Uncertainty management of a hydrogeological data set in a greek lignite basin, using BME

Abstract

The occurrence of chemical pollutants in ground water is an issue of considerable interest. In the case of Ptolemais lignite opencast mining area in Greece, ammonium, nitrites, nitrates, iron total and total manganese concentrations, as well as various other elements have been monitored since the early 2000s through a borehole network. The continuous, though, alteration of the surface topography due to the intensive mining works, limits the life span of the water boreholes and results to irregular spatiotemporal distribution of the samples. Regarding the problem of mapping the water contamination, the coarse and irregular sampling pattern, combined with absence of seasonal variations and temporal trends, does not facilitate spatiotemporal processing of the data. On the other hand, a mere spatial analysis requires the attribution of the whole set of monitored values for each borehole, to a single point in space. The objective of this work is to develop a methodology to cope with the problem of uncertainty caused due to the above assumptions. The proposed solution is based on the consideration of interval data. The Bayesian maximum entropy (BME) theory (Christakos, Modern spatiotemporal geostatistics. Oxford University Press, New York, 2000) is an essential component of this methodology. The main reason for this is that it is the only theory in the framework of Geostatistics that offers powerful tools to merge the uncertainty sources with the rest of conventional measurements. This ability is a result of its generalized view on the problem of estimation that focuses on the knowledge of a natural variable and not the variable itself. The application of the proposed methodology led to sharper posterior distributions, an indication of the increased certainty in estimated values which is induced by the broader utilization of data.