Two thunderstorms in one week (8 and 13 October), a 111 mm downpour at Tobalaba in between and, according to the forecasts, another one on its way. People are already saying that “this year is weird”, «never seen before», «it’s all el niño/climate change». I know little about thunder and a bit more about data, so I decided to measure it: are we in a particularly electrified year (or month)? Spoiler: the answer has fine print, and the fine print is the best part.
I used three things: the lightning seen by the GOES satellites (GLM) since 2018, the reports from Santiago airport (SCEL, Pudahuel) since 1973, and rainfall from the official table of the Chilean Weather Service (DMC). Everything is raw data, nothing interpolated.
First, the lightning: 13 October has almost no competition
I took the 92 days since 2018 with 10 mm or more of rain at Quinta Normal or Tobalaba, and for each one I counted all the GLM flashes inside a 144 × 111 km box around Santiago. The result is quite clear:

13 October adds up 3,091 flashes between 13:37 and 19:26 local time (almost half of them in a single hour, between 4 and 5 pm). It is followed by the summer storm of 30 January 2021 with 2,768 and then nothing comparable: 8 October with 524 and everything else below 400. Between 13 Oct and 30 Jan 2021 there is only a 12 % difference, so I would talk about “the two most loaded storms”, not an undisputed champion.
Looking at the map, the discharges line up in two nearly vertical bands, one west and one east of SCEL, with the airport right in the middle. That matters for what comes next.
| Day | GLM flashes | What SCEL reported that day |
|---|---|---|
| 13 Oct 2026 | 3,091 | Thunderstorm (TS) |
| 30 Jan 2021 | 2,768 | Showers (no storm) |
| 8 Oct 2026 | 524 | Thunderstorm (TS) |
| 6 May 2026 | 392 | Showers |
| 18 Aug 2021 | 318 | Showers |
| 10 Nov 2023 | 313 | CB/TCU clouds |
| 12 Nov 2018 | 287 | Showers |
| 20 May 2021 | 163 | Showers |
| 30 Aug 2026 | 140 | Nearby thunderstorm (VCTS) |
| 20 Sep 2025 | 135 | Showers |
Table 1. The 10 days with the most flashes and what the airport reported over the same interval (12 to 12 UTC). TS: thunderstorm at the station; VCTS: thunderstorm in the vicinity.
Two warnings. First, the airport is a single point: only 3 of these 10 days have a thunderstorm in the SCEL METAR. On 30 January 2021, with 2,768 flashes, the airport reported heavy showers right at the lightning peak, but never wrote “thunderstorm”. Second, I picked the days by rain, not by lightning, and the box includes the Andes, so I may be leaving out days with lots of lightning and little water (and I do not separate GOES-16 from GOES-19, which changed in April 2025). If anyone feels like downloading the whole GLM series over Santiago, we can corroborate or refute this.
Second, the airport as a thermometer: 54 years of thunder (with an asterisk)
To go back further than 2018 we have to look at the SCEL reports since 1973. I defined a thunderstorm day as a day (12 to 12 UTC) with at least one thunderstorm report: the manual present-weather code written by the observer (ww 13, 17, 27, 29, 87–99) or a TS or VCTS signal in the METAR text, whether it is 1 report or 20.

First: 2026 has 13 thunderstorm days up to 13 October, spread over six months (2 in January, 1 in April, 1 in July, 3 in August, 2 in September and 4 in October; eight with a storm over the airport and five only nearby). That ties the series maximum, 1976. Here comes the asterisk: in 1976 there are reports in only 81% of the hours, so its 13 days are a minimum and could be more. Third place is 1986, with 11.
Other metrics put 2026 on top: 20 thunderstorm hours (previous maximum 14, in 1977) and 6 days with thunderstorms in at least two different hours (previous maximum 3). And October 2026 has 4 thunderstorm days, while the 53 Octobers from 1973 to 2025 added up to 10, spread over only 9 Octobers. It is a very stormy October.
Second, the trend: from 8.7 and 5.7 days per year in the 70s and 80s we go to 3.0, 1.6 and 0.7 in the following decades. Megadrought? The truth is that the decline started earlier. Days with 10 mm or more at Quinta Normal stayed at ~11 per year until 2009 and dropped to ~7.5 during the megadrought (−33%), while thunderstorm days had already fallen from 7.0 per year in 1973–89 to 2.3 in 1990–2009. The megadrought may explain part of the lower rainfall, but not the decline of thunderstorms by itself.
And third, the fine print: how a thunderstorm gets recorded has changed. Until 1998 the manual code came in about half of the reports; between 1999 and 2004, in almost all of them; since 2006 it drops to between 12 and 34%, and since August 2025 the source (IEM) carries only the METAR text. On top of that, in 75 of the 188 days the strongest code is 17 (“thunderstorm, no precipitation at the time”), and before 1999 only 38% of thunderstorm days have rain at Quinta Normal or Tobalaba that same day (75% between 1999 and 2004 and 87% since 2005). So I tried three definitions:
| Definition | 1973–79 | 1980–89 | 1990–99 | 2000–09 | 2010–19 | 2020–25 | 2026 (to 13 Oct) | Previous max. (year) |
|---|---|---|---|---|---|---|---|---|
| All signals | 8.7 | 5.7 | 3.0 | 1.6 | 0.7 | 0.7 | 13 | 13 (1976) |
| With rain that day or CB clouds | 4.4 | 2.9 | 2.4 | 1.6 | 0.7 | 0.7 | 13 | 7 (1987) |
| Strict (ww 91–99 or TS/VCTS text) | 2.7 | 1.9 | 1.4 | 1.6 | 0.7 | 0.7 | 13 | 6 (2000) |
Table 2. Thunderstorm days per year, decadal means, with three increasingly demanding definitions. The last column is the annual maximum before 2026.
The honest reading: 2026 stands out under all three definitions (13 days against a previous maximum of 13, 7 and 6). How large the decline since the 70s is depends on the definition (−92 %, −84 % and −74 % from the 1970s to the 2010s), but it is large in all cases. What changes is its shape: with the strict definition the series is roughly flat between the 80s and 2009 and halves in the 2010s, during the megadrought. Before 1999 most days rest on weak manual codes (thunder heard or lightning seen), and only about a third coincide with rain at Quinta Normal or Tobalaba, against eight in ten afterwards. That may reflect a different observing practice, but I cannot tell from these data whether those were real thunderstorms. And 2026 “tying” 1976 only holds with the most generous definition.
Third, storm or downpour? Intensity and the blame on El Niño
Here comes what I like the most. The two October storms were very different things. I measured what share of the day’s rainfall fell in the most intense hour, at Quinta Normal:

13 October was a downpour: at Quinta Normal 12.2 mm fell in 60 minutes out of a daily total of 13.4 mm. That is 91% of the day in one hour. As a clock hour it was 10.6 mm (79%), the most intense clock hour in the DMC record since 2019 (the previous maximum was 10.2 mm on 15 June 2025, a 30 mm day), and the 60-minute rolling window (12.2 mm) is also above the previous maximum of 10.2 mm. It is not unique in concentration, though: on 10 September 2019, 8.5 of 9.0 mm fell in one hour (94 %), but a much weaker hour. And its daily total (13.4 mm) was smaller than the annual maximum daily rainfall of each of the 76 previous years. 6 October was the opposite: 53.8 mm in the day (111 at Tobalaba, the largest daily total in its 55 valid years), but the strongest hour contributed only 17%. Long, spread-out rain, almost no lightning.
With so many El Niño years around, the obvious question is whether El Niño favours thunderstorms. I tried it with the June–August ONI:

The 8 El Niño years in the series have 3.3 thunderstorm days per year, the same as the rest (3.3), and no signal shows up either if I split by season. With only 8 years the sample is small, but just look at the strong El Niños:
| Year | ONI (JJA) | Thunderstorm days (SCEL) | Rain from 1 Jan to 13 Oct (Quinta Normal, mm) |
|---|---|---|---|
| 1987 | +1.30 | 7 | 684 |
| 1997 | +1.48 | 7 | 704 |
| 2015 | +1.44 | 1 | 196 |
| 2026 | +1.80 | 13 | 340 |
Table 3. The four strong El Niños of the series. The median rainfall from 1 Jan to 13 Oct between 1973 and 2025 is 276 mm.
Same “strong El Niño”, results from 1 to 13 storms. And 2026 is especially odd: lots of thunderstorms with little water. Between January and June only 22.7 mm fell at Quinta Normal; the rest came since July. Compared with 1987, which was also a strong El Niño with intense daily rain, 2026 has almost twice as many thunderstorm days (13 versus 7) with half the accumulated rainfall (340 versus 684 mm), and its storms are concentrated between August and October (9 of 13), while those of 1987 are spread between March and October.
My conclusions
- 13 October was one of the two most lightning-loaded storms since 2018 (3,091 flashes, together with the summer of 2021), and it was a downpour: 91% of the day’s rain in one hour.
- 2026 ties the maximum number of thunderstorm days at the airport (13, like 1976) and beats it in hours and in days of prolonged thunderstorms. It is exceptional under all three definitions I tried. October, with 4 days, is the stormiest of the series. Also, a new storm might come in the next days, so this numbers could increase.
- The “decline since the 70s” is real in part, but exaggerated by the record: who writes down what changed, and before 1999 many days depend on a single manual code.
- El Niño does not explain the number of thunderstorms. El Niño years are not stormier than the rest, and 2026 looks more like a case of its own than a typical El Niño.
- What I do not know: the cause. Hypotheses to test: the phase of the Pacific (the PDO/IPO was positive in 1977–1998 and negative afterwards), the type of system (cut-off lows and cold air aloft), and how well the airport represents all of Santiago. One more: that we are going back, for a while, to the rainy and stormy years of the 70s and 80s. For now it is just one year, not a decade.
About the data: the airport is a single point and the GLM box includes the Andes; before 2005 present weather was written by hand; the 1976 series has reports in only 81 % of the hours; Tobalaba 2012–2015 was excluded because of cumulated windows in the official table. If anyone has data from other stations, a different approach, or the will to download the full GLM, let me know.
Sources: GLM on GOES-16/19 (NOAA/NASA); NOAA NCEI Integrated Surface Database; Iowa Environmental Mesonet; Chilean Weather Service (DMC: precipitation tables, 1-minute automatic station and public rain gauge); INIA Agromet (a collaboration between INIA and the institutions that hold agreements with INIA); ONI, NOAA CPC; Natural Earth.



























