One of the main problems with how we as a community have approached the study of river dynamics is that we have created unreasonable expectations for those that are charged with managing rivers. For example, we describe thresholds for channel scour and migration, but fail to mention that the thresholds are fuzzy, and that rivers are unpredictable, chaotic systems. When we make a a prediction that channel change should occur for a flood of magnitude X, and then nothing much happens during such an event, we look foolish and the managers lose confidence in the underlying science.
Dynamic Equilibrium Paradigm
In my view, this largely stems from the fact that we have adopted a research paradigm predicated on the idea that rivers are in a long-term equilibrium with the average flow and sediment supply conditions, constrained by the properties of the rivers’ boundaries. One could argue that the concept of river regime arose simply to account for the fact that rivers are clearly NOT at equilibrium with respect to the annual flood cycle (i.e. when confronted with the fact that river dimensions do not adjust the the equilibrium state associated with each flood event, we fell back on the argument that the rate at which rivers adjust is slow enough that equilibrium cannot be established except over very long time spans).
While we have produced statistical correlations between the size of a stream and some representation of the formative flow for that stream, many studies of stream channel response to disturbance indicate that changes are rapid, not gradual, and that rivers often reshape themselves over a period of days and then may remain stable for decades (In writing this, I realize that my obsession with gravel bed rivers shines through, and that many of my thoughts are probably not applicable to meandering sand bed streams with cohesive banks). Furthermore, the state of the river at the time of the disturbance often determines whether or not the river will respond or not. While we have been busy trying to stuff this inconvenient fact into the dynamic equilibrium paradigm, we have overlooked one very important fact: rivers are complex systems with highly non-linear governing equations with strong feedback loops…they are far more likely to be chaotic systems than stable ones.
This implies to me that an equilibrium approach is fundamentally inappropriate. However, it is so deeply ingrained in our minds (or it is in mine, at least) that, even when trying to think outside the equilibrium box, we are inevitably drawn back within its comforting and mentally constraining boundaries.
Stochastic Dynamics Paradigm
An alternative approach is to embrace variability, rather than to seek to control it, to make variability (in both the forcings and the channel state) the subject of study, rather than averaging it out. We need to think in terms of distribution properties, not single values. Fundamentally, we will need to change the very nature of the questions we ask in our research if we are going to switch to a research paradigm that recognizes that rivers exhibit stochastic dynamics, rather than equilibrium behavior.
Consider the case of a channel migration threshold introduced above. The standard approach is to associate a flood with a given magnitude with the threshold for channel widening; when floods reach or exceed this magnitude, lateral channel migration is supposed to occur. However, in her model STOCASIM, Sarah Davidson shows that a very simple representation of the physics of channel adjustment (i.e. a threshold shear stress for bank erosion and a constant annual rate of re-vegetation of exposed bar surfaces) produces very different behavior: in her stochastic model, a 100-year flood may produce extensive lateral erosion (indeed it is likely to do so), but it may also do virtually nothing, if it follows close on the heels of a flood of equal magnitude. In fact, the relation between bank erosion magnitude and flood magnitude is particularly scattered for all floods less than the 100-year event, and the scatter suggests a probability-based description of the likelihood of erosion for a given flood magnitude is more appropriate. That is, the process of bank erosion is fundamentally stochastic, driven by both the flood magnitude and the channel state. This is just one example of how we might modify our paradigm for thinking about river dynamics, but I imagine many aspects of river behavior are similarly conditioned by their geomorphic history. Indeed many geomorphologists have stated this in the past, but because it is impossible to accommodate those ideas within the equilibrium paradigm, we have done little quantify the effects of history.
To make the switch to a stochastic dynamic paradigm, we need data…lots of data, probably far more data than we can reasonably collect in the field. This means the advances in this field will be driven by numerical modelling efforts (where channel morphodynamics are numerically approximated) and by experiments using physical model. Since most of our existing ideas are firmly based upon equilibrium thinking, the models that we construct and test within this paradigm are unlikely to teach us much about stochastic river behavior, particularly because disagreements between the model predictions and our expectations-of-equilibrium usually result in the model being discarded or modified. Therefore, physical models are much more likely to lead the way, since the physics governing these models are never approximated, and cannot be manipulated to make the model match our expectations. It is for this reason that our informal group is called the Experimental Rivers Network. Where we go from here is up to you.