5.10 Uncertainty of Historical Flood Data

Although recorded stage and discharge data are desired, even historical stage data obtained before the start of actual gaging records can be of value. However, before using these data, the engineer should attempt an evaluation of the accuracy of the historical data.

Before the Great Flood of 1993 in the Midwestern United States, the two largest floods at St. Louis were believed to be the 1844 and 1903 events. Published records by the USGS included a discharge estimate for the 1844 flood under the "Extremes Outside Period of Record" section of the gage records. The 1903 flood is included in the actual records, but there were no discharge measurements during the 1903 flood. The published values for the 1993 and the historic flood events are listed in the table below.

Although these discharges have been in the records for years, only the 1993 flood was actually computed from velocity and depth measurements of the actual flood. The stages of the earlier floods are presumed to be accurate; however, the discharges were never measured, but were estimated after the flood. In fact, the 1844 peak discharge was not estimated until over 60 years later, based on the 1903 event. The 1903 flood discharge was only computed at the Chester, Illinois gage about 70 miles (113 km) downstream of the St. Louis gage. The river between the two sites was very different in 1903, compared to 1844, with the clearing of a large portion of the floodplain for agriculture and a shortening of the river by several miles from a natural river cutoff in 1881. In addition, the velocity meters used in 1903 overestimated the velocities for higher discharges compared to today's standard, the Price current meter. It has been long suspected that the estimated discharges for both the 1844 and 1903 flood events are overly conservative.

Consequently, as part of some physical model tests using the Mississippi Basin Model in the late 1980s, channel and overbank conditions representing those of 1844 and 1903 were incorporated into the model. For both historic flood events, the MBM was used to determine the steady flow discharge that approximated about a dozen high water marks recorded for each flood through the St. Louis reach. The results of the MBM study found that peak discharges of 892,000 ft3/s (25,280 m3/s) and 782,000 ft3/s (22,160 m3/s) closely matched the high water marks for the 1844 and 1903 floods, respectively. It was conservatively estimated that the maximum discharge for the historic 1844 flood of record was no more than 1,000,000 ft3/s (28,350 m3/s). Following the review of the Corps' work, the USGS modified their records to reflect the revised discharge for the 1844 event for the water-records publication in 1999. The 1993 flood discharge is almost certainly the greatest flow at St. Louis over at least the last 200 years (Dieckmann and Dyhouse, 1998).

This example demonstrates the need for caution when evaluating historic data, especially flood discharge data from the nineteenth or early twentieth centuries, when the gaging techniques and methods were less accurate than those used today. Published values of historic flood discharge are simply best estimates by an individual or agency and may be very different from the discharges that actually occurred.

Date of Flood Peak	Discharge, ft3/s (m3/s)	Stage, ft (m)
1785	unknown	42.0 (12.1)
June 27, 1844	1,300,000 (36,800)	41.3 (12.6)
June 1903	1,019,000 (28,880)	38.0 (11.6)
August 1, 1993	1,070,000 (30,320)	49.58 (15.1)