El Nino Watch
Each El Nino is different, the sample is 6-7 events, ERA5 is a 30 km grid cell that under-reads the island's rain, and the climate has warmed since 1982: read the analogs as a range of what happened, never as a prediction.
Koh Samui watch
The list of reasons the analogs page can mislead if read too literally.
1. Each El Nino differs. The 1997-98 and 2015-16 events peaked in the central-eastern Pacific in December; 2023-24 was weaker in the relative index (RONI +1.42 °C vs ONI +1.99 °C); 2009-10 was a central-Pacific event. The Gulf coast of Thailand gets most of its rain from the NE monsoon, which also depends on the Indian Ocean Dipole, the MJO and the cold surges of the winter: an El Nino tilts the odds, it does not set the outcome. The takeaways therefore count "N of M events", not "El Nino means". 2. Small sample. Six or seven events since 1950 (the ERA5 and ONI records both reach back to 1950 here). A median of six numbers moves a lot when one event is added, so the page shows every event's number, not just the summary. 3. One grid cell. ERA5 is a ~30 km reanalysis: it smooths the island's hills and sea breezes, under-reads the rain measured at Samui's own gauges (the live water factor says the same) and gives cooler maxima than a thermometer in Chaweng. Percentages of normal are computed ERA5 against ERA5, so the bias mostly cancels; absolute millimetres and degrees are cell values. 4. Warming trend. 2026 is about 1 °C warmer globally than 1982. An old analog understates today's heat: the "days at or above 39 °C feels-like" of 1998 would be more numerous with the same weather pattern today. Compare the dry-season heat of the recent events (2016, 2024) rather than the oldest ones. 5. Documented impacts depend on reporting. Water shortages make the news when hotels complain; a 1998 rationing on a much less developed island may simply not have been written up online. The page lists what was looked for and not found instead of filling the gap. 6. Water supply has changed. Samui's demand (about 34,000 m³ a day in 2026, per the PWA) is far above 2016's 30,000 m³ and 1998's; the undersea pipeline and desalination plants did not exist for the early events. A rain deficit like 1998's would bite differently today.
Use the analogs to size the risk ("Feb-Apr rain under 60% of normal in N of M events; dry spells of up to X days") and the live factors and seasonal forecast to see whether 2026-27 is tracking those ranges.
The owner's decisions (fill the cistern, book a departure threshold) should rest on the live numbers; the analogs set expectations and stop both complacency ("2015-16 was fine") and panic ("1998 was a disaster").