Service Communautaire d'Information sur la Recherche et le Développement - CORDIS

Final Activity Report Summary - SPATSTAT (Development of Spartan spatial random field models for geostatistical applications)

This project focussed on the development of innovative methods based on Spartan spatial random fields (SSRFs) for the statistical analysis of spatial information.

The research that was conducted during the project led to flexible, accurate, reliable and computationally fast statistical models of spatial variability. These models could be used for analysing and mapping environmental data with either Gaussian or non-Gaussian distributions, e.g. pollutant concentrations in environmental media, climatic variables, geological structures of economic interest including oil reservoirs and ore bodies, geographic distribution of environmental health factors and disease across Europe. The proposed Spartan models were frugal in terms of parameter requirements and admitted updating in case new information became available. In addition, the Spartan model parameters had a clear physical meaning and interpretation.

The developed approaches were interdisciplinary in nature, since the project incorporated spatial statistics, statistical physics and environmental applications. The outcomes of this research had applications in fields such as environmental science and engineering, geographic information systems (GIS), oil reservoir engineering, environmental health and image analysis that required analysis and extraction of information from spatially distributed samples. The proposed Spartan spatial random field methods were suitable for use in 'automatic mapping systems'. Such systems offered significant advantages for the visualisation and practical use of spatial information, e.g. for decision making processes.

The methods developed during the project could lead to efficient mapping tools that would require only minimum information from the users. This would help users who might not be experts in geostatistics, and thus unable to make informed choices. In fact, it was widely recognised that a major obstacle to the widespread use of spatial analysis was the complexity of the decisions, e.g. choice of variogram model, non-systematic analysis of anisotropy, selection of the appropriate kriging methodology etc., required by the users of geostatistical software. The innovative analytical techniques that we developed for spatial analysis could also lead to improved risk assessment tools in fields like the environment and public health and to more accurate GIS advanced functions for managing geographical information. The methodological developments were accompanied by the development of computer code, in Matlab programming environment, towards which the experienced researchers and the coordinator contributed.

This Marie Curie project, in addition to its considerable scientific output, also contributed to the integration of the European research area. The project attracted scientific talents from Europe, i.e. France and Slovakia, and the United States of America (North Carolina) to the host, i.e. to the Technical University of Crete (TUC). The host faces considerable difficulties in recruiting qualified graduate students and post-doctoral researchers due to its remote island location, and this Marie Curie project greatly helped in this direction. The project also helped TUC to further establish its reputation as a competitive research University in the Mediterranean region, thanks to the dissemination of scientific results that were obtained during this project. Graduate students and colleagues at TUC also benefited from the scientific interaction with the incoming researchers and enjoyed the short course in Bayesian statistics delivered by an international expert in the field. The Marie Curie fellows of this project, including two experienced researchers and one more experienced researcher, were engaged, by the time of this initiative completion, in continuing collaborations with TUC beyond the end of their fellowship. The development of advanced capabilities in modelling and simulation of scattered spatial data would finally be useful in other research activities in the Department of Mineral Resources Engineering, such as environmental monitoring, analysis of remote sensing data, geostatistical studies, desertification effects on the island of Crete, and in activities in fields of potential interest, including oil reservoir simulation, subsurface-flow and porous-media modelling.

Reported by