Demo · real storm:
Solvayu combines data from several official observation and forecast sources, processes it continuously and calibrates it for each point of interest using the site's own storm history. So every indicator on the dashboard reflects local conditions, not a regional average.
Lightning is detected by satellite, including flashes inside the cloud, day and night, and filtered to discard signals that aren't lightning. Each flash is located to within about ±5 km, so the 8 km ring is close to the resolution limit. Satellite detection records most flashes, not all (typically 70 to 90%). Data usually arrives in under a minute, and the top bar shows how old it is.
Pulso is a thunderstorm warning system: it reports the status and the reason. Under IEC 62793:2020, any action taken on this information is the responsibility of the system's user, in line with applicable regulations, and is defined in their emergency plans. In Peru, D.S. 024-2016-EM (art. 382 e) requires storm detection and warning equipment, lightning rods and shelters, and refers to the operator's own procedures.
| Ring | Distance | Status | Holds for |
|---|---|---|---|
| Regional | up to 150 km | Watch | 30 min after the last strike within 150 km |
| Prepare | up to 32 km | Prepare | 20 min after the last strike within 32 km |
| Monitoring | up to 16 km | Alarm | 30 min after the last strike within 16 km |
| Surrounding | up to 8 km | Alarm | 30 min after the last strike within 16 km |
The monitoring and surrounding rings correspond to the monitoring area (MA) and surrounding area (SA) of IEC 62793:2020; the regional and prepare rings are part of the site protocol. Distances are set for each operation. The alarm ends after 30 minutes with no lightning within 16 km (the standard's all-clear time), and every new strike restarts the count. Prepare holds for 20 minutes after the last strike within 32 km, and Watch for 30 minutes after the last within 150 km.
No data: if lightning data stops arriving for more than 3 minutes, the screen stops showing All clear or Watch and shows NO DATA. If Prepare or Alarm was active, it stays: without data there's no way to confirm the lightning has stopped.
The blue bars show the hourly chance of rain from numerical forecast models; use them for timing. The dotted line is not a forecast: it's what's usual at the site, the share of days with at least one strike within 16 km at that hour, by season (wet, dry or transition). Rain doesn't mean lightning, so the storm warning uses a different number: the calibrated chance of lightning within 16 km in the afternoon (11:00 to 19:00). It comes from a correction trained on the site's own storm history. The yellow band marks the hours when a storm is expected: it appears when that chance passes the threshold, over the afternoon's wettest hours.
Tested one day ahead and always on months the model hadn't seen, it caught 9 in 10 afternoons with lightning within 16 km of the site, and 3 in 4 warnings were right. In the dry season, when storms catch people off guard, it anticipated 9 of 15 afternoons. It's a planning tool: the alarm is triggered by observed lightning, not by the forecast.
| Season | Cases | POD | Lead time (median) | FAR |
|---|---|---|---|---|
| Wet season (Jan 2026) | 31 | 45 % | 27 min | ≈ 0 % |
| Wet season (Feb 2026) | 23 | 61 % | 24 min | ≈ 0 % |
| End of dry season (Aug–Sep 2026) | 12 | 33 % | 27 min | ≈ 0 % |
POD: cases with a strike in the surrounding ring that already had an alarm. Lead time: time from the start of the alarm to the first strike in the surrounding ring. FAR: alarms with no storm in the monitoring area; here it's practically zero because every alarm is triggered by a detected strike. Distances measured from the site. In the dry season, 7 of 12 storms formed right over the site, with no earlier lightning within 40 km (8 of 31 in January and 4 of 23 in February). Lightning detection can't anticipate those cases: Pulso triggers the alarm with the first nearby strike. The figures are updated every season as the sample grows.
At this site, storms form with very little convective available energy: in hours with lightning within 16 km of the site, median CAPE was 65 J/kg, and only 2 of 156 hours exceeded the 1,000 J/kg that classic thresholds look for. That's why it isn't used.
Pulso doesn't show whether the storm is approaching or when it would arrive. Working that out from the nearest strike is misleading: when a new cell forms closer, the distance jumps and the storm looks like it's moving even when it isn't. Estimating it properly means tracking each storm cell separately, which is in development.
Each ring has its own pattern so you can tell the distance without looking: more pulses means closer (regional 1, prepare 2, monitoring 3, surrounding 3 + 2). Within 16 km you also hear thunder, about as far as real thunder carries. When the alarm clears, two soft falling pulses play. Two sounds never play at once: the most urgent wins. Sound is on by default; if your browser blocks it during Prepare or Alarm, the sound button turns red and one tap fixes it.
If Pulso is behind another tab or window, it keeps calculating the status. When the level rises, the alarm clears or data stops, it shows a system notification and the tab title blinks. You need to allow notifications the first time. With the phone locked or the browser closed, there's no alert.