Interannual variability of Arctic Ocean temperature and salinity fields for fifties-eighties by spectral analysis method

Lack and unevenly location of oceanographic observing sites at Arctic Ocean was a principal cause of absence of related variability studies. Spectral analysis was developed to overcome this problem. The basic idea of spectral analysis is to represent all available observational information in compressed form to minimise losses in accuracy. A well known spectral method is principal component analysis (PCA)[ Hotteling method ]expressed in terms of two-dimensional empirical orthogonal functions (EOF's). The combined spatial and temporal interrelationship of observational data can be expressed as a bi-orthogonal expansion closely related to SVD (singular value decomposition) of the observational matrix. The EOF's and the corresponding expansion coefficients are usually referred to as the principal components (PC) and describe the temporal modes. For each temporal mode a separate set of EOF's are calculated. Its values in missed grid points had been interpolated by kriging technique. Spectral method assumes a priori covariances of expansion coefficients to be known. The estimated a posteriori error covariances of expansion coefficients and the estimated a posteriori error covariances at the grid points are determined by some non-linear relationships contained observation error covariances and a priori covariances of expansion coefficients as well as EOF's values at observational sites and grid points. Reconstructed and error fields of water temperature and salinity for 20 depth levels in Arctic Ocean for 40 year data set are discussed. In the present application, intensive statistics were derived for the 70's decade, which provided good temporal and spatial sampling. Statistics obtained from the spectral analysis procedure were then applied to the other three decades (50's, 60's, and 80's) which all contain areas under-sampled during all years of the decade. Early products from spectral objective analysis have been included in the climatic and digital atlases (http://ns.noaa.gov/atlas/html/met/met_int.htm). Relationship between Atlantic inflow anomalies and related peculiarities of Arctic circumpolar patterns has been analysed. Annual patterns of NAO index were classified by fuzzy logic approach. It was appeared that there are some relationships between annual NOA index anomalies and temperatures of ascending Atlantic warm waters from Nordic seas to Arctic basin. Some linkages between NAO index anomalies and Arctic circulation patterns had been revealed. EOF's temperature and salinity field analysis had been performed for variability study. First EOF explains 17-19% of STD (standard deviation) for salinity fields at 0-300 meter depth band and 11-13% of STD for temperature fields at the same depths. Temporal dependencies of EOF's expansion coefficients had been calculated. First EOF coefficient time series reveals fast biennially oscillations at the beginning of 50's and at the end of 60's for the surface level. Therefore, 50's and 60's surface temperature variability exceeded the same values for other decades. Slow 5-9 year oscillations give a substantial impact on inter-annual variability of surface fields. In contrary, the oscillation of 3-5 year scale dominates at 100-300 meter levels for 50's and 60's. The 5-7 year oscillations are also revealed at time series for 100-300 meter levels.