Strategies to improve information extraction from multivariate geophysical data suites

Abstract

This paper discusses strategies for improving interpretation outcomes from large, multivariate data suites. It is shown that a systematic approach using simple spatial analysis techniques implemented in a geographic information system reveals information that is not readily evident using conventional, manual interpretation procedures. A combination of simple exploratory analysis, principal components analysis and unsupervised classification is used to represent the landscape as a series of geophysical/terrain units. A comparison of these classified units with the distribution of surface salinity reveals that the classification is not just an arbitrary division of the landscape, but one clearly related to the properties of the landscape relevant to salinity. This result is important for two reasons. First, it demonstrates a fast, systematic, robust and repeatable process for successfully extracting information from the multivariate data suite. This result would not be achievable using the conventional interpretation approach. Second, it provides a useful summary and clear representation of multivariate geophysical data, which when applied appropriately could significantly improve the utility of the data for land management planning.