The development of a sampling design for optimising sampling site locations collected from a coastal marine environment has been the purpose of the present work; application of statistical analysis and spatial autocorrelation methods have been carried out. The dataset included data collected from 34 sampling sites spaced out in the Strait of Lesbos, Greece, arranged in a 1 x 1 NM grid. The coastal shallow ecosystem was subdivided into three zones, an inner one (7 stations), a middle one (16 stations) and an offshore zone (11 stations). The standard error of the chlorophyll-a concentrations in each zone has been used as the criterion for the sampling design optimisation, resulting into reallocation of the sampling sites into the three zones. The positions of the reallocated stations have been assessed by estimation of the spatial heterogeneity and anisotropy of chlorophyll-a concentrations using variograms. Study of the variance of the initial dataset of the inner zone taking into account spatial heterogeneity, revealed two different sub-areas and therefore, the number of the inner stations has been reassessed. The proposed methodology eliminates the number of sampling sites and maximises the information of spatial data from marine ecosystems. It is described as a step-by-step procedure and could be widely applied in sampling design concerning coastal pollution problems.