Spectral description of low-frequency oceanic variability

A synthesis is made of simple dynamics with a wide variety of observations to produce a zero-order approximate analytical spectral description of low-frequency oceanic variability in the Northern Hemisphere oceans. Because the spatial inhomogeneity is so great, one must account for it at lowest order, rendering a power density spectrum only a first step toward to a full statistical description. The fundamental hypothesis is that there exists, for each vertical mode of variability, n, a function Phi (k, l, omega, n, phi, lambda), where (k, l) are local horizontal wavenumbers; omega is frequency; and (phi, lambda) are latitude and longitude, respectively, and that can, as a first approximation, be represented in a simple factored form. Data from altimetry, moored current and temperature sensors, acoustic tomography, and XBTs are used to find a first guess form for Phi( k, l, omega, n, phi, lambda), which is at least semiquantitatively accurate. A useful model spectrum proves to be representable as a product of separate factors for wavenumber, frequency, mode number, and a function of latitude and longitude. The results raise dynamical questions concerning the forms that emerge, and present a challenge for improvement of the representation by existing and future observations. Numerous improvements can be made to the detailed structure. A number of illustrative applications are then made, including calculation of an unobserved spectrum (velocity wavenumber) and the detection of climate-scale shifts in ocean property fluxes.