The existence of apparent cycles or periodicities in long time series of meteorological variables helps understanding the physical processes that control the climate and the identified cycles may serve as tools for regional climate prognostics. With this insight, the Hodrick-Prescott (HP) filter was applied to the series of Southern Oscillation (SOI) and North Atlantic Oscillation (NAO) indices covering the 1900 to 2010 period. The application of the HP filter to a time series results into two new time series, one of the trend and other of the cyclic/random signal of the variable. Afterwards, the two series were submitted to the Morlet wavelet transform analysis for identifying the apparent cycles. The SOI trend presented cycles of 6, 9, 17 and 28 years, the first associated to the El Niño/La Niña Cycle and the next two possibly related to the 18 year Lunar Nodal Cycle (Saros Cycle), and the remainder, to the Pacific Decadal Oscillation (PDO) cycle. The NAO trend showed 6, 9 and 33 year cycles, the last two possibly related to the Saros Cycle. The linear correlation coefficients between the cyclic/ random SOI series, with 6 month lead, and the Brazilian rainfall totals presented high values, up to 0.72 and -0.67 for the country's northern and southern regions, respectively, during the 1948-1999 (52 years) period, with a statistical significance level above 99%. This period covers the two recent PDO phases. Considering that climate models still do not have still adequate skill for regional forecasting in Brazil, it is suggested that the climate forecasting technique, using up to 6 month leading time of the cyclic/random SOI series, be exploited and, eventually, validated.
Climate diagnostics and prognostics; climate cycles; PDO
