A Spatio-Temporal Mining Approach for Enhancing Satellite Data Availability: A Case Study on Blue Green Algae

Satellite imagery provides geospatial data, generally terabytes in size. Due to their cost effectiveness and scalability, they are used in various large scale applications related to ecological monitoring. However, satellite data is prone to problems of data incompleteness owing to a number of issues such as cloud cover, fog, etc. In this paper, we conduct a detailed study on the 10 years of satellite data using Blue-Green Algae as a case study. Blue-Green Algae (BGA) are a toxic phytoplankton which are now a worldwide phenomena and a concern to various public authorities. We first illustrate the data incompleteness problem in our dataset with respect to BGA, and then formulate this problem using time-series spatio-temporal data mining approach that harnesses similarities among lakes with respect to BGA manifestation. We evaluate our approach through experimental studies from 99 lakes in the southeastern United States. Our experiment shows that our approach is an effective method to address data incompleteness in the BGA domain.