USING THE MONTE CARLO METHOD TO ESTIMATE ERRORS IN CALCULATIONS OF RIVER FLOW CHARACTERISTICS
V.A. SHELUTKO^{1}, E.S. URUSOVA^{1}, E.S. ANDREEVA^{2}
^{1}Russian State Hydrometeorological University, 192007, St. Petersburg, ul. Voronezhskaya, 79, Russia shelutko@rshu.ru ^{2}Don State Technical University, 344000, RostovonDon, pl. Gagarina, 1, Russia espmeteo@yandex.ru
Keywords: ряд наблюдений, кривые обеспеченности, линеаризация и нормализация связей, речной сток, характеристики рассеивания, series of observations, probability curves, linearization and normalization of relations, river flow, characteristics of the dispersion
Abstract
We examine some issues related to the application of the Monte Carlo method for estimating of errors in calculations of numerical characteristics of the river flow from available series of observations. As a result, it is shown that the adopted Monte Carlo algorithm, which served as the basis of official recommendations for calculating the numerical characteristics of the flow, leads to a significant exaggeration of the negative bias of the numerical characteristics of the dispersion. It is found that the implementation of the above algorithm does not take into account a number of questions, and these authors suggest that special attention should be given to them. Among them are the following four: the first question is related to the fact that the tables of ordinates of the binomial probability curve are used for statistical testing of samples of different durations, which are often not applicable for a large number of tests due to extrapolation beyond the data given in the tables. The second question implies that, in some cases, the Pearson type III probability distribution curves, constructed for time series of the annual flow, negative values are obtained from modeling, which contradicts the physical essence of the river flow. The third question is related to the applica tion of the method of normalization and linearization of G.A. Alekseev’s connections without taking into account the smoothing effect of this method, which leads to an increase in the negative shift of the scattering characteristics. The fourth question is due to the lack of research on the issue of accounting for the spread of empirical points relative to theoretical curves, and the influ ence of the above effect on the final result of modeling. In this context, it is concluded that there is a need for substantial clari fication of these points set out in the four questions as well as for the development of appropriate recommendations.
