We conclude with a technical note, briefly describing analyses of space-time data and Geographical Information System (GIS) methods – regarding point data, type of interpolation, and so on – used by the authors in producing the maps. Sometimes symbols with contrasting information appear to overlap, due to representation of spatially very close information which is not to scale. An important characteristic of the study of the Lagoon – and of environmental systems in general – is the involvement of several disciplines (physics, chemistry, biology, geology, ecology, environmental sciences, engineering, economics, archaeology, etc.), whose methods of measurement and analysis vary widely. These disciplines produce masses of heterogeneous data, which enable deeper understanding of the relations and functions of environmental systems only if they are interpreted in an integrated and unified way. Evaluations and predictive models require integrated analysis of data and interdisciplinary reading of the processes – only in this way can management systems and methods able to support political decision-making be created. Complexity and uncertainty exist in all forms of environmental analysis: from the study of the distribution and/or evolution of environmental phenomena to the construction of indices for evaluating environmental quality. Statistical methods are a powerful tool in this field, because they tackle the issue of uncertainty directly and provide effective techniques for reduction of data and model-based treatment of complexity. GISs are an additional tool for analysing territory. This type of software makes it possible to position and analyse objects and events which exist and are verifiable. As an instrument for organising data relating to land and sea, a GIS enables researchers to integrate the results of measurements, models and statistical analyses, and to generate geographical analyses accompanied by tables, documents and maps. Regarding environmental studies, the two main areas of use are visual representation of patterns on the geographical level (descriptive use) and spatial analysis (analytical use). GISs can be of help in setting up simulation models and, combined with the use of expert systems, can help to predict the intensity of pollution or coastal erosion in various places and in dynamic situations (the seasonal pattern of chlorophyll, for example). Visualisation of spatial phenomena is a highly effective descriptive method. In terms of environmental analysis and evaluation, a GIS is a very useful tool for producing thematic maps of the spatial distribution of various phenomena (for example, precipitation, surges in river flows, morphological changes, pollution, nutrients, salinity, diseases). With access to updated archives, new maps can be created almost instantly to represent the current situation, and the spatial dynamics of the phenomena in question can be monitored continuously. The level of information is controlled by the user, who can decide whether to focus on individual, local cases, or to represent indices and measurements of frequency, intensity, prevalence etc., on the level of the geographical, ecological or administrative unit. Spatial analysis carried out with GISs includes a large number of operations. In this field, three types of spatial data can be analysed: points, surfaces and geostatistical patterns. The first problem is the correlation of a small dataset within a large space and the creation of spatial variation models. Geostatistical analysis makes it possible to correlate data from spatially dispersed points and to determine how they vary in space and what errors may derive from interpolation models. The second issue is the evaluation of data collected at different times, in an environment which varies widely from year to year. Here too, statistics can help, by calculating the resulting errors, thus providing a measure of the reliability of the analysis. The maps based on the interpolation of point data were produced using the Inverse Distance Weighted method (IDW) as a mathematical interpolation technique and the ordinary kriging method as a geostatistical interpolation technique. IDW predicts the variables in points which were not directly sampled, assuming that the similarity of values measured in two points is inversely proportional to the spatial distance between them. Kriging makes it possible to quantify the spatial autocorrelations between sample points, creating visualisation surfaces according to the statistical properties of the distribution of values. Its ability to compute and assess error and uncertainty helps to quantify the soundness and reliability of the method in specific contexts of study. The kriging method has two precise phases:
1. determination of the spatial structure of the data;
2. creation of estimates, i.e., prediction of the interpolated surface.
The former involves calculating the semivariance (the variation in the similarity of the data with increasing distance between points of the dataset). Its values are represented in a semivariogram, a graph with the distances between sampled points on the horizontal axis and the semivariance on the vertical axis. The resulting semivariogram is independent of the geographical position of the measurements and depends only on the distance between them. The subsequent phase entails the creation of a model (exponential, circular, gaussian) that seeks the best approximation of the distribution of the points in the semivariogram. This is critical for the whole kriging process, since it directly affects interpolation and hence data output. The best approximation of the distribution of the points in the semivariograms was almost always provided by the exponential model. Interpolation by kriging requires the definition of many parameters so that spatial correlations can be studied in the best possible way and approximated with a suitable model. At times, poor correlation between the measurements of a dataset may cause intractable problems, with results that are not always satisfactory. The maps reproduced here were obtained with ArcGis 8.3 and 9, using the specific Geostatistical Analyst extension for processing with the kriging method.