|
Subject of the paper is the presentation of the potential of use of multispectral remote sensing data for the investigation
of water quality of large water basins on the example of the monitoring of the Baltic Sea with MERIS data. An
interpretation and inversion scheme for optical satellite data over water has been developed to be used in several national
and international projects to monitor different aspects of water quality. The resulting "Principal Component
Interpretation" algorithm allows an optimized estimation of water constituents: chlorophyll pigment concentrations,
suspended matter concentration and yellow substance concentration as well as optical properties of the water body. From
these are derived secondary parameters like water transparency. In the frame of the international ESA MARCOAST
project this interpretation scheme was developed for a regular (daily) monitoring of the Baltic Sea. Results are uniformly
mapped images and concentration maps of the Baltic Sea area from which are additionally derived weekly, monthly and
seasonal means. The Principal Component Interpretation belongs to the class of model based multivariate interpretation
schemes and is closely related to Neural Networks techniques, but bases on a completely different training procedure. It
makes use of an optimal information redistribution between the spectral bands and relates them to the water constituents.
This kind of estimation allows an simultaneous estimation of expected global estimation accuracy. The regular
monitoring is accompanied by the survey of in-situ ground measurements, which can be used for validation.The paper
will present the bio-optical model which is used for the interpretation of Baltic Sea water.
The basics of the interpretation scheme basing on principal component analysis will be explained and results of the
monitoring of different products will be discussed on examples of a time series in 2008, showing the development and
movement of algae blooms, together with other constituents. The obtained results are critically compared with available
ground measurement.
|
