Spatio-temporal Aggregates over Streaming Geospatial Image Data

Geospatial image data obtained by satellites and aircraft are increasingly important to a wide range of applications, such as disaster management, climatology, and environmental monitoring. Spatio-temporal aggregations are some of the most important operations over such data. Because of the size of the data and the speed at which it is generated, computing such aggregates over geospatial image data is extremely demanding. Due to the special characteristics of the data, existing spatio-temporal aggregation models and evaluation approaches are not suitable for computing aggregates over such data. In this thesis, we analyze the characteristics of streaming geospatial image data and outline the key challenges of spatio-temporal aggregate computations. By showing that traditional aggregation models do not always provide an accurate view of the data, we propose new spatio-temporal aggregation models that infuse a more meaningful semantics into a query. More importantly, our experiments show that existing approaches do not evaluate these queries efficiently. Existing approaches do not take advantage of the characteristics of a raster image, such as the rectangular gridded structures of raster images, or similar or identical point values among neighboring points. As a result, these approaches usually have (1) high storage requirements, and (2) high construction or maintenance costs so that they can not catch up with the data arrival rate. In this research, we aim to design IO-efficient and space-efficient evaluation approaches for spatio-temporal aggregate queries. Several novel data struc-