Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
137 CHAPTER 4 CONCLUSIONS AND SUGGESTED RESEARCH CONCLUSIONS The Problem Approximately 84 percent of the 575,000 bridges in the National Bridge Inventory (NBI) are built over streams. A large proportion of these bridges span alluvial streams that are continually adjusting their beds and banks. Many, especially those on more active streams, will experience problems with scour, bank erosion, and channel migration during their useful life. The magnitude of these problems is demonstrated by the estimated average annual flood damage repair costs of approximately $50 million for bridges on the Federal aid system. Highway bridge failures caused by scour and stream instability account for most of the bridge failures in this country. About $75 million were expended annually up to 1973 to repair roads and bridges that were damaged by floods (226). Extrapolating the cost to 2003 makes this annual expenditure to roads and bridges on the order of $300 to $500 million. This cost does not include the additional indirect costs to highway users for fuel and operating costs resulting from temporary closure and detours and to the public for costs associated with higher tariffs, freight rates, additional labor costs and time. The indirect costs associated with a bridge failure have been estimated to exceed the direct cost of bridge repair by a factor of five (227). A study of 373 bridge failures in 1973 indicated that 72 percent of the failures involved abutment damage. A more extensive study in 1978 showed about 50 percent of the failures were from abutment problems, in which lateral channel migration may have been a contributing factor. Although it is difficult to be precise regarding the actual cost to repair damage to the nation's highway system from problems related to channel migration, the number is obviously very large. In addition, the costs cited above do not include the extra costs that result from over design of bridge foundations (deeper foundation depths, unnecessary or over designed countermeasures) that result from our inability to predict of stream instability and channel migration. This lack of knowledge often results in overly conservative design. A practical methodology to predict the rate and extent of channel migration could help reduce the cost of design, construction, repair, rehabilitation and countermeasures for lateral channel instability. A screening procedure to identify stable meandering stream reaches would ensure that engineering and inspection resources are not allocated to locations where there is little probability of a problem developing. The basic objective of this research was to develop a practical methodology to predict the rate and extent of channel migration (i.e., lateral channel shift and down valley migration) in proximity to transportation facilities. The methodology developed will enable practicing engineers to evaluate and determine bridge and other highway facility locations and sizes and ascertain the need for countermeasures considering the potential impacts of channel meander migration over the life of a bridge or highway river crossing.
138 The Methodology Literature Review The propensity for flowing fluids to meander indicates that this behavior is inherent to shear flows and cannot be attributed solely to local non-uniformity of sediment transport or bank erosion, although both are necessary for meandering in alluvial rivers. While there is still much to be explained about the fundamental causes and mechanisms of meandering, it is clear from the literature that meandering is a natural attribute of most alluvial streams. It follows that meandering behavior should be expected in alluvial streams and must be accounted for in the design, siting, and inspection of highway bridge crossings on alluvial streams. The review of the literature on meander growth and migration indicates that while the occurrence, patterns, and sequences of meander growth and migration have been well- documented, it is very difficult to predict the magnitude, direction, and rate at which changes will occur. It is clear that prior screening of bends to exclude those that are part of a multi- channel system, although they display meander-like behavior, is essential to successful predictions. Also, classification of the type of sinuosity present in single-thread, meandering streams greatly enhances predictive confidence. At the very least, the recognition that bends of equal channel width are relatively stable in contrast to meanders with variable width, should be of significance to the highway engineer. This simple observational criterion could eliminate many rivers from concern. Many attempts have been made to model flow and sediment processes at bends, and the fundamental approaches that can be adopted have been fully reviewed in the literature summary of Chapter 2. A number of authors have attempted to produce simple models suitable for engineering applications by modifying more complex models. An example is the work of Garcia et al. (46) who developed the model of Ikeda et al. (47) specifically to provide a tool for stream management and engineering. The basic approach was to use spatial distribution of depth- averaged velocity to drive a morphological model capable of predicting bed scour and the spatial distribution of bank retreat. However, the practical utility of Garcia et al.'s model comes at the price of accepting limiting assumptions that rule out its application to many alluvial rivers. Johns Hopkins University (48) used historical records of bend movement for 26 study sites selected from the Brice collection to test the utility of Garcia et al.'s analytical model of bend migration. Their findings were not encouraging and they recommended against attempting to apply analytical models to make routine prediction of meander movement. In spite of evidence that the prediction of meander shift using numerical models is possible in principle, many difficulties remain unresolved with this approach. Most models require field calibration that demands unrealistic lead times before predictions can be obtained. Also, the input data required is simply unavailable for most streams. Few models consider all of the processes known to be involved in meander migration and those that do are impractical for routine use due to their complexity and need for very accurate field data. In any case, sedimentary and geologic controls within the floodplain that cannot be detected in advance may interrupt progressive meander migration and cause deformation of the bend. In addition, changes of the meander pattern itself can complicate the bend behavior and, finally, human activities can have significant impacts. As a result, a river may be composed of reaches of very
139 different morphology, which requires that each meander must be described quantitatively, and predictions made for a single meander may not be transferred directly to another meander. The conclusion to be drawn from the literature review is that the only complete model of a river is the river itself. While the past behavior of a meandering reach is not necessarily indicative of its future behavior, at least the historical record integrates the effects of all the relevant variables as they operate in that location. If changes in flow regime, sediment availability, bank materials or human activities are known to have occurred during the period of record, the response of the river in the past can indicate how the river may respond to continued changes in the future. It appears that, provided the planform evolution of the study reach can be accurately chronicled using aerial photographs and GIS techniques, a reliable basis exists for prediction by extrapolation on the basis of meander class and style of change, adjusted where appropriate, to account for changes known to have occurred during the period of record or believed to be likely to occur during the period of prediction. Analysis Options The conclusions from the literature review are supported by an evaluation of empirical and deterministic approaches to predicting meander migration. The study by Johns Hopkins University (48) was designed specifically to investigate the use of empirical and analytical approaches to provide solutions to the problem of predicting meander migration. The two approaches to prediction evaluated were: (1) the use of empirical (statistical) relationships between planform characteristics and controlling variables such as discharge, sediment loads, stream or valley gradient, and (2) the use of flow-based computational meander migration models. The Johns Hopkins study concluded that the multidimensional variability of the meander process cannot be captured in a simple regression equation. While several useful empirical relationships were developed for 26 study sites, local erosion direction was accurately predicted, on average, for only 62 percent of a given meandering reach. Using the much larger enhanced data base assembled for this project leads to the same conclusion even when multi-variate regression analysis techniques are tried. Three modes of meander movement were considered: expansion of the radius of curvature, extension across valley, and translation down valley. For bend tightness (bend radius of curvature divided by channel width) and time, the best fit equation for the data yields an R2 = 0.23, indicating that while there was a trend, there was significant scatter around the equation. Attempts to improve the predicted radius of curvature by including discharge, unit discharge, slope, stream power, unit-stream power, grain size, and percent silt-clay did not yield increased R2. Statistically significant relationships were also not forthcoming for the two other modes of meander migration, translation and extension. As noted above, after testing a bend flow model for 26 of the meandering sites in the Brice data set, the Johns Hopkins study concluded that both the accuracy and applicability of the bend-flow meander migration model are limited by a number of simplifying assumptions. Among the most important of these are the use of a single discharge and the assumption of constant channel width, both of which prevent the model from successfully forecasting the spatial and temporal variability that appears to be inherent in the process of bend migration.
140 It was also concluded that much of the discrepancy between the predicted and observed distributions of erosion can be accounted for by the fact that meander migration is modeled as a smooth, continuous process. In reality, erosion occurs predominantly in discrete events, and varies greatly both temporally and spatially along the channel from bend to bend. The Johns Hopkins study noted that the identification of local factors that influence the amount of bank erosion that occurs is a subject that will require further investigation and that further refinements in bend-flow modeling will not improve our predictive capability until we find a more rational way to wed the flow model to a bank erosion model. In 1999, the Federal Emergency Management Agency (FEMA) published a report which evaluated the feasibility of mapping Riverine Erosion Hazard Areas (REHA) and assessed the economic impact of erosion and erosion mapping on the National Flood Insurance Program (NFIP). The conclusions from this study were that despite decades of research into the physical processes associated with riverine erosion (which includes channel migration), knowledge of the subject is still imperfect, and much work remains to be done. Accurate mathematical representation of these processes has not been achieved yet, and available tools produce results surrounded by varying degrees of uncertainty. Nevertheless, there are analytical procedures that can be used to characterize riverine erosion and that, depending on the application, can yield reliable results. For example, because of limitations in data availability and model capabilities, it is extremely difficult to reproduce detailed time variation of stream movement; however, the FEMA study concluded, it is entirely feasible to analyze channel history and infer trends in the stream alignment and average migration rates. Review of the literature, evaluation of analysis options, and consideration of data needs for empirical and deterministic (physical process mathematical modeling) approaches to predicting meander migration support the conclusion that, at present, empirical approaches are more likely than deterministic approaches to yield a practical methodology that will be useful to practicing engineers. Thus, the research approach for this project emphasized enhancing and using empirical data bases to develop photogrammetric comparison techniques as a basis for predicting meander migration. The Handbook A Handbook for using aerial photographs and maps to predict meander migration accompanies this Final Report. The Handbook contains applications guidance and examples for the analytical products of this research, map/aerial photograph comparison techniques and guidelines to predict channel migration in proximity to transportation facilities. This methodology will be useful in reconnaissance, design, rehabilitation, maintenance and inspection of highway facilities, particularly since the Handbook provides the methodology in a stand-alone package to facilitate ease of application. The end result will be a more efficient use of highway resources and a reduction in costs associated with the impacts of channel migration on highway facilities. An essential first step in applying the methodology is screening and classifying the river reach(s) under consideration. Brice (49) attempted to discriminate qualitatively between very stable and less stable channels. He discovered that channels that do not vary significantly in width were relatively stable, whereas channels that were wider at bends were more active. High
141 sinuosity equal-width streams were the most stable, whereas other equal-width streams of lower sinuosity were less stable, and wide bend streams had the highest erosion rates. As presented in the Handbook, this simple stratification of meanders will be of value to the bridge engineer as a screening procedure, allowing preliminary identification of meanders that are very stable. Brice's conclusions were validated by regression analyses using the expanded data base assembled for this project. Of the approaches reviewed, the classification system of Figure 11 (Chapter 2) was adopted as the most applicable for the purposes of this project. As shown in Figure 11, nine screening and classification categories can be used to represent the full range of meandering rivers encountered in the field. As noted above, equiwidth rivers, such as A, B1, and G1, can be screened as stable, but one class, the "wandering" river, should be screened as potentially so unstable and unpredictable that further evaluation would not be likely to produce a meaningful result (in terms of predicting meander migration). All other meandering rivers can be classed as one of the remaining five categories, and analyzed by the photogrammetric comparison techniques presented in the Handbook. As with any analytical technique, aerial photograph comparison technologies have limitations. The accuracy of photo comparison is greatly dependent on the period over which migration is evaluated, the magnitude of internal and external perturbations forced on the system over time, and the number and quality of sequential aerial photos and maps. The analysis will be much more accurate for a channel that has coverage consisting of multiple data sets (aerial photos, maps, and surveys) covering a long period of time (several decades to more than 100 years) versus an analysis consisting of only two or three data sets covering a short time period (several years to a decade). Predictions of migration for channels that have been extensively modified or have undergone major adjustments attributable to extensive land use changes will be much less reliable than those made for channels in relatively stable watersheds. Overlay techniques require the availability of adequate maps and aerial photos that cover a sufficient period of time to be useful. It is the ready availability of aerial photography resources that make the methodologies presented in the Handbook powerful and practical tools for predicting meander migration. Historical aerial photos and maps can be readily obtained from a number of federal, state, and local agencies and the Handbook provides specific guidance for archive and Internet-based search. In general, both air photos and maps will be required to perform a comprehensive and relatively accurate meander migration assessment. Since the scale of aerial photography is often approximate, contemporary maps are usually needed to accurately determine the true scale of air photos without the use of sophisticated photogrammetric instruments. In addition to scale adjustment and distortion problems that are inherent in the use of aerial photography for comparative purposes, there are a number of physical characteristics of the river environment that can complicate the prediction of meander migration impacts on transportation facilities. Countermeasures to halt bank erosion or protect a physical feature within the floodplain can have an impact on the usefulness of the overlays and these features should be identified prior to developing the overlays. Anomalous changes in the bend or bankline configuration or a major reduction in migration rates may suggest that bank protection
142 is present, especially in areas where the bankline is not completely visible or on images with poor resolution. Geologic features, such as clay plugs or rock outcrops, in the floodplain can also limit the usefulness of the overlays because they can have a significant influence on migration patterns. Bends can become distorted as they impinge on these features and localized bankline erosion rates may decrease significantly as these erosion resistant features become exposed in the bank. In reaches where geologic controls are exposed predominantly in the bed of the channel, migration rates may dramatically increase because the channel bed is not adjustable, which may cause the channel to migrate rapidly across the feature. A fundamental assumption of the overlay techniques based on aerial photo or map comparison is that a time period sufficient to "average out" such anomalies will be available, making the historic meander rates a reasonable key to the future. These limitations not withstanding, the results of internal and external Beta testing of the methodologies presented in the Handbook support the conclusion that map and aerial photograph comparison techniques represent the most practical methodology currently available to enable State DOT engineers to predict and plan for the potential impacts of meander migration. Testing and evaluation of the manual overlay technique, computer assisted technique, and GIS-based approach in the Handbook by six State DOTs strongly support the following conclusions: ⢠The Handbook is well organized, well written, and generally easy to follow. The step-by- step approach on examples was well received. ⢠The Handbook provides useful methodologies that are easy enough to apply in practice to be used by DOTs on a regular basis to support design, rehabilitation, and maintenance decisions. ⢠All reviewers were comfortable with the basic manual overlay technique, which provides a good fall-back approach for any analysis. In fact, this fundamental approach may be the preferred methodology for a DOT with only a few sites to analyze or where meander predictions are required only infrequently for specific projects. As an additional internal test of the methodology, the ArcView GIS-based meander migration predictor was applied to the evaluation of 43 active, freely meandering bends with three time periods of photography. The results indicate that: (1) bank erosion direction was predicted within 0 to 30 degrees in nearly 80 percent of the cases; (2) maximum bank migration magnitude was predicted within an accuracy of one percent of channel width per year over the time period covered by the prediction; and (3) this level of accuracy was comparable to the variability of channel width. A qualitative assessment of the procedure indicates that the majority of the predictions were reasonable and compared well with the actual channel migration.
143 The Archive Data Base Another deliverable for this project is an archive of the data base compiled on CD-ROM to include all meander site data acquired for this study. The CD-ROM archives contain the Excel workbooks, MicroStation files, 1990s and historic (where applicable) aerial photos, and the topographic maps for each site in digital file format. The data base includes 141 meander sites containing 1,503 meander bends on 89 rivers in the U.S. The data for each meander site is compiled in Microsoft Excel workbooks. There are multiple spreadsheets within each of the workbooks. The first spreadsheet, designated General Data, contains the general information compiled from various sources and an aerial photo showing the site limits and the included meander bends. Each meander bend is numbered from upstream to downstream. There are individual spreadsheets, designated by the bend number, which contain detailed historic data for each of the bends of the site. There is also a spreadsheet, designated Discharge Data, that has the mean daily and annual peak discharge data for the gage nearest to the site. Finally, a summary spreadsheet contains all the measured data for all the bends of the site. The Excel workbook file includes four spreadsheets that cross-reference each data site by the (1) source of the data, (2) stream classification, (3) river name, and (4) state in which the site is located. This Excel spreadsheet format permits cross-referencing and provides a simple and useable approach to searching the data base. With this archive data set, future researchers will have a readily accessible data base in a very useable format for a variety of studies. These studies could include additional empirical analyses, more complex regressions based on the archive data, and research to develop more practical deterministic models of the meandering process. SUGGESTED RESEARCH It is apparent from the literature review and evaluation of analysis options in Chapter 2, as well as from the conclusions presented in this chapter, that much work remains to be done before the potential impacts of meander migration on transportation infrastructure can be predicted with certainty and ease using statistical or deterministic methods. While the results of this research, comparative analysis based on maps and aerial photography, could be viewed as an interim approach, it is not likely that this approach will be replaced by more sophisticated analytical techniques in the near future. The techniques presented in the Handbook will always be useful at the reconnaissance level or as a "reality check" on other approaches to solving the problem of predicting meander migration. At present, it appears that advances in analyzing and predicting meander migration will take one of the two traditional avenues evaluated in this study: empirical (primarily statistical/ regression) and deterministic (numerical physical process modeling). As identified in the literature review and conclusions, progress on either avenue will require a substantial investment of research resources. While the archive data base assembled for this project will provide a significant resource for an approach along statistical lines, neither the single variable approach with a limited data base attempted by Johns Hopkins University or the multi-variate approach with a greatly expanded data base yielded significant results. Conceivably, an expanded data base with broader
144 geographic distribution could permit segmenting the data by geographic or geomorphic region rather than by meander class, leading to regional regression equations for meander migration. However, this approach was considered carefully at the outset of this project, and rejected on the grounds of practicability and budget. It is by no means certain that the meandering process would exhibit regional, as opposed to river class, characteristics. A deterministic (numeral modeling) approach clearly faces substantial obstacles. The FEMA study in 1999 concluded that despite decades of research into the physical processes associated with riverine erosion, our knowledge of the subject is still imperfect, much work remains to be done, and mathematical representation of these processes has not yet been achieved. The Johns Hopkins University study in 1996 noted that identification of local factors that influence bank erosion is a subject that will require further investigation. It was also concluded as a result of that study that further refinements in bend-flow modeling will not improve our predictive capability until we find a more rational way to "wed" the flow model to the bank erosion model. The data needs alone to develop and apply a purely deterministic or process-based model are formidable. While advances will be made in deterministic modeling of geomorphic processes such as meander migration, at least from a research perspective, configuring the resulting model to provide a practical tool for application by DOTs will remain a challenge. Here again, the archive data base developed for this project could support progress by providing field data for model development, calibration, and testing. Several improvements could be made to the GIS-based measurement and prediction techniques. While the Panel suggested combining the Data Logger and Channel Migration Predictor ArcView extensions, the budget did not permit this enhancement. The data acquisition and prediction steps of the comparative methodology would be streamlined if these tools were combined. In addition, the ArcView extensions are developed as avenue scripts in ArcView 3.2, but should be in ArcGIS 8.3 or greater to take advantage of advances in GIS technology and to ensure continued support by the GIS software developer. Finally, the prediction tool itself could be improved to handle the more complex case of change in the direction of bend migration where three time periods of photography are available for comparison. The predictor currently applies a straight line extrapolation of the direction of migration established by the two most recent time periods.
