Reconstruction of the aa index on the basis of spectral characteristics

Abstract

Spectral analysis and reconstruction of the geomagnetic activity aa index using the wavelet transform and an iterative regression method (ARIST) is presented. The aa annual average data (1868-1998) was decomposed in frequency levels using the best wavelet packet decomposition tree. ARIST was applied at each level to identify the main periodicities and amplitudes and phases. The most relevant frequencies at the 95% confidence level were found at periods between 2-5 years, between 5-10 years, and around 11-12 years, 23 years, 31 years, 44 years, 100 years and 156 years. The last level of wavelet decomposition is the longterm trend, and a linear fit was applied to this level. Using significant frequencies, a sum of sine waves was calculated and added to the long-term trend, to obtain a reconstructed aa index time series. The original and reconstructed aa time series have a correlation coefficient r = 0.72 in the period 1868-1998. The reconstruction was extended beyond the data interval, for the period 1800 – 2050 AD. The reconstructed past aa values (1800-1867) show low values, similar to the period 1868-1920. The forecasted values of aa for years 2000-2050 show geomagnetic activity levels similar to the present ones, and a small decrease in the period 2010-2025. The same technique was applied to the 3-year running average of the aa index and the correlation of this data with the reconstruction was higher (r = 0.91) because of the removal of high-frequency fluctuations. The long term trend of the reconstructed series is similar to the one obtained with aa annual averages.