DATA-DRIVEN HYPOTHESES TO ADDRESS CORAL DISEASE AND REEF HEALTH

Coral reefs once covered shallow-water marine ecosystems across the entire western Atlantic and Caribbean. In the 1970s, reports began of diseases affecting the coral reef inhabitants and, within less than 50 years, there is now a broad, unyielding, and ongoing decimation whose complex etiology remains elusive despite decades of traditional investigation. Recent advances in computer technology have made storage relatively cheap and computation fast, offering scientists access to a scale and diversity of data that, until a few years ago, was infeasible. Access to this data shows promise to answering questions of commensurate size and complexity. Our research examines coral disease using a data-driven approach by assembling, integrating, and analyzing relevant disease-climate data. More specifically, our system draws from repositories storing large oceanographic datasets and coral disease catalogs, containing tens of millions of sea surface temperature records, together with coral species names and diseases for nearly 300 geographic locations in the Atlantic and Caribbean. Our modular system gives scientists the ability to now (1) Study virtually any collection of locations with their respective climate and disease states over any period of time for which data has been recorded; (2) Manually build hypotheses as rules and combinations of rules; (3) Computationally build hypotheses (pattern discovery) as rules and combinations of rules; (4) Produce data that either verifies, falsifies, or demonstrates no discernible relationship with the rules; (5) Search through the data; (6) Bridge across missing data using either several well-known or tailored approaches. Our system requires no ancillary experience and is usable by researchers who are not computer scientists. Substantiating our work are findings consistent with widely accepted hypotheses on thermal stress and coral health. This research forces an understanding of the processes underlying reef degradation, and focuses on resilience-based science to inform on management and the complexity of reef governance. We propose intervention of computational approaches in this resilience-based science initiative to further research on climate/coral disease, and advocate that Big Data can address societal-relevant climate issues.