2. Data and Methodology
The SLP data north of 20°N are analyzed by using the MWF. The SLP data used here is an updated version of the SLP data set of Trenberth and Paolino (1980), and has monthly and 5 °×5° degree grids from January 1899 to July 2000. Since the previous study reported that the bidecadal oscillation is prominent in the wintertime (Minobe 2000), we focus our attention on winter (Dec.-Feb.) data. The SLP data are available at most of grid points through the century, but the data in high latitudes are missed in the early part of the record. For a meridionally isolated unavailable datum, which is sandwiched by available data both at the northern and southern grid points, the datum is filled with a mean value of the northern and southern data. For the unavailable data that cannot be filled by the meridional interpolation, an EOF technique used for filling the data values. The number of the interpolated data is only a few percent of the total data number.
A gridded SST data is produced from observations of newly devitalized Kobe-collection together with the observations in Comprehensive Ocean-Atmosphere Data Set. The features of the newly digitized Kobe collection were described by Manabe (1999). Monthly climatology and standard deviations are calculated on a monthly 2°×2°grid box using the data from 1960-1990, and the observations that have a larger difference than 3.5 times of the standard deviation from the climatology are not used for further procedures. In gridding, the observations passed the quality control within the monthly 2°×2°grid box are averaged to form monthly mean SSTs. Then, the monthly anomalies are calculated based on the monthly 2°×2°gridded SST data, and the monthly anomalies are further averaged spatially and seasonally to produce 4°×10° grids in latitude and longitudes during the cold-season (November-May) in order to obtain relatively better coverage of the gridded data. The use of the winter SLP data and cold-season SST is consistent with the seasonality of the BO signature shown by Minobe (2000); the BO signature in the SLP field was only found in the winter (Dec.-Feb.) season, but the BO in SST field is observed not only in the winter season but also in the spring (Mar-May) season with weaker amplitudes probably due to the lagged response of the ocean. The resultant yearly sampled data are used in the present paper.
The monthly 5 °×5° gridded land-surface air-temperature (LAT) data from Jan. 1880 to Jun. 1999 is the gridded version of the Global Historical Climate Network (Vose et. al., 1992), and gridding method was described by Baker et al. (1994).
MWF procedure consists of two steps; First step is to detect temporally-varying periods that give the maximal absolute amplitudes of wavelet coefficients as a function of time for a specific region. Second step is to apply a wavelet filter to global data, with the maximal filter gain at the detected periods. The maximal gain, thus, changes temporally in a consistent manner with the observed changes of the period of the interested phenomenon. In the present study, we set the averaging area as the region where North Pacific Index (NPI) (Trenberth and Hurrel 1994) is calculated (30-65°N, 160°E-140°W), since previous studies indicated that the BO observed in the NPI exhibited interesting phase and amplitude modulations (Minobe 1999; 2000).
