ABSTRACT : Geoinformation approach to spatial and spatio-temporal data mining in earth sciences

Exploration of geological processes requires studying dynamic and static multidisciplinary measurements. Earthquake catalogues, time series of geo-measurements and aerospace images refer to the dynamic data. Linear structures, stationary geological and geophysical fields refer to the static data.

Geoinformatic approach to analysis of multidisciplinary geo-data is based on the uniform representation of spatial and spatio-temporal properties of geological environment. To do so all types of dynamic data are transformed into 3D grid-based fields with two spatial and one temporal coordinates, all types of stationary data are transformed into 2D grid-based fields. Thus exploration of geological processes is reduced to comprehensive analysis of 2D and 3D grid-based data.

There are the following basic spatial and spatio-temporal analysis techniques in geoinformatics:

1. Visual exploration to discover patterns of stationary and dynamic process.
2. Analytical transformations to extract new spatial and spatio-temporal properties of the process under consideration. For example: transformations of 2D and 3D grid based layers into 2D or 3D grid-based layer (raster filtering and raster operators), transformations of a point layer into 2D or 3D grid-based layer (functions of the earthquake catalogues), transformations of a line layer into 2D grid-based layer (functions of the linear structures), transformations of time series into 3D grid-based layer, transformations of vector layers into a vector layer, and transformations of grid-based layers and a vector layer into the vector layer attributes.

3. Inference of new transformations to forecast and detect spatial objects or spatio-temporal phenomena (e.g. operators for detecting earthquake precursors).

The following examples of spatial and spatio-temporal seismological data mining are considered: the analysis of a seismic network sensitivity (Greece), analysis of noise in the earthquake catalogue (Western Turkey), web-GIS analysis for seismic potential zone delineation (Bulgaria), analysis of earthquake precursors under several characteristics of the earthquake catalogue (Central Asia), analysis of earthquake precursors under multidisciplinary time series (North-East China). The examples demonstrate:

1. spatial and spatio-temporal data mining give effective techniques to comprehensive analysis of geological process and to detection of earthquake precursors,
2. integration of multidisciplinary data allows to find out more reliable decisions,
3. spatio-temporal analysis of earthquake catalogues allows to estimate sensitivity of a seismic network and to reveal sources of the noise.