Forecasting the hourly average Bz near the solar minima
Several unpredictable factors may influence the Bz magnitude during periods of high solar activity making impossible to make forecasts of the magnitude of Bz. The present study [Núñez and Alonso, 2007, 2008] proposes a model for predicting the hourly average of Bz. The approach was validated with solar data near the solar minima (1995-1996 and 2006-2007) at 1 AU by making predictions with several months in advance. We think that this model may be useful for making mid-term predictions of auroral processes or estimating future time intervals of favourable/unfavourable conditions for the occurrence of magnetic connections between the Sun and the Earth. This study presents our mid-term predictions of the latitude and moment of appearance of an auroral process carried out with six months in advance and compares them with the real data during the last two solar minima.
Evaluation
The developed model, SimulaBz, forecasts the hourly average of the component Bz near the solar minima. Figure 1 presents SimulaBz's output (in blue) and the real hourly average of the component Bz during 2006 and 2007, years of solar minimum (in red). Note that SimulaBz´s output follows the same frequency and phase of the hourly average of Bz. Regarding the magnitude, there is a majority of intervals with similar values (real vs. simulated) and there is a minority with different magnitudes.
Figure 1. This figure shows the time series corresponding to the negative values of the hourly average of real Bz (in red) and the model's output SimulaBz (in blue) for the period between January, 1st and September, 14th of 2007.
Figure 2. Root Mean Squared Error of the SimulaBz's forecasts. The error is mostly smaller than 0.5 nT. Although the error is low in some intervals, there are periods with a high error (near 0.8nT), but acceptable keeping in mind the complexity of the problem, and that there is no alternative model.
Note 1 (see figure 2): The reason of the high forecast error during the third month of 2006 is unknown. Note 2 (see figure 2): The reason of the high forecast error at the end of 2007 was probably the sequence of strong events occurred during December 2006 (flares, CMEs, solar proton events and geomagnetic storms) that perturbed the whole interplanetary medium.
An application:
The POES satellites orbit the poles of the Earth measuring the auroral power. The purpose of this experiment consisted in comparing the evolution of the auroral power measured by the POES satellites and the southward Bz component forecasted by SimulaBz. Figure 3 shows the results of this experiment.
Figure 3 shows the auroral power of the north hemisphere in tens of Gigawatts (in light blue), the negative values of real Bz (in red) and the negative values of Bz forecasted by SimulaBz (in blue) for 2007. When the forecasted Bz is between 0 and -1nT, there is no correlation between the auroral power and the forecasted Bz. However, when the forecasted Bz is lower than -1nT, a peak of high power is observed. Note that Bz forecasts were issued with several months in advance.
Another application that was developed was the prediction of auroral processes with several months in advance. Firstly, we predicted the hourly average value of Bz, generated by our model, SimulaBz. Then, we used the Kozelov model (Kozelov et al., 2003) to simulate auroral processes. The Kozelov model is a cellular automata that simulates the substorm activity of the magnetospheric-ionospheric system from Bz.
