Efficiency of streamflow modeling and forecasting / Borsch S.V., Simonov Y.A., Khristoforov A.V. // Hydrometeorological Research and Forecasting, 2020, no. 1 (375), pp. 176-189.

Various performance indicators of river flow generation models and methods for predicting its characteristics are analyzed. It is shown that the widely used indicator of a model’s effectiveness – Nash-Sutcliffe – does not fully take into account the specifics of the subsequent use of these models for operational hydrological forecasting.

Examples from the operational practices of streamflow forecasting presented in the paper demonstrate the mismatch between the conclusions with using the Nash-Sutcliff indicator and the results of evaluating the effectiveness of forecasting methods according to the indicators adopted at Roshydromet. On this basis, it is recommended not to be limited to this criterion and use domestic rules for assessing the applicability of forecasting techniques, depending on the relationship between their error and the error of the alternative forecast.

It is shown that to evaluate the methods of daily short-term forecasting of discharges and water levels throughout the year or a separate phase of the water regime, the method of linear extrapolation of the hydrograph can be used as an alternative forecast.

An improved indicator of the applicability of the river flow forecasting technique is proposed, which takes into account not only the compared errors of the tested method and alternative forecast, but also the number of test forecasts, the correlation between the series of these errors and the autocorrelation of each of these series.

Keywords: quality of the river flow generation model, applicability of the forecasting technique, alternative climatic and inertial forecast, linear extrapolation of the hydrograph, efficiency indicator

DOI: https://doi.org/10.37162/2618-9631-2020-1-176-189

Tab. 1. Ref. 26.