Smoothing algorithms for statistical solar data preprocessing

Abstract

Errors of the moving average and the conditions under which the use of the moving average method in the analysis of solar activity distorts the important features of this process was investigated. It was shown that in the approximation of solar cycle oscillations with the period significantly greater than 13 months on the basis of 13-month moving average there is only a small systematic error. Magnitude of these oscillations decrease somewhat, but the whole process is not distorted. However, oscillations with a period in the range from 6 to 12 months are inverted by 13-month moving average, i.e. convex “wave” is replaced with a concave one and vice versa. At the boundary points, when the oscillation period is 6 or 12 months, a 13-month average leads to a complete loss of these oscillations, i.e. reduces their value to zero. This is of fundamental importance for the study of the short-term (months) variability of the solar activity.