Data with which to calibrate a model are often hard to come by. Stage and/or dis-charge records are the best calibration data but are often unavailable for smaller streams or streams in urban areas.
Gage Data. Where gages are present, the engineer should obtain the largest dis-charges and/or stages recorded at the site, along with the corresponding dates. These data are usually published annually for federal gage sites. For studies using the entire hydrograph, the engineer will often have to obtain the raw data (continuous stage readings over the time of interest) from the appropriate agency and convert stage to discharge using the established rating curve for the gage.
Calibrating a model to only a single cross section has limited value. Output from the floodplain modeling program should be calibrated to several actual discharge and stage data pairs representing different runoff events. If multiple gage sites or flood highwater marks are available, they should be used to calibrate the model to several sites along the stream. The most effective way to calibrate the model is to approximate the entire rating curve, or at least that portion representing the higher in-bank flows and any flood records available. The change in channel n with increasing depth can therefore be evaluated and the appropriate value of n selected, depending on the range of discharges that are of interest in the study. HEC-RAS can incorporate varia-tions in n as water surface elevavaria-tions increase. An appropriate target for comparing the peak stages developed by the floodplain model to the prototype is ±0.5 ft (0.2 m).
Highwater Mark Data. If one or more large floods have occurred, highwater marks are often available from the USGS or possibly from a local agency. While recorded discharge data may not be available, recorded flood elevations at several locations throughout the study reach would allow the engineer to approximately match the model output to field conditions. But what if the computed and observed profiles are dissimilar? In this case, the discharge and n values would most likely con-tain the most uncercon-tainty, and the engineer would have to decide how to adjust one or
both (within reason) to better approximate the prototype data. The engineer must be aware that highwater marks can contain some error. If the maximum elevations were obtained from debris lines that were possibly affected by wind-driven waves, the deposited debris would be at a somewhat higher elevation than the water could reach without the wind. Highwater marks obtained from field interviews depend on a per-son’s memory and whether the individual witnessed the flood level on the recession side rather than at the peak. Debris pileups on the upstream side of bridges can result in higher water levels than the program can compute, unless the debris is simulated (discussed in Chapter 6). This happens because the debris partially blocks the flow, reducing bridge cross-sectional area and increasing the upstream water surface eleva-tion.
Field Interviews/Newspaper Records. If no formal highwater mark surveys were performed, the engineer can survey local residents to obtain these estimates directly. The problems identified in the preceding paragraph still apply here. News-paper records are often a good source of highwater mark data, as references to flood level on a local landmark are often printed. Many papers maintain a “morgue” (data-base) of different subject headings, and reviewing these newspaper files could facili-tate finding data on floods in the study area.
The field interviews and records research should look for indications of a road or bridge overtopping, especially if the road was designed for a certain flood frequency.
If the low chord of a bridge was designed to be 2 ft (0.6 m) above, for example, the 10-year design water level, then it could be expected that the hydraulic model would show the 10-year water levels being less than the low chord. Similarly, one might expect the 25- to 50-year average recurrence-interval flood to overtop the structure.
Section 5.8 Calibration and Verification Needs 161
Bankfull capacity is often taken as the one- to two-year recurrence interval event.
Therefore, the one- to two-year peak discharges could be used to determine whether the hydraulic model is approximating the bankfull stage for this discharge. Similarly, if the largest actual flood in recent memory (possibly over the past 20 to 40 years) has lower elevations than those the engineer is obtaining for the 10-year average recur-rence interval flood, the simulated discharge and/or the selected n values may be too high. The following, taken from USACE, 1993b, lists some additional considerations when gathering flood data in the field:
• Obtain as many highwater marks (HWM) as possible after any significant flooding, no matter how close together and how inconsistent they are with nearby HWMs. Describe each HWM location so that surveys may be obtained at a later date.
• Obtain HWMs upstream and downstream of bridges, if possible, so that the effects of these obstructions can be estimated and so that bridge modeling pro-cedures may be confirmed.
• Check on bridge or culvert debris blockages with local residents. For urban streams, check with residents and newspaper files on occurrences of bridge opening blockages by automobiles or debris.
• For historical flooding, check on land use changes, both basinwide and local, since the flood(s) occurred.
• What changes have occurred to the stream channel since the last flood? Have channel modifications been undertaken by the local community? Channel changes, or just natural erosion or deposition that may have occurred since historic floods, if significant, will render calibration with todayʹs channel con-figuration invalid.
• If HWMs are taken from debris lines, remember that wave wash can result in the debris line’s being higher than the HWM, particularly for pools.
• Is the observer giving a biased HWM? A homeowner may give an excessive HWM if he thinks it might benefit a potential project to better protect the prop-erty. Similarly, the owner with a house for sale may give a low estimate or indi-cate that no flooding occurs if he thinks it will affect the sale.
Hydrologic Comparisons. The peak discharge output for a selected recurrence interval flood from hydrologic modeling programs can be compared to hydrographs from gage data and/or the results of regional regression equations for peak discharges of the same frequency.
If one or more discharge gages are available in the study watershed, actual rainfall data for each runoff event should be obtained. The runoff hydrographs for each event can be computed with the hydrologic program, based on the historic storms. The most important outputs from the hydrologic model during calibration are the dis-charge hydrographs at each gage site. The computed hydrographs are compared to the actual hydrographs, with time to peak, peak discharge, hydrograph shape, and hydrograph volume being the important parameters to compare. Calibrating to match hydrograph shape and time of peak generally results in a more defensible model than one that concentrates only on the peak stage or discharge. If the comparisons are poor, adjustments to loss-rate parameters, transformation coefficients (usually unit hydro-graphs), and routing parameters can be made to better approximate the known
hydrograph. Once the calibration is complete, hypothetical rainfall (such as the 100-year return period storm) is used in the simulation to generate the 100-100-year return period runoff hydrograph. The peak discharge from this event can be compared to the peak discharge computed with a regression equation for the parameters of the water-shed. If the comparisons are significantly different, selected watershed parameters (normally infiltration values) are adjusted to achieve a better match between the model simulation and the regression equation calculation. All adjustments should be reasonable and defensible to a technical reviewer. Calibration of hydrologic models is further discussed in Chapter 8.
When a study area has no stream gages, the calibration of the output from the hydro-logic program may only be to the peak discharge computed from a regression equa-tion. Adjustments to the output of the program are normally made by increasing or decreasing infiltration losses to better approximate the results of the regression equa-tion. There is, however, much uncertainty in the regression equation results. For the 100-year average return-interval flood, the predicted peak discharge from a typical regional equation often contains a standard error of 30 percent or more. The adjust-ment of the results of a hydrologic model would be most appropriate if the model results were consistently higher or lower than the results from the prediction equation at several sites. This comparison is appropriate for any watershed but is probably most often used for urban streams.
Verification Data. Verification data are either gage records or highwater marks that were not used in the calibration process. This amount of data is often not available unless the floodplain modeling study is for a large river having long-term gage records. Where such records are available, the engineer would select calibration and verification events. This process could, for instance, be a calibration to the second- and third-largest floods, with model verification on the largest event. Verification is sim-ply the operation of the model(s) using the recorded information for the verification event or events and comparing the actual records to the resulting discharge and/or stage simulation. No model parameter adjustment is done in the verification stage.
The engineer is simply looking for further confirmation that the model reflects the processes of the actual watershed. Chapter 8 discusses verification in detail.
Calibration and verification are extremely important steps in a hydraulic study and should not be ignored or overlooked. Calibrating and verifying a model gives the modeler confidence in the results, which should give further confidence in design ele-ments associated with the study. In the absence of calibration data, the modeler should perform a sensitivity analysis by adjusting those variables that have the most uncertainty associated with them. The purpose of the sensitivity analysis is to give the modeler an intrinsic feel for the behavior of the hydraulic system in response to changes, especially for flows and roughness. Based on the results of the sensitivity analysis, one may find that the water surface profiles do not change that much with higher discharge but rather are very sensitive to changes in roughness values. Conse-quently, the modeler can concentrate more effort on obtaining suitable values for channel roughness coefficients and still have some confidence in the results of the modeling effort.
Section 5.9 Chapter Summary 163
5.9 Chapter Summary
Data needs for floodplain modeling comprise a very large portion of the work required of the engineer. Developing the cross-section geometric and discharge data throughout the stream length typically accounts for most of the time and expense of data gathering. However, roughness coefficients, expansion and contraction values, obstruction geometry, sediment data, and future watershed changes are also important.
A wide variety of data sources may be available, including data published by the fed-eral government, floodplain models developed in earlier studies, and field interviews of people living near a river or stream. Geometric data development carries the most acquisition cost but is typically the most accurate data used for the study. Discharge data may be obtained through several different techniques, but contains much uncer-tainty unless extensive discharge and stage records are available. Roughness coeffi-cients are normally regarded as having the largest uncertainty of any of the hydraulic data. Several techniques are available to assist the modeler in estimating roughness values for a floodplain modeling effort. All data used in a floodplain model should represent reasonable and defensible values that can withstand an independent review.
Calibration and verification of the hydraulic model are very important and can best be performed when actual stage and discharge data are available. In the absence of such data, other techniques may be employed, but the calibration process would be consid-ered less rigorous and thus less accurate. Regardless of whether extensive, or even any, calibration data are available, sensitivity tests on key variables such as peak dis-charge and Manning’s n values should be performed to estimate the importance of these variables on the final makeup of the project or on the final water surface profiles.
Problems
5.1 English Units – Construct an HEC-RAS model using the data in the following tables and answer the questions that follow. The channel discharge is 1200 ft3/s along its entire length, the flow regime is subcritical, and the water surface eleva-tion at river staeleva-tion 1.0 is 163.5 ft mean sea level. Note that the left and right bank stations of the main channel are indicated by the shaded boxes.
a. What is the computed water surface elevation at river station 2.0?
b. What is the average velocity in the main channel at river station 3.0?
c. What is the head loss due to friction between sections 1.0 and 2.0?
d. What are the conveyances for the left overbank, main channel, and right over-bank at river station 4.0?
e. What is the energy grade elevation at river station 2.0?
f. What is the energy correction factor (α) at river station 3.0?
SI Units – Construct an HEC-RAS model using the data in the following tables and answer the questions that follow. The channel discharge is 34 m3/s along its entire length, the flow regime is subcritical, and the water surface elevation at river station 1.0 is 49.8 m mean sea level. Note that the left and right bank sta-tions of main channel are contained in the shaded boxes. Answer quessta-tions (a) through (f) above.
River Station
Left Overbank Main Channel Right Overbank
Manning’s n Length, ft Manning’s n Length, ft Manning’s n Length, ft
1.0 0.035 N/A 0.02 N/A 0.04 N/A
2.0 0.035 250 0.02 250 0.04 250
3.0 0.035 330 0.02 340 0.04 350
4.0 0.035 500 0.02 500 0.04 500
River Station 1.0 River Station 2.0 River Station 3.0 River Station 4.0 Station, ft Elevation, ft Station, ft Elevation, ft Station, ft Elevation, ft Station, ft Elevation, ft
200.0 165.0 200.0 165.4 200.0 165.8 189.9 167.3
206.1 163.3 206.1 163.6 206.1 164.1 216.7 164.4
220.9 161.0 235.2 163.4 220.9 161.8 220.9 162.2
237.8 160.6 237.8 161.0 251.9 161.4 237.8 161.8
250.0 158.0 250.0 158.4 254.6 160.2 250.0 159.2
267.9 158.8 269.1 158.2 266.9 158.7 267.9 159.5
278.1 158.5 278.3 157.7 278.3 158.5 278.3 159.4
290.0 158.0 290.0 158.5 290.0 158.8 290.0 159.2
298.9 161.0 298.9 161.4 298.9 161.8 298.9 162.2
305.7 163.3 314.2 162.1 316.1 162.7 320.3 164.1
310.0 165.0 319.8 166.4 334.8 164.9 332.4 167.3
River Station
Left Overbank Main Channel Right Overbank
Manning’s n Length, m Manning’s n Length, m Manning’s n Length, m
1.0 0.035 N/A 0.02 N/A 0.04 N/A
2.0 0.035 76.2 0.02 76.2 0.04 76.2
3.0 0.035 100.6 0.02 103.6 0.04 106.7
4.0 0.035 152.4 0.02 152.4 0.04 152.4
River Station 1.0 River Station 2.0 River Station 3.0 River Station 4.0 Station, m Elevation, m Station, m Elevation, m Station, m Elevation, m Station, m Elevation, m
61.0 50.3 61.0 50.4 61.0 50.5 57.9 51.0
62.8 49.8 62.8 49.9 62.8 50.0 66.1 50.1
67.3 49.1 71.7 49.8 67.3 49.3 67.3 49.4
72.5 49.0 72.5 49.1 76.8 49.2 72.5 49.3
76.2 48.2 76.2 48.3 77.6 48.8 76.2 48.5
Problems 165
5.2 English Units – Construct an HEC-RAS model for the system shown in the figure, and answer the questions that follow. The channel is trapezoidal with a bottom width of 49.2 ft and 3:1 side slopes. The channel slope is 0.002, the material is con-crete (Manning’s n = 0.013), and the length is 492.1 ft. A flow of 11,300 ft3/s enters the channel from the broad-crested weir, as illustrated in the figure.
a. Is the flow in the channel subcritical or supercritical?
b. What is the depth of water at the upstream end of the channel?
c. What is the depth of water at the downstream end of the channel?
d. What should be the minimum depth of the channel?
e. Does uniform flow occur within the channel?
f. What type of flow regime does this profile depict?
SI Units – Construct an HEC-RAS model for the system shown in the figure, and answer the questions that follow. The channel is trapezoidal with a bottom width of 15 m and 3:1 side slopes. The channel slope is 0.002, the material is concrete (Manning’s n = 0.013), and the length is 150 m. A flow of 320 m3/s enters the chan-nel from a broad-crested weir, as illustrated in the figure. Answer questions (a) through (f) above.
81.7 48.4 82.0 48.2 81.4 48.4 81.7 48.6
84.8 48.3 84.8 48.1 84.8 48.3 84.8 48.6
88.4 48.2 88.4 48.3 88.4 48.4 88.4 48.5
91.1 49.1 91.1 49.2 91.1 49.3 91.1 49.4
93.2 49.8 95.8 49.4 96.3 49.6 97.6 50.0
94.5 50.3 97.5 50.7 102.0 50.3 101.3 51.0
River Station 1.0 River Station 2.0 River Station 3.0 River Station 4.0 Station, m Elevation, m Station, m Elevation, m Station, m Elevation, m Station, m Elevation, m
Bridges are the most common obstruction that modelers must address in water sur-face profile computations. Obtaining accurate profile estimates of rivers and streams with bridges can require much time and effort. A variety of flow situations through a bridge are possible, including subcritical low flow, pressure and weir flow, and super-critical flow. This chapter gives guidance for modeling bridge flow and discusses the various types of flow conditions that are possible through bridges. Modeling bridges with HEC-RAS is discussed along with the critical simulation of contraction and expansion of flow into and out of the bridge. The ineffective flow area concept, which is one of the most common sources of error in bridge computations, is presented in detail. The procedures and methods for modeling different types of bridges are described with examples to illustrate the key points.
6.1 The Effects of a Bridge on Water Flow
Bridges are constructed over waterways, resulting in the bridge structure and its ele-ments obstructing the natural water flow. For a bridge to be structurally sound, it must have supports, such as piers and abutments. These supports are normally located within the waterway.
An obstruction to flow typically forces water surface elevations on the upstream side of the structure to be higher than they would be if the obstruction werenʹt present.
Water surface elevations can be affected for some distance upstream of the structure.
When designing a bridge, it is extremely important to analyze its adverse effects on upstream flow.
Under subcritical flow conditions, the effects of a bridge may be observed well upstream of the structure as the width of flow across the valley contracts to pass C H A P T E R