Experiment on Long-Term Forecasting of Geomagnetic Activity Based on Nonlocal Correlations

An experiment was performed on using advanced macroscopic nonlocal correlations to forecast slow random oscillations of the Dst index of geomagnetic activity. The global maximum of the correlation of Dst with the electrode detector signal reaches 0.97, which is sufficient for forecasting, and its time shift corresponds to advance of the detector signal relative to Dst for 329 days. The large magnitude of the time shift is due to the slow diffusion mechanism of entanglement swapping between the detector and the source. In this case, the position of the global maximum of the correlation function coincides with the position of the global minimum of the entropy independence function, which confirms that it is not distorted by possible nonlinearity of the relationship and determines the optimal lead time of the forecast. Long series of test Dst forecasts have been calculated based on data from a nonlocal correlation detector with a fixed lead time using three methods: current regression, current impulse transient response, and current neural network. The forecast accuracy is sufficient for all practical purposes.