Knowledge Discovery from Disparate Earth Data Sources

Advances in data collection and data storage technologies have made it possible to acquire massive Earth science data sets. In principle, these data sets could be transformed into great scientific discoveries. However, due to the heterogeneous nature and to the scale of the available Earth science data, traditional analysis methods are challenged and much of these data remain largely unexplored. We have developed a general strategy for transforming machine learning algorithms for learning from a single data source into algorithms for learning from disparate, semantically heterogeneous data sources. We believe that our strategy could be adapted and used for data exploration and knowledge discovery from Earth science data.