Representation and Analysis of Large Scale Atmospheric and Earth Observation Data

Abstract

For every computational activity, whether technical or non-technical, and in every area of knowledge creation, data representation has essentially increased. In order to efficiently understand and communicate knowledge, numerous algorithms are developed for the modeling of this data and its analysis. In a wide range of mesh processing applications in computer graphics, geometric modeling, scientific visualization, geographical data science, and finite element analysis, efficient mesh data structures play a crucial role. Considering the ease with which simple problems can be modeled on small, low-dimensional meshes, interesting phenomena may only be observed on much bigger meshes and in higher dimensions. Flexibility is needed in order to handle increasingly complex meshes, such as those defined by irregularly connected heterogeneous and/or multidimensional cell types discretizing areas with complex topology. The goal of this thesis is to examine and evaluate vast earth observation data that has been presented as mesh-like simplicial complexes. Additionally, it seeks to investigate and develop a novel data representation for simplicial complexes that fits smoothly with distributed computing pipelines and parallel processing environments. In this thesis, an efficient approach for decomposing and distributing mesh data across parallel nodes is identified and put into practice.