HR: 13:40h
AN: H42J-01    [PDF]
TI: Lessons From the Largest Historic Floods Documented by the U.S. Geological Survey
AU: * Costa, J E
EM: jecosta@usgs.gov
AB: A recent controversy over the flood risk downstream from a USGS streamgaging station in southern California that recorded a large debris flow led to the decision to closely examine a sample of the largest floods documented in the US. Twenty-nine floods that define the envelope curve of the largest rainfall-runoff floods were examined in detail, including field visits. These floods have a profound impact on local, regional, and national interpretations of potential peak discharges and flood risk. These 29 floods occured throughout the US from the northern Chesapeake Bay in Maryland to Kauai, Hawaii, and over time from 1935-1978. Methods used to compute peak discharges were slope-area (21/29), culvert computations (2/29), measurements lost or not available for study (2/29), bridge contraction, culvert flow, and flow over road (1/29), rating curve extension (1/29), current meter measurement (1/29), and rating curve and current meter measurement (1/29). While field methods and tools have improved significantly over the last 70 years (e.g. total stations, GPS, GIS, hydroacoustics, digital plotters and computer programs like SAC and CAP), the primary methods of hydraulic analysis for indirect measurements of outstanding floods has not changed: today flow is still assumed to be 1-D and gradually varied. Unsteady or multi-dimensional flow models are rarely if ever used to determine peak discharges. Problems identified in this sample of 29 floods include debris flows misidentified as water floods, small drainage areas determined from small-scale maps and mislocated sites, high-water marks set by transient hydraulic phenomena, possibility of disconnected flow surfaces, scour assumptions in sand channels, poor site selection, incorrect approach angle for road overflow, and missing or lost records. Each published flood magnitude was checked by applying modern computer models with original field data, or by re-calculating computations. Four of 29 floods in this sample were found to have errors resulting in a change of the peak discharge of more than 10%.
DE: 1821 Floods
DE: 1860 Runoff and streamflow
SC: Hydrology [H]
MN: 2003 Fall Meeting