Predictive Modeling in Geomorphology: An Oxymoron?

Abstract

Four basic types of model are commonly used in geomorphology: Conceptual, Physical, Analytical, and Numerical. Many research projects start with a conceptual model that needs to be tested. Physical models allow one to see, visually, in a laboratory setting, how certain geomorphic processes operate. Analytical models are appropriate when the system is sufficiently simple, or simplifiable, that equations describing it can be written and solved in closed form. The required simplifications, however, particularly of the boundary conditions, often limit the predictive value of the model. Numerical models allow one to include more complicated boundary and initial conditions, and also permit solution of differential equations that cannot be integrated in closed form. While projects often begin with a conceptual model, testing with other types of model commonly leads to new insights, and hence to revision of the original conceptual model. Thus, geomorphological research commonly involves feedback between various types of model and, of equal importance, between models and field observations. Despite the usefulness of models in geomorphological research, their predictive value is limited owing to the lack of sufficiently detailed knowledge of variations in forcing, boundary conditions, and material properties. Thus, many predictions must be prefaced by a statement of the probability of a certain outcome. Whether such statements qualify as predictions or not depends on one's definition of predict, but for the general public more precise predictions may be required before appropriate action is taken.