A few minutes before three in the afternoon on May 31, 1889, after two days of rain heavier than any the region had seen since the dam went up, the South Fork Dam above Johnstown, Pennsylvania, gave way. It had impounded the Little Conemaugh River's headwaters since 1853. The flood that followed killed more than 2,200 people.
The dam had already been repaired once, by men who were not engineers. Built by the state, it passed to the Pennsylvania Railroad in 1857, then to John Reilly in 1875, who conveyed it in 1880 to the South Fork Hunting and Fishing Club of Pittsburgh, which still had to close a breach from 1862. The club's contractor filled the gap with dumped stone and clay, packed by cart wheels, not the layered fill the design specified. The five cast-iron pipes that once let the builders draw the lake down were never replaced, the crest came out about two feet low, wide enough now for a road, and the club screened the spillway to keep its fish in. No engineer checked what any of it cost the dam's capacity to pass a flood.
The American Society of Civil Engineers appointed a committee five days later: James B. Francis, ASCE's own president from 1880 to 1882, with Max J. Becker, its sitting president; Alphonse Fteley, its vice president; and William E. Worthen, a past president. They surveyed the ruins, interviewed witnesses, gathered rainfall readings from fifty miles around, and worked out by hand how much water a dam built to the 1853 plan could pass against how much the altered one actually could. Their report, signed in January 1890, did not appear in the Society's own Transactions until June 1891, two years after the flood, and it cleared the club of the one finding most people wanted most: by the committee's own arithmetic, a dam at full original height, drainage pipes intact, would have overflowed and failed that same afternoon anyway, only a little later than the one that broke.
The finding held for nearly a century. An ASCE engineer named Frank challenged it in 1988, pointing to the same overlooked spillway on the dam's western abutment and arguing the original, unaltered dam would not have overtopped, but published no discharge calculations to prove it. In 2016, the geologists Neil Coleman and Uldis Kaktins and the historian Stephanie Wojno supplied those calculations: with a laser survey of the embankment and a model of how fast an earthen dam that size drains, they found the committee had also overstated the storm's inflow, discounting on-scene observers who reported the lake's feeder streams had already crested hours earlier. A dam rebuilt to the original 1853 specification, they concluded, could have discharged more than double what the altered one managed, and probably would not have overtopped.
The 1891 committee was not a panel of amateurs: its four members held the society's top offices between them. Their report was careful by the standards of 1891 hydraulics, and still missed a spillway standing in the wreckage they surveyed, and leaned on an inflow estimate the watchers on the scene could have corrected. A dam either passes a flood or it doesn't. Finding out which one actually failed first took someone willing to reopen a signed verdict, not just more time.