Scour at bridges is a major concern in the design of new bridges and in the evaluation of structural stability of existing bridges. Equations for estimating pier, contraction, and abutment scour have been developed from numerous laboratory studies using sand-bed flumes, but little verification of these scour equations has been done for actual rivers with various bed conditions. This report describes the results of reconnaissance and detailed scour assessments and a sediment-transport simulation for selected bridge sites in South Dakota. Reconnaissance scour assessments were done during 1991 for 32 bridge sites. The reconnaissance assessments for each bridge site included compilation of general and structural data, field inspection to record and measure pertinent scour variables, and evaluation of scour susceptibility using various scour-index forms. Observed pier scour at the 32 sites ranged from 0 to 7 feet, observed contraction scour ranged from 0 to 4 feet, and observed abutment scour ranged from 0 to 10 feet. Thirteen bridge sites having high potential for scour were selected for detailed assessments, which were accomplished during 1992-95. These detailed assessments included prediction of scour depths for 2-, 100-, and 500-year flows using selected published scour equations; measurement of scour during high flows; comparison of measured and predicted scour; and identification of which scour equations best predict actual scour. The medians of predicted pier-scour depth at each of the 13 bridge sites (using 13 scour equations) ranged from 2.4 to 6.8 feet for the 2-year flows and ranged from 3.4 to 13.3 feet for the 500-year flows. The maximum pier scour measured during high flows ranged from 0 to 8.5 feet. Statistical comparison (Spearman rank correlation) of predicted pier-scour depths (using flow data col- lected during scour measurements) indicate that the Laursen, Shen (method b), Colorado State University, and Blench (method b) equations correlate closer with measured scour than do the other prediction equations. The predicted pier-scour depths using the Varzeliotis and Carstens equations have weak statistical rela- tions with measured scour depths. Medians of predicted pier-scour depth from the Shen (method a), Chitale, Bata, and Carstens equations are statistically equal to the median of measured pier-scour depths, based on the Wilcoxon signed-ranks test. The medians of contraction scour depth at each of the 13 bridge sites (using one equation) ranged from -0.1 foot for the 2- year flows to 23.2 feet for the 500-year flows. The maximum contraction scour measured during high flows ranged from 0 to 3.0 feet. The contraction- scour prediction equation substantially overestimated the scour depths in almost all comparisons with the measured scour depths. A significant reason for this discrepancy is due to the wide flood plain (as wide as 5,000 feet) at most of the bridge sites that were investigated. One possible way to reduce this effect for bridge design is to make a decision on what is the effective approach section and thereby limit the size of the bridge flow approach width. The medians of abutment-scour depth at each of the 13 bridge sites (using five equations) ranged from 8.2 to 16.5 feet for the 2-year flows and ranged from 5.7 to 41 feet for the 500-year flows. The maximum abutment scour measured during high flows ranged from 0 to 4.0 feet. The abutment-scour prediction equations also substantially overestimated the scour depths in almost all comparisons with the measured scour depths. The Liu and others (live bed) equation predicted abutment-scour depths substantially lower than the other four abutment-scour equations and closer to the actual measured scour depths. However, this equation at times predicted greater scour depths for 2-year flows than it did for 500-year flows, making its use highly questionable. Again, limiting the bridge flow approach width would produce more reasonable predicted abutment scour.