Geostatistical Modelling Applied To Coastal Gis: An Environmental Case Study

Alexandra Helena Morgado1 and Fernando Costa Gomes2

1 Divisão de Geologia Marinha, Instituto Hidrográfico (PT)
2 Centro de Dados Técnico-científicos, Instituto Hidrográfico (PT)

Geostatistical analysis is a part of statistics that studies the spatial distribution of phenomena based on the "First Law of Geography" (TOBLER, 19701, p. 236), which states that "everything is related to everything else, but near things are more related than distant things", or either, in general closer samples tend to be more similar between themselves then those which are more distant. The obvious implication of this statement is that spatially distributed data present interdependence and therefore the traditional statistics tools are not quite adequate to treat them. This perception led to the creation of tools like spatial self-correlation and opened way to the development of new methodologies applied specifically to spatial data investigation. From the available information on the characteristics of a spatial phenomenon, geostatistical methods search, through the inference in the unsampled space, to quantify the space unfamiliarity of the phenomena through the application of models.
Nevertheless, in geoestatistical analysis the methods of visualization of data and results are still somewhat underestimated. The Geographic Information Systems (GIS) constitute a basic element to assist geostatistical analysis, exploration and presentation of results. Thus, GIS supplies a set of potentialities, e. g. the georreferencing of the data, the possibility of spatial analysis which supply a spatial context to the interpolations and simulations, as well as the capacity to use a wide set of tools of editing and visualization of data.
This work intends to, using geostatistical modelling, evaluate the spatial variability of the relative amount of lead in the sediments in the River Sado Estuary. The study crossed descriptive statistics, construction and analysis of experimental semivariograms, spatial interpolation using ordinary kriging and error evaluation associated with the estimates from the crossed validation and regression analysis. The main objective is to evaluate the spatial variability of the quantity of lead in the sediments, in a particularly sensible area, under an ecological point-of-view.
The study area is located in the south center region of Portugal, in River Sado Estuary. This region is a zone of transition between fluvial (Sado river) and marine (Atlantic Ocean) environments, being forced not only by fluvial currents, but also by marine dynamics, associated with tide fluctuations.

References:

• Tobler, W. (1970) - A Computer Movie Simulating Urban Growth in the Detroit Region. Economic Geography, 46, 2 (1970), pp. 234-240. International Geographical Union, Commission on Quantitative Methods. Ann Arbor, Michigan, USA.