lightning over an event. I did wish it was parallax corrected though. I was kind of guessing on the images and with the ENTLN lightning data when the last lightning strikes were.

 

Figure 3: GOES-19 SPoRT Lightning Stoplight and ENTLN 5 minute lightning data

One way to improve the data is to use the top right color scale with the stoplight product. This keeps the color flow smooth but adding a dark red for the very most recent lightning strikes proved to be the most helpful to me.

 

Figure 4: GOES-19 SPoRT Lightning Stoplight Product with 4 different color scales.

 

Dry Thunderstorms

Tags: None

Overview: Week 3, Day 4

The final operations day of the testbed began with our last daily debrief session, and we talked a bit about the applications sounding information with respect to the future GeoXO mission. The forecast for today was…meh? We had storms in the forecast, but the environment during our operations was marginal for severe weather overall. One office was localized to Billings, MT (BYZ) to capture developing thunderstorms closer to the jet streak, while a second office was localized to Topeka, KS (TOP) close to a remnant MCV that was kicking off convection already by 16 Z. Both offices got two mock-DSS events again today. If thunderstorms were looking unlikely to form in the Billings CWA by ~20Z we planned to move them to the Miami, FL office. Thank you Florida for being our safety net for thunderstorms.

Two MDs were issued by SPC during the afternoon for our forecast areas, with one resulting in a severe thunderstorm watch.

 

 

With Jonny’s help I made a ‘test RGB’ from the Synthetic GXI imagery. There’s no physical basis to the RGB (yet!), but it was a proof of concept that we could do it. An example and the recipe is shown below. What features can we pull from this new imagery relative to water vapor at low levels and through the entire column? Maybe we can figure it out in future testbeds!

 

 

During operations, I showed forecasters another example of SZA imagery over California with the impacts of the marine layer with low clouds and fog. Forecasters also compared this imagery against the Nighttime Microphysics to see how long the ‘gap’ was between the two, and what features you could identify with each product. Additionally I found a case from a fire from sunrise through the lens of the Day Fire RGB.

 

Kevin

Tags: None

LTG DSS with LTG Cast and Stoplight

Upon starting the shift we had a supercell that was about 1 hr out from our first event. In operations our DSS forecaster would have immediately informed our partner that the cell (which had 1” hail report) is barreling in from the southwest.

Both versions of the LTG cast were providing a 70% chance of LTG within the next hour as well, with convection out ahead of the main supercell the heads up notification would also have included this information.

45minutes later we had our first lightning strike onsite ahead of the main supercell.

——————————————————————————————————————————

Activity began to wind down by 2030Z

An all clear call would have been made at 2100Z with the stoplight product largely vacant of any lightning detection. MRMS-10C was also outside of the 8mile range ring with little build up expected upstream. Lightning cast had also dropped to around 10%.

IsthataTOR

Tags: None

Every Cloud Has a Silver Lining – Satellite-Based Mesoanalysis

Satellite-based observations have become a larger part of an operational forecaster’s toolshed with each passing decade. Over the past 10 or so years since the launch of the GOES satellites over North America, forecasters now have access to extremely high-resolution, high-quality data. That data can be used for a wide array of potential benefits, and this satellite HWT is designed in large part to show us how we can widen out that repertoire even more.

With that in mind, I’d like to start out this Hump Day blog post by discussing one of the products that has not been featured as prominently in my previous blogs – GXI water vapor data. In essence, the next generation of weather satellites (GEO-XO) will have the kind of sensitivity that lets us really drill down into the absorption bands to see some neat stuff. That kind of capability won’t arrive until the 2030s, so for this week we are taking a look at what the data could look like via HRRR simulated satellite data.

Figure 1: WVT Ratio from the HRRR (left) compared to Split Window Moisture on GOES-East (right).

As you can see above, there are limitations to this analysis. For one thing, even perfectly initialized models are going to struggle to carry cloud cover forward given its sensitivity. And for another, models are not going to perfectly initialize.

Still, if you squint and look across the eastern portions of NWS Aberdeen’s CWA, there is an area of somewhat lighter grey feeding into the cloud band on the HRRR. This is suggestive of a potential moist pool in the region. Actual observations of this would help forecasters dominate the mesoanalysis space like never before.

Perhaps one of the most powerful uses of mesoanalysis tools came from the OCTANE speed-direction tools today. I have spoken at length about those tools, so won’t spend too much time on them. This gif just does a great job of summarizing what we might be able to do:

Figure 2: OCTANE Speed (top left), Direction (top right), Cloud Top Cooling (bottom left), and Day Cloud Phase (bottom right)

Here we have a storm on the north end of the cluster (yellow OCTANE speed, purple OCTANE cloud tops) and developing updrafts to its south-southwest. Those updrafts are occurring in an area of boundary-layer cumulus (shown well by their northeast or yellow motion in the Direction panel). Further to the east, there are clouds oriented along two axes: an area of HCRS (red in the Direction tool), and an area of stable billow clouds (yellow in the Direction tool). Knowing your mesoanalysis, this provides a tell that the northern updraft is likely to wither as it enters a stable boundary layer, which it did. It’s also a tell that further south, updrafts won’t have the same issue. As of the time of this writing, a supercell has developed out of that southern cluster.

This author would be remiss if they didn’t mention the in-person IDSS potential offered by the Lightning Stoplight tool. This has also been discussed previously, so I won’t belabor the details too much. But the ability to display a dashboard from your browser with basically a color-coded area showing how long it has been since the last lightning strike will go a long way toward helping partners understand when DSS activities may restart.

Figure 3: Lightning Stoplight in its web-browser-based glory.

Sabrina Carpenter

Tags: None

FGF CTC CTD Testing

Today’s DSS event brought us to the wonderful state of North Dakota in the FGF CWA. Unfortunately with no lightning at either DSS event today, I did not get to use the GOES Stoplight product, so I used Octane again. I continue to be highly impressed by the utility of the OCTANE products. You know the saying “A picture is worth a thousand words”? Well, to me, that is exactly what the cloud top cooling and divergence product embodies. There are so many features that can be picked out from this. Near the bottom left, there are multiple updraft attempts that try but most fail. It shows multiple storms developing in McPherson and Dickey Counties with cooling cloud tops and increasing cloud top divergence. Further to the west in Emmons county it shows a cloud top cooling signal but it ends up disappearing and subsequently the cloud top divergence decreases. Pairing all of the imagery with a visible band really gives some nice textures to the picture where you can easily point out quickly developing features. I find myself continually migrating back to the Octane products for situational awareness as storms approach the DSS event.

Figure 1: East Meso1 Cloud Top Cooling and Cloud Top Divergence with ENTLN lightning data overlaid.

Dry Thunderstorms

Tags: None

GEOXO storm development and LightningCast DSS

The GEOXO Sim 5.15um band and WVT ratio showed supercell or at least thunderstorm development near Gettysburg by 21Z. However the initialization of the upper level cirrus was underdeveloped and the main severe warned thunderstorm was farther north than observed. The lower level cumulus also appeared to be underdeveloped. There is a clear region in the sims where higher moisture is present and dryer air exists on either side, with the thunderstorms developing on the western boundary line.

Figure 1: GEOXO Sim 5.15um band and WVT ratio on top with observed GOES-19 IR an visible imagery on bottom. First frame shows Observations at around 20Z compared to the model and the second frame is in the future on the top an hour later.

While not severe or as intense as the Sim suggested, storms did develop and provided an interesting borderline case where LightningCastv1 showed barely 10% and v2 showed barely 30% chance of lightning in the next hour within 5 miles of the Gettysburg DSS location. A lightning strike occurred roughly miles from Gettysburg that was detected by the ENTLM and the Stoplight tool maybe 2 miles to the east. In this case, the addition of MRMS was clearly helpful in detecting the initial nearby convection.

Figure 2. LightningCast comparison between version 1 and 2 and parallax vs no parallax adjustment.

Cloudius

Tags: None