Indonesia · Flores island, East Nusa Tenggara · the magnitude 7.7 earthquake of Saturday 15 August 2026 · AT LEAST 53 DEAD on both chains after both pages moved again on Monday — and NAGEKEO, the regency nearest the epicentre and absent from every national count since the earthquake, is finally in the breakdown, while NGADA has vanished from it · the agency's own components close at their floor for the second time in this story's life, 25 plus 20 plus 4 plus 2 reaching 53 on two at-least-ones, with four stated figures now where three stood on Saturday · about 5,000 displaced against components summing to 5,434, and 113 injured on a health ministry count the disaster agency has never adopted

A magnitude 7.7 earthquake strikes Indonesia's Flores island — at least 38 dead on the national count, and the region closest to the epicentre has not been reached

A magnitude 7.7 earthquake struck off Indonesia's Flores island early on Saturday 15 August, killing at least 53 people on the national disaster agency's count as of Monday. The United States Geological Survey puts the quake at 05:58 local time — 21:58 GMT on Friday — with an epicentre about 68km (42 miles) north-northwest of the city of Ende in East Nusa Tenggara province, at a depth of 10km; Indonesia's own meteorology and geophysics agency, BMKG, gives the depth as 15km, and we print both. The BBC renders the local time as shortly before 05:00 against its own GMT figure of 22:00 Friday, which do not reconcile in a province on UTC+8; the two publishers' GMT stamps agree within two minutes, so the hour is settled and its local rendering is not. Some 341 aftershocks have followed.\n\nTHE TOLL HAS BEEN PUBLISHED FIVE WAYS ACROSS THREE DAYS AND WE PRINT EVERY ONE. At dawn on Saturday: at least 20, attributed to rescuers around the port town of Maumere, against at least 38 on the count of Indonesia's disaster mitigation agency BNPB. By lunchtime a rescue official gave 40 dead and 50 injured, a local figure above the national one. By mid-afternoon BNPB stood at at least 47, on Sunday at at least 51, and on Monday at AT LEAST 53 — the last of those carried by both of this story's established chains after both pages were rewritten again.\n\nAND THE AGENCY'S REGIONAL BREAKDOWN HAS MOVED FOR THE FIRST TIME SINCE SATURDAY, IN THREE WAYS AT ONCE. It now reads: 25 fatalities in Manggarai, 20 in East Manggarai, four in Sikka, two in Ende, and at least one each in NAGEKEO and West Manggarai. Every named figure has risen, Ende has firmed from 'at least one' to a stated two — and NAGEKEO, the regency described as closest to the epicentre and absent from every national count since the earthquake, is in the ledger at last. NGADA, which this story has carried in that same list since Saturday, has vanished from it with no correction from anybody. Twenty-five plus twenty plus four plus two is fifty-one, and the two at-least-ones carry it to fifty-three at their floor — so the components close against the stated total, an arithmetic this desk did rather than relayed. WE SAY WHAT KIND OF CLOSE THAT IS RATHER THAN CALL IT A FIRST: this page's own record shows the ledger balancing the same way at 47 on Saturday, at its minimum rather than to the unit, after a morning in which twenty-four deaths sat outside the regional returns, and four sat outside again on Sunday against a stated 51. What is new tonight is a tighter floor — four of the six places now carry stated figures where three did on Saturday. ONE LIMIT BELONGS ON ALL OF IT: this breakdown exists on one of our two chains only. The other carries no regional figures of any kind.\n\nWHAT NAGEKEO'S ENTRY DOES NOT MEAN. Nobody has said the regency has been reached. Landslides blocked the roads and communications were disrupted, per Fathur Rahman, head of the local search and rescue agency in Maumere, and a team was attempting to get in by ferry; about 2,000 villagers there fled their homes to temporary shelters, per BNPB head Lieutenant General Suharyanto, who says three helicopters and a rescue vessel have been deployed given a province of numerous islands with significant transport and access challenges. None of those names appears anywhere on the current BBC page. One death recorded in a column is the national ledger admitting the place exists; it is not a survey, a headcount or a rescue, and 53 remains a floor.\n\nTHE INJURED ARE COUNTED BY THREE BODIES AND THE SMALLEST HAS GONE QUIET. Deputy Health Minister Benjamin Paulus Octavianus gave 36 seriously injured and 77 with minor injuries in East Nusa Tenggara — 113 in total, carried unchanged by both chains. BNPB's own figure of at least two seriously and eleven slightly injured, which this story published three times as unmoved while the dead rose, appears on neither page; nobody has withdrawn, superseded or reconciled it. A rescue official's fifty sits between them.\n\nAROUND 5,000 PEOPLE ARE OUT OF THEIR HOMES AND THE COMPONENTS EXCEED THE TOTAL: 3,356 displaced in Sikka and a further 2,078 evacuated in Manggarai is 5,434 from two regencies alone, before Nagekeo's roughly 2,000 and the roughly 2,000 who self-evacuated during the tsunami warning are counted at all. More than 3,300 people in Sikka either fled their homes or were left stranded, taking shelter in a sports arena; in Maumere some residents camped outside collapsed homes under tarpaulin tents. What was destroyed is now counted three ways rather than four: more than 900 homes severely damaged; more than 1,300 damaged on the other chain; and at least 157 houses destroyed with nearly 200 damaged on Suharyanto's count. The 'more than 1,000 damaged homes' figure this story recorded six paragraphs below the first has been withdrawn from the page that carried it. Destroyed, severely damaged and damaged are three different measures and nobody publishes the boundaries. Alongside them: dozens of schools, healthcare facilities and government offices, among them Ruteng Regional Hospital in Manggarai, where images show operating theatre and clinic ceilings caved in and patients treated in tents. Its spokesperson, Rista Mari, told BBC Indonesian: 'We heard that medical assistance would be provided, but nothing has arrived yet.' Against that, BNPB says 10,500 aid packages have arrived at base operations, that meeting basic needs is a top priority and that tonnes of emergency supplies have been dispatched to Manggarai; more than 3,500 military and police personnel have been deployed overall, per Cabinet Secretary Teddy Indra Wijaya; twenty petrol stations remain shut due to power outages, per the state energy firm Pertamina; and officials are considering declaring a state of emergency for East Nusa Tenggara. Considering is not declaring, and dispatched is not delivered.\n\nWhat is established on the ground has not changed. Emanuel Melkiades Laka Lena, governor of East Nusa Tenggara, told a news conference that several of the victims had been killed in their sleep by collapsing rubble. Video taken at the port in Maumere showed large chunks of concrete breaking off the terminal building onto passengers waiting to board an inter-island ferry, with some thrown into the sea when the gangway to a ship collapsed; nothing we have walked says whether anyone at that port died or was recovered. Rescue teams are described as working to free people trapped under rubble; nothing we walked describes anyone being got out. A tsunami warning issued by BMKG was lifted about three hours after the initial jolt when no significant sea level rise was detected, after some 2,000 people had self-evacuated. USGS estimates suggest more than half a million people on Flores are likely to have experienced very strong shaking, with moderate to strong shaking felt by several million more across the wider region. A separate earthquake struck elsewhere in Indonesia later on Saturday and is given three ways we do not reconcile. This is described as Indonesia's deadliest earthquake since 2022, when hundreds were killed in West Java. One precedent sits under all of it: the same stretch of coastline was struck by a similarly powerful earthquake in December 1992, when the quake and the tsunami it triggered killed more than 2,500 people on Flores.

Medium confidence REBUILT ON 17 AUGUST AND SHORTER THAN THE VERSION IT REPLACES, AND WE SAY WHAT THAT MEANS FIRST. This note had grown to carry four days of walks and their corrections; it is rebuilt here at about half its length because tonight's re-walk overtakes several of the page-state comparisons it was built on. Every limit it carried survives below, some compressed from a paragraph to a clause, and the compression is an editorial act rather than an accident. THIS STORY HAS NOW BEEN WALKED AT SOURCE ELEVEN TIMES ACROSS TWO ADDRESSES, and its standing discipline is to stamp-check both established pages before opening anything. That discipline earned its keep tonight: BOTH HAD MOVED AGAIN and neither had been touched by anybody here since Sunday morning. BBC News, Emma Rossiter, is now at dateModified 16 August 12:37:21.296 GMT, four hours, thirteen minutes and forty-two point nine six five seconds past the 08:23:38.331 this story published. Al Jazeera, by Al Jazeera Staff with AFP, AP and Reuters inputs, is now at dateModified 16 August 15:24:26 GMT against a stamp this story published as identical to its 05:14:00 publication — ten hours, ten minutes and twenty-six seconds of movement — and its headline has gone from at least 51 to at least 53. Both intervals were computed by both seats rather than relayed. The consequence matters more than the arithmetic: THE NEW TOLL IS TWO-CHAIN, where every previous movement in this story was one publisher moving to another's figure. Medium, and the reasons have changed shape. FIRST, THE ARITHMETIC CLOSES AT ITS FLOOR AND WE RECORD WHAT KIND OF CLOSE THAT IS. Twenty-five plus twenty plus four plus two is 51, and two at-least-ones reach the stated 53 exactly. THE COPY FILED HERE CALLED THAT THE FIRST CLOSE IN THIS STORY'S LIFE AND THE VERIFIER SEAT STRUCK IT: this page's own published record shows the same floor-close at 47 on Saturday afternoon — 'the ledger balances at its minimum rather than to the unit' — after a morning in which twenty-four deaths sat outside the returns, and four sat outside again on Sunday against a stated 51. So the series is a gap, a floor-close, a gap, and now a floor-close again, with the difference that four of the six places now carry stated figures where three did. It balances at its minimum and not otherwise, which is the same caveat we attached when it did not balance at all: one extra death in either Nagekeo or West Manggarai and the components exceed the total. A superlative the record does not support is exactly what this desk struck on the Novorossiysk 'largest' claim and on the Ebola convergence, and we would rather print the narrower true thing. SECOND, AND IT IS THE OTHER CORRECTION MADE BEFORE PUBLICATION RATHER THAN AFTER: ONE PROVINCE ENTERED THE LEDGER AND ANOTHER LEFT IT ON THE SAME DAY, AND ONLY ONE CHAIN HAS A LEDGER. Nagekeo now returns one occurrence on the BBC's page, against the none three separate seats had established there; Ngada returns none. The filed copy reported Al Jazeera's two Nagekeo occurrences and zero Ngada as 'matching the other chain on both the province that entered the ledger and the one that left it'. IT DOES NOT MATCH ANYTHING. The verifier seat searched that page for Abdul Muhari, for 'East Manggarai', for 'Manggarai province' and for the stated per-province figures, and every one returns ZERO — Al Jazeera carries no regional breakdown at all, exactly as this story's own source line recorded on 16 August. Its two Nagekeo mentions are Suharyanto's shelter figure and Fathur Rahman's not-reached account, both carried here since Saturday; its zero Ngada is absence from a page that never had the field. This is the desk's own metadata lesson one shelf further out: a negative is scoped to the field you searched, and a word appearing on a page that never carried the category is not a second chain. The ledger movement rests on the BBC alone and now says so. THIRD, EVERY FIGURE BELONGS TO AN OFFICIAL BODY — BNPB and its named deputy Abdul Muhari, its named head, BMKG, the USGS, a deputy health minister, a cabinet secretary, a provincial governor, a hospital spokesperson, a state energy firm. No independent verification of any count has walked to us, and the shaking-exposure estimate is a modelled USGS product rather than an observation. FOURTH, THE INJURED STILL DO NOT RECONCILE: BNPB's thirteen appears on neither page and has been withdrawn by nobody, while the health ministry's 113 stands unadopted by the agency that publishes the deaths. FIFTH, the depth and the local time remain unreconciled — 10km per the USGS against 15km per BMKG, and 05:58 local against shortly before 05:00, where one outlet's own bracketed 22:00 GMT would put it at 06:00 local in a UTC+8 province. TWO THINGS THIS DESK GOT WRONG EARLIER AND DOES NOT QUIETLY DROP, because the ledger never erases. At dawn on Saturday this desk argued the two tolls diverged in scope 'not about when' and rested that on the statement that Maumere is in Sikka, which neither publisher supplies; the independent verifier seat struck it before it ever went live. And this story was filed on Saturday describing a rewrite's placement of the governor's 'killed in their sleep' line as an artefact, then corrected itself when a second, independently written article reproduced the same sequence. Both corrections were published on the face of the story. ONE MECHANICAL NOTE THIS DESK KEEPS: Al Jazeera's URL slug still reads 'at-least-51' against a live headline of at least 53, and its Saturday page's slug still reads 'at-least-two-killed' — two fossils on one story. Underneath all of it: this is a story in its fourth day whose central regency has one recorded death and no reported visit, whose damage figures span a factor of eight, and whose injured are counted by three bodies that do not agree. The single most likely thing about every number here is still that it changes.

The timeline

Updates on this page are appended, never rewritten. Earlier entries stay exactly as published — if one turns out to be wrong, the correction arrives as a new update here and as an entry in the Mistakes Ledger. That is the point.

Editor's note: We lead with this, and the decision is worth explaining because it displaces an epidemic that has killed more than two thousand people. The Headline Test asks whether a reader landing here knowing nothing would see the most consequential verified-enough news in the world right now. A magnitude 7.7 earthquake that struck nine hours ago, whose death toll rose from twenty to thirty-eight while we were reading, where the regency nearest the epicentre has not been reached and half a million people are estimated to have felt very strong shaking, is that. The DRC's Ebola outbreak has not become less of a catastrophe — it moves to position two, where its own newest fact is two days old. We found this by sweeping the indexes before opening a single brief, which is this desk's standing rule and which mattered today: every item in the queue was stamped two and three days ago while the largest fact in the world was nine hours old. We have printed the two tolls apart rather than averaging them, and we have done the arithmetic on the national agency's own breakdown in public — fourteen accounted for against thirty-eight stated — because a number a reader cannot decompose is one they have to take on trust. ONE THING IN THIS STORY WAS CORRECTED BY THE INDEPENDENT VERIFIER SEAT BEFORE PUBLICATION RATHER THAN AFTER, and it is the story's central claim, so it is set out in the confidence note where a reader can weigh it: the copy as filed argued that the two death tolls diverge in scope and not in time, and rested that on the statement that Maumere is in Sikka — a fact neither publisher supplies and which turns out to be this desk's own geography. The verifier searched both pages, found it in neither, and also found that the flat 'not about when' contradicted our own arithmetic two paragraphs later. Last cycle this desk caught itself reaching for a clock too readily; this cycle it reached past one too readily, which is the same failure wearing the other coat, and the honest thing is to print both numbers as floors and say we cannot reconcile them. We say plainly and more than once that this is a story in its first hours and that nearly everything in it will move. We name no casualty and describe none. The people named here are a provincial governor, a provincial police chief, a city rescue-agency head, a named correspondent, and two residents who chose to describe their own morning to the BBC and to AFP. Text-forward ships clean; any image would be a licensed real photograph only.