| A study of iceberg scour data suggests that a combination of GIS, EDA and inductive modelling provides a comprehensive approach to the understanding of the spatial relationships between iceberg scour variables and seabed information. The objective of this study is to provide scientists with a deductive and inductive tool whereby they can analyze regional trends associated with iceberg scour data archived in the 'East Coast Iceberg Scour Database'.;GIS provides a deductive tool which permits the integration and spatial analysis of point, line and area data affiliated with iceberg scour variables and seabed information. EDA provides resistant summary statistics that are not influenced by extreme value ranges in the data. The results of EDA can be presented graphically as stem-and-leaf and box-and-whisker plots or numerically in the form of letter summaries. An integration of the EDA results and the mapping capabilities of a GIS provide a method whereby the spatial distribution of extreme events and resistant summary statistics can be visually examined.;Inductive modelling through classification and regression tree algorithms, permits simultaneous multivariate analysis of nominal, ordinal and interval iceberg scour and seabed data. This method of data analysis is not constrained by the limitations associated with the multivariate linear models. The results of inductive modelling are expressed as statistically significant decision trees and natural language 'IF ... THEN' rules. These rules, which describe multivariate relationships are used to define the conditions for mapping.;The methodology was used to examine the relationships between iceberg scour characteristics and seabed information. The regions examined were the Baffin Shelf, Saglek Bank and the Grand Bank. The results of the EDA and inductive modelling indicate that there exists between and within regional variations. Furthermore, the results demonstrated that extreme values, for iceberg scour variables, are present in all three areas. The results of the multivariate analysis suggest that simple bivariate relationships cannot be used to explain regional variations in iceberg scour frequencies and dimensions. Finally, the results demonstrate that the variables in the 'East Coast Iceberg Scour Database' should be evaluated using the more resistant statistics available in EDA techniques. |
