Fisher, K.E.1 and P.H. Wiebe2
1Los Alamos National Laboratory, Los Alamos, NM 87545
2Biology Department, Woods Hole Oceanographic
Institution, Woods Hole, MA 02543
3TBA
The Greene Bomber tow body was towed at 3 meters depth between
stations on a number of US GLOBEC broad-scale surveys, providing
millions of observations of temperature, salinity, fluorescence, and
acoustic backscatter. Our goal is to provide an overview of
multiscale analysis, as applied to this alongtrack data for
characterization of patch structures. This approach is closely
related to the work going on at Los Alamos National Laboratory that
aims to provide physics-based characterization of non-linear
systems, specifically to compare simulations and observations of
stochastic processes. Multiscale approaches require large amounts
of data, and characterize patterns statistically; they are limited
in scope by the scales of resolution ranging from the separation of
observations (minima) to that of the survey (maxima); they provide
possible links of pattern to process that must then be carefully
assessed using ancillary information. We use wavelet analysis as
our initial assessment tool. The wavelet transform localizes
contributions to variance for (mono) fractal analysis. The first
steps are: 1) use transform values to construct power spectra; 2)
use local slope of spectra to get local fractal dimension; and 3)
use local fractal dimension to produce stochastic simulation. This
approach works for some applications but using variance alone does
not take advantage of information contained in higher order moments.
Contributions of intermittency are understood by considering
increasingly higher order moments, which characterize the power
contained in events within the signal that are increasingly rare,
and increasingly extreme.
A powerpoint version and an html version of this presentation are available on-line.