U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-RD-03-052
Date: May 2005
Field Observations and Evaluations of Streambed Scour At Bridges
ANALYSIS OF REAL-TIME DATA
During record flooding in the Minnesota River Basin in April 1997, USGS, in cooperation with FHWA, deployed the USGS bridge scour data collection team to collect real-time scour (contraction and local) measurements at contracted bridge openings. An analysis of two sites that were surveyed during the April 1997 flooding is presented. Both contracted bridges span the Pomme de Terre River, where an estimated 200-year discharge was measured at the USGS Appleton gauging station (05294000) located approximately 19 km downstream of the U.S. Route 12 bridge. The compiled field data (channel and floodplain bathymetry, water discharge, water-surface elevations, roughness, and bridge geometry) were used to calibrate a step-backwater model at each site. The hydraulics and predicted depth of scour based on the calibrated model were compared with the field measurements.
U.S. Route 12 over the Pomme de Terre River
U.S. Route 12 crosses the Pomme de Terre River about 20 km west of Danvers, MN. The single-span steel-truss structure was constructed in 1933 with a maximum span length of 26.9 m. The bridge has vertical wall abutments with wing walls; each abutment and wing wall rests on concrete footings supported on timber piling. Neither abutment was riprapped, nor were there any other scour-protection measures. A field investigation conducted in 1995 by a local engineering firm, BRW, Inc., revealed no evidence of significant scour at the abutment face.(127)
The upstream floodplain consists of a mixture of open agricultural land with scattered trees and brush, with a park on the upstream left bank. The area downstream of the bridge is more heavily wooded and is classified on the USGS topographic map as a wetland area. The streambed material near the bridge generally consists of fine-grained, organic, silty sand with some gravel. A sieve analysis of a surficial bed material sample indicated a median diameter of 0.15 mm. Based on the soil borings and blow counts documented in the bridge plans, the bed material appears to become harder and denser as depth increases. Because samples were not collected and analyzed, it is difficult to ascertain the makeup of the soils at depths below the surface.
During the April 1997 flood the bridge experienced both contraction and abutment scour. A large scour hole developed at the right abutment, scouring below the abutment cutoff wall and resulting in failure of the fill material behind the abutment. Slumping of the embankment slope and some deformation of the approach highway were observed. Although scour measurements showed a scour hole 2 m below the footing of the left abutment, no deformation was observed near the left abutment. These conditions resulted in closure of the bridge. Because of the age and scheduled replacement of the bridge, it was not repaired but was replaced with a new structure after the flood.
Discussion of Field Data
Data were collected during the flood (on April 5 and 9, 1997) at U.S. Route 12 over the Pomme de Terre River. A crewed boat was deployed during the initial visit on April , 1997. The use of the crewed boat and an ADCP allowed bathymetry and three-dimensional velocities to be measured at the bridge and in the approach and exit sections extending about 100 m upstream and 70 m downstream. Heavy vegetation and submerged obstructions in the floodplains limited data collection to the main channel. Measurements on April 9 were limited to data collected from the bridge deck. Channel bathymetry was measured along the upstream and downstream faces of the bridge and at selected locations beneath the bridge using an echo sounder deployed on a knee-board. Velocity magnitudes and water discharge were measured using a vertical axis current meter. Water-surface elevations were measured by taping down from the top-of-curb on the bridge both upstream and downstream, near the left abutment. On April 5, the water-surface elevation was 310.70 m above National Geodetic Vertical Datum (NGVD) of 1929 at the upstream edge of the bridge, and the total discharge was 141.6 m3/s. By April 9, the water-surface had risen to an elevation of 311.5 m above NGVD of 1929, and the discharge had increased to 162.8 m3/s.
The direction of flow through the bridge was controlled by the configuration of the upstream floodplains. The channel upstream of the bridge was straight, but the left floodplain was much wider and carried considerably more flow than the right floodplain. The flow from the left floodplain skewed the flow through the bridge opening to about 50° on average. Figure 33 is a sketch of spot elevations and the flow direction on April 9, which shows the severe skew of the flow to the bridge opening.
The appropriate reference surface was determined from an analysis of cross sections collected by BRW, Inc., on June 5, 1995 and by the USGS during the flood on April 5, 1997. On these two dates, cross sections collected approximately 90 m upstream from the bridge showed only about 0.15 m difference in the channel-bottom elevation. The flood cross section was the lower of the two. Downstream from the bridge, the cross section surveyed on June 5, 1995 (approximately 23 m downstream) had less than 0.3 m in variation when compared with the cross section surveyed on April 5, 1997 (approximately 61 m downstream), but was 0.5 m higher than the cross section surveyed on April 5, 1997 (approximately 30 m downstream). It is possible that the April 5, 1997, cross section could have been affected by the scour at the bridge section; thus, it was not considered in setting the reference surface. The WSPRO bridge section surveyed by BRW, Inc., on June 5, 1995, showed from 0.3 to 0.6 m of abutment scour in the cross section; however, the center of the channel at the bridge appeared to be representative of consistent channel slope from the upstream section to the downstream section. Because little general scour was observed at the upstream and downstream sections, the mean elevation of the unscoured portion of the WSPRO bridge section (elevation 307.9 m) was used as the contraction scour reference surface.
A summary of the contraction scour data is shown in table 16. The contracted section on April 5, 1997, was measured under the bridge from data collected by an ADCP. The maximum erosion of the streambed was 2.3 m from the defined reference surface; however, when the entire streambed below the bridge was averaged the depth of contraction scour was only 0.9 m. The hydraulic data presented for April 5, 1997, were collected with the ADCP. The ADCP data had a significant amount of invalid data that were estimated in final processing. Channel banks were not clearly delineated in the approach section, creating a degree of uncertainty in the approach discharge. Overall, it is suspected that the approach discharge is ±20 percent, and the total discharge is ±10 percent. Measurements made with a sounding weight on April 9, 1997, were collected during the discharge measurement along the upstream face of the bridge, and no approach data are available. An echo sounder mounted on a knee-board was also used to make measurements on April 9, 1997. The board was floated from upstream to downstream under the bridge; the measurements reflect the depths at the upstream or downstream face of the bridge.
The reference surface used to determine the depth of abutment scour was the concurrent ambient bed; therefore, the depth of abutment scour reported is additional local scour below the depth of contraction scour (table 17). The data collected on April 5, 1997, were collected with an ADCP using a weighted average of all four beams as the measured depth. Because a weighted average was used, it is possible that the local abutment scour was not accurately measured, and no values are reported. The cross sections measured on April 9, 1997, all showed a similar pattern with abutment scour holes on each side and a sharp mound in between the scour holes but skewed toward the left bank (figure 34). It appears that the abutment scour holes may have overlapped. The highest elevation in the center of the cross section was subtracted from the reference surface to obtain the depth of contraction scour. The abutment scour was reported as the depth below the highest elevation in the center of the cross section. All velocities presented in table 17 were from the discharge measurement made along the upstream side of the bridge. Although no abutment scour was observed on April 5, 1997, the velocities at the abutments were much higher (left = 1.6 m/s, and right = 1.8 m/s).
The HEC-RAS model, a one-dimensional step-backwater model, was calibrated to represent the field hydraulics as accurately as possible. The bathymetry from the April, 1997, flood was used to build the calibration models for the two sets of data (April 5 and 9, 1997). Because bathymetry data on April 9, were limited to the upstream and downstream edges of the bridge, the cross sections collected on April 5, were used to build the HEC-RAS model for April 9. The majority of the floodplain bathymetry utilized in building the models was taken from a full valley section found in the original bridge plans and adjusted to be consistent with topographic maps.
The water-surface elevation observed in the field rose 0.76 m between April 5 and 9. The model only showed a 0.3 m change and was unable to reproduce the observed change without unreasonable changes to the model input. This large hydraulic variation may be attributed to the U.S. Route 12 bridge reach being under a backwater condition because of some unidentified downstream condition. Large ice drifts were observed during both site visits, indicating the potential for the formation of a debris and (or) ice dam downstream of the data collection area. Analysis of the Appleton gauging station records was of little assistance because the gage was washed out on April 6, 1997, by the failure of a small upstream dam. The water-surface elevation at the upstream side of County Route 22 located about 10 km upstream changed only 0.2 m over the same period; therefore, the model was considered adjusted despite the apparent discrepancy with the water-surface elevation observed on April 9.
One of the most important factors in using one-dimensional models at contracted bridges is the ability for the model to accurately represent the velocity distribution laterally across the stream and floodplain. Figures 35 and 36 depict how the velocity distribution varied between the model
and field measurements, using the geometry from April 5 and 9, 1997. The distribution shown in figure 35 reveals that the flow in the field was indeed skewed toward the right abutment. HEC-RAS did not duplicate this skewed flow pattern, but rather computed a relatively uniform flow distribution across the cross section because the model assigned flow tubes of equal conveyance through the geometrically uniform bridge section. For the scoured channel bathymetry, HEC-RAS did a better job of reproducing the observed velocity distribution (figure 36), although the model does not recognize the region of reverse flow that occurred adjacent to the left abutment. The 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.
Assessment of Scour ComputationsThe calibrated model was used to assess how accurately the scour for this flood could have been predicted. The bathymetry in the calibrated model was replaced with the original bathymetry extracted from the BRW, Inc., WSPRO model, which represented the pre-flood condition. The discharges from both April 5 and 9, were then run through the HEC-RAS model with the original bathymetry to determine the hydraulic parameters required to compute scour at the bridge. The contraction scour was computed in HEC-RAS by allowing the model to use the default equation (live bed or clear water) depending upon the hydraulic conditions. Table 18 compares the observed contraction scour to that computed by the model.
The computed depth of contraction scour was less than the observed value for all measurements. The contraction scour observed on April 9, 1997, may not be typical live bed contraction scour because depth of contraction scour could be affected by overlapping abutment scour holes. The abutment scour was computed in HEC-RAS by both the Froehlich equation and the Hydraulic in the River Environment (HIRE) equation, which are the two equations recommended in HEC-18. The HIRE equation is only applicable (but not required) if the embankment length-to-flow depth ratio at the abutment is greater than 25. In this case, the embankment to flow depth (L/a) ratio is 33. Table 19 compares the observed abutment scour to that computed by the model.
The data summarized in table 19 show the overprediction of scour that is common for abutment scour computations. Although the abutment scour equations overpredicted the local scour and the contraction scour equation underpredicted the contraction scour (table 18), when added together they predicted the total scour with reasonable accuracy and actually underpredicted the scour observed at the upstream edge of the bridge on April 9, 1997. These are somewhat surprising results that should be viewed with caution because the skew of the flow through the bridge could not be accounted for in the one-dimensional model, and the individual components were both in error. The agreement, therefore, may be somewhat coincidental.
Swift County Route 22 over the Pomme de Terre River
Swift County Route 22 crosses the Pomme de Terre River near Artichoke Lake, MN, and is located 10 km upstream from the U.S. Route 12 bridge. This bridge has two piers in the main channel with the abutments set at the edge of the main channel. The spill-through slopes at the abutments were protected by riprap and formed the banks of the main channel. The bridge is located in a very sinuous reach of the river with two large meanders immediately upstream and downstream of the bridge (figure 37). The floodplains are composed of farmland and forest.
During the flooding in April 1997, the USGS visited this site three times. During all three visits, the floodplain flow was concentrated in the right floodplain. This concentration of flow in the right floodplain is likely caused by the channel alignment upstream of the bridge. No defined point of reattachment along the right embankment was found during the flood. Flow was toward the main channel along the entire length of the right embankment. The flow separated from the right embankment, nearly perpendicular to the main channel flow, and joined the main flow just left of the rightmost pier (figure 38). During the visit on April 5, the flow from the right floodplain was so intense that a standing wave formed upstream of the bridge where the floodplain and main channel flow began mixing. The area from the rightmost pier to the right abutment was primarily slack and reverse flow. The depth of flow at the right abutment progressively deepened from 4.5 m on April 4, to 6 m on April 9. On April 9, a portion of the right embankment slumped, forcing Swift County officials to temporarily close the bridge until riprap was placed to protect the bridge.
Discussion of Field Data
Data collection efforts were restricted to data that could be collected from the bridge deck for all three site visits during the flood (April 4, 5, and 9, 1997). All bathymetry data were collected by floating an echo sounder attached to a knee-board across the river; the sounder was controlled by a hand line from the bridge. The board was allowed to float downstream; streambed elevations were collected as far as 30 m downstream from the bridge. Data collected upstream of the bridge were restricted to the upstream edge of the bridge deck and the area around the upstream end of the right wing wall. Data could not be collected in the floodplains because of heavy vegetation. Velocity magnitudes and water discharge were measured during two of the three site visits using a vertical-axis current meter deployed along the upstream edge of the bridge. Water-surface elevations were measured at the upstream edge of the bridge from the top of the bridge deck between the left most pier and the left abutment. Table 20 summarizes the hydraulic data collected during the flood. Additional bathymetry data were collected 21 m upstream from and 30 m downstream from the bridge after the flood during a low-water site visit on July 15, 1997. Figure 39 shows the elevation and geometry changes experienced by the streambed at the bridge during the period of data collection.
The rightmost pier may have had some effect on the depth of scour at the right abutment, yet it is difficult to determine the pier's effect on the depth of local abutment scour. Limited measurements upstream of the rightmost pier showed the scour hole extended beyond the influence of the pier. The effect of the abutment is believed to be the dominant scouring factor; therefore, all scour is credited to the abutment with none reported for the pier. The observed velocity in the area at the right abutment dropped considerably as the scour-hole depth increased. The velocity at the left abutment held steady through the data collection period, as did the depth and shape of the scour hole. All abutment scour measurements were collected from the upstream edge of the bridge.
Contraction scour is typically computed as the difference in average bed elevation between the uncontracted and contracted sections, adjusted for bed slope. Because field measurements could not be collected in the uncontracted section during the flood, a cross section collected in 1991 and included in the bridge plans was used as a reference surface. All contraction scour measurements were made along the upstream edge of the bridge. As shown in figure 39, there is less than 0.3 m difference in the bed elevation near the center of the channel (beyond the limits of the abutment scour holes) between the 1991 cross section and those collected during and after the 1997 flood. A value of zero for contraction scour is reported.
The reference surface used to determine the depth of abutment scour was the concurrent ambient bed; therefore, the depth of abutment scour reported is additional local scour below the depth of contraction scour, which for this site was negligible. A reference surface at 313.7 m above NGVD of 1929 was used to measure local abutment scour. A summary of the abutment scour data is presented in table 21.
The data collected on April 5 and 9, 1997, and July 15, 1997 were utilized to build and calibrate the HEC-RAS model. Because no bathymetry data were collected during the flood in either the approach or exit sections, low-flow cross sections measured before and after the flood were used. The bathymetry data collected on July 15, 1997, along with geometry taken from the bridge plans, were the basis for the cross sections upstream and downstream of the bridge crossing. Despite the added hydraulic complexities introduced by the meander of the channel near the County Route 22 bridge, the HEC-RAS model predicted the water surface at the bridge within 0.06 m of what was measured in the field on April 5 and 9. When an ineffective flow area representing the recirculation zone between the right abutment and the rightmost pier was included, the model predicted the water-surface elevation at the bridge within 0.03 m of what was observed in the field.
The velocity distributions from the model and the field compared favorably, although the one-dimensional model could not replicate the two-dimensional features of the flow field. Figures 40 and 41 show the velocity distributions for the model, using the geometry from April 5 and 9, and field measurements collected with a vertical-axis current meter along the upstream edge of the bridge. The one-dimensional model results did not compare well with the April 5, 1997, observations (figure 40). Although the model estimated the peak velocity near the rightmost pier reasonably well, the model velocities were too high near the right bank and in the center of the main channel and too low along the left bank. Figure 41 shows that the model did a better job redistributing the flow after the scour had fully developed. The errors displayed should be expected when using a conveyance method to distribute flow that is complex and dominated by two-dimensional effects of the contraction. Since data were not available for the approach section, no comparisons could be made upstream from the bridge.
Assessment of Scour Computations
The calibrated model was used to assess how accurately the scour for this flood could have been predicted. The original geometry of the bridge section was taken from the bridge plans and input into the calibrated HEC-RAS model. The approach-and-exit cross sections were modified to be consistent with the streambed elevations from the bridge plans. The ineffective flow area between the rightmost pier and the right abutment was assumed to be effective since it is unlikely that it would have been assumed ineffective without field observations. The discharges from both April 5 and 9, 1997, were then modeled with the original bathymetry to determine the hydraulic parameters needed for scour computations. The analysis did not include the data collected on April 4, 1997, because no hydraulic measurements were made during that site visit.
The contraction scour was computed in HEC-RAS by allowing the model to use the default equation (live bed or clear water) depending upon the hydraulic conditions computed by the model. The model correctly predicted little or no contraction scour for the prescribed discharges.
Abutment scour was computed in HEC-RAS by both the Froehlich and the HIRE equations. The data contained in table 22 show that the Froehlich equation did a good job predicting abutment scour, when compared to the fully developed scour holes on April 9, 1997. Because the equations predict maximum depth of scour, the Froehlich equation correctly overpredicted the depth of scour, when compared to the scour holes at the right abutment measured on April 5, 1997, which had not fully developed. The HIRE equation overpredicted scour for all situations.