ABOUT OF FORECASTING POSSIBILITY OF HYDROLOGICAL TIME SERIES

Abstract

It is offered forecast the natural time series (hydrological time series) by methods the deterministic chaotic dynamic. At analysis of the time series reveal the hidden regularities at raw data’s. Hereinafter revealled regularities use for realization of the forecast of input data. Hydrological time series either as chaotical time series possible forecast only at determined number step onward. For hydrological time series exists the limit of forecasting (forecasting horizon). The identical prediction possible unless come behind of the forecasting horizon. The aim of the work there is design of the procedure the investigations of the natural time series on possibility of the forecasting. As raw data’s were used the mean day data of consuption for the large flat river of the Ukraine (length is 4*365 points). For forecasting of the natural time series it was used designed an autors modification of the method of Local Approximation. The forecasting horizon of time series was defined by means of factor the linear correlation (on how much points onward (maximum) possible forecast the time series for conservation of the factor the correlation within the range of 0.7 … 1.0). The explored dependency of the forecasting horizon from parameter DIM (Dimensionality Reconstrusted the phase space). The results of the research say for essential determinism of time series (hydrological) – the time series is forecasted on 15 … 22 points onward. Software was designed for investigations of the natural time series (hydrological and hydrochemistry) for forecasting (to find forecasting horizon). The called on investigation has shown that the method of Local Approximation more efficient than classical methods the forecast (for classical methods the adequate forecast possible only on 1 … 2 points onward).