Analyzing the Influence of African Dust Storms on the Incidence of Coral Disease in the Caribbean Sea using Remote Sensing and Association Rule Data Mining
Date
2014-06-02
Authors
Hunter, Heather E
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
It is estimated that approximately 11% of the historical extent of coral reefs is already lost, and an additional 16% severely damaged (Gardner et al., 2003). In the Caribbean Sea, alone, approximately 90-95% of a major reef-building coral suffered mass mortalities in the last 40 years (Garrison et al., 2003), contributing to a dramatic decrease in coral cover across the globe (Selig et al., 2006), primarily caused by increased sea surface temperatures and the subsequent thermal stress and bleaching. In conjunction with the observed increase in coral mortalities, the Caribbean region has also experienced an increase in the number of coral disease cases, hypothesized to be caused by pathogens carried by airborne African dust. It is also hypothesized that these pathogens are opportunistic; that is, they take advantage of the corals when the corals suffer physiological stress due to the negative influence of environmental conditions (Rypien, 2008). While threats attributed to humans can be managed by policy changes or behavior xii modification, without a better understanding of the relationships between the environment, disease, and coral reefs, the means by which to mitigate the corals’ response to disease remain indeterminate. Given that coral reefs in this region play a vital role in the region’s economy, as well as in the marine ecosystem, this is a worthy challenge to pursue. Satellite remote sensing is well-suited to enable understanding, at vast temporal and spatial scales, of the local Caribbean climate, how it varies, and how African dust storms contribute to further variations that may cause coral disease to thrive. While the data available from satellite sensors can measure a breadth of geophysical parameters and provide an unprecedented wealth of information about geophysical processes, the data sets are vast, and it is increasingly necessary to use computational techniques to enable the automatic extraction of useful information and patterns. Data mining of association rules is a technique that combines methods from machine learning and statistics to uncover interesting patterns in large databases. The goal of this thesis is to show that the use of an association rule data mining algorithm on a combination of satellite remote sensing and in-situ data can produce meaningful, qualitative results that show correlations between the Caribbean climate, African dust storms, and observations of coral disease.
Description
Keywords
Coral reefs, African dust storms, Remote sensing, Coral disease, Data mining