RN-5 : Partial duration series model

Abstract

The basic requirement of flood frequency analysis is to know the probability with which a flow Q is exceeded during a stated design life of a particular project. The partial duration series models provide an alternative method of estimating floods from the number and magnitude of exceedances over a suitable threshold. The other models are annual peak series models and time series models.

Various partial duration series models have been proposed in literature. Most of the models assume Poisson distribution for the distribution of number of exceedances in a year and exponential distribution for the magnitude of exceedances. These models mainly differ from each other in the way the number of peaks every year over a threshold are treated. In the simplest model for partial duration series the variation between years and between seasons, in the number of peaks exceeding the threshold is ignored and a constant number of exceedances is assumed to occur each year. There are models which are capable of accommodating varying number of occurrences of events within the year and different distribution of peak magnitudes during different seasons of the year.

In the view note a theoretical comparison of annual maximum series models and partial duration series models have been presented. The literature review reveals that the assumptions regarding distribution of flood magnitudes are important as they influence the estimates of the floods of high return periods very significantly.