Quick answer
Lightning maps are useful for recent detected activity, but they are not exact ground-strike records. Lightning Tracker markers are optical-flash centroids, not exact ground-strike locations. An empty map is never an all-clear: sensor coverage, processing, publication, and page refresh can all limit what is visible.
Five meanings of “accurate”
- Completeness: whether a sensor detected an event in its field of view.
- Location precision: how closely a marker represents observed activity.
- Classification: whether the system measures total lightning or a particular ground-impact class.
- Latency: time from event through product, pipeline, and display.
- Aggregation window: whether the map shows minutes, a rolling day, or a closed local day.
These questions can have different answers on the same map. A system can quickly show a broad storm cluster while still being unsuitable for declaring which tree, roof, or field was struck. A rolling 24-hour map can accurately describe its chosen window while being the wrong tool for a question about a single minute. Before comparing services, identify the sensor and the decision you need to make.
| Question | Useful interpretation | Unsafe shortcut |
|---|---|---|
| Was activity detected? | Recent electrical activity is present in the product | No marker means no lightning is possible |
| Where was it? | Marker is an observed optical-flash centroid | Marker is exact ground impact |
| How current is it? | Check timestamp, source, and display refresh | Product latency equals app latency |
| What is counted? | Read whether events, groups, or flashes are used | Every count means ground strikes |
What NOAA GLM measures
NOAA’s Geostationary Lightning Mapper observes optical emissions from total lightning. Its products organize events, groups, and flashes; a displayed point is a radiance-weighted centroid, not a surveyed impact point. The GOES-R product documentation describes approximately 8 km resolution near nadir and 14 km near the field edge. Those are pixel-resolution characteristics, not a guaranteed marker error for every storm.
NOAA product material describes characteristic detection performance such as 70–90% under stated conditions. That is not a per-storm guarantee, and it cannot be converted into a promise about every marker on this site. Read the GOES-R baseline product and official GLM product pages for product definitions.
GLM’s strength is broad, continuous optical observation from geostationary orbit. Its output is especially useful for seeing whether a storm is electrically active and how activity is distributed over time. It does not change a total-lightning observation into a confirmed cloud-to-ground strike record. That distinction matters in search snippets, map labels, and emergency decisions.
Events, groups, flashes, and centroids
GLM begins with an optical event: a pixel-level signal observed in a 2 ms frame. Adjacent or simultaneous events can be combined into a group. Groups close in time and space—within less than 330 ms and 16.5 km in the product definition—form a flash. That hierarchy is why a “flash” count is not the same thing as a count of individual pixels or confirmed ground contacts.
The marker location is a radiance-weighted centroid: a useful center of observed optical activity, not a claim about a point on the ground. This is especially important for cloud flashes spanning an area. The official NOAA GLM product documentation describes these event, group, and flash products.
Parallax, latency, and windows
Satellite viewing geometry can shift the apparent position of high clouds relative to the surface: parallax matters, especially away from nadir. GLM product latency can be about 20 seconds, but that is not Lightning Tracker end-to-end latency. Our pipeline, network delivery, browser refresh, and device conditions add time; see the methodology. A last-24-hour map and a closed daily-count page also answer different questions.
Lightning Tracker polls GLM data every 30 seconds and browser maps refresh every 90 seconds. Those are operating cadences, not latency guarantees: upstream publication, processing, network delivery, and device conditions can add delay. Treat timestamps as recent-context evidence, never as a promise that every current flash is already displayed.
For example, a flash high in a cloud may be observed along a line of sight that does not point to its surface projection. A city label under a marker is therefore context, not a verified impact address. Similarly, a page refreshed every minute cannot make every upstream stage instantaneous. Preserve the timestamp, use the map as a recent-activity layer, and defer to official warnings for actions.
Aggregation changes the meaning of a number
Lightning Tracker has live views and closed-day views. A live state page uses a moving recent window; a yesterday page describes one completed local calendar day. Counts from neighboring state bounding boxes can overlap, and they cannot be summed into a unique national total. This is not a defect in a table: it is a boundary and measurement definition that must remain visible.
A useful reading habit is to state the unit with the number: “detected optical flashes in this state during this window,” rather than “strikes in this state.” That wording preserves the sensor, geography, and time range. It also stops an old number from being compared casually with a live map or with a different provider’s radio-detection count.
Satellite maps versus ground RF networks
Ground radio-frequency networks and satellite optical sensors measure lightning differently. Neither generic label proves superiority: coverage, geometry, processing, and the question being asked matter. Lightning Tracker publishes satellite optical-flash context and describes its limits rather than converting it into a ground-strike claim.
A ground network may be designed around radio emissions and location algorithms; a space-based optical sensor observes light from cloud tops. Their units, coverage edges, and latency paths are not interchangeable. Comparing a single accuracy percentage without the product definition, time period, geometry, and target class can mislead more than it informs.
Examples of correct and incorrect map use
Useful: “Several recent optical flashes are clustered west of town; I will check the official warning and move indoor plans earlier.” Not supported: “That marker proves the exact ground strike was at this address.” The first uses a map for situational awareness; the second asks it to provide a precision claim the product does not make.
Useful: “The last update is older than expected, so I will not rely on the map alone.” Not supported: “No dots means the field is safe.” A data feed can be delayed, outside coverage, or simply not show a detection relevant to a person’s immediate risk. Audible thunder and official instructions override map interpretation.
Use maps safely
Use a map to understand recent activity, then check official forecasts and warnings. If thunder is audible, shelter immediately in a substantial building or enclosed hard-topped vehicle. Do not wait for a marker, a push notification, or a claimed precision value to make that decision.
- Read the map’s source and window before interpreting a count.
- Use the newest timestamp as context, not a safety guarantee.
- Compare activity over time instead of treating one dot as a surveyed impact.
- Follow official warnings and audible thunder, even if a map appears quiet.