Comments: download file (pdf
)
Ref.: SiPLAB Report 02/06, FCT, University of Algarve,2006.
Abstract:
The space-time oceanographic mapping of given areas assumes an obvious
importance in environmental and acoustic assessment, from regional to
large scales. Since the 70’s, deterministic models for oceanographic
processes have been coupled with statistical interpolation tools, in
order to cope with the inevitable sparsity of oceanographic
measurements. These tools treat physical quantities as random, and
assume one from a collection of possible models for their
inter-correlation. One of these models, a modified Gaussin with
space-time dependence, is here fitted to temperature data correlations
from the acoustic-oceanographic MREA ’03 sea trial, in order to
characterize the temperature space-time distribution and evolution. The
model assumes the temperature shift from the mean as anisotropic,
homogeneous and stationary. The correlation model is parameterized by
four correlation lengths and a scaling factor. The
multiple-instrumental sea trial setup allowed the estimation of the
unknown parameters, by fitting the model with two differently
originated data correlations. The first is from data measured by a
spatially fixed single instrument or by CTD casts in the area. The
second is from stacked data measured by all temperature monitoring
instruments. The results show that the actual temperature perturbation
field is not homogeneous, as expected. Significant differences exist
for the parameter estimates, comparing that obtained with shallower
instruments with those obtained by instruments with a deeper covering
array aperture. The question about the usefulness of combining the
information coming from several instruments has not a trivial answer.
If the objective is to interpolate along depth and time, the
information coming only from instruments localized on the interest area
is likely to give better results. Instead, for horizontal
interpolations, it is essential to combine various instruments,
considering the data at hand. This was the unique possibility to obtain
a reasonable number of data correlation lags, to be fitted by the
model, and give realistic horizontal correlation lengths.
ACKNOWLEDGMENT: this work was partially supported by
project NUACE, funded by FCT, Portugal.