The problem appears to lie in the lack of recognition of different types of rivers with different
meander behavior (Figure 5.30) and the assumption that all meanders are represented by
Analysis Options. A study by Johns Hopkins University (Cherry et al. 1996) for the U.S.
Army Corps of Engineers Waterways Experiment Station investigated the use of both
empirical and analytical approaches to provide solutions to the problem of predicting
meander migration. This study evaluated empirical and analytical (computational) methods
for forecasting planform change and bankline migration in flood control channels, using data
originally assembled and analyzed by Brice (1982) to assess stream channel instability
problems at bridges for the FHWA. Twenty-six sites were used to evaluate the predictive
capabilities of bend-flow meander migration computer models. The computational bend-flow
meander migration model used by Cherry et al. was developed by Garcia et al. (1994).
The Johns Hopkins study recognized that a meandering river is a complex system involving
relations among many variables. The erosion rate for a meander bend is determined by the
balance between the erosive forces applied to the channel bank and the resistance to
erosion provided by the bank material and bank vegetation. Erosive force is a complex
function of discharge, channel cross section geometry, sediment load, bed roughness,
related to the properties of the bank material, the bank geometry (slope, height, shape), the
presence of vegetation, and the state of the pore water in the bank (Cherry et al. 1996).
Although simplified, single valued correlations between a number of variables were
established empirically and expressed as power functions, Johns Hopkins concluded that
they did not adequately describe meander behavior.
In regard to computer modeling, a number of authors have developed versions of the bend
flow model (e.g., Parker et al. 1982; Beck and Melfi 1984; Hasegawa 1989; Odgaard and
Bergs 1987; Garcia et al. 1994) and although the models have typically been tested with
plots of predicted vs. observed channel form for a limited number of channels, there has
been little general testing of these models over a range of hydrologic and geologic conditions
(Cherry et al. 1996).
After testing the 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.
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
(Nanson and Hickin 1983). 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