Page 25 - Geochemical Anomaly and Mineral Prospectivity Mapping in GIS

20 Chapter 1

computational functions that express empirical relationships of the predictor variables
with the target variable. Knowledge-driven empirical methods, on the one hand, usually
employ logical functions (e.g., AND and/or OR operators; see Chapter 7) for sequential
integration of predictor maps through so-called inference networks (see Chapter 7). An
inference network depicts knowledge about the inter-play of processes represented by
individual predictor maps. Data-driven empirical methods, on the other hand, usually
employ mathematical functions for simultaneous integration of predictor maps
regardless of knowledge about the interactions of processes depicted by each predictor
map (see Chapter 8). Some data-driven methods employ functions representing logical
operations (e.g., Dempster’s (1968) rule of combining of evidential belief functions) for
sequential integration of predictor maps through an inference engine. Likewise, some
knowledge-driven methods apply mathematical functions for simultaneous integration of
predictor maps.
The way by which data are integrated in a GIS is controlled precisely by the spatial
topology and linkage of data at every location to their map coordinates (see Chapter 2),
although the topology and map coordinates do not directly play a role in a computational
function applied to integrate data. Note that topology and map coordinates also provide
precise control in data analysis involving at least one map. The choice of a spatial data
model (vector or raster; see Chapter 2) could affect, however, computation during data
integration. Although a vector model represents geometry of geo-objects better than a
raster model, data integration using raster maps is faster and more precise than using
vector maps (Brown et al., 2005). That is because raster maps represent continuous
variables, such as element concentrations, better than vector maps. In addition, because
of the one-to-one coordination of pixels referring to the same location in every raster
map, building of topology of so-called unique conditions geo-objects (see Chapter 2)
during data integration with raster maps is simpler than during data integration with
vector maps (Mineter, 2003). GIS technology is still advancing, however, toward
achieving routine capability to integrate data in both vector and raster maps (Winter,
1998; Winter and Frank, 2000). Vector maps are nevertheless preferable to raster maps
in visualisation of many types of spatial data or geo-information.

Visualisation of spatial data or geo-information
Displaying spatial data or geo-information on-screen is perhaps the most exploited
functionality of a GIS. Exploration geochemists usually ‘eye-ball’ the data for patterns
of interest before actually performing quantitative analysis of the data. Graphical,
especially interactive or dynamic, display of spatial data or geo-information is especially
useful in the early stages of predictive modeling of geochemical anomalies (Haslett et
al., 1991). Most GIS software packages contain, however, only a few dynamic graphical
display functionalities. Visualisation of spatial data in a GIS is also useful in selective
query, retrieval and analysis of certain data in a database (e.g., Harris et al., 1999).
Finally, a GIS provides capability for mapping (i.e., preparing analogue maps in contrast
to modeling). More than two decades ago, Howarth (1983a) has explained the types of
useful geochemical maps and the techniques for preparing such maps. Recently,