The delineation of ocean regions with similar water quality characteristics is an all important component of the study of
marine environment with direct implications for management actions. Marine eutrophication constitutes an important
facet of ocean water quality, and pertains to the natural process representing excessive algal growth due to nutrient
supply of marine systems. Remote sensing technology provides the de-facto means for marine eutrophication assessment
over large regions of the ocean, with increasingly high spatial and temporal resolutions. In this work, monthly
measurements of sea water quality variables – chlorophyll, nitrates, phosphates, dissolved oxygen – obtained from the
Sea-viewing Wide Field-of-view Sensor (SeaWiFS) with spatial resolution 0.125 degrees for the East Mediterranean
region over the period January 1999 to December 2010, are used to define regions or zones of similar eutrophication
levels. A novel variant of the K-medoids clustering algorithm is proposed, whereby the spatial association of the
different variables (multivariate textural information) is explicitly accounted for in terms of the multivariate variogram;
i.e., a measure of joint dissimilarity between different variables as a function of geographical distance. Similar water
quality regions are obtained for various months and years, focusing on the spring season and on the qualitative
comparison of the traditional and proposed classification methods. The results indicate that the proposed clustering
method yields more physically meaningful clusters due to the incorporation of the multivariate textural information.
Spatial interpolation methods are widely applied in marine studies to evaluate the spatial distribution of oceanographic
parameters and inter-compare time series of maps of selected variables. A variety of such methods are nowadays
available and therefore, selection of the most appropriate for a specific case study is not an easy task. Within geography
and other spatially oriented disciplines, and most of the times in the framework of a Geographical Information System,
several attempts have been carried out to assess the efficiency of various spatial interpolators using diverse
methodologies. In this paper, an attempt was carried out to evaluate the accuracy of spatial interpolators for mapping the
distribution of organic carbon (weight %), an important indicator of marine sediments in the marine environment.
Measurements of organic carbon were carried out in a network of 20 sampling sites in the Gulf of Gera, which is
representative of a semi-enclosed and shallow marine ecosystem at the south-eastern part of the island of Lesvos,
Greece. For the interpolators under study, the cross-validation error was calculated at each sampling station and
calculation of the RMSE (root-mean-square error), the MAE (mean-absolute error) and the MBE (mean-bias error) was
carried out to assess the accuracy of their performance. The results revealed the most appropriate interpolator for the
given dataset which was then applied to develop the thematic map of the spatial distribution of organic carbon.
Discussion on the potential increase of the surface accuracy is also carried out.