Just before two in the morning on April 1, 1946, the sea floor broke on the northern slope of the Aleutian Deep, south of Unimak Island, Alaska. Within the hour a wave struck the Scotch Cap lighthouse, a reinforced-concrete tower the Coast Guard had built six years earlier, and left only its foundation standing. All five men stationed there died. The same earthquake had, within minutes, been recorded on two instruments more than two thousand miles away: seismographs at the University of Hawaii in Honolulu and the Hawaiian Volcano Observatory on Kilauea.
The reading was accurate from the first instant, and for nearly five hours it did nobody in Hawaii any good. The wave crossed the Pacific at about 490 miles an hour, and Honolulu's tide gauge did not register its first rise until about 6:33 that morning. At Hilo, the water pulled back from the bay, then returned as a wave that broke twenty-five feet above normal at Coconut Island. By the time the waves stopped, a hundred and fifty-nine people were dead across the islands, most of them in Hilo, and property damage totaled about twenty-five million dollars.
Nothing about this was a failure of measurement. The ground motion was there on the seismograph record in Honolulu and at Kilauea almost as soon as it happened. But neither station had an alarm wired to its own instrument in 1946; a seismograph's recording was a fact sitting on a drum, not a signal that rang a bell or woke anyone. G. A. Macdonald, F. P. Shepard, and D. C. Cox, who spent months measuring the wave's path, closed their report, published the following January, by proposing a network of stations able to turn a reading like that into a warning before the water arrived.
On August 12, 1948, the Coast and Geodetic Survey and the armed forces agreed on a communication plan, and the Seismic Sea Wave Warning System began operating from a Honolulu observatory that took in readings from stations as far off as Alaska and Japan. New instruments installed in 1947 and 1948 gave a continuous, visible trace instead of film developed later, but the alarm itself, the part that would ring with nobody watching, waited until summer 1950: an amplifier whose circuit tripped the moment a major earthquake came in. When an earthquake set off a tsunami on March 9, 1957, that alarm sounded within six minutes, and tide observers in its path were alerted within the hour. A government bulletin tallying the next decade's tsunamis called 1946 the only one of three that arrived unannounced, and credited the system with zero deaths in both 1952 and 1957.
A true reading does nobody any good until a person acts on it before the thing it measured arrives, and a drum turning in an empty room does not close that gap by itself. The instruments in 1946 did their job in minutes. Wiring an alarm to that same kind of instrument, so a reading turned itself into a warning without waiting for someone to walk in and look, took four more years and a hundred and fifty-nine deaths that a seismograph, left unread, had already failed to prevent.
The ground had already told the truth on April 1, 1946. Hawaii just had no way to hear it in time.