Ph.D. THESIS PROPOSALTITLE: From spatio-temporal data to functional weighted networks: methods and applications in climate science, neuroscience and ecology.Ilias FountalisSchool of Computer ScienceCollege of ComputingGeorgia Institute of TechnologyDate: Wednesday, December 3, 2014Time: 11:00 AM - 1:00 PMLocation: KACB 3100Committee:----------Prof. Constantine Dovrolis, School of Computer Science, GeorgiaTech (Advisor)Prof. Mostafa H. Ammar, School of Computer Science, GeorgiaTechProf. Annalisa Bracco, Earth and Atmospheric Sciences Department, GeorgiaTechAssistant Prof. Bistra Dilkina,  School of Computational Science and Engineering, GeorgiaTechAssociate Prof. Shella Keilholz, Wallace H. Coulter Department of Biomedical Engineering, GeorgiaTech and Emory University School of MedicineProf. Athanasios Nenes, Earth and Atmospheric Sciences Department, GeorgiaTechAbstract:----------There is an abundance of spatio-temporal data today from diverse complex systems such as the Earth's climate, the human brain, or the mobility patterns of migratory species. By analyzing such data, scientists are able to discover the key modules of the corresponding system, and to investigate their dynamics and inter-dependencies.Spatio-temporal data are typically embedded in a two- or three-dimensional grid, and the dynamics of each grid cell are represented by a time-series. Common computational analysis methods for such data include standard time series analysis, spatial clustering, and principal/independent component analysis. These techniques, although valuable in specific contexts, are not able to directly identify the latent functional components of the system and how these components interact with each other. This objective can be met more naturally with a framework that is based on network analysis.The emerging field of network analysis incorporates a broad range of models, metrics and algorithms to study complex nonlinear dynamical systems; its main premise is that the underlying topology or network structure of a system has a strong impact on its dynamics and evolution.We propose a novel network-based analysis framework for the study of spatio-temporal data. First, we cluster grid-cells into "areas", defined as spatially coherent regions that are highly homogeneous in terms of dynamics. The proposed algorithm identifies a parsimonious set of latent functional components, and it relies on a single parameter that is set based on a target statistical significance level. In a second step, we identify edges between areas. The strength of the edge between two areas is given by the covariance of their cumulative anomaly time series. Each edge is also characterized by the lag at which the cross-correlation between the two areas is maximum, in absolute sense.The proposed framework has been applied successfully in Climate Science to evaluate state-of-the-art climate models and to assess their performance. Further, we have investigated future projections of these models' trajectories under increased greenhouse gas emission scenarios. We are going to also apply the proposed method on functional MRI data to construct dynamic functional brain networks. Finally, we will apply the proposed framework in the context of Ecology, to investigate bird migration patterns.