The National Disaster-Warning Case Study on heain-core The Rain Flood in Bangkok, September 2026 Replaying the Real Event on a 39-Node Simulation: Data at the City, the Button at DDPM, Water Across a Provincial Line, with Core C5

Abstract Bangkok's flood of late September 2026 came from days of rain that TMD warned of from 22 September, yet the first Cell Broadcast went out at 09:22 on 26 September, after flooding had begun the day before [9][10]. Reports say the data sat with the city (BMA) while sending sat with DDPM, and the steps between them were unclear [11]. On 29 September water from Pathum Thani crossed the Khlong Hok Wa dike into Khlong Sam Wa, and three districts were still flooded on 16 October [9][12]. This Volume 4 of the case study [1] reports the replay of this event (series E2). The author replayed 21 September to 16 October on a 39-node layout adding Pathum Thani province with the Hok Wa dike, a rain station at Khlong Sam Wa, and Sai Mai and Nong Chok districts, in two runs: "as it happened" and "with HEAIN". Requirements R15–R18 test a hazard from outside the warner's area, Push Through sending, a gate that affects two provinces (a synthetic scenario), and weeks of warnings without alert fatigue. The first run on core C4 passed every check with one expected gap: evidence from the dike gauge in Pathum Thani stopped at its own office, so Khlong Sam Wa waited about 3 hours for its own gauge. Core C5 (ed74e31) added evidence.reach, a rule at the Primary that lets evidence naming a Zone outside its subtree climb to the Primary and come down to the Zones allowed. After the fix the HEAIN run passed 26 of 26 checks with no gap. The first CAP went out 0.03 hours after TMD's warning instead of 105.42; the dike's evidence reached Khlong Sam Wa in 3 seconds; the siren sounded 0.08 hours and the warning went out 0.52 hours after the overtopping, instead of 8.83; and zone-days under water without a current warning fell from 30 to 0 of 42. Keywords: HEAIN Framework, heain-core, Bangkok Flood 2026, Event Replay, Push Through, Cross-provincial Evidence, evidence.reach, Alert Fatigue, CAP 1.2, Cell Broadcast