Kīlauea Episode 56: USGS Forecasts Lava Fountains Between November 3 and 10

On the morning of 31 October, Hawai'i time, the US Geological Survey's Hawaiian Volcano Observatory published a forecast that looks deceptively simple: Kīlauea's next lava-fountaining episode, number 56 in the eruption that began on 23 December 2024, is most likely to begin between 3 and 10 November. The summit eruption at Halema'uma'u has been paused since episode 55 on 17 October, and the ground above the magma chamber has been slowly swelling ever since. That swelling is the whole forecast. What makes it interesting is that the last episode broke the pattern the forecast is built on, so this window tests the method as much as the volcano.

The method is elegant, and it has worked for nearly two years. When a fountain drains magma from the summit chamber, the ground above it sinks and tilts inward; as fresh magma refills the chamber during the pause, the ground bulges back out. Since episode 5 in January 2025, observatory scientists have used that symmetry — roughly, the tilt lost in one episode equals the tilt regained before the next — to publish forecast windows. This time the numbers are moving in the right direction: the Uēkahuna tiltmeter has recovered about 16.2 microradians since episode 55 ended, including about 0.8 microradians in the past day. Overnight glow and spatter strengthened, seismic tremor became more continuous after midnight, and the observatory now believes the bright glow previously attributed to the west vent comes from a merged southwest vent, where the old south and west vents have joined. Seven small earthquakes below magnitude 1.5 were recorded at the summit in the past day. None of this is an eruption yet, but it is the familiar shape of a volcano refilling.

The honest reason to hold this window a little loosely is episode 55 itself. It arrived after a gap of nearly eight weeks since episode 54, far longer than the pauses of two or more weeks typical of this eruption, and its buildup looked nothing like the ones before it. Summit tilt rose beyond the levels usually seen before fountaining, then stalled or sagged for extended stretches instead of climbing steadily. Before the fountain, lava overflowed intermittently for 39 days — the longest precursory overflow period of the eruption — several short-lived vents opened inside the crater, and a shallow dike was emplaced beneath the caldera. The observatory says plainly that, given that unusual buildup, uncertainty remains over whether summit deformation will follow a consistent enough pattern for precise forecasting. The window has already moved as new tilt data arrived: an update on 25 October pointed to 30 October–5 November, press reports of Friday's forecast cited 2–9 November, and the 31 October daily update now says 3–10 November. A forecast that slides by days as it is revised is not a failure; it is the model telling you, in public, how much it does not know.

That is also why the stakes here are modest and the science is not. Fountaining episodes in this eruption generally last less than 12 hours, and the activity stays inside a closed area of Hawai'i Volcanoes National Park. The hazards that reach people are mostly downwind: volcanic gas and vog, which light trade winds were carrying southwest toward Ka'ū on 31 October, and tephra that fell mostly within the closed park area during episode 55. The alert level remains ADVISORY and the aviation colour code YELLOW, signalling unrest above background but no threat to communities. What is at stake is whether one of the most closely watched volcanoes on Earth still behaves the way two years of data say it should.

My own reading is that the window will hold, at about 65% confidence — not because the observatory's word is final, but because the refill signal has reasserted itself: steady inflation, strengthening glow and tremor, and tilt recovery approaching the levels that preceded earlier episodes. The 35% against is the episode 55 lesson: a magma system that has just emplaced a dike and spent 39 days overflowing may not refill on the old schedule, and a window that has already shifted twice could shift again, past 10 November or into an overflow that never becomes a fountain. If the fountain arrives inside the window, it confirms that the tilt-refill rule survived its strangest test. If it does not, that miss will be more useful than another correct window, because it will show the rule needs revising after the dike.