Scour at Contracted Bridges (2006)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

7 CHAPTER 2: FINDINGS EXECUTIVE SUMMARY OF LITERATURE REVIEW AND FIELD DATA Scour at contracted bridge openings results from the complex interaction of hydraulics, bridge geometry, soils characteristics, and other site-specific conditions. The current understanding of bridge-scour processes, largely have been derived from laboratory investigations consisting of physical models in straight, rectangular flumes with uniform non- cohesive bed material. These models simplify the complexities of field conditions, but allow researchers to gain insights about scour processes under controlled conditions. Observations from these studies have led to the development of simplified concepts of scour processes and various methods for evaluating scour at bridges. The concepts and methods for evaluating scour derived from these studies may not accurately account for scour processes in the field, because of the simplifications inherent in these laboratory investigations. The literature divides bridge scour into various components that are considered independent and additive. The most common components include long-term streambed aggradation or degradation, contraction scour, and local scour. Most research has focused on the last two components, and a summary of contraction and local scour as it relates to this study follows. Contraction scour is the erosion of material from the bed and banks across all or most of the channel width, resulting from the contraction of flow area. The literature presents various

8 methods for estimating contraction scour including (1) regime equations, (2) hydraulic-geometry equations, (3) numerical sediment-transport models, and (4) contraction-scour equations. Regime and hydraulic-geometry equations are empirical equations that are used to assess changes in channel geometry for given hydraulic conditions. Although originally developed to assist in the design or assessment of channel shape, these methods can be used for estimating contraction scour at bridges. The assumption implied by use of these equations is that changes in unit discharge cause a unique change in channel depth. These equations must be calibrated with local or regional field data, which limits their application to sites with characteristics similar to those used for calibration. Numerical sediment-transport models combine various sediment-transport equations with numerical hydraulic models to simulate scour processes in streams. Hydraulic conditions estimated with these models are used to drive the sediment-transport equations. The literature shows that the various sediment transport equations provide significantly different estimates of sediment discharge for the same site. Given adequate topographic and channel data, numerical models have been shown to provide reasonable estimates of hydraulic parameters at some sites. Adequate representation of sediment transport and scour requires selection of specific sediment- transport equations developed for the specific conditions of the site and may require site calibration. To assure that the results from the sediment-transport numerical model are reasonable, the model should be calibrated and verified with observed field data. Sediment transport models are rarely used to estimate contraction scour because of the time and cost associated with data collection necessary to construct, calibrate and verify these models.

9 The literature describes a number of semi-empirical contraction-scour equations that were developed by use of conservation of flow and sediment in a control volume, in conjunction with laboratory derived concepts of sediment transport. These equations can be readily applied to a given site, which may account for their common use. Laboratory researchers have found that the transport or lack of transport of sediment in the flow approaching an obstruction or contraction is critical in assessing scour at bridges. Contraction scour has traditionally been classified as live-bed or clear-water, which reflects the bed material sediment-transport conditions of approaching flows. Researchers have used similar approaches to derive the various equations. In the case of live-bed scour, the common assumption is that scour will cease when the load of sediment transport into the contraction is equal to the load transported from the contraction. The major difference in the various equations stems from the use of different sediment-transport relations. Though differences exist within the derivations, the format and exponents of the various live-bed equations generally are similar. In the case of clear-water scour, the common assumption is that scour will cease when the bed- shear stress in the contraction equals the critical shear stress for the bed material. The critical shear stress is typically determined from Shield’s diagram that represents a laboratory-derived shear stress for incipient motion of uniform, noncohesive sediments. The Shield’s relation and other similar relations represent laboratory-derived shear stress for incipient motion of uniform, noncohesive sediments. Other common assumptions used in the derivation of live-bed and clear- water contraction-scour equations include steady-uniform flow, noncohesive bed material, and sufficient time to achieve equilibrium conditions. To the degree that field conditions deviate

10 from these and other assumptions, it is likely that the contraction-scour equations may not provide reasonable scour depths under field conditions. Local scour is the removal of bed material from around flow obstructions such as piers, abutments, spurs, and embankments caused by the local flow field induced by a pier or abutment. Analytical equations for predicting abutment scour primarily have been derived from observations obtained from small-scale physical-model studies conducted in laboratory flumes. As with contraction scour, abutment-scour equations have been classified as live-bed or clear- water, reflecting the approaching sediment-transport conditions. The equations can be subdivided further into empirical and semi-empirical equations. The empirical equations were developed from envelope curves or regression analysis of dimensionless variables obtained from laboratory investigations. The semi-empirical equations were derived in a similar manner to the contraction- scour equations by use of conservation of flow and sediment in a control volume in conjunction with laboratory-derived concepts of sediment transport. Abutment-scour depth is often assumed to be a function of contraction-scour depth and the contraction-scour equation is adjusted to reflect the increased scour potential at the abutment. In addition to laboratory-derived equations, there are several abutment-scour equations derived from field observations. These field-derived equations were developed from limited data sets for site-specific conditions; therefore, they may not be applicable to other sites. Numerical sediment-transport models also have been used to investigate abutment scour, and results from these models are subject to the same limitations described for contraction scour. Complete and reliable field data sets are rare, although there have been more than 100 laboratory studies in which detailed and complete data sets have been published (Melville and

11 Coleman, 2000). A survey of the literature located 30 references with potential field data for abutment and contraction scour. Of the 30 references reviewed, 4 are potential sources of data for abutment scour and 22 are potential sources for contraction scour. Most of the scour data presented in these references were collected during post-flood investigations, and flow conditions that created the scour were estimated from hydraulic models. Nearly all of the sites identified in the literature review required the compilation of raw data and additional analysis to obtain complete abutment and contraction-scour data sets. An exception to this is data collected by the U.S. Geological Survey (USGS) at 146 bridges in South Carolina. Hydraulic models were developed for these sites and hydraulic variables were compiled into a database and associated with field observations of scour. This database was developed to assess clear-water contraction and abutment-scour equations. It should be noted that the South Carolina data were not just post- flood measurements, but were often remnant scour after several years or decades of recovery and there was often no knowledge of what flood event caused the scour. Studies found in some of the references compare field observations with computed scour. Contraction- and abutment-scour comparisons frequently predict scour depths greater than those observed and often this bias can be three to four times the measured scour depth; however, some comparisons indicate that there are conditions under which some equations will predict scour depths less than those observed. These comparisons indicate that the current methods for predicting contraction and abutment scour at bridges are unreliable.

12 SCOUR AT BRIDGE CONTRACTIONS Field observations of scour at many bridges indicate that conceptual separation of contraction and abutment scour as described in Hydraulic Engineering Circular-18 (HEC-18) (Richardson and Davis, 2001) is problematic because the hydrodynamic mechanisms that induce the individual scour components work together. It is clear from the field observations of this study that the scour that occurs near the ends of the abutment is the result of a complex combination of flow contraction and flow curvature. Scour-prediction methods published in HEC-18 indicate that contraction and abutment scour are separate and additive for all contracted bridge openings. HEC-18 follows a conservative approach of adding the scour components to create a scour prism for design and assessment purposes, because of an insufficient amount of field data to develop an understanding of the interaction of scour components. Therefore, to compute the total scour at an abutment, the individual components of long-term streambed change, contraction scour, and abutment scour within the abutment region must be estimated and then summed. Isolating the effect of an individual scour component is difficult because the various components interact in the development of the total depth of scour. Laboratory investigations typically have focused on understanding each scour component in isolation, necessitating the approach for estimating total scour as outlined in HEC-18. Analyses of field observations, in conjunction with the theory of flow patterns in short contractions, indicate that this view of scour in the abutment region may be inappropriate.

13 Consideration of contraction and local-abutment scour as independent and separate processes in the abutment region is a particular concern. The assessment of contraction scour is often based on the simplifying assumption of uniform-flow distributions within a long contraction. By simplifying the hydraulics in a contraction to uniform flow, the current patterns are assumed rectilinear and equations for predicting scour can be derived following the procedure applied by Laursen (1963), which utilized the concept of critical bed-shear stress for rectilinear flow. As the abutment-to-abutment width of the contracted bridge increases for a given depth, local flow patterns through the structure trend toward becoming approximately rectilinear and the assumptions used to develop Laursen’s (1963) equation are more appropriate. The highly curvilinear velocities near the abutment promote vortexes, which are the primary mechanism for the development of scour in the abutment region (Dongol, 1993). The absence of rectilinear flow patterns in the abutment region indicate that it may be reasonable to assume that scour produced by this flow pattern is absent as well. Following this logic, contraction scour produced by rectilinear flow should not be considered a component of total scour within the abutment region; therefore, the total scour in an abutment region should consist of long-term streambed change, local abutment scour generated from the highly curvilinear-flow patterns, and local scour generated from any piers within this same flow field (Benedict, 2003). Reference conditions from which the abutment scour and contraction-scour depth can be measured at field sites are difficult to determine because, unlike the planer initial conditions of laboratory experiments, field sites have a complex bed topography that reflects site history including bridge and roadway construction. When attempting to compare field observations and scour predictions from various abutment-scour equations derived from laboratory data, it is

14 necessary to understand how the depth of abutment scour or contraction scour was measured in the laboratory. In the laboratory, observed contraction scour that has occurred beyond the abutment region is subtracted from the total observed scour depth at a simulated abutment. For example, laboratory investigations by Dongol (1993) measured contraction scour at a flume wall opposite from an abutment, and subtracted it from the total scour at the abutment in an attempt to isolate the scour created by the abutment alone. A parallel reference condition rarely exists under complex field conditions making comparisons of data difficult. Analysis of field data has shown that the interaction between what has been called contraction and abutment scour is highly complex. It is sometimes difficult to separate the two components in the region adjacent to abutments. Using data collected at 146 sites in South Carolina, Benedict (2003) found that bridges approximately 240 ft or less in length tend to form a large, single scour hole (Figure 1) rather than separate left and right abutment-scour holes (Figure 2). This phenomenon may be caused by the highly rotational flow separating from the channel boundary at the left and right abutments. Scour holes at these shorter bridges could be classified as abutment scour hole, because highly rotational flow is typically associated with abutment scour; however, the scour hole could instead be classified as contraction scour because it is separated from the abutment and is a single scour hole near the center of the channel. Scour holes at bridges shorter than 240 ft were classified as abutment scour by Benedict (2003), although the bed was degraded across the entire channel, characteristic of contraction scour. Field data collected throughout the United States for NCHRP Project 24-14 did not show the same consistent breakpoint at 240 ft as was observed in South Carolina; however, the data collected

15 Figure 1. Example of single scour hole at shorter bridges, as shown at U.S. Route 301, crossing Douglas Swamp in Florence County, South Carolina, July 31, 1996. (.5 feet = .152 meters)

16 Figure 2. Example of separate left and right abutment-scour holes at longer bridges, as shown at Road S-87 bridge, crossing the Coosawhatchie River in Jasper County, South Carolina, November 12, 1997. (1 foot = .305 meters)

17 showed some bridges with abutment scour and no contraction scour (Figure 3) and other bridges (Figures 4) with apparent contraction scour and no abutment scour. Thus, development of contraction and (or) abutment scour is highly dependent up on the site and approach flow conditions (see Appendix A, case studies No.1 and No. 5, for a discussion of site and approach conditions for the bridges shown in Figures 3 and 4), and the presence of a contracted bridge opening does not guarantee that either or both types of scour will occur. Although the overall effects of flow contraction and the local flow curvature that occurs around abutments can be conveniently separated conceptually, the resulting scour pattern cannot be separated into contraction- and abutment-scour components. The cause of the specific scour patterns is believed to be highly sensitive to local field conditions. The field observations collected during this study are not adequate to develop a definitive classification system based on site characteristics that could indicate the expected scour pattern. For the purposes of this report, we will consider scour that occurs adjacent to an abutment to be abutment scour and any other general scour not occurring adjacent to a pier or abutment to be contraction scour or some other type of scour resulting from unique site characteristics. The depth of abutment scour is always measured from the adjacent bed unaffected by local scour, which will include any contraction scour that occurred at the site. The depth of contraction scour will be measured relative to an uncontracted surface usually defined by cross sections collected upstream and downstream from the bridge beyond the hydraulic affect of the bridge.

18 Figure 3. Plot of abutment scour measurements at the C.R. 22 bridge over the Pomme De Terre River near Fairfield, Minnesota in April 1997. 180 190 200 210 310 311 312 313 314 315 316 317 4/4/97 4/5/97 4/9/97 BRIDGE PLANS STATION, IN M EL E V AT IO N , I N M PIERPIER EL E V AT IO N , I N M ETERS E TE R S

19 B Figure 4. Plot of contraction scour measurements at (A) Conehoma Creek at State Highway 35, near Kosciusko, Mississippi, April 1979 and (B) Beaver Creek Overflow at US 2, 7 miles West of Saco, Montana.

20 CONTRACTION SCOUR Contraction scour traditionally has been classified as live-bed or clear-water. The live- bed condition is characterized by bed material being transported into the contracted opening from upstream of the bridge. Live-bed scour is typical of scour that occurs in the main channel portion of a waterway in high-flow conditions. Clear-water contraction scour occurs when the flow conveyed to the bridge crossing is not transporting bed material; thus, all material that is transported from the contracted section is sediment being scoured. Scour occurring on vegetated floodplains may be classified as clear-water scour despite the potential for the shear stress in the approach section to be greater than the critical shear stress of the material comprising the floodplains. Discussion of Live-Bed Contraction-Scour Equations Straub (1935) was the first to develop an approach to predict contraction scour that most others would follow. He assumed that the bed in the contracted section would scour until it reached a depth at which the local transport capacity was equal to the amount of material being supplied from upstream (sediment-discharge continuity). He selected the DuBoys sediment- transport equation to compute the amount of material supplied to the reach and the local transport capacity in the contraction. Straub estimated the energy-dissipation rate (friction slope) in the contracted and uncontracted reaches by use of Manning’s equation. This assumption is reasonable where flow curvature is small and pressure gradients are small compared to boundary stresses. The hydraulics in a short contraction, such as a bridge crossing, require the

21 consideration of additional energy losses not accounted for in a roughness coefficient based on the channel composition and configuration (Matthai 1968; Schneider and others, 1977; Shearman and others, 1986). Straub’s equation, based on sediment discharge-continuity, water-discharge continuity, and the Manning equation has the general form of equation 1: )(f n n b b Q Q y y nbQ E 1 2 E 2 1 E 1 2 1 2 τ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛ ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛ ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛= (1) where y2 is the depth in the contracted section; y1 is the depth in the uncontracted section; b1 is the bottom width in the uncontracted section; b2 is the bottom width in the contracted section; n2 is the Manning’s n in the contracted section; Q2 is the discharge in the contracted section; n1 is the Manning’s n in the uncontracted section; Q1 is the discharge in the uncontracted section; τ represents one or more shear-stress variables; EQ is the exponent on the ratio of discharges; Eb is the exponent on the ratio of bottom widths; and En is the exponent on the ratio of roughness coefficients. For most applications the shear stress based function is assumed equal to unity.

22 TABLE 1. Summary of live-bed contraction scour equation exponents. Equation (Q2/b2)/ (Q1/b1) b1/b2 Q2/Q1 Straub (1935) 0.43 0.86 Straub (1935) .642 Griffith (1939) .637 Neill (1973) 0.67-0.85 Laursen (1962) .6 – .7 .86 Komura (1966) .85 Komura (1966) .667 Culbertson and others (1967) .667 The approach to developing live-bed contraction-scour equations is similar among all researchers and differs primarily by the method of determining the sediment-transport capacity. A summary of the exponents of the ratios common to the equations developed by select researchers is shown in Table 1. There is good consistency in the exponents, considering that each researcher used a different sediment-transport equation. In the derivation of the live-bed contraction-scour equations the sediment-transport equation is applied to both the contracted and uncontracted sections and only the difference in the transport rates between these sections affects the computed depth of scour. Thus, the depth of computed contraction scour does not appear to be sensitive to the selection of the transport equation (Mueller and Wagner, 2002).

23 Richardson and Richardson (1994) modified Laursen’s live-bed equation by removing the ratio of Manning’s n in equation 1. They concluded that Laursen’s equation did not correctly account for the increase in transport that would occur if a plane bed existed in the contracted opening with a dune-bed configuration in the approach section. For this situation, Laursen’s equation would predict less scour than if the roughness coefficients were equal. The Manning’s n ratio in Laursen’s equation does, in fact, behave properly. The basic principle of estimating contraction scour is the assumption of achieving equilibrium sediment transport. With a plane bed configuration more sediment can be transported at a reduced depth than for a dune-bed configuration; therefore, equilibrium sediment transport can be achieved at a shallower depth. To achieve a plane bed configuration, the streambed had to progress through the dune bed configuration in the contracted section. A deeper scour may have occurred at a lower flow with a dune configuration in the contracted section than at a higher flow with a plane-bed configuration; therefore, to predict the maximum depth of scour for design purposes, a constant Manning’s n should be assumed in the approach and bridge sections. This yields the same result as that proposed by Richardson and Richardson (1994). Discussion of Clear-Water Contraction-Scour Equations Clear-water scour occurs where the boundary-shear stress in the uncontracted section is less than or equal to the critical tractive force of the bed material, thus, preventing the supply of material into the contracted section. Laursen (1963) assumed that the maximum limit of clear- water scour occurs when the boundary shear stress is equal to the critical tractive force. This assumption is common among all of the proposed clear-water contraction equations. The critical

24 shear stress of noncohesive sediments with a specific grain size is commonly estimated from the Shield’s diagram. The critical velocity for incipient motion can be computed from the Shield’s parameter by substituting the Manning equation for the slope term of the shear-stress equation and then using Strickler (1923) to approximate Manning’s n. By setting the velocity in the contracted section equal to the critical velocity and solving for depth, the following generic clear-water contraction-scour equation (2) is obtained: 3 3/2 m 2 u 2 2 2 D)08.31K( Vy ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛ θ= (2) where: y2 is the average equilibrium depth in the contracted section after scour, in m; V2 is the average velocity in the contracted opening in m/s; θ is the Shield’s parameter for the bed material; Ku is a constant (.025 in SI units, .0077 in English units); and Dm is the mean grain size of the bed material, in m. Based on research on the effective size of bed material for riprap design and resistance to erosion presented in Richardson and others (1990), Richardson and Richardson (1994) suggest that 1.25 D50 be used for Dm. The critical velocity or critical shear stress and the corresponding scour is not well established for channels having cohesive bed material, bed material that varies with depth, heavily vegetated floodplains, previously developed scour holes, or armored beds. The clear- water scour equations were developed by use of flow variables obtained under initially flat bed

25 conditions; however, flow variables can change substantially after a scour develops. As a result large changes in flow area and backwater effects can affect the accuracy of scour-prediction equations. Clear-water scour prediction is highly sensitive to the critical conditions of the bed material; therefore, accurate representation of the bed material is essential. Bed-material samples should represent both the surface and subsurface material. Various agencies use different methods for determining the critical velocity. These inconsistencies can have a significant impact on clear-water scour estimates, especially in fine grained soils. Summary of Observations The data presented in this section were observed at 13 real-time sites included in the BSDMS and 146 sites in South Carolina (Benedict, 2003). The observed data represent a wide range of site characteristics (i.e., non-cohesive and cohesive soils, upstream channel meanders, spur dikes, varied geomorphic settings, and complex bridge configurations). A detailed description of the techniques and equipment used to collect real-time scour measurements can be found in Mueller and Wagner (2002). The procedure that was used to collect the post-flood scour measurements is detailed in Benedict (2003). Comparison of Observed and Computed Scour Depths In this study, observed depths of contraction scour were compared with computed depths based on the approach presented in HEC-18. Only three sites had sufficient data to allow computation of contraction scour based on field data alone. Thus, all computed scour used in the comparison

26 is based on hydraulic parameters from a one-dimensional model calibrated to the discharge and water-surface elevations observed in the field. The equations for live-bed contraction scour consistently overpredicted the observed scour (Table 2). This overprediction ranged from two to four times the observed scour. The equations for clear-water contraction scour substantially overpredicted the observed depths of scour (Table 2). Situations that fit into the clear-water scour regime are more likely to be in areas with cohesive bed material. Cohesive soils greatly reduce actual scour depths, and therefore, result in even larger overpredictions of scour when applying clear-water equations. Contraction scour was not always observed in the field, even at sites with considerable contraction ratios and high velocities. Benedict (2003) observed no scour at about 25 percent of the sites in South Carolina; however, the HEC-18 approach predicted scour at all of these sites (Figure 5). Bed-material size and gradation, cohesion, armoring potential, and road overflow are common factors that can limit or prevent contraction scour, but other highly site-specific preventative factors have been discovered in the analysis of field measurements. Real-time measurements during an approximate 50-year event at the 247th Street bridge over the James River near Mitchell, South Dakota revealed no scour through the bridge opening and around .3 meters (m) of contraction scour 9 to 15 m downstream of the bridge. Intense velocities through the bridge, a considerable contraction of the left floodplain (about 610 m) and a silty bed material would indicate the occurrence of contraction scour using HEC-18 prediction techniques. Analysis of the data and bridge plans revealed that although there was minimal road-overflow, remnants of an old bridge under the existing structure were acting as scour protection.

27 Channel alignment upstream and through the bridge opening is another factor that can greatly affect the occurrence of contraction scour. The increase in flow conveyance caused by scour at an abutment can be sufficient to reduce overall velocities and shear stresses through the bridge opening, limiting scour to the area near the abutment. This phenomena has been observed

TABLE 2. Comparison of observed and theoretical contraction-scour depth (clear-water and live-bed) for bridge sites in the National Bridge Scour Database (BSDMS). Contraction scour Site name Scour type Date Computed (meters) Observed (meters) Chehalis River at Galvin Road Overflow, Centralia, Washington Clear- water 2/9/1996 9.32 0.92 Beaver Creek Overflow 7 miles West of Saco, Montana Clear- water Unknown 39.6 1.37 Beaver Creek Overflow 9 miles West of Saco, Montana Clear- water Unknown 21.0 .99 Bitterroot River at US 93 near Darby, Montana Clear- water 6/11/1996 0 .21 Minnesota River at SR 25 near Belle Plaine, Minnesota Live-Bed 4/17/2001 10.8 4.57 Conehoma Creek at SR 35 near Kosciusko, Mississippi Live-Bed 4/5/2001 5.79 1.83 James River at SR 37 near Mitchell, South Dakota Live-Bed 4/15/2001 4.48 1.68 Conehoma Creek at SR 35 near Kosciusko, Mississippi Live-Bed 4/12/1979 5.79 1.22 Pomme De Terre River at US 12 near Holloway, Minnesota Live-Bed 4/5/1997 1.40 .94 Cedar River at S.R. 218 near Janesville, Iowa Live-Bed 7/23/1999 .76 .61 Pomme De Terre River at CR 22 near Fairfield, Minnesota Live-Bed 4/9/1997 .00 .00 28

29 Figure 5. Relation of observed and theoretical clear-water contraction-scour depth for the 100- year flow, in the Piedmont of South Carolina. (Theoretical contraction scour calculated with the Laursen (1963) equation.)

30 at sites with long bridges that have extremely high rotational velocities at the abutments as a result of upstream meanders. Appreciable embankment skew and (or) expansive floodplains also have been shown to limit or exclude contraction scour in some situations. The County Route 22 bridge over the Pomme De Terre River near Fairfield, Minnesota is an example of a site where the upstream channel alignment and corresponding abutment scour precluded any contraction scour (see Appendix A, Case Study No. 1). Thus, development of contraction and (or) abutment scour is highly dependent upon the site and approach flow conditions. The presence of a contracted bridge opening does not guarantee that either or both types of scour will occur. Location of Scour at Contracted Bridges Analysis of the field data also has revealed that the location of scour in a contracted bridge opening is highly variable and does not follow the patterns typically reported from laboratory experiments. The longitudinal location of contraction scour can be dependent upon factors such as the configuration of scour protection, guide banks, bridge length, channel alignment, and bed material. The detailed data from two of the real-time scour sites clearly shows the variation in longitudinal location of contraction scour. The contraction scour at State Route 218 over the Cedar River near Janesville, Iowa is located upstream of the S.R. 218 Westbound Bridge in the region where the floodplain flow combines with the main channel flow (Figure 6) (also see Appendix A, Case Study No. 8). A spur dike extending upstream from the right abutment directs the right floodplain flow into the main channel approximately 35 m upstream of the bridge. The configuration of the scour pattern observed at State Route 25 over the Minnesota River at Belle Plaine, Minnesota (Figure7) (also see Appendix A, Case Study No. 3) is caused by a

31 combination of upstream approach alignment and riprap around the left abutment and piers. Abutment-scour holes were observed at both abutments and contraction scour was observed across the entire bridge section, but the deepest scour was observed downstream between the left abutment and the leftmost pier. The downstream location of the scour hole may be attributed to scour protection that was in place on the left abutment and leftmost pier that prevented degradation of the bed in the bridge opening. The uncontracted approach section is commonly assumed to be one-bridge length upstream from the bridge. Field observations show that the location of the uncontracted approach section also can be highly variable depending upon site specific characteristics such as flow structures, geomorphic setting, floodplain topography and land cover, and upstream channel configuration. The berm and geomorphic setting of the S.R. 37 Bridge over the James River near Mitchell, South Dakota (see Appendix A, Case Study No. 4) induced hydraulic conditions that could not have been represented using the standard guidelines for locating the approach cross-section (approximately one-bridge length upstream). Inspection of the "approach" section (one-bridge length upstream) revealed a large discharge (334 cubic meters per second (m3/s)) relative to that of the contracted opening (394 m3/s). It was discovered that the blockage caused by the roadway embankment forced the majority of the left floodplain flow back into the main channel at the typical approach section (Figure 8). A cross section made further upstream showed much less discharge (191 m3/s), which was consistent with channel discharge

Figure 6. Example of scour-hole low point located upstream of S.R. 218 over the Cedar River near Janesville, Iowa. Flow S.R. 218 Westbound Bridge S.R. 218 Pier Locations 32 Eastbound Bridge

Figure 7. Example of scour-hole low point located downstream of S.R. 25 over the Minnesota River near Belle Plaine, Minnesota. 33

Figure 8. Site configuration, flow patterns and approach section location for S.R. 37 over the James River near Mitchell, SD 34

35 downstream of the bridge opening. Data from an ADCP section that cut-off the left floodplain flow accounted for all but 14 m3/s of the difference in discharge between the typical approach section and the section further upstream. The section furthest upstream was used as the uncontracted section because it was most representative of the flow naturally carried by the main channel had the roadway embankment not been present. A similar situation is present at the Minnesota River at Belle Plaine site in which an upstream bend forces much of the overbank flow into the main channel before the section located one bridge width upstream (see Appendix A, Case Study No. 3). A summary of the longitudinal scour location for the contracted bridges surveyed in South Carolina (Benedict, 2003) showed that the location of maximum contraction scour varied appreciably (Figure 9). For the shallow scour depths (1.4 m or less) in the clayey soils of the Piedmont, the low point of the scour hole was typically found along the centerline of the roadway directly under the bridge (Figure 9A). For the Coastal Plain sites, the longitudinal location of the scour hole varied appreciably for 100-year-flow bridge top widths of 91.5 m or less (Figure 9B); however, beyond 91.5 m, the longitudinal location was close to the roadway centerline directly under the bridge. Although data for the deeper scour holes in the Piedmont were not as abundant as data in the Coastal Plain, a similar longitudinal location of the scour holes was observed (Benedict, 2003); however, for relatively short bridge lengths, bridges with 100-year-flow top widths approximately 91.5 m or less, Figure 9 illustrates that the low point of the scour hole may form outside of the bridge limits. In some cases, the low point of the scour hole was largely removed from the bridge so as to not significantly threaten bridge foundations (Figures 10 and 11). Interaction of highly rotational flow coming from the left and right

36 (A) (B) Figure 9. Relation of longitudinal location for the low point of the abutment-scour hole and the 100-year-flow top width at the bridge for (A) shallow and deep scour holes in the Piedmont of South Carolina, and (B) sites in the Coastal Plain of South Carolina. PIEDMONT SITES

37 Figure 10. Example of scour-hole low point located upstream of Road S-299, crossing Cannons Creek in Newberry County, South Carolina, November 24, 1997. Figure 11. Example of scour-hole low point located downstream of S.C. Route 41, crossing Maiden Down Swamp in Marion County, South Carolina, December 3, 1996.

38 abutments likely causes complex, unsteady flow patterns for shorter bridge lengths and creates the scatter within the longitudinal pattern. As bridge length increases beyond approximately 91.5 m, the interaction of flow from left and right abutments is diminished and steadier flow patterns are established around each separate abutment. The steady flow patterns promote the creation of the classical abutment-scour hole pattern where the low point falls longitudinally near the bridge (Benedict, 2003). Although Benedict (2003) and others have discussed factors that contribute to the position of scour holes, no method for predicting the location has been developed. Consequently, the present scour-prediction methods found in HEC-18 recommend that the scour- hole low point be located at the bridge. Additional research and data collection is needed to determine the factors that control scour and to develop a method for predicting the location of scour holes. Discussion of One-Dimensional Backwater Model Applications The discharge in the contracted and uncontracted sections is typically determined by use of a one-dimensional step-backwater computer model such as Water-Surface Profile Computations (WSPRO) (Shearman, 1990) or Hydrologic Engineering Center–River Analysis System (HEC-RAS) (U.S. Army Corps of Engineers, 1998). Stream-tube and one-dimensional models distribute the flow in a cross section based on conveyance alone. The flow distribution upstream of a bridge has no effect on the flow distribution within the bridge opening, because one- dimensional backwater models compute solutions in the upstream direction. The observations in this study show that the flow distribution in a contracted opening is highly dependent upon flow distribution of the approaching flow in addition to roughness and topography of the approach

39 flow reach. Models based on cross-section conveyance distributions alone can be highly inaccurate where non-uniform approach-flow conditions exist at bridges. Experience in applying backwater models in this study indicates that distribution of the flow by conveyance may lead to overestimating the depth of contraction scour. At the uncontracted approach cross section, a conveyance-based model distributes more flow in the floodplain than was measured, resulting in the computed main channel flow being too low. Increased flow in the floodplain may result because (1) the one-dimensional model assumes a constant friction slope for the whole cross section when, in reality, the downstream water-surface slope varies across the section; (2) the one-dimensional model does not account for flow-path lengths between main channel and the floodplains; and (3) the one-dimensional model does not account for the lateral resistance of the flow moving from the main channel to the floodplains. Conversely, the conveyance-based flow distribution may place too much flow in the main channel at the bridge because the conveyance tubes fail to represent the accelerating curvilinear flow separating from abutments and (or) road embankments. This problem could be partially addressed if ineffective areas were used to represent the flow-separation regions in the upstream and downstream ends of the bridge opening. The combination of a reduction of the main channel flow in the uncontracted section coupled with an increase of main channel flow at the bridge section could lead to an overprediction of depth of contraction scour. Analysis of the one- dimensional model outputs for contracted bridge sites of this study show that computed discharge ratios are not representative of field conditions. Too much flow is distributed to the floodplains in the approach and in the main channel through the bridge opening resulting in consistent overprediction of the observed contraction scour (Table 2).

40 ABUTMENT SCOUR The current knowledge on prediction of scour at abutments is derived from regime theory equations, equations used to estimate the depth of scour for spur dikes, and equations developed from small-scale physical-model studies conducted in laboratory flumes. Unfortunately, none of these approaches have resulted in a satisfactory prediction equation. The inability of these approaches to accurately predict scour at abutments is a result of the simplifying assumptions on which the research is based and the complexity of abutment scour in field conditions. The configuration of bridge abutments and associated embankments is complex when placed in the context of river hydraulics. Field Conditions The geometric configuration of the bridge crossing, floodplain, and channel greatly affects the way flow is directed around the abutments. The abutment may be located in the channel, at or near the top bank, or on the floodplain. The configuration of the abutment may be vertical, have wing walls at various angles, or have a spill slope protected with riprap or some other armoring material. Although abutments with spill slopes are usually protected, the armoring can fail or be undermined by scour causing the abutment configuration to change during a flood. The embankments may not be perpendicular to the approach flow but may be angled either upstream or downstream. Drainage ditches along the toe of the embankment are common and complicate the flow patterns around the abutment.

41 The natural flow distribution in a river and its floodplain also can have an appreciable effect on the depth of scour at an abutment. The distribution of the approach flow blocked by the embankment is dependent upon the roughness and topography of the floodplain and alignment of the main channel. The flow distribution and direction can change appreciably during a flood hydrograph. Such complexity and the variability of these conditions between sites are major obstacles in developing a reliable method for predicting scour at abutments. Discussion of Laboratory-Based Equations Although some predictive equations are based on field observations, such as the HIRE (Richardson and others, 1990) equation (based on spur dikes on the Mississippi River), most of the equations for predicting scour at abutments are based on small-scale physical model studies. Literature documenting the laboratory experiments reveals that the approach section of the flume usually had a constant depth with a uniform velocity distribution except for a small boundary layer along the flume walls. The roughness of the channel also was typically uniform throughout the approach and bridge sections and the bed material generally was composed of uniform sand. The abutments were represented by solid, non-erodible obstructions protruding from the sides of the flume with ends that varied in configuration to represent typical shapes of embankments and abutments at contracted bridge openings. These conditions are different from the conditions that occur in the field. Recently (2000), several researchers have attempted to account for some of the conditions commonly found in the field. Dongol (1993), Melville (1995), and Sturm and Janjua

42 (1994) have used models that incorporate the floodplain, some channel-geometry effects, and non-uniform flow distributions. Unfortunately, the amount of field data on abutment scour that can be used to evaluate the validity of the laboratory studies is limited. The laboratory research, although not in agreement, typically has used some combination of the following variables to predict scour at abutments: (1) embankment length, (2) abutment shape, (3) depth of flow, (4) velocity, (5) sediment size, and (6) discharge. The variables in equations developed from these experiments are ambiguous when these equations are applied to field conditions, because of the simplicity of many of the laboratory experiments. In simple flume studies the flow depth is uniform everywhere, and there is no way to define what depth is controlling the depth of scour (the depth at the abutment, the depth in the approach upstream of the abutment, or an average depth of flow blocked by the abutment). In the field these all may be different values; however, in the laboratory with a uniform bed they are all the same. The representative or reference velocity also is a good example of potential ambiguous variables in the field. The local velocity in the contracted opening adjacent to the abutment (which would represent the peak-flow velocity near the region of highest flow curvature) would be different from the unobstructed approach velocity upstream of the abutment or the average velocity of the approach flow blocked by the length of the embankment. Flume studies often use the total length of the embankment from the flume wall to the abutment as the embankment length; however, this approach fails to account for the flow separation and recirculation zone that forms along the upstream edge of the embankment (Figure 12). The effective length of the embankment on the depth of scour is dependent upon the distribution of the approach flow and the floodplain roughness and geometry (Mueller and Wagner, 2002). It is important that laboratory research

43 emulates the conditions in the field so that the equations developed are more representative of field conditions. Laboratory research on abutment scour has focused on equilibrium scour in non-cohesive materials. Floodplain soils typically contain fine-grained sediments sufficient to provide a degree of cohesion that makes them more resistant than non-cohesive soils. Vegetation also can appreciably increase soil resistance to scour. Cohesive sediments are present in many main channel streambeds. S A AS A S DIRECTION OF FLOW CIRCULATIONA - FLOW ATTACHMENT POINT S - FLOW SEPARATION POINT Figure 12. Illustration of flow contracted by an embankment constructed in a floodplain.

44 Research on scour in cohesive materials shows that, in general, cohesion increases the resistance of soils. In addition, the time required for maximum scour-depth conditions in cohesive soils is substantially longer than would occur with non-cohesive soils. Scour equations derived from small-scale laboratory experiments are incapable of providing accurate equations for cohesive soils because of scaling problems associated with fine-grained sediment entrainment. In many cases, cohesive fine-grained soils reinforced by vegetation root mats overlay non-cohesive sands or gravel and sand mixtures. Abutment scour in these complex but common soil conditions has not been addressed in model studies. Summary of Observations The data collected at the 12 sites with abutment scour for the NCHRP Project 24-14 have shown that the magnitude and location of abutment scour are highly site specific and dependent upon the geometry of the bridge crossing and the roughness of the channel and floodplains. The South Carolina data (Benedict 2003) indicate that scour through short contracted bridge openings is more often dominated by scour processes associated with “curvilinear flow” around the abutments rather than “rectilinear flow” associated with HEC-18 defined contraction scour. Although the length of the bridge and channel geometry are important factors in the development of abutment scour, there are a multitude of other parameters that must be considered when evaluating and (or) predicting scour in the abutment region including the following: 1. cohesion of soils; 2. geometric contraction ratio; 3. approach flow-velocity distribution including the effects of channel bends;

45 4. floodplain roughness and topographic variation; 5. floodplain-flow obstructions; 6. valley geometry, including valley width variation and slope; 7. roadway crossing geometry – roadway profile, embankment geometry and orientation, and bridge length; 8. embankment protection; and 9. duration and frequency of flood flows. Appreciable limitations of extrapolating laboratory results to conditions in the field can be illustrated in a comparison of the relation between embankment length and scour depth, normalized by flow depth, for laboratory (Melville, 1992) and field data collected in South Carolina (Figure 13). The field data shown in Figure 13 exhibit a similar trend to the laboratory relation, but there is more scatter within the field data and the asymptotic limit of the field data (3.4) is appreciably lower than the Melville (1992) laboratory data limit (10). A review of the primary scour factors specified by Dongol (1993) in the field data indicates that many of the factors likely have a minimal affect on abutment-scour depths for the prevailing field conditions in the NCHRP Project 24-14 sites and in South Carolina. Dongol (1993) classified the variables affecting abutment scour into the seven categories listed below: (1) Variables describing the channel channel width

46 (A) (B) Figure 13. Relation of observed clear-water abutment-scour depth and the 100-year-flow embankment length, normalized by the 100-year-flow depth near the abutment toe, for the Piedmont and Coastal Plain of South Carolina with (A) a complete horizontal axis, and (B) a truncated horizontal axis.

47 channel slope channel geometry (2) Variables describing the abutment embankment length skew abutment shape (3) Variables describing the flow flow depth mean approach velocity energy slope gravitational acceleration (4) Variables describing the bed material median size specific gravity gradation fall velocity particle shape factor angle of repose cohesiveness dimensionless critical-shear stress particle Reynolds number (5) Variables describing the fluid density

48 dynamic viscosity (6) Temperature (7) Time Inspection of the data compiled for the NCHRP Project 24-14 sites revealed that approach and contracted flow velocity, geomorphic setting (i.e., upstream channel alignment and valley configuration), bed-material cohesion and size, and geometric-contraction ratio are the factors that have the most affect on the measured scour. A review of the data collected in South Carolina revealed that the factors having the most appreciable effect on abutment-scour depth are embankment length, geometric-contraction ratio, flow velocity, and soil cohesion. The envelope curves plotting observed abutment scour versus embankment length for the South Carolina field data enveloped the observed scour at all the selected NCHRP Project 24-14 sites (Figure 14). Comparisons of Observed Scour to HEC-18 Abutment-Scour Predictions Abutment scour at 8 of the 12 NCHRP Project 24-14 sites was calculated using a one- dimensional model and 4 of the abutment-scour prediction equations (Froehlich, modified Froehlich, Sturm and HIRE) documented in HEC-18. The results of the predictive methods were compared to the real-time abutment-scour observations documented at the NCHRP Project 24-14 sites. Of the four NCHRP Project 24-14 sites not included in the comparison, two sites did not have models, one site had flow only in the channel, and one site had scour caused primarily by debris. The following comparisons should be viewed as comparisons of scour-prediction methods rather than simply comparisons of specific scour equations. The methods include the

Figure 14. Comparison of the embankment-length envelope for field observations of abutment-scour depth in South Carolina with the observed abutment scour for selected sites from the National Bridge Scour Database (BSDMS). 49

50 application of one-dimensional hydraulic models and the selection of scour-prediction variables from these models. Errors in the hydraulic models or the selection of scour parameters from hydraulic models can lead to appreciable error in the scour predictions. The results and comparisons of the scour-prediction method using Froehlich equations are shown in Figure 15. Recent (2001) adjustments to the embankment length term in the Froehlich equation have resulted in the modified Froehlich equation. An evaluation of the performance of the scour-prediction method using the Sturm equation, shown in Figure 16, reveals that the Sturm relation both over- and under-predicted the observed abutment scour at the sites in the BSDMS. An equation that excessively over-predicts scour is not a good tool and leads to excessive bridge-construction costs; however, an equation that under-predicts scour is a worse tool for design purposes because it can lead to bridge failures and potential loss of life. The data illustrated in figures 15 and 16 also are presented in Table 3. A scour-prediction method that excessively over-predicts scour is not a good tool and leads to excessive bridge construction costs. The analysis of the Sturm, Froehlich, modified Froehlich and HIRE abutment-scour equations predictions of scour for the NCHRP Project 24-14 sites indicates that all four relations can appreciably over-predict scour when used in combination with standard one-dimensional models for the selection of hydraulic parameters. Variability in Scour Predictions A series of comparisons were developed to illustrate the relative effects of channel geometry, hydraulic parameters derived from one-dimensional models, and selected predictive

51 Figure 15. Comparison of field observations of abutment-scour depth with the theoretical abutment-scour depth computed with the original Froehlich (1989) and modified Froehlich (2001) equations for selected sites from the National Bridge Scour Database (BSDMS). 51

Figure 16. Comparison of field observations of abutment-scour depth with theoretical abutment-scour depth computed with the HEC- 18 (2001) Sturm equation (with and without the safety factor) for selected sites from the National Bridge Scour Database (BSDMS). 52

TABLE 3. Comparison of observed abutment scour with scour calculated with the HEC-18 (2001) Froehlich and Sturm prediction equations using HEC-RAS modeled hydraulics for eight abutment scour sites in the National Bridge Scour Database (BSDMS). (-- not computed, m - meters) Site Name Location of Scour Froehlich Scour Prediction (m) Modified Froehlich Scour Prediction (m) Sturm Scour Prediction (m) Field Observed Scour (m) James River at SR 37 near Mitchell, South Dakota Left Abutment 5.9 -- 9.9 1.2 Minnesota River at SR 25 near Belle Plaine, Minnesota Left Abutment 12.3 -- 12.4 5.5 Minnesota River at SR 25 near Belle Plaine, Minnesota Right Abutment 9.1 -- 5.3 1.2 Pomme De Terre River at US 12 Holloway, Minnesota Left Abutment 4.0 -- 2.1 1.8 Pomme De Terre River at US 12 Holloway, Minnesota Right Abutment 4.6 -- 2.1 3.4 Pomme De Terre River at CR22 Fairfield, Minnesota Left Abutment .88 -- .31 .61 Pomme De Terre River at CR22 Fairfield, Minnesota Right Abutment 3.3 1.65 3.0 Bitteroot River near Belle Crossing, Montana Left Abutment 6.8 5.3 16.7 1.5 Bitteroot River near Belle Crossing, Montana Right Abutment 7.4 6.4 12.4 1.4 Beaver Creek Overflow 7 miles West of Saco, Montana Left Abutment 4.1 3.9 10.4 1.4 Beaver Creek Overflow 7 miles West of Saco, Montana Right Abutment 4.2 4.0 9.6 1.4 Beaver Creek Overflow 9 miles West of Saco, Montana Left Abutment 3.0 2.9 7.2 1.0 Beaver Creek Overflow 9 miles West of Saco, Montana Right Abutment 3.0 2.9 7.5 1.0 Gallatin River at I-90 near Manhattan, Montana Left Abutment 1.2 1.2 0 0 Gallatin River at I-90 near Manhattan, Montana Right Abutment 3.5 1.7 2.4 .91 53

54 equations when calculating scour by methods documented in HEC-18. Comparing the observed to computed depths of scour using field measured hydraulic parameters in the selected predictive equations allows evaluation of the accuracy of the predictive equations. Comparing the observed to computed depths of scour using hydraulic parameters derived from one-dimensional models allows an overall assessment of the accuracy of the methods recommended in HEC-18. A prerequisite to predicting scour from a one-dimensional model simulation is selecting the channel geometry to be modeled. The channel geometry that exists at the time of the flood may be appreciably different from the channel geometry that existed when the survey of the channel for modeling purposes was made. Scour resulting from use of both pre-flood and flood-channel geometry are compared to evaluate the effect of channel geometry on computed depth of scour. Analysis of these comparisons will highlight any deficiencies in the modeling approach and (or) the selected predictive equations. The hydraulic parameters measured in the field were used in the selected prediction equations to directly evaluate the accuracy of the equations. Only two sites had detailed hydraulic measurements both through the bridge opening and in the approach section (Table 4), because of the difficulty of collecting data in the approach section and floodplain during major floods. The application of the HIRE (Richardson and others, 1990) equation showed the problems of using equations developed for simple geometries at bridge sites with complex geometry. The HIRE equation uses the velocity and depth of flow “at the abutment.” Current guidelines in HEC-18 indicate that when using one-dimensional models and the HIRE equation, it is acceptable to use the conveyance tube closest to the abutment for determining the velocity and depth at the abutment. However, the flow velocity in the conveyance tube closest to the

TABLE 4. Comparison of observed abutment scour with scour calculated by use of the HEC-18 (2001) prediction equations using hydraulic parameters measured in the field for two sites in the National Bridge Scour Database (BSDMS). (-- not applicable, L’/y < 25) Field Distributed Hydraulics Scour Calculations Site Name Location of Scour Date (Froehlich) (meters) (HIRE) (meters) (Sturm) (meters) Field Observed Scour (meters) James River at SR 37 Mitchell, SD Left Abutment 4/15/2001 9.7 10.8 16.8 1.2 Minnesota River at SR 25 Belle Plaine, MN Left Abutment 4/17/2001 15.6 20.5 17.8 3.0 Minnesota River at SR 25 Belle Plaine, MN Right Abutment 4/17/2001 6.0 -- 3.0 1.2 55

56 abutment is always low compared to the rest of the flow field around the abutment and does not represent the acceleration of the high curvature flow especially near the point of flow separation (Table 5). The use of the field-measured velocity as defined by HEC-18 for the HIRE equation, results in gross over-prediction of abutment-scour depths. The HIRE equation was derived from scour measured at dikes on the Mississippi River using hydraulic variables measured in the flow approaching the dikes (Richardson and others, 1990); therefore, a more representative field- velocity measurement for use in the HIRE equation should be taken upstream of the abutment tip, near the approach section. The HIRE abutment-scour values found in Table 4 were calculated using velocity values measured at the approach section rather than directly adjacent to the abutment. None of the equations accurately predicted the scour using the measured hydraulic parameters at these sites. Comparing the predicted depths of abutment scour using measured hydraulics and modeled hydraulics provides an evaluation of the adequacy of the one-dimensional modeling approach. Tables 6 and 7 show that the predicted depth of scour generally was larger for measured hydraulics than for modeled hydraulics. The hydraulics for the two sites used in this comparison are dominated by the alignment of the channels upstream of the bridges (Figures 8 and 17). These channel alignments and their effects are not captured in one-dimensional models that only extend one bridge-length upstream and downstream. Although the abutment-scour equations over-predicted the observed scour for both measured and modeled hydraulic parameters, it is possible that the modeled hydraulics could have resulted in underpredictions because of the failure of the one-dimensional model to account for the highly two-dimensional hydraulics occurring at these sites if the equations provided accurate scour predictions.

TABLE 5. Comparison of measured and modeled abutment-tip velocities for use in the HIRE abutment scour prediction equation (m/s, meters per second). Site name Location of scour Date Modeled abutment tip velocity (m/s) Field abutment tip velocity (m/s) James River at SR 37 Mitchell, South Dakota Left Abutment 4/15/2001 0.20 1.8 Minnesota River at SR 25 Belle Plaine, Minnesota Left Abutment 4/17/2001 .35 4.1 Minnesota River at SR 25 Belle Plaine, Minnesota Right Abutment 4/17/2001 .62 1.7 57

TABLE 6. Comparison of observed abutment scour with scour calculated by use of the HEC-18 (2001) Froehlich and HIRE prediction equations using modeled and field hydraulics for two abutment scour sites in the National Bridge Scour Database (BSDMS). (-- not applicable, (L’/y < 25); m, meters) Site name Location of scour Date Modeled calculated scour (Froehlich) (m) Field hydraulics scour calculation (Froehlich) (m) Modeled calculated scour (HIRE) (m) Field hydraulics scour calculation (HIRE) (m) Field observed scour (m) James River at SR 37 near Mitchell, South Dakota Left Abutment 4/15/2001 5.2 9.7 2.6 10.8 1.2 Minnesota River at SR 25 Belle Plaine, Minnesota Left Abutment 4/17/2001 11.6 15.6 10.9 20.5 3.0 Minnesota River at SR 25 Belle Plaine, Minnesota Right Abutment 4/17/2001 9.0 6.0 -- -- 1.2 58

TABLE 7. Comparison of observed abutment scour with scour calculated by use of the HEC-18 (2001) Sturm and Maryland prediction equations using modeled and field hydraulics for two abutment scour sites in the National Bridge Scour Database (BSDMS). (m, meters) Site name Location of scour Date Modeled calculated scour (Sturm) (m) Field hydraulics scour calculation (Sturm) (m) Field observed scour (m) James River at SR 37 near Mitchell, South Dakota Left Abutment 4/15/2001 6.3 16.8 1.2 Minnesota River at SR 25 Belle Plaine, Minnesota Left Abutment 4/17/2001 11.6 17.8 3.0 Minnesota River at SR 25 Belle Plaine, Minnesota Right Abutment 4/17/2001 9.2 3.0 1.2 59

Figure 17. Aerial photograph of State Route 25 over the Minnesota River near Belle Plaine, Minnesota Flow 60

61 The effect of changes in channel geometry on the computed depth of scour is evaluated by comparing depth of scour computed from modeled hydraulic parameters for pre-flood and flood-channel geometry. Because this comparison does not require detailed measured flow data in the approach section and floodplain, additional sites can be used in this evaluation. The pre- flood geometry was approximated through inspection of bridge plans as well as upstream- and downstream-channel bathymetry. The flood geometry was taken directly from the detailed real- time scour measurements on the specified dates. The results of this comparison clearly show that channel geometry has an effect on the computed depth of scour, but the equations were unable to accurately predict the observed depth of scour using either geometry (Tables 8 and 9). The comparisons presented in this section show that although geometry and the one- dimensional modeling approach can cause variations in the depth of predicted scour, the accuracy of the selected equations currently is the largest source of error in abutment-scour predictions. The channel and floodplain geometry both near the bridge and upstream play an important role in the distribution of flow at the bridge. Although these effects are not accounted for in the one-dimensional model, the use of field-measured hydraulics did not improve the accuracy with which the selected equations could predict the observed scour. Therefore, although the one-dimensional approach proposed in HEC-18 has limitations, the data, assumptions, and experiments on which the selected equations are based do not adequately represent complex field conditions and render the equations inaccurate.

TABLE 8. Comparison of observed abutment scour with scour calculated by use of HEC-18 (2001) Froehlich and HIRE prediction equations using flood and pre-flood geometry for four abutment scour sites in the National Bridge Scour Database (BSDMS). (-- not applicable (L’/y < 25); m, meters) Site name Location of scour Date Pre-flood geometry modeled scour (Froehlich) (m) Flood geometry modeled scour (Froehlich) (m) Pre-flood geometry modeled scour (HIRE) (m) Flood geometry modeled scour (HIRE) (m) Field observed scour (m) James River at SR 37 near Mitchell, South Dakota Left Abutment 4/15/2001 5.9 5.2 3.4 2.6 1.2 Minnesota River at SR 25 Belle Plaine, Minnesota Left Abutment 4/17/2001 12.3 11.6 9.5 10.9 5.5 Minnesota River at SR 25 Belle Plaine, Minnesota Right Abutment 4/17/2001 9.1 9.0 -- -- 1.2 Pomme De Terre River at US 12 Holloway, Minnesota Left Abutment 4/9/1997 4.0 3.0 5.2 4.6 1.8 Pomme De Terre River at US 12 Holloway, Minnesota Right Abutment 4/9/1997 4.6 4.3 10.8 4.3 3.4 Pomme De Terre River at CR 22 Fairfield, Minnesota Left Abutment 4/9/1997 .88 1.8 -- -- .61 Pomme De Terre River at CR 22 Fairfield, Minnesota Right Abutment 4/9/1997 3.3 4.4 4.2 1.4 3.0 62

TABLE 9. Comparison of observed abutment scour with scour calculated by use of the HEC-18 (2001) Sturm prediction equation using flood and pre-flood geometry for four abutment scour sites in the National Bridge Scour Database (BSDMS). (m, meters) Site Name Location of scour Date Pre-flood geometry modeled scour (Sturm) (m) Flood geometry modeled scour (Sturm) (m) Field observed scour (m) James River at SR 37 near Mitchell, South Dakota Left Abutment 4/15/2001 9.9 6.3 1.2 Minnesota River at SR 25 Belle Plaine, Minnesota Left Abutment 4/17/2001 12.6 11.8 5.5 Minnesota River at SR 25 Belle Plaine, Minnesota Right Abutment 4/17/2001 5.3 9.2 1.2 Pomme De Terre River at US 12 Holloway, Minnesota Left Abutment 4/9/1997 2.1 1.6 1.8 Pomme De Terre River at US 12 Holloway, Minnesota Right Abutment 4/9/1997 2.0 2.4 3.4 Pomme De Terre River at CR 22 Fairfield, Minnesota Left Abutment 4/9/1997 .33 1.1 .61 Pomme De Terre River at CR 22 Fairfield, Minnesota Right Abutment 4/9/1997 1.7 3.4 3.0 63

64 SCOUR WITH DEBRIS In general, very little research has been done regarding bridge scour that is directly associated with woody-debris accumulation despite the fact that debris accumulations have contributed to one-third of all bridge failures in the United States (Chang, 1973). Woody-debris accumulations (also referred to as a debris raft) affect the scour depth and patterns at bridges in several ways: (a) they increase the effective size of piers and (or) abutments, which may lead to deeper local scour; (b) if they are large enough, they can contract the flow through the bridge sufficient enough to induce contraction scour; and (c) they can deflect flow into a neighboring pier or abutment causing deeper scour at nearby foundations. Dongol (1989) experimentally investigated the effects that debris rafting had on scour depths at bridge piers and developed a procedure to estimate the equilibrium local scour depth associated with debris accumulation. The study provided a method for determining the effective diameter to be used in computing local scour depth for piers with floating-debris accumulation. Single cylindrical piers are considered the least likely to accumulate large debris rafts relative to other pier shapes and configurations. Piers consisting of multiple piles can be especially susceptible to appreciable debris accumulations because the free space between columns is not typically wide enough to pass floating debris and provides an excellent place for debris to lodge. Predicting the probability and size of debris rafts is an issue of great importance for those responsible for the maintenance of bridges; however, little work has been done on this subject.

65 Current design guidelines treat debris rafts as a detriment to bridges, but do not provide methods for estimating the likelihood and size of debris accumulation. The majority of published information regarding debris scour is subjective and qualitative; although this information is useful, it is difficult to apply in bridge design. A report prepared for the Federal Highway Administration (Diehl and Bryan, 1997) attempted to provide more quantitative criteria on the likelihood and size of potential debris accumulation at bridges by reviewing published literature on drift, analyzing data from 2,577 reported drift accumulations, and conducting field investigations of 144 debris accumulations. Summary of Observations The current NCHRP Project 24-14 data set only contains one real-time measurement of scour appreciably affected by a debris accumulation (State Route 129 over the Chariton River near Prairie Hill, Missouri; see Appendix A, Case Study No. 10). A review of flood- measurement notes indicates that this site does not experience substantial scour of any form when there is no debris accumulation; however, for floods where a debris accumulation forms on the central pier, the streambed elevations drop by as much as 6.1 m in what appears to be a combination of contraction scour (caused by the reduced flow area as a results of the debris accumulation) and local scour effects caused by the debris and pier. The procedure for estimating scour at piers with a debris accumulation (Dongol, 1989 and Melville and Dongol, 1992) combined with contraction scour estimates using hydraulic parameters from WSPRO was compared to a series of five separate scour measurements at this site.

66 The local scour associated with the debris accumulation for each of the five measured floods was calculated using Melville and Dongol (1992) wherein the effect of a debris accumulation is converted to an effective pier diameter based on the thickness and diameter of the accumulation. The effective diameter was then substituted for the diameter term in the HEC- 18 equations to predict the total scour. The total scour computed using the Melville and Dongol approach and the total scour observed in the field compared very closely (Table 10). Although data from one site is an insufficient basis for an overall validation of the proposed technique, it indicates that this technique has promise and that comparisons with additional field data would be useful. The design debris-accumulation width criteria developed by Diehl and Bryan (1997) is based on the width of the approach channel. The channel width at this site was unchanged during the period of data collection so only one design debris-accumulation width was computed. The comparison of the design width to observed widths of the debris accumulation show close agreement (Table 11). As with the approach for determining scour depth, one site is insufficient for an overall validation of the proposed technique, but it indicates that this technique has promise and that comparisons with additional field data would be useful. APPLYING NUMERICAL MODELS FOR SCOUR ANALYSIS The use of one-dimensional models is currently (2004) the standard method to estimate the bridge hydraulics for scour computations by State and Federal highway agencies; however,

TABLE 10. Comparison of Melville’s debris-scour-estimating procedure with HEC-18 procedures and observed debris scour at S.R. 129 over the Chariton River near Prairie Hill, Missouri. (m, meters) Pier scour Contraction scour Total scour Date Melville (m) HEC-18 (m) Melville/HEC-18 (m) Melville (m) HEC-18 (m) Observed (m) 3/29/1960 4.7 2.9 0.37 / 0.21 5.0 3.1 5.2 4/22/1973 5.2 2.6 0 / 0 5.2 2.6 5.2 5/8/1978 5.9 3.0 0 / 0 5.9 3.0 6.1 7/8/1993 6.4 3.1 .12 / 0 6.6 3.1 6.1 5/24/1995 3.9 2.9 0 / 0 3.9 2.9 3.6 67

TABLE 11. Comparison of debris width-design-criteria relationship (Diehl and Bryan, 1997) and measured debris raft diameters for S.R. 129 over the Chariton River near Prairie Hill, Missouri. Measurement Date Approach Channel Width (m) Design Debris Width (m) Measured Debris Width (m) 3/29/1960 74.7 24.0 13.4 4/22/1973 74.7 24.0 21.3 5/8/1978 74.7 24.0 21.3 7/8/1993 74.7 24.0 23.2 5/24/1995 74.7 24.0 9.1 68

69 flood flow contracting through bridge openings is an inherently two-dimensional and many times three-dimensional hydrodynamic situation. One of the most important factors in using numerical models at contracted bridges is the ability for the model to accurately represent the velocity distribution laterally across the stream and floodplain. Summary of Observations One-dimensional hydraulic models were developed for most of the sites included in the BSDMS. The hydraulic parameters estimated by the models were used to predict scour depths using the HEC-18 methods and to build comparisons with scour measurements. A two- dimensional hydrodynamic and sediment-transport model also was developed for a site with real- time detailed scour data. The results of the two-dimensional simulation were compared to the one-dimensional results and the field measurements. One-Dimensional Numerical Models One-dimensionally modeled velocity distributions for two contracted sites over the Pomme De Terre River in Minnesota are compared with real-time field measurements of velocity made during the 1997 flooding. The velocity-distribution comparisons at both sites were made for channel geometry measured on 4/5/97, which was on the rising limb of the flow and for channel geometry measured on 4/9/97, which was just after the peak. The velocity distribution at U.S. Route 12 on 4/5/97 indicates that the flow in the field was skewed toward the right abutment (Figure 18A).

70 HEC-RAS did not duplicate this skewed flow pattern but rather computed a uniform flow distribution across the cross-section caused by the model assigning flow tubes of equal conveyance through the geometrically uniform bridge section. HEC-RAS was more accurate in reproducing the observed velocity distribution for the scoured channel geometry (Figure 18B). The one-dimensional model is not able to reproduce the region of reverse flow that occurred adjacent to the left abutment. HEC-RAS computed velocities are greater near the deeply scoured region adjacent to the right abutment because the slope and roughness are constant across the cross section, so the conveyance becomes dependent upon the depth of flow. The one- dimensional model results did not compare well with the 4/5/97 observations at County Route 22 (Figure 19A) because of its inability to replicate the two-dimensional features of the measured flow field. Although the model estimated the peak velocity near the right-most pier reasonably well, the modeled velocities were too high near the right bank and in the center of the main channel and too low along the left bank. The model once again did a better job redistributing the flow after the scour had fully developed (Figure 19B). Although the current amount of field data in the approach sections of the surveyed bridges were inadequate to provide a comprehensive evaluation of the ability of a one- dimensional model to represent complex two-dimensional flow fields, the comparisons that could be made showed the limitations of the one-dimensional modeling approach. Where conveyance dominates the hydrodynamics, such as for fully developed scour-hole conditions, a one-dimensional model is able to provide a reasonable estimate of the velocity distribution; however, where two and three-dimensional effects caused by flow accelerations dominate the

71 (A) (B) Figure 18. Comparison of observed- and model-velocity distributions at U.S. Route 12 over the Pomme de Terre River, Minnesota, for (A) April 5, 1997 and (B) April 9, 1997.

72 (A) (B) Figure 19. Comparison of observed- and model-velocity distributions at County Route 22 over the Pomme de Terre River, Minnesota, for (A) April 5, 1997 and (B) April 9, 1997.

73 flow field, such as at the beginning of a flood and during the scouring process, the one- dimensional model is severely limited in its ability to accurately distribute the flow. An analysis of scour computations in HEC-RAS revealed that the approach channel alignment is not accounted for in calculations of abutment scour. Default HEC-RAS hydraulic parameters used for abutment-scour calculations can provide erroneous predictions based on incorrect projection of the bridge opening to the approach section. The HEC-RAS default scour parameters for the two Pomme De Terre river sites and the State Route 25 Bridge over the Minnesota River had to be adjusted. Without this adjustment the parameters such as the blocked discharge (Qe), area of flow blocked (Ae) and average depth of blocked flow in the approach (Ya) were inaccurately estimated because of upstream channel bends that were not considered in the HEC-RAS algorithms for determination of scour variables. Two-Dimensional Numerical Models A two-dimensional hydrodynamic model (Resource Management Associates – 2 (RMA- 2)) also was developed for the County Route 22 site over the Pomme De Terre River near Fairfield, Minnesota in order to evaluate the flow distribution relative to the one-dimensional model and field measurements. The site has a large bend directly upstream of the bridge, which has a large affect on the flow distribution and scour processes at the bridge (Figure 20). Two separate measurements were made at County Route (C.R.) 22 during an appreciable flood event. The measurements produced velocity magnitudes and distributions primarily because of the

74 Figure 20. Plan view of topography and channel alignment for County Route 22 over the Pomme De Terre River near Fairfield, Minnesota (elevation referenced in feet above seal level; 1 ft = 3.2808 m). Flow

75 formation of a large scour hole at the left abutment during the time between the measurements. A large standing wave and area of reverse flow was witnessed during the 4/5/97 measurement (Figure 21) because of the interaction of the main channel and floodplain flow contracting through the bridge opening. The modeled flow field for the 4/5/97 conditions is illustrated in Figure 22 and is representative of the field observations. The reverse flow and standing wave were absent during the 4/9/97 survey and modeled flow field. The calibration of the two-dimensional model (Figure 23) to the field measurements was limited because of flow conditions through the bridge opening being inherently three dimensional, especially around the right abutment. On 4/9/97, flow was shifted from the left to the right abutment in the two-dimensional model relative to the field measurements (Figure 24). Detailed data were unable to be collected at the site during the flood because of heavy vegetation on the floodplain and near pressure-flow conditions at the bridge. Bathymetry data throughout the model reach was not collected until October 2001, over 4 years after the measured flood. Uncertainty in the geometry upstream of the bridge also is a likely a large contributor in the difference between the field measurements and the two-dimensional model results. A two-dimensional sediment-transport model of C.R. 22 site also was developed utilizing the calibrated hydrodynamics from 4/5/97 and 4/9/97. The sediment-transport model was run for the period between measurements (4/5-4/9/97) at the site to evaluate its ability to replicate the observed scour through the bridge opening. To replicate the conditions that produced the observed scour, the discharge measured on 4/5/97 was run for 48 hours followed by 48 hours of the 4/9/97 discharge. Although steady-flow conditions were obviously not what occurred in the

76 field, the lack of data between 4/5/97 and 4/9/97 prevented a more accurate representation of the hydraulics. The bed elevations of the sediment-transport model relative to scour measurements made on 4/9/97 are illustrated in the difference map found in Figure 25. The model predicted more scour at the left abutment and less scour at the right abutment than were measured in the field (red contours indicate the model underpredicted scour, blue contours indicate the model overpredicted scour). The errors in the modeled-scour patterns can be directly associated with the differences in the modeled- and measured-velocity distributions. The model conveyed too much flow near the left abutment and not enough flow near the right abutment. Figure 21. Sketch of the hydrodynamics observed during bridge scour measurements at County Route 22 over the Pomme de Terre River on April 5, 1997. AREA OF MIXINGSTANDING WAVE BOILS AND REVERSE FLOW Left BankRight Bank

Figure 22. Modeled flow field for County Route 22 over the Pomme de Terre River for conditions on April 5, 1997. 77

78 Figure 23. Computational mesh for the two-dimensional model of County Route 22 over the Pomme de Terre River. Flow C.R. 22

79 (A) (B) Figure 24. Comparison of the velocity distribution for the two-dimensional model and field measurements at the upstream bridge face of County Route 22 over the Pomme de Terre River on (A) 4/5/1997 and (B) 4/9/1997.

80 Discussion of One-Dimensional and Two-Dimensional Model Comparisons Comparison of the output from the one- and two-dimensional models yielded surprising results. Despite the calibration complexities induced by geometry uncertainty and three- dimensional flow, the two-dimensional model was able to reproduce the hydraulics in the bridge opening for the conditions measured on 4/5/97 more accurately than the one-dimensional model (Figure 26). However, for the fully developed scour-hole condition on 4/9/97, the one- dimensional model provides a slightly better representation of the velocity distribution than does the two-dimensional model. A comparison of the HEC-18 scour estimates using the one- dimensional and two-dimensional models relative to the observed scour depths was complicated by the automatic approach section selected by HEC-RAS and the section selected from the two- dimensional model based on the modelers evaluation of the flow lines (Figure 27). HEC-18 scour computations also were computed from the two-dimensional model hydraulics with the HEC-RAS selected and the manually selected approach cross-section and compared to scour depths calculated from the one-dimensional model, sediment-transport model, and those observed in the field (Table 12). The contraction scour and right-abutment scour estimated using the two-dimensional model hydraulics improved by adjusting the location of the approach cross section, however there was an increase in the difference between Froehlich’s prediction and the observed depth of scour at the left abutment. The comparison shows that the one-dimensional model more accurately estimated the observed scour than did any of the two- dimensional model results. The equations in HEC-18 were developed and mostly based on data collected in a laboratory setting, which does not replicate the complex, site-specific hydraulic

Figure 25. Difference in bed elevation (in meters) between two-dimensional sediment-transport model output and field data collected during flood conditions on the Pomme de Terre River at County Route 22, April 4-9, 1997. Flow 81

82 (A) (B) Figure 26. Comparison of the velocity distribution for the two-dimensional model, one- dimensional models and field measurements at the upstream bridge face of County Route 22 over the Pomme de Terre River on (A) 4/5/1997 and (B) 4/9/1997.

Figure 27. Comparison of the original one-dimensional and two-dimensional model approach section location with an approach section in a location more representative of the actual blocked and main channel hydraulics at County Route 22 over the Pomme de Terre River. (Velocity vectors are overlain to illustrate flow patterns through the bridge contraction for conditions on April 9, 1997). 83

TABLE 12. Comparison of HEC-18 scour estimates (Froehlich, HIRE, and Live-bed equations) from the one-dimensional and two- dimensional models (original and adjusted approach section locations) relative to the sediment transport model results and observed scour at County Route 22 over the Pomme de Terre River on April 9, 1997. (-- not applicable; m, meters) One-dimensional model results Two-dimensional model results (original approach) Two-dimensional model results (adjusted approach) Sediment transport model Scour Type/Location Froehlich (m) HIRE (m) Live- Bed (m) Froehlich (m) HIRE (m) Live- Bed (m) Froehlich (m) HIRE (m) Live- Bed (m) Sed2D Results (m) Observed Scour (m) Left Abutment 0.95 -- -- 6.8 -- -- 2.9 -- -- 5.5 0.61 Right Abutment 3.3 4.1 -- 7.2 4.5 -- 9.1 4.5 -- 4.5 3.0 Contraction -- -- 0.40 -- -- 8.7 -- -- 2.5 0.94 0 84

85 processes that are present in the field. Field hydraulic measurements and two-dimensional models are more representative of the processes that induce scour than one-dimensional models; however, the scour depths at the selected field sites were more accurately estimated when using HEC-18 equations and the output from the one-dimensional models, which simulate hydraulic conditions similar to those of laboratory experiments. ERODIBILITY AND GEOTECHNICAL PROPERTIES OF MATERIALS Several factors contributed to the resistance to erosion and slope failure that was observed in the soils and sediments examined in this study. Unlike coarse-grained soils for which erodibility is primarily a function of grain-size distribution and secondarily a function of grain shape and packing, erodibility of fine-grained soil is dominated by other factors such as apparent or true cohesion, porewater pressure, and root reinforcement. True cohesion may occur because of cementation and (or) attractive forces developed in soils rich in very fine particles of clay minerals. Apparent cohesion is caused by development of negative pore-water pressure changes in all soils but is most important in fine-grained clay and silt soils. Roots provide reinforcing effects and growth of vegetation lowers pore-water pressures. All three of these factors affected the location and dimensions of scour holes observed in this study. These effects are not observed in laboratory flumes where cohesionless sands are tested without any vegetation present. The increases in resistance to both erosion and mass movement attributed to vegetation and soil strength appear to control many important aspects of scour. Despite the variability of fine- grained soil characteristics from site to site in floodplains (Benedict 2003), the increase in resistance of vegetated fine-grained floodplain soils over the resistance of coarse-grained

86 cohesionless soils has a profound affect on the initiation of scour, the development of scour holes, and the final geometry of scour holes near bridges. Generally, at the sites examined in this study, the characteristics of sediments composing the streambed were different from those composing the upper layers of the floodplain and streambanks. An increase in soil resistance caused by vegetation and fine-grained soil effects appeared to affect the location, depth of scour and scour pattern. Most frequently, the combined effects of vegetation and fine-grained soil inhibited the initiation of scour holes on the surface of floodplains where erosion was anticipated; e.g., immediately adjacent to the sides of abutments. However, sections of floodplain adjacent to the main channel appeared to collapse into scour holes extending laterally from the main channel under resistant surface layers. Our observations indicate that scour was frequently initiated in (1) soils unprotected by vegetation under the shadow of the bridge, and along streambanks where coarse, cohesionless, and unvegetated soils could be eroded from the bank toe; or (2) at flow separation points at piers, at abutment walls, or near riprap edges. Scour appear to be initiated at (1) Non-vegetated and unprotected areas in shadows under bridges; (2) Stream banks in the flow field of bridge abutments; (3) Around piers, especially those located in the high-velocity flow near bridge abutments; and (4) Locations where bends cause high-velocity zones in the bridge opening.

87 Scour propagates away from the initial erosion point by (1) Undermining of vegetated soils and fine-grained soils strengthened by cohesion effects; (2) Mass failure of scour protection because of undermining; and (3) Changes in flow patterns as scour holes developed. Although main-channel sediments typically are treated as cohesionless soils, fine-grained sediments in the silt-size range or smaller were obtained from the channel bed in several of the river channels in this study (see Appendix A, Case Studies No. 1, 2, and 4). The sediments in three of the rivers examined in this study were derived from fluvial-glacial processes including reworked deposits from lakes formed by glacial moraines. Borings from some of these sites indicated the presence of silt layers beneath the streambed surface. Samples from all other sites indicated only small amounts silt- and clay-size particles on the surface of the streambed. The effects of soil cohesion, both real and apparent, and of vegetation on soil erosion were observed at all sites examined. Scour patterns on floodplains and the shape of scour holes appear to have been affected by the increased soil strength and erosion resistance afforded by soils containing fine-grained particles and covered by vegetation. Increased resistance of soils to erosion and mass movement derived from undrained strength of dried or preconsolidated fine- grained soil (silt and clay) is termed soil cohesion here.

88 Examination of soils on floodplain surfaces, in scour holes, and in streambanks showed the affect of several factors that contributed to resistance to scour. Fine-grained soils such as silts and clays were observed near the surface of floodplain alluvium at all sites. Both woody and (or) herbaceous vegetation covered all floodplains. The soils encountered at the sites visited during this research in all cases were layered, with appreciable variations in grain size and texture from layer to layer. At most of the sites, much of the soil profile consisted of layers containing fine-grained silts and clays. Also, at most sites, the fine-grained layers had been subjected to preloading from desiccation effects (soil drying) and (or) overburden pressures from pre-existing strata. Such preconsolidation effects caused those fine-grained soils to have medium to high undrained shear strengths (to be firm to very stiff). The time for pore-water pressure equilibration during shearing (during a scour episode) for these fine-grained soil layers far exceeded the time during which shear would have occurred, and in most cases would have exceeded the time during which a flood event would occur. Empirical evidence for the appreciable shear strength of the fine-grained soil layers at these sites was the exposure of nearly vertical faces in fine-grained layers whereas adjacent coarse-grained layers typically displayed gently sloping faces. In the long-term situation, drained strengths in these same layers would be much lower than the undrained strengths because of the low values of effective confining pressure at the shallow soil depths in the zone where scour typically occurred. Vegetation also contributed to the scour resistance at these sites by providing tensile reinforcement (roots) at shallow depths (0 to 1 m) and by developing soil suctions to contribute to both drained and undrained shear strengths.

89 The upstream edge of a scour hole formed during Hurricane Floyd at the U.S. 70 Bridge over Bear Creek in North Carolina (Appendix A, Case Study No. 6) is illustrated in Figure 28. The scour hole appeared to be initiated in the main channel and propagated across the stream and upstream. The fine-grained surface soils reinforced by the roots of woody vegetation did not appear to have been scoured; however, the toe of the bank appeared to have been scoured and the undermined soil block containing the fine-grained surface material failed (toppled) into the scour hole. The orientations and positions of trees in Figure 28 indicate the mechanism of mass failure. Complex flow patterns caused by the rapid vertical expansion of flow over the vegetated floodplain surface at the upstream end of the scour hole and into the scour hole may be responsible for erosion of the more vulnerable basal soils that typically containing higher components of coarse-grained sediments. Erosion of the basal soils at the edge of the scour hole caused mass instability of the upper layers eventually leading to their collapse and the upstream propagation of the scour hole. The processes governing scour-hole formation are dependant on the vertical variation in soil characteristics and are not represented in laboratory experiments in which uniform and non- cohesive sediments are used to model floodplain soils. The scour processes involved in the scour of soils with non-uniform characteristics also may affect the distribution of scour including

Figure 28. Looking upstream from the east bound bridge deck of higway 70 over Bear Creek near Lagrange, North Caroina during low-flow. 90

91 the location and depth of the maximum point of scour and, more importantly, the scour around pier and abutment foundations. The upstream and lateral extension of the scour holes is one mechanism by which the non-uniform soil characteristics of floodplain sediments affect scour- hole formation processes; other effects of the variation of soil characteristics also probably exist.