A residual noise model for streamflow prediction

Abstract

A nonlinear, conceptual hydrologic model is formulated for use in predicting flows under conditions of controlled input. The model has time-invariant parameters and a residual noise model is incorporated for sequential prediction of hourly flows using a closed loop feedback scheme with hypothetically controlled inputs which are adapted to the residuals. The residual model derived comprises error terms which are correlated in time and variable.It is concluded that the residual noise model approach using an input control adjustment algorithm with adaptation to the model residual, produces satisfactory improvement in flow prediction compared with deterministic predictions using physically measured inputs. Keywords: flow prediction, nonlinear model, residual, correlated noise, input control, adjustment, algorithm, adaptation, measurement, satisfactory improvement.