Abstract:
ainfall based flood estimation techniques are common in hydrologic practice. The currently used
methods are based on the design event approach; they use a probabilistic rainfall depth in combination
with representative values of other inputs and then assume that the resulting flood has the same
frequency as that of the rainfall depth. In many cases this assumption is unreasonable and the arbitrary
treatment of various inputs is likely to introduce significant bias in flood estimates for a given
frequency. This study examines the limitation of the design event approach and proposed Joint
Probability Approach.
The proposed approach treats the key design inputs (rainfall intensity, duration, temporal pattern, and
initial loss) and the flood output as random variables, and takes account of the correlations between
these variables in the flood generation process. In this study, the initial loss - continuing loss model
was adopted for computing rainfall excess, and a non-linear runoff routing model for computing
design flood hydrographs. To compute the probability distribution of design floods, Monte Carlo
simulation was adopted. The interaction of random variables involved in the design was taken into
account by using conditional probability distributions.
The proposed model was tested on Harie River catchment. A storm definition was first developed to
extract significant stochastic storm events from rainfall records. The correlations between the
stochastic inputs were then examined, with an emphasis on the dependence of the temporal pattern,
storm duration, and depth. The conditional probability distributions of the stochastic inputs were next
derived and other fixed design inputs determined. Monte Carlo simulation was then used to generate
synthetic flood events. The derived flood frequency curve was finally determined using a frequency
analysis method.
To evaluate the proposed model, design floods estimated by the model were compared with those
obtained by direct flood frequency analysis. It was shown that the proposed Joint Probability Model
provided reliable flood estimates for ARI between 1 and 1 OOyears but it underestimates the flood
peaks for larger ARI. But before concluding firmly on the approach it is necessary to make a test on
more catchments with longer record data.
