By Marsha M. Hilmes and J.E. Vaill
Appendix A: 1995 Surveyed Cross Sections in Relation to Resulting Model Cross Sections
Frontispiece. Photographs showing people standing on
a recently completed bridge at Mesquite, Nev., during the flood of February 9, 1932,
and the remains of a bridge at Riverside, Nev., after the 1932 flood.
Figure 1. Map showing location of bridge sites, study reaches, and drainage area of Virgin River upstream from Littlefield, Ariz., gaging station
Figure 2. Graph showing recorded March 1995 hydrographs and synthesized 100- and 500-year flood hydrographs for Virgin River at Littlefield, Ariz., and Riverside, Nev., used in model
Figure 3. Graph showing flood-frequency curve for Virgin River at Littlefield, Ariz.
Figure 4. Graph showing cross sections of channel-bed altitude for 1996 (initial altitudes) and for peak of 100- and 500-year floods at Mesquite and Riverside, Nev., bridges
Figure 5. Graph showing relation of stage, mean annual discharge, and peak discharge for Virgin River at Littlefield, Ariz., for 1929-95
Figure 6. Photographs showing Mesquite, Nev., bridge in 1924 and 1996
Figure 7. Photograph showing Riverside, Nev., bridge circa late 1920's and in 1996
Figure 8A and B. Aerial photographs showing Mesquite, Nev., bridge area in 1938 and 1954
Figure 8C and D. Aerial photographs showing Mesquite, Nev., bridge area in 1978 and 1995
Figure 9A and B. Aerial photographs showing Riverside, Nev., bridge area in 1938 and 1953
Figure 9C and D. Aerial photographs showing Riverside, Nev., bridge area in 1978 and 1995
Figure 10. Graph showing mass curves of precipitation at St. George, Utah, and streamflow at Virgin River at Littlefield, Ariz.
Figure 11. Graph showing annual peak discharge and annual mean discharge for Virgin River at Littlefield, Ariz.
Figure 12. Landsat Thematic Mapper images of Riverside, Nev., bridge area May 10, 1988, before Quail Creek Dam failure, and April 11, 1989, after dam failure
Sea level: In this report, "sea level" refers to the National Geodetic Vertical Datum of 1929 (NGVD of 1929, formerly called "Sea-Level Datum of 1929"), which is derived from a general adjustment of the first-order leveling networks of the United States and Canada.
Water year: A water year is the 12-month period from October 1 through September 30. The water year is designated by the calendar year in which it ends.
U.S. GEOLOGICAL SURVEY
GORDON P. EATON, Director
Any use of trade names in this publication is for descriptive purposes only and does not constitute endorsement by the U.S. Government
For additional information write to:
U.S. Geological Survey
333 West Nye Lane, Room 203
Carson City, NV 89706-0866
Copies of this report can be purchased from:
U.S. Geological Survey
Branch of Information Services
Denver, CO 80225-0286
Orders should specify WRIR 97-4073
The calibrated model then was used to estimate the contraction, channel, pier, and total scour for synthesized hydrographs for 100- and 500-year floods at the two bridge sites. The estimated maximum total scour at the Mesquite bridge was 1.30 meters for the 100-year flood and 1.32 meters for the 500-year flood. The maximum total scour at the Riverside bridge was 1.90 meters for the 100-year flood and 2.01 meters for the 500-year flood.
General scour was evaluated using stage-discharge relations at nearby streamflow-gaging stations, 1993-95 channel-geometry data, and channel-geometry data for the 100- and 500-year floods. On the basis of stage and discharge at the Littlefield, Arizona, gaging station, no long-term trend in aggradation or degradation was found. However, several cycles of aggradation and degradation had occurred during the period of record; the difference between the highest and lowest stage was 0.87 meter for a chosen low-flow discharge of 5.66 cubic meters per second for 1929-95. The value of 0.87 meter is probably the best estimate of general scour. The cross sections had an average scour depth of 0.07 meter between 1993 and 1994 and 0.16 meter between 1994 and 1995. The model simulated little general scour for the 100- and 500-year floods at the cross sections and did not give a good estimate of general scour, probably because the duration (days) of the floods used in the model was relatively short when compared with the duration (months or years) of geomorphic processes that influence long-term aggradation or degradation.
Historical geomorphic changes of the Virgin River at the bridge sites and the causes of those changes were documented using aerial photographs from 1938-95 and other historical information. The Virgin River has become narrower and more sinuous through time, the vegetation on the flood plain has increased, and the channel has shifted laterally many times. The processes associated with these channel changes were found to be long-term changes in precipitation and streamflow; the duration, magnitude, and timing of floods; sediment-transport characteristics; channel avulsion; changes in density of vegetation; and anthropogenic influences.
The U.S. Geological Survey (USGS), in cooperation with NDOT, began a study in April 1996 to evaluate scour at the Mesquite and Riverside bridges on the Virgin River. Data for 1994 (prior to the flood of March 1995) were used as input to BRI-STARS and the model was calibrated to the documented channel changes at and near the two bridge sites resulting from the March 1995 flood. Then the calibrated model was used to compute scour depths for 100- and 500-year floods. An additional part of the study was to describe the geomorphic processes in the study area. This information will be used by NDOT to aid in the design process for the new bridges.
The flood of March 12, 1995, on the Virgin River caused major channel changes (lateral migration of the thalweg, aggradation, and degradation). These changes were documented by channel-geometry data collected in water years 1994 and 1995 by the USGS. The instantaneous peak flow recorded at the USGS streamflow-gaging station at Littlefield, Ariz. (number 09415000), hereafter referred to as the Littlefield gaging station, was 581 m3/s. This peak flow was approximately a 25-year flood based on a log-Pearson Type III analysis (Interagency Advisory Committee on Water Data, 1982) of 64 years of record.
Historical aerial photography, channel-geometry data, bed-material particle-size distributions, sediment-discharge data, and continuous streamflow data are available for the Virgin River within the study area. The amount of information provides a unique opportunity to apply the BRI-STARS model to compute channel aggradation and degradation during a flood and to compare the results with documented channel changes. These historical data also help to define the geomorphic processes acting within the study area.
The USGS operates two streamflow-gaging stations on the Virgin River. The Littlefield gaging station (number 09415000) has a drainage area of 13,183 km2 (fig. 1) and has been a continuous-recording site since 1929. The mean annual flow for the period of record is 6.9 m3/s. The highest instantaneous peak flow was 1,730 m3/s on January 1, 1989, as a result of the Quail Creek Dam failure near St. George, Utah. The highest natural instantaneous peak flow was 997 m3/s on December 12, 1966. The instantaneous peak flow for water year 1995 was 581 m3/s on March 12. The sediment record at Littlefield, Ariz., consists of daily sediment data for 1947-68, periodic data for 1977-79 and 1986-92, and daily sediment data for 1992-95. At Littlefield, Ariz., the mean annual air temperature is 18°C and the mean annual rainfall is 16 cm (National Climatic Data Center, 1992).
The Virgin River near Riverside, Nev., gaging station (number 09415190), hereafter referred to as the Riverside gaging station, has a drainage area of 15,488 km2 and was operated from December 1970 to September 1974 and from October 1992 to December 1995. The gage is approximately 1 km downstream from the Riverside bridge (B-85; fig. 1). The instantaneous peak flow recorded during the period of record was 498 m3/s on September 20, 1972. The instantaneous peak flow on March 12, 1995, was 425 m3/s. The sediment record at Riverside, Nev., consists of daily sediment data for 1992-94 and periodic data for 1994-95. Historical climate data are not available at Riverside, Nev.; however, climatic conditions are similar to those at Littlefield, Ariz.
Several reaches along the Virgin River from near Littlefield to downstream from Riverside (reaches 1 through 7 on fig. 1) have available channel-geometry information, bed-material particle-size data, historical aerial photography, and other geomorphic data. The upstream bridge (B-89) at Mesquite is approximately 1.5 km downstream from reach 3. The downstream bridge (B-85) near Riverside is within reaches 5 and 6, which overlap each other (fig. 1).
Owing to a lack of data during natural phenomena and to operating conditions in physical experiments, these variables may have a large range of acceptable values. Accuracy of the model depends on how close the values assigned to variables are to true values of the river system.
Molinas (1990, p. 73) indicates the relative importance of the variables used in sediment routing models. The importance is also a measure of the sensitivity of the variables. If a variable has a high level of sensitivity, the model output will differ significantly with small changes in the input variable and the input variable value will need to be close to the value that occurs naturally. A variable that has a low level of sensitivity has a wider range of input values that can be used in the model before substantial changes in the model output result. Unlike many other sedimentation modeling projects, an observed data base for the Virgin River was available for use in this study to calibrate the model. Actual data values were available for all variables identified by Molinas (1990) as having a high degree of importance in increasing the degree of accuracy for model results. Therefore, results of applying the BRI-STARS model in the Virgin River should compare favorably with results from other BRI-STARS applications.
Channel geometry can be determined from field surveys of cross sections or from top ographic maps. Cross sections can be synthesized at different locations in the study reach by extending survey data using valley slopes from field surveys or computed from topographic maps. Bed-material size distributions are determined from sieve analyses of samples collected at the study sites or from particle counts from the study sites. Recorded hydrographs from streamflow-gaging stations can be used in the model or synthetic hydrographs may be used after they have been divided into periods of equal discharge for a given time step (discretized). Sediment hydrographs based on recorded data can be used in the model or the model can be allowed to create a sediment-inflow hydrograph using a specified sediment-transport equation and data on the size distribution of bed material. Initial water-surface altitudes are determined from stage-discharge relations developed for the initial cross section or from a separate computation analysis of water-surface profiles. Water temperatures can be estimated or a measured value can be used in the model. Roughness coefficients for the channel and flood plain can be estimated during a field inspection of the study sites or from inspection of aerial photography and ground photographs of the sites.
To supplement the historical cross sections, additional cross sections were synthesized by extending surveyed data using valley slopes computed from topographic maps and from the shape of surveyed cross sections. Cross-section data from an indirect determination of discharge at the Mesquite bridge in January 1989 (following the Quail Creek Dam failure) were included to supplement the historical cross-section data. Cross sections at the downstream opening of each bridge site were surveyed in May 1996. Cross sections one bridge width upstream from the bridge (approach section) and one bridge width downstream from the bridge (exit section) were determined from 0.61-m contour maps provided by NDOT.
Two subreaches that include the bridge sites and historical cross sections were selected for use in BRI-STARS. The length of each subreach selected and the cross sections used differed depending on the cross-section spacing and location. The subreaches were selected to include as many historical cross sections as feasible on the basis of the amount of synthesized data required between surveyed sites and the increase in computational time required for model completion.
The subreach selected for model input for the Mesquite bridge (B-89) near Mesquite, Nev., extended from 13,716 m upstream from the bridge to 5,182 m downstream from the bridge and included five historical cross sections in reaches 2 and 3. The subreach selected for model input for the Riverside bridge (B-85) near Riverside, Nev., extended from 6,096 m upstream from the bridge to 2,286 m downstream from the bridge and included six historical cross sections in reaches 4, 5, and 6.
A 2-hour time step was selected for use in the model for the Mesquite bridge and discharge values were selected from the recorded hydrograph for March 1-20, 1995, from the Littlefield gaging station. A 1.5-hour time step was selected for use in the model for the Riverside bridge and discharge values were selected from the recorded hydrograph for March 10-16, 1995, from the Riverside gaging station. Initial water-surface altitudes at the most downstream cross section for each discharge value selected in each channel reach modeled were determined by a slope-conveyance computation at the cross section.
Magnitudes of the 100- and 500-year floods were computed from a log-Pearson Type III analysis (Interagency Advisory Committee on Water Data, 1982) of 64 years of record at the Littlefield gaging station (1930-94) that excluded the 1989 Quail Creek Dam failure flood. No log-Pearson Type III analysis was made for the Riverside gaging station due to the short period of record (less than the 10 years recommended by the Interagency Advisory Committee on Water Data, 1982). The flood-frequency curve for the Littlefield gaging station is shown in figure 3.
The magnitude of the 100-year flood at the Littlefield gaging station was determined to be 969 m3/s and the 500-year flood was 1,597 m3/s. On the basis of the flood-frequency analysis for the Mesquite bridge, the December 1966 flood (997 m3/s) was approximately equal to the 100-year flood. The 1989 Quail Creek Dam failure flood (1,730 m3/s) was approximately equal to the 500-year flood. Therefore, the 1966 and 1989 flood magnitudes only were used to represent the magnitudes of the 100 and 500-year floods, respectively, in the hydrograph synthesis.
Attenuation of peaks between the Littlefield and Riverside gaging stations was found to be approximately 30 percent on the basis of comparison of concurrent recorded hydrographs at the two gaging stations. For the Riverside bridge, the 1972 flood peak recorded at the Riverside gage was approximately equal to the 100-year flood when the 30 percent attenuation is applied to the flood frequency curve (fig. 3) for the Littlefield gaging station.
Synthesized hydrographs were developed for the 100- and 500-year floods by applying a ratio to the recorded March flood hydrographs. The value used to increase the recorded March hydrographs was computed as the ratio of the design-flood magnitude (100- and 500-year) and the magnitude of the peak on March 11-12. The ratios used at the Mesquite bridge were 1.72 for the 100-year flood hydrograph and 2.98 for the 500-year flood hydrograph. The ratios used for the Riverside bridge were 1.71 for the 100-year flood hydrograph and 2.96 for the 500-year flood hydrograph. This approach was used to simulate flood peaks that are representative of an actual storm and of antecedent conditions common to the time of year when most flooding takes place. These synthesized flood hydrographs are shown in figure 2.
The synthesized flood hydrographs described above were compared to flood hydrographs developed using a method described by Eychaner (1976). This method is based on analysis of 364 floods from 18 continuous gaging stations in the Colorado River Basin of southern Utah. The drainage area for the Littlefield gage is considerably greater than the range of drainage areas used to develop the Eychaner method (13-777 km2) and may be a limitation of using this method. The shape and duration of the synthetic hydrographs developed using the method described by Eychaner (1976) were evaluated by comparison with the hydrographs for the peak flows of record at the Littlefield gaging station in water years 1967 (997 m3/s) and 1989 (1,730 m3/s). In general, the hydrographs developed using the method from Eychaner (1976) resulted in a shorter duration of the peak flow than recorded by hydrographs for naturally occurring floods. Therefore, on the basis of this analysis, the synthetic 100- and 500-year flood hydrographs developed by applying a ratio to the March 11-12 flood hydrograph were used in the model.
No previously determined sediment-transport equations were available for the study sites. On the basis of a calibration analysis of the equations available for use in the model, the 1973 Yang method was selected for the Mesquite bridge study reach where the predominant bed-material size was sand and the Meyer-Peter Mueller method was selected for the Riverside bridge where the predominant bed-material sizes were relatively coarse sand. The methods used in this study were chosen for the model because of their accuracy and the short computational times associated with them. The calibration analysis was done by determining which equation resulted in the best comparison of the 1995 data with the model results. BRI-STARS was allowed to develop sediment-inflow hydrographs at both sites using the specified equations and the available bed-material data.
A minimization routine is available in BRI-STARS that allows the model to change channel widths or altitudes after each computational time step based on the Minimum Stream Power Theory (Chang, 1988, p. 229). The theory states that
An option in the minimization routine of BRI-STARS allows imposition of physical limitations on the amount of lateral and vertical change in a cross section if prior knowledge exists of any vertical constraints (bedrock, boulders, diversion weirs, or other anthropogenic features) or lateral constraints such as levees or riprap. A few features were noted within the study reaches but were not at the surveyed cross-section locations. To allow the model to simulate observed changes at the surveyed cross sections, no limits were imposed in the vertical or lateral directions for either study subreach.
A part of a model for Water-Surface PROfile computations (WSPRO; Shearman, 1990) has been incorporated as an option in BRI-STARS to allow analysis of sediment transport through the bridge reach under bridge-backwater conditions. Bridge backwater is the amount of backwater caused by encroachment on the flood plain by the bridge and the upstream extent of bridge-affected water-surface altitudes exceeding those of unconstricted flow. If the use of WSPRO bridge-hydraulics routines is requested, the WSPRO data are prepared separately and stored in a different data file. This associated data file must comply with restrictions imposed in WSPRO. These restrictions include location of a cross section one bridge-width upstream from the bridge (approach section), a cross section at the downstream bridge opening that includes bridge geometry (bridge section), a cross section at the downstream side of the bridge that represents natural channel conditions without the bridge in place (full valley section), and a cross section one bridge-width downstream from the bridge (exit section) for use in the bridge hydraulics routines. Including the cross sections dictated by WSPRO (approach, bridge, full valley, and exit) can substantially affect the BRI-STARS computations by decreasing the flow length between cross sections and decreasing the amount of sediment transport during each time step of the model.
Separate input files were made for each bridge according to the limitations imposed by WSPRO. The geometry for the approach and exit cross sections was determined from 0.61-m contour maps provided by NDOT and from data of an indirect discharge measurement at the Mesquite bridge in January 1989. Field-surveyed data from May 1996 were used for the bridge section.
Several computational options, which include selection of a sediment-transport equation, the time step, the number of sediment computations for each time step, the number of streamtubes used, and the thickness of the active layer, were evaluated to obtain the optimal combination. The optimal combination showed the best correspondence of the resulting model cross sections to the 1995 surveyed data. The sediment-transport equations selected for each bridge reach (see section "Sediment Hydrographs") are not the equations that might be chosen on the basis of bed-material size distributions in the bridge reaches; however, the best results were obtained using these selected equations. Specifying one sediment computation for each time step was sufficient due to the relatively short time steps (1.5 and 2 hours) selected. Increasing the number of sediment computations for each time step did not improve the comparison of the resulting model cross sections with the 1995 surveyed data and also greatly increased the computational time of the model. Two streamtubes were selected for each bridge reach. Increasing the number of streamtubes did not improve the comparison of the model results with the surveyed data. The thickness of the active layer was adjusted to correspond to conditions observed during the field surveys and during the collection of bed-material samples.
Thalweg altitudes of the resulting model cross sections and of the 1995 surveyed cross sections, and the differences between them, are given in table 2. Plots of the 1995 cross sections in relation to the resulting model cross sections are shown in appendix A. The thalweg altitudes for the resulting model cross sections as compared to the thalweg altitudes for the 1995 cross sections were ±1 m at all cross sections (table 2, appendix A) except cross section 3-3, which is immediately downstream from the drop structure at the Bunkerville diversion. For this cross section, the difference between model-simulated and documented (1995) scour was greater than ±1 m. This difference is because of the "unnatural" features (concrete blocks, old cars) in the channel at this cross section that reduce velocity and protect the channel bed that cannot be accounted for in the model. Although the thalweg altitudes compared reasonably well, the model results did not indicate the widening and lateral shifting of the channel that was observed in 1995 at many of the cross sections, especially in reaches 5 and 6 near the Riverside bridge. This difference is because the model does not include a mass-wasting routine and cannot account for the complex channel avulsion(footnote 3) that occurred in these reaches. In summary, the comparison between the calibration results and the observed channel changes between 1994 and 1995 indicates that the model provides a good overall estimate of aggradation or degradation at the bridges.
Relations of stage, mean annual discharge, and peak discharge with time for the Littlefield gaging station are shown in figure 5. A low-flow discharge of 5.66 m3/s was chosen for this analysis(footnote 4) because this flow would be confined within the channel. This analysis was repeated for a slightly higher discharge (11.33 m3/s) with similar results. Figure 5 shows no general, long-term trend of aggradation or degradation;(footnote 5) however, several cycles of aggradation and degradation occurred during 1929-95. The periods of aggradation are generally associated with periods of lower discharge; the periods of degradation are generally associated with periods of higher discharge, except for a high peak flow in 1967 that resulted in aggradation. On the basis of rating curves(footnote 6) in use on October 1 of each year of record (1929-95), which have been adjusted to a common datum, the difference between highest and lowest stage for the chosen low flow (5.66 m3/s) is 0.87 m. This value (0.87 m) is the best estimate of general scour.
As shown in figure 5, approximately 20 years of gage-height data are unavailable for use in the stage-discharge analysis. However, some interpretations about what may have occurred during the period of missing data can be inferred from the streamflow statistics. A series of aggradation and degradation cycles during this period probably resulted in the 0.51 m of degradation between 1937 and 1959. The first major period of degradation probably was associated with the high annual mean discharge during the 1941 water year. A period of gradual aggradation probably followed during the relatively dry period of the 1940's and early 1950's. According to the data, the highest gage height may have been in about 1950. Another period of degradation probably occurred during water year 1952 when the annual mean discharge was above average. This may have been followed by a few years of relative stability and then another period of degradation associated with higher streamflows in 1958. Estimates from the data indicate that the lowest gage heights of record may be associated with the degradation cycle in 1958.
The reach with the greatest amount of scour was reach 5, which is immediately upstream from the Riverside bridge. Between 1993 and 1994, all four cross sections in this reach scoured. Between 1994 and 1995, three of the four cross sections scoured, with cross section 5-2 having the largest amount of scour (1.14 m) of all the cross sections. The large amount of scour at cross section 5-2 was accompanied by channel widening of approximately 58 m on the left bank, resulting in a shift of the channel thalweg to the left bank. Some of the scour in this reach may be linked to the lateral instability of the river in this section. The instability appears due in part to the natural constriction at the downstream end of this reach, which is also the site of the Riverside bridge. No other localized causes of scour were noted.
Reach 3, upstream from the Mesquite bridge, showed relatively little scour. This reach is at the Bunkerville diversion dam. Two of the cross sections are upstream from the diversion and two are downstream. The two cross sections upstream from the dam (cross sections 3-1 and 3-2) showed more scour than those below (cross sections 3-3 and 3-4). Thus, the drop structure on the diversion dam does not have much influence on channel scour downstream.
The dynamics of the river are seen by the large variability in the scour depths within the reaches. For example, in reach 7 between 1993 and 1994, cross section 7-1 at the upstream end of the reach in a straight section was scouring (0.24 m), cross section 7-2 on a meander showed only a minor amount of fill (0.02 m), and cross section 7-3, the most downstream cross section and also in a straight section, was filling (0.41 m). No field evidence suggests a localized cause for this variability.
These data show that the model simulated little scour for the 100- and 500-year floods at most of the cross sections. This is probably because of the short duration hydrograph that was used for the 100- and 500-year floods. This technique does not give a good estimate of general scour because the duration (days) used in the flood model was relatively short when compared with the duration (months or years) of geomorphic processes that influence long-term aggradation or degradation.
In an attempt to gain the longest record of channel change, a variety of historical information, including aerial photography, was used to document the channel changes of the Virgin River through time. Understanding why the channel has changed is difficult. Possible causes of geomorphic change include long-term changes in streamflow, climate, sediment loads, or land use; extreme floods; vegetation encroachment on the flood plain; diversions or dams; or other anthropogenic influences.
Early explorers who described the lower Virgin River include Jedediah Smith, John C. Fremont, William Henry Jackson, and Lieutenant George M. Wheeler. In 1826, Jedediah Smith briefly described the lower Virgin River as being muddy and a little brackish (Fletcher, 1980). In May 1844, John C. Fremont camped near present day Littlefield, Ariz., and described the river in more detail:
In January 1867, William Henry Jackson, the well-known western photographer, wrote in his journal describing the area near Littlefield, Ariz., at the con-fluence of Beaver Dam Wash and the Virgin River:
A few years later, in 1869, Lieutenant George M. Wheeler, on a reconnaissance of southern Nevada, described the lower Virgin River near the confluence of the Muddy River (before Lake Mead) as:
Wheeler's description suggests that in 1869 the river was dynamic with many channel changes, especially channel widening, occurring as the result of floods. On the basis of his description, the channel was tightly meandering, possibly similar to the planform of the river today in some areas. The quicksand banks Wheeler described suggest saturation similar to visual observations made after the series of winter floods in 1995.
Early land surveys of the Virgin River were done in 1881 by the General Land Office. Descriptions of the vegetation and measurements of channel width and bank height were obtained from the field notes. Vegetation mentioned included cottonwood, mesquite, branch grass, greasewood, and willow. Measurements of the channel width varied considerably, but in some places it was noted as being more than 305 m wide. Compare this to the widest areas today at approximately 200-230 m. Directly comparing measurements of channel width and bank height from these notes to present conditions is difficult because little detail was given as to what was considered part of the channel and what was considered the top of bank. In addition, the accuracy of some of these early surveys is questionable. A survey done in 1912-13 of the area near the Nevada-Arizona border appears to be more accurate and detailed than the 1881 survey. The channel in the 1912-13 survey was in some places twice as wide as at present. Vegetation in the 1912-13 survey was similar to that noted in 1881.
Ground photographs were obtained for the Mesquite and Riverside bridges that pre-date the available aerial photography. Figure 6A was taken in 1924 at the Mesquite bridge. The bridge shown pre-dates the current bridge and was slightly upstream. At that time, the channel was wide and shallow. A 1996 photograph (fig. 6B) shows a decrease in channel width, an increase in the amount of vegetation on the flood plain, and an increase in urban development. The current bridge at Mesquite was completed in 1931.
Figure 7A is a pre-1932 view of the bridge at Riverside, Nev. A flood in 1932 destroyed this bridge. Figure 7B is a photograph in 1996 showing the present bridge at the same site. Comparing these two photographs shows a decrease in channel width and an increase in the amount of vegetation.
By the early 1970's, agriculture had increased on the right bank downstream from the bridge. Sewage treatment ponds had been built on the right bank, immediately upstream from the bridge. Both the increase in agriculture and the construction of the treatment ponds encroached on the flood plain and further confined the channel as it went through the bridge. At this time, the channel was relatively straight with some braiding. During 1976 to 1977, a distinct low-flow channel had formed and was relatively straight approaching the bridge. This low-flow channel spanned approximately one-eighth of the bridge. Prominent channel banks had formed, indicating vertical scour. In 1978, additional treatment ponds were built upstream from the bridge extending further into the flood plain (fig. 8C), further constricting the channel upstream from the bridge. A meander formed on the upstream edge of the treatment ponds and the channel widened and became more sinuous. Vegetation continued to become dense and occupy more of the flood plain.
In the early 1980's, the channel upstream and downstream from the bridge shifted laterally several times and became more sinuous. However, near the bridge, the channel was confined not only by the bridge but also by development on the right bank flood plain, which restricted the river from meandering as much as in areas where the channel was not confined. In this confined zone, the valley width available for the channel to migrate across was approximately one-quarter of the width of the valley upstream and downstream. By the 1990's, the channel had become dramatically narrower (fig. 8D) with vegetation covering any area that was not part of the active channel. Also, by the early 1990's, the right-bank flood plain changed from agricultural to urban use. In 1993 and 1995, migration and avulsion of the channel increased in the unconfined areas of the valley upstream from the Bunkerville diversion and downstream from the Mesquite bridge. In the confined area near the bridge, a meander formed immediately upstream and threatened the treatment ponds. Channelization was done after the flood of March 1995; the channel was straightened through the bridge opening in an attempt to save the treatment ponds. In May 1996, a bank stabilization project, which included filling and riprapping of the right bank, was completed at the bridge.
The angle of attack on the bridge piers, which can be measured from the aerial photographs (table 7), has changed through time. From 1938-77, the angle of attack was fairly constant at 10-17 degrees southwest. An exception was in 1973, when the angle was 15 degrees northwest because a mid-channel bar had formed. In the late 1970's and early 1980's, the channel was relatively straight and was not attacking the piers at an angle. The 1990's showed an increase in the angle of attack primarily because of the meander that had formed upstream from the bridge.
Several changes occurred between 1953 and 1973. Several large meanders formed upstream from the bridge and two meanders had been cut off, decreasing the sinuosity of the river. The timing of these changes is unknown, but multiple floods during the period may have been responsible as indicated by the differences in vegetation density in the abandoned meanders. The agricultural area seen in 1953 upstream from the bridge was gone by 1973. By 1978 (fig. 9C), the channel had widened in several places and several of the meanders had changed shape and moved upstream. In 1980, several channel changes were noticeable. The most distinctive was that the channel had straightened, cutting off the large meanders and decreasing the sinuosity once again. The channel also widened in some areas. By 1984, some of the areas that had been cut in 1980 were stabilizing with vegetation, thus decreasing the channel width. In the late 1980's, a new meander began to form in the area where the old meander was cut off.
In 1994, multiple channels had formed and the channel appeared to have widened recently, as evidenced by a lack of vegetation on the channel margins. This widening was probably the result of high flows in 1993. The large meander upstream from the bridge appeared ready to be cut off by another meander through the center. A smaller meander had formed immediately upstream from the bridge. The 1995 photo (fig. 9D) shows some dramatic changes upstream from the bridge, with another meander being cut off and a major shift of the channel toward the left bank, accompanied by channel widening. Two smaller meanders had formed in place of the larger meander. Downstream from the bridge, just beyond the edge of the photograph (fig. 9D), the channel moved toward the left bank approximately 150 m, leaving the Riverside gaging station abandoned.
The angle of attack on the bridge piers, which can be measured from the aerial photographs (table 7), has changed through time. In the 1938 and 1953 photos, the angle of attack was 45 and 49 degrees southwest, respectively. Between 1973 and 1980, the angle fluctuated back and forth from a northwest angle to a southwest angle. Fluctuation continued in the angle of attack during the 1980's and 1990's, with variations between 0 and 25 degrees to the northwest. This fluctuation is associated with the changing meandering pattern of the river and is indicative of the lateral instability at this site.
On the basis of figure 10, the following observations are made. The late 1920's (beginning of data) until the mid 1940's had relatively high precipitation and streamflow. Ground and aerial photographs indicate that the channel was wide and shallow during this period, with sparse vegetation on the flood plain. A second period, mid-1940's to the mid-1970's, is characterized by lower precipitation and streamflow, although the largest natural peak flood (997 m3/s) occurred during this period. Also, during this period, the channel narrowed considerably, probably due to aggradation and to vegetation becoming established on the channel bars and the flood plain. A third period, the mid-1970's to mid-1980's, is marked by an increase in precipitation and streamflow that resulted in major channel changes, including channel widening and channel avulsion. During a fourth period, the mid-1980's to mid-1990's, precipitation and streamflow decreased again, and the curves have slopes similar to the second period. This decrease resulted in the channel again narrowing as vegetation was able to reestablish along the channel margins that had been widened during the previous wet period. Water years 1993 and 1995 were above average for streamflow, and major channel avulsions also occurred during these 2 years. Some vegetation along the channel margins was removed as a result of the channel widening during the floods. These trends in channel change related to precipitation and streamflow over time are similar to those noted by Hereford (1995, p. 61-62) for the upper Virgin River in Zion Canyon.
Additional evidence supporting the statement that peak flow is not a good indicator of geomorphic change on the lower Virgin River involves the flood resulting from the failure of Quail Creek Dam. On January 1, 1989, the dam on Quail Creek Reservoir, an off-stream storage reservoir near St. George, Utah, failed, resulting in a peak flow of 1,730 m3/s at the Littlefield gaging station. This was nearly twice the discharge of the largest natural flood on record (December 5, 1966). Enzel and others (1994) determined that the magnitude of the flood associated with the Quail Creek Dam failure was similar in magnitude to natural paleofloods on the lower Virgin River during the last 1,000 years. The flood associated with the dam failure was of short duration, remaining near peak stage for 1.5-2 hours (Carlson and Meyer, 1995). The Quail Creek Dam was rebuilt after the 1989 flood.
The erosion and deposition that resulted from the Quail Creek Dam failure are described by Carlson and Meyer (1995). Major channel changes occurred for approximately 80 km downstream from the dam. For the lower Virgin River, the effects were less evident. Carlson and Meyer (1995) noted that the channel filled 0.03 m with sand at the Littlefield gaging station and that the flood plain had both aggradation and degradation that decreased downstream. They did not discuss any major lateral shifts in the channel position.
Landsat Thematic Mapper (TM) images taken before and after the Quail Creek Dam failure for the area near the Riverside bridge are shown in figure 12. As previously discussed, this section of the river historically has been one of the most geomorphically dynamic reaches in the lower Virgin River, especially with respect to lateral migration. However, the Landsat images show little evidence of major lateral movement of the channel as a result of this flood. Other areas of the lower Virgin River were examined on the Landsat images and no major lateral shifts of the channel were observed. This short-duration, but extremely high peak flow did not result in major geomorphic change in the study area.
The lack of lateral geomorphic change may be due to the short duration of the flood. This hypothesis is supported by Costa and O'Connor (1995), who found minimal geomorphic changes downstream from two small dam breaks in Oregon and Washington. They attributed the lack of lateral and vertical geomorphic change to the short duration of the flood and not to insufficient stream power (footnote 8) (Nanson and Hean, 1985) or extraordinary stabilization of surfaces by vegetation (Zimmerman and others, 1967). In the lower Virgin River, major channel changes, such as in 1980, have historically resulted from floods that had much lower peak flows and thus lower stream power than that associated with the 1989 flood from the Quail Creek Dam failure. During some of these lower peak discharge floods, new channels have been cut through the vegetation or the channel has been widened resulting in major scour of a vegetated bank. Therefore, lack of lateral geomorphic change as a result of the 1989 Quail Creek Dam failure was not the result of insufficient stream power or stabilization by vegetation.
Duration of the flood is more important than the magnitude of the flood when examining the geomorphic effectiveness on the Virgin River. Longer duration floods of moderate magnitude result in more geomorphic change than short-duration, high-magnitude floods. This is because, for the longer duration floods, the average flood stream power per unit area is high, and total energy expended by the flood is large (Costa and O'Connor, 1995, p. 54). This has the potential to result in tremendous changes to alluvial channels, such as the Virgin River, because stream power remains above the alluvial erosion threshold (Costa and O'Connor, 1995, fig. 11) for a longer time.
Two small floods (peak flows, 41 m3/s and 24 m3/s, as measured at the Littlefield gaging station) in January 1995 were followed by a larger flood in February (309 m3/s), and then two larger, closely spaced floods (544 m3/s and 581 m3/s) at the beginning of March. Relatively little channel change was observed during the January and February floods. On the basis of visual observations during and after the floods, some deposition on the flood plain, removal of vegetation, and minor bank collapse occurred near the bridge sites. The banks were saturated and in some places quicksand was a hazard. The first flood in March resulted in increased deposition on the flood plain in some areas and some widening and shifting of the channel upstream from the Riverside bridge. The meander upstream from the Mesquite bridge continued to cut into the right bank near the sewage treatment ponds. The banks were saturated after this flood and quicksand was abundant. However, most of the major changes in the channel happened during the second March 1995 flood.
A series of moderate, longer duration, flood-inducing storms, which generally occur in winter, is necessary to cause major channel change. Earlier floods saturate the banks, thus allowing future floods to easily cut a new channel, which in turn results in channel avulsion. A high spring snowmelt runoff, as in March 1995, helps to further establish the channel in its new position. The period of the late 1970's and early 1980's, when good aerial photographic coverage is available to document channel changes, also shows a large amount of channel avulsion resulting from a series of moderate, longer duration winter floods, specifically in 1978 and 1980.
Hilmes (1996) also compared loads from water year 1993 to the historical record of water years 1948-68 and found that the Quail Creek Dam failure did not deplete or increase the amount of sediment available for transport in the lower Virgin River. Annual suspended-sediment loads before and after the Quail Creek Dam failure were commonly greater than 4 million metric tons at the Littlefield gaging station. The highest annual suspended-sediment load on record was 6,360,000 metric tons in 1955. These data also suggest that the construction of Quail Creek Reservoir in 1985 did not result in decreased sediment loads. Because Quail Creek Reservoir is an off-stream impoundment and does not interfere with the high sediment loads carried by floods on the Virgin River, it was not unusual to find no change in the sediment loads of the mainstem river.
On the basis of historical research of many southwestern streams, Christensen (1962) found that tamarisk spread rapidly from approximately 1925 to 1960, with the greatest invasion from approximately 1935 to 1955. This increase corresponds with the observed changes in vegetation density seen in the historical aerial photography for the lower Virgin River. In the 1938 photos (figs. 8A and 9A), vegetation is sparse on the flood plain, but by 1953 (figs. 8B and 9B) the vegetation density has greatly increased. Although the type of vegetation cannot be positively determined from the aerial photography, tamarisk presumably dominates.
By 1953, the vegetation occupied areas that had been part of the old channel and this coincides with the period when channel narrowing was observed. Thus, the vegetation encroachment and channel-narrowing process were related and probably occurred simultaneously as a result of decreased discharges during the mid-1940's to mid-1970's. Aggradation of the flood plain also probably occurred when the river overflowed the banks. These geomorphic processes are consistent with those observed on many southwestern streams (Graf, 1978, 1987; Hereford, 1984, 1986, 1995).
Tamarisk has continued to play an important role in the more recent channel stability of the river. The density of vegetation on the flood plain continues to increase, although it is often destroyed in localized areas along the channel margins during higher streamflow, such as in the late 1970's-early 1980's. The only time that large amounts of flood-plain vegetation are destroyed is when channel avulsion occurs. However, tamarisk quickly recolonizes exposed bars and begins to stabilize them, resulting in continued channel narrowing and deepening. Recently, more native plants, such as willows, have begun revegetating some of the newly exposed bars and margins of the channel where widening was localized during the wet years of 1993 and 1995. The native vegetation may continue to dominate these areas or the tamarisk may eventually outcompete the native vegetation, as in the past. Regardless of the type of vegetation now occupying the channel margins, the channel is likely to remain narrow and meandering unless a long-term increase in streamflow results in the removal of the vegetation on the flood plain.
The first anthropogenic change to the flood plain was from agriculture. To irrigate the agricultural area, several irrigation diversions were built along the river and many are still in operation, having been rebuilt after almost every major flood. These diversions generally control the position of the river at least to some degree, but during periods of high streamflow the river often destroys these structures and cuts a new channel. Through time, many owners of these agricultural areas have tried to use some form of bank stabilization (old cars and tires, for example) in an attempt to keep the Virgin River from eroding the land. Stabilization efforts were often unsuccessful, especially during larger floods. During periods of low to moderate streamflow, the bank stabilization has constricted the channel and not allowed the river to utilize the entire flood plain as it had historically, resulting in a narrow, deep channel with confined meanders and steep channel gradients.
In addition to agriculture encroaching on the flood plain, more recent urban developments, such as the sewage-treatment ponds, have further confined the valley width and the area in which the channel is able to migrate. This confinement appears to be a major factor associated with the scour problem at the Mesquite bridge. The channel reach that includes the bridge is naturally constricted compared to channel reaches upstream and downstream. The natural constriction is further enhanced by the encroaching agriculture and sewage-treatment ponds. By forcing the river to go through a narrower opening, the river must find a way to dissipate its energy. The two most obvious ways for the river to dissipate its energy are to meander, reducing the channel gradient, or to downcut, resulting in scour. Beginning about 1978, according to the aerial photographs, the river began forming a meander that threatened the sewage treatment ponds. This meander has been physically removed several times by channelization, but continues to reform as the river minimizes its energy in the constricted zone. By limiting lateral migration and further confining the channel by continued development on the flood plain, the expected result will be an increase in scour, causing increased problems for the structural stability of the Mesquite bridge.
A similar response has been observed at the Riverside bridge, although this study reach has not been as greatly influenced by anthropogenic changes. Agricultural areas have encroached on the flood plain immediately downstream from the bridge. The natural valley constriction at the Riverside bridge is prominent. The response of the river has been to form large meanders upstream from the bridge. The meandering process has not been altered and is thought to be a main reason for the reduced scour problem at this bridge as compared with scour at the Mesquite bridge.
The calibrated model then was used to estimate the contraction, channel, pier, and total scour or fill for synthesized hydrographs of the 100- and 500-year floods for the two bridge sites. The maximum contraction and channel scour or fill at the Mesquite bridge was 0.20 m of fill for the 100-year flood and 0.22 m of fill for the 500-year flood and at the Riverside bridge was 0.41 m of scour for the 100-year flood and 0.47 m of scour for the 500-year flood. The maximum pier scour at the Mesquite bridge was 1.50 m of scour for the 100-year flood and 1.54 m of scour for the 500-year flood and at the Riverside bridge was 1.49 m of scour for the 100-year flood and 1.54 m of scour for the 500-year flood. The maximum total scour at the Mesquite bridge was 1.30 m for the 100-year flood and 1.32 m for the 500-year flood and at the Riverside bridge was 1.90 m for the 100-year flood and 2.01 m for the 500-year flood. The simulated scour for the 100- and 500-year floods was less than anticipated at the bridges on the basis of the scour computed for the lower magnitude floods used in the calibration process. The cross sections also showed little scour for the 100- and 500-year floods. The low scour may be attributed to the large amount of sediment being transported and how the model was designed.
General scour was evaluated using stage-discharge relations, 1993-95 channel cross sections, and channel-geometry data for the modeled 100- and 500-year floods. No long-term trend of aggradation or degradation was determined from analysis of stage and discharge at the Littlefield gaging station. However, several cycles of aggradation and degradation were observed and the difference between the highest and lowest stage was 0.87 m for 1929-95 for a chosen low-flow discharge of 5.66 m3/s. This value is probably the best estimate of general scour. The cross sections had an average depth of scour of 0.07 m between 1993 and 1994 and 0.16 m between 1994 and 1995. The model results for the 100- and 500-year floods showed little scour at the cross sections and did not give a good estimate of general scour, probably because the duration (days) of the floods used in the model was relatively short compared with the duration (months or years) of geomorphic processes that influence long-term aggradation or degradation.
Channel morphology of the Virgin River near the Mesquite and Riverside bridges has changed dramatically, as evident from the historical aerial photography. The Virgin River has become narrower and more sinuous through time, the vegetation on the flood plain has increased, and the channel has shifted laterally many times. To understand why the river has changed through time and how this relates to the current bridge-scour problem, all the geomorphic processes and factors need to be considered.
Climate and streamflow records indicate that the early 1990's was a relatively dry period, with 1993 and 1995 being abnormally high streamflow years. Historically, the greatest channel changes have occurred during longer duration floods that were generally part of a series of floods in which the first few do not result in much geomorphic change, but subsequent ones result in greater channel change, probably due to saturated channel bank conditions. Channel avulsion, a result of sediment plugging the old channel, appears to be the process that results in the most noticeable channel changes. Sediment loads are higher for winter floods because winter floods are generally of longer duration and have a greater water discharge than the summer floods. Construction of Quail Creek Reservoir near St. George, Utah, did not result in decreased sediment loads in the Virgin River because the reservoir is off-stream. When Quail Creek Dam failed on January 1, 1989, the resulting flood did not cause much geomorphic change to the lower Virgin River, probably because of its short duration. Vegetation and anthropogenic influences, especially those related to flood-plain encroachment, have played important roles in channel stability and are probably key factors in the vertical scour at the bridge sites.
If any of these geomorphic factors change in the future, accompanying changes are likely in the stability of the river. Some of these factors, such as long-term trends in streamflow as a result of climate change and the channel avulsion process, cannot be easily managed. However, anthropogenic influences such as encroachment on the flood plain and regulation of flow, can be controlled. The geomorphic processes acting on the Virgin River are among the factors to be considered when designing bridges and other structures.