The bottom left frame depicts a well defined outflow boundary pushing ahead of the convection in northeast FL. KJAX 0.5 degree Z is the only product in which this feature is visible, and it provides a lot of additional context as to what’s going on in the mesoscale and how that will impact the downstream convection. Will note that GREMLIN correctly identified this overall trend of weakening outflow dominant convection and was more obvious in its presentation than the RALA or base Z (which had very similar reflectivity presentations).

MRMS RALA (top left), GREMLIN (top right), base reflectivity (bottom left and right from JAX and MLB, respectively)

– millibar

CONUS did better than the Meso sector with initiation throughout the majority of the event.

OCTANE did phenomenal with detecting the stronger storms, once again. I felt that I had ample lead time after hearing back regarding LSR’s from the storms I warned on. Following this, we noticed some pretty noticeable parallax issues with GREMLIN with respect to the warnings I issued.

I wonder how this would impact warning operations in areas where radar coverage is limited or nonexistent. Would this result in a higher POD and FAR for those offices as they’d likely have to draw larger polygons to account for this? Can this be fixed?

Ryan Cooper

For today, we were assigned to the Central Illinois (ILX) CWA, close to the low pressure center. CAPE wasn’t as robust, but there was already ongoing convection that was expected to further strengthen during the afternoon with a bit of additional heating in some clear skies east of the low.

One feature that was very noticeable for our area compared to further south in the warm sector was the low-topped nature of our convection. Even our supercells were fairly low-topped. As a result, that influenced some of the satellite-based tools. Most notably, GREMLIN returns were much weaker than one should have expected, even once lightning production got going in a cell.

Fig. 1: 4-panel loop of GREMLIN data from 2025z to 2105z. Counterclockwise from top-left: GREMLIN East Meso-2; GREMLIN East Meso-1; GREMLIN CONUS; MRMS -10C Reflectivity.

As you can see in Fig. 1, the two Mesosector-based GREMLIN simulations on the left show very weak returns compared to the -10C returns. The CONUS-based GREMLIN (bottom right) shows stronger simulated returns, including some that get into the red above 50dBZ, but it still generally underdoes ‘reality’.

Getting a bit later in the day, this would become even more pronounced at times, as seen in Figure 2. Returns over 60dBZ (whites and pinks) are frequently evident on the MRMS Composite Reflectivity, but the East Meso-2 GREMLIN essentially never showed anything over 50dBZ. And while the CONUS GREMLIN does better, as noted above, and shows returns over 50dBZ a bit more frequently, it still largely is undergoing the strength of the storms.

Fig. 2: Loop from 2005z to 2200z showing, clockwise from top left: GREMLIN East Meso-1, MRMS Composite Reflectivity; GREMLIN CONUS;  GOES East Meso-1 Band 13.

Contrast this to further south (Fig. 3), in and around the Memphis (MEG) CWA where the other team was focused for the day. Here, cloud tops are much higher and colder. The East Meso-1 GREMLIN shows simulated returns over 50dBZ much more frequently compared to in Illinois. It is much closer to the CONUS-based GREMLIN, and in some cases shows simulated returns higher than the latter, and both are much closer to the reality shown in the MRMS data.

Fig. 3: Loop from 2015z to 2210z showing, clockwise from top left: GREMLIN East Meso-2, MRMS Composite Reflectivity; GREMLIN CONUS;  GOES East Meso-2 Band 13.

Returning to the case in Illinois, it is not entirely clear why the CONUS performed marginally better than the Meso-based GREMLIN simulations, but perhaps with the algorithm having much weaker influence from the IR satellite data, having 5 minutes of lightning data was weighted a bit more heavily vs. one minute of lightning data. It seems like the low-topped convection and warmer cloud tops rather significantly hindered the performance of GREMLIN. There is likely no easy fix or workaround for that, so forecasters using the product just need to be aware of that limitation, and know that a day with low-topped convection may make GREMLIN less useful than some other tools.

– Marko Ramius

When initially digging into the GREMLIN product (since I had never seen or worked with it before), I wanted to see how it looked in comparison to reflectivity products that I use daily. Loading a 4P with MRMS RALA in the top left, GREMLIN in the top right, and KTLX/KSRX base reflectivity in the bottom left and right, respectively, helped me visualize the differences. One thing I noticed is that the GREMLIN product had an odd jump in dBZ into Okfuskee County at 19:50Z Mon 19-May-25 that was not reflected by the base reflectivity products. This appears to be because GREMLIN’s view is from the top down rather than the top up, so it’s sampling more of the lightning in the anvil blowing off downstream to the north, causing an apparent downstream jump in reflectivity. I think this would be useful in anticipating where downstream convection is going to grow based on the lightning presence and anvil presentation on the satellite view with GREMLIN.

– millibar