IWRA Proceedings

AbstractThe study of the statistical structure of runoff series in the low-flow range is a key problem of engineering hydrology. Use of stochastic models for river runoff fluctuations to describe low-flow and hydrological drought will allow us to evaluate the risk, related with water. The basis for the obtained results is the threshold level approach proposed by V. Yevjevich and involved investigation of the behavior of runoff processes in reference to some specified quantity (a threshold). In essence, this threshold approach means the investigation of the distribution for excursions of the process below some specified runoff level. The investigation of such distributions can be possible with help of stochastic models for river runoff fluctuations for different averaging periods. A simple Markov chain specified by nonlinear regression equation with marginal gamma-distribution, describes adequately the distribution of series of low-flow years. Seasonal (intra-annual) low flow periods are reproduced properly by a non -stationary periodic PARMA model. Having an adequate stochastic model for the runoff fluctuation process, one is able to obtain the characteristics of interest either analytically or numerically (by means of simulation experiments). The latter approach is preferable since the emerging analytical difficulties are considerable while, on the other hand, there is a wealth of experience in hydrology of constructing stochastic models. As the main result of investigation we have received that the distribution of variables such as sojourn time for a river runoff (time of staying below some specified threshold level) is characterized by a multi-modal function caused by genetic non-homogeneity of runoff formation conditions in different seasons of a year.