SHORT-TERM FORECASTING OF DISASTROUS HIGH WATERS AND FLOODS

V. A. Kuzmin

Main causes of a poor quality of short-term forecasts of extreme floods and high waters are determined and explained. It is shown how to control the deficiencies attributed to the limited number of models used and the hydrological characteristics observed and to the lack of information on minor and runoff-forming factors which are not measured in principle. An objective criterion for selecting optimal prognostic procedures is given. A procedure on current param¬etrization is suggested which can be performed with a mechanism of stochastic self-training used for calculating the characteristics of the probability density distribution that allows a correct presentation of the random nature of runoff-forming process and a substantial improvement of its prediction.