Sensitivity analysis is a widely applied tool used to investigate the
predictions of numerical models of environmental processes. This paper
illustrates the importance of undertaking a sensitivity analysis that
considers the spatially distributed nature of model predictions, rather
than simply assessing the sensitivity of one or more model parameters
that are assumed to represent the distributed behaviour of the system.
A spatially distributed sensitivity analysis is applied to the output
from a distributed model of turbulent river flow, used to simulate the
flow processes in a natural river channel bifurcation. Examples are
provided for an input parameter which illustrates the importance of
sensitivity analysis with respect to model assessment and error
analysis. The distributed nature of the analysis suggests the
importance of spatial feedback in environmental systems that more
traditional approaches to sensitivity analysis cannot reveal.