The National Disaster-Warning Case Study on heain-core The Hat Yai Flood, November 2025 Replaying the Real Event on a 34-Node Simulation: As It Happened versus with HEAIN, and Two Core Rounds (C4, C7)

Abstract The Hat Yai flood of late November 2025 followed about 630 mm of rain in 72 hours [9]. Station X.90 on Khlong U-Taphao reported Extreme at 20:11 on 21 November, but the full evacuation order came at 11:05 on 24 November, after the water had already overflowed into the city [8][10]. Warnings gave no shelter, route or time, power cuts silenced telecoms, and over 20,000 calls for help were duplicated or lacked a location [8]. This Volume 5 of the case study [1] reports the replay of this event. The author replayed the event along its reported timeline (1 hour = 1 minute) on a 34-node layout with three canal stations, a hospital Micro Node, a shelter Zone at Prince of Songkla University (PSU), 12,345 synthetic callers and 200 rescue resources. Each round ran twice: "as it happened", with decisions at their real times, and "with HEAIN", with rules approved in advance acting. Requirements R1–R14 were drawn from what failed in the real event. The first run on core C3 found 4 gaps and 2 findings: a planned role move that never happened, a single route to the parent, a Micro Node unable to change Zones, no input for a low-power warning, an app on the standby refused 19,883 times, and a standby that did not go STANDALONE when cut off seconds after taking over. After two core rounds (C4 033ee05 and C7 cb86f51), the HEAIN run passed 34 of 34 checks with no gap. The first evacuation CAP went out 1.1 hours after the station reached Extreme instead of 62.9; messages without a shelter, route or time fell from 17 to 0; the same resources reached 7,778 people instead of 5,631; missions spent on duplicates fell from 2,088 to 0; and the office handed its role to its standby 4 hours before the power failed. Keywords: HEAIN Framework, heain-core, Hat Yai Flood 2025, Event Replay, Early Warning, CAP 1.2, Evacuation, Rescue Requests, Standby Failover, Power Step-down, Network Partition