Scintilla: watching thunderstorms from space (and from under them)

In March 2024 I took a NASA ARSET short course on lightning observations. ARSET — Applied Remote Sensing Training — is NASA's free online teaching program for earth observation. I take their courses regularly, not because I need them for specific projects but because they're a reliably good way to broaden my mental map of what's available in remote sensing. This particular series, taught by researchers from Goddard, Marshall, and NOAA, walked through how spaceborne lightning sensors work, what data products they produce, and where to find them.

It left me wanting to build something.

Two years later, that something is scintilla — Latin for spark. It's a Python pipeline that downloads public NASA lightning data, clips it to a region of interest, and renders an animation. It just went public on GitHub.

The pipeline pulls from two complementary sensors. GLM — the Geostationary Lightning Mapper — rides aboard a GOES weather satellite sitting at a fixed longitude 35,786 km above the equator and sees cloud-top optical flashes continuously across most of a hemisphere. ISS LIS — the Lightning Imaging Sensor — flew on the International Space Station from 2017 to late 2023 and produced uniform-quality flash detections in a narrow swath directly beneath the orbit. In the renders below, GLM detections are the animated cloud-top flashes, and individual ISS LIS flashes are overlaid as magenta star markers.

The Manitoba storm

On June 4, 2023, a long-lived thunderstorm complex moved through southern Manitoba. I pointed scintilla at it using GOES-18 — the GOES-West satellite:

Watch the northeast corner of the frame. It's mostly dark. But when the ISS flies overhead, magenta flashes appear there — standing alone, no GLM lightning under them. The spaceborne sensor is seeing flashes the geostationary sensor isn't.

Why? Geometry. GOES-18 sits over the Pacific at −137° longitude. Manitoba is nearly 40° of longitude east of that. From GOES-18's vantage the northeast corner of this storm is near the edge of what it can see — the satellite is looking sideways across the atmosphere, through more air, past neighboring cloud tops that physically block its view into the lightning-emitting volume. Its detection efficiency collapses in that corner.

The fix is to use the other satellite. GOES-16 sits at −75° longitude, essentially over the eastern US, with a much cleaner view of central Canada. Same storm, same time window, different vantage:

The northeast corner fills in. The ISS LIS flashes that were standing alone now have GLM lightning under them — both sensors agreeing on the same flashes in the same cells at the same times. Aggregated across the 4.5-hour window, GOES-16 caught 80% more active pixel-minutes and 2.4× more total optical energy than GOES-18.

This isn't a bug. It's what happens when you ask a geostationary satellite to watch something near the edge of its disk. Without the ISS passing overhead and leaving magenta breadcrumbs in the dark quadrant, the effect would have been invisible — the "storm fading over the evening" I thought I was watching in the GOES-18 video turned out to be partly the satellite losing sight of the storm, not the storm losing intensity.

A storm from underneath

One more. A Tucson monsoon, July 31, 2023. A Reconyx trail camera sitting inside the storm's area of interest recorded the afternoon from the ground:

Near the end, the storm tips the camera over.

GOES-18 was watching the same system from geostationary orbit. Here's the cloud-top view:

Same storm, two vantages. One from 35,786 km up. One from a fence post in the desert.

scintilla is on GitHub; the README walks through a runnable demo that works on a fresh clone with no NASA credentials. The storm gallery in the repo lists other renders — including an August 2023 Tucson pairing that does catch an ISS LIS overflight off to one side of the AOI.

If you want to find your way to the same ARSET course that started all this: NASA's ARSET program publishes recordings and slides for every course they run. Every one I've taken has been equally worth the time.