Local Temporal Compression For (Globally) Evolving Spatial Surfaces

The advances in the Internet of Things (IoT) paradigm have enabled generation of large volumes of data from multiple domains, capturing the evolution of various physical and social phenomena of interest. One of the consequences of such enormous data generation is that it needs to be stored, processed and queried - along with having the answers presented in an intuitive manner. A number of techniques have been proposed to alleviate the impact of the sheer volume of the data on the storage and processing overheads, along with bandwidth consumption - and, among them, the most dominant is compression. In this paper, we consider a setting in which multiple geographically dispersed data sources are generating data streams - however, the values from the discrete locations are used to construct a representation of continuous (time-evolving) surface. We have used different compression techniques to reduce the size of the raw measurements in each location, and we analyzed the impact of the compression on the quality of approximating the evolution of the shapes corresponding to a particular phenomenon. Specifically, we use the data from discrete locations to construct a TIN (triangulated irregular networks), which evolves over time as the measurements in each locations change. To analyze the global impact of the different compression techniques that are applied locally, we used different surface distance functions between raw-data TINs and compressed data TINs. We provide detailed discussions based on our experimental observations regarding the corresponding (compression method, distance function) pairs.