A linear programming approach to fitting rainfall runoff models based on finite mixture hydrographs: a potential for flood forecasting?

Abstract

It is recognised that the transformation of rainfall to discharge is essentially nonlinear and it would seem logical to construct parsimonious nonlinear models to approximate the runoff process. Lumped nonlinear rainfall-runoff models are simple to construct but difficulties may be created in the calibration process if suboptimal fits to data arise from local minima in the objective function measuring goodness of fit, where small fit values indicate improved fit.