A Multiple-input single-output time-invariant, non-linear model, based on a
black-box systems approach, was used for how forecasting during monsoon fl
ood events using daily data. The model consists of a non-linear component r
epresenting the immediate and moderately delayed responses and a linear com
ponent representing the delayed response of the catchment. It is based on a
structure originally proposed by Muftuoglu, R.F. (New models for nonlinear
catchment analysis. J. Hydrol. 73 (1984) 335-357; Monthly runoff generatio
n by non-linear models. J. Hydrol. 125 (1991) 277-291). The spatial variati
on in rainfall amounts is incorporated in the model by treating the rainfal
l as separate lumped inputs. The model, in its non-parametric form, is firs
t calibrated on various data sets stacked together and then verification te
sts are performed on different data sets. The results show that the model i
s capable of forecasting flows with high efficiencies for the data used, in
dicating that it provides a low cost alternative model for use in sparse-da
ta scenarios and for small-one-off projects. (C) 1999 Elsevier Science B.V.
All rights reserved.