Study On Spatiotemporal Visualization Of Geoscience Observation Data

Using information visualization or visual analytics to perform traditional scientific research has become a common practice, which can expand scientific research scope and mine the potential data values that closely related to people’s daily activities. Information visualization of scientific observation data becomes a necessity for advancing scientific discovery and improving the quality of life. Although many classical techniques and applications have been proposed, they have several drawbacks in the aspects of(1) data analysis capability,(2) representing multiple dimensions,(3) generality and scalability,(4) cognition impact on viewers and(5) evaluation. To further advance research in these aspects, this dissertation proposes a suite of visualization techniques that are generic and scalable. Major contributions are summarized as follow:1. A visualization framework consisting of three views for the analysis of station observation data. The main view can simultaneously show the high level characteristics of the entire dataset in terms of the space, time and attribute. As the complement, the two other views are used to analyze time trend and detect anomaly cases based on a triangle metaphor and ScatterPlot dimension reduction. The framework integrates three types of interactive operations, e.g. showing the interrelated information, adjusting the display region of the map, and interactively generating a Fisheye view in polar coordinates based on the Focus+Context model. The three views are highly interrelated, and the analysis results of any view can be used as the input to the two other views. Furthermore, this framework can cover long temporal intervals and accommodate any number of stations throughout the world. It can also be used to analyze geo-related statistical data of other domains, such as economy and healthcare.2. Case studies on multiple datasets of different domains, such as meteorology, oceanography, environment protection, traffic and economy, as well as the evaluation experiment based on eye-tracking. By setting Areas of Interest(AOIs), and analyzing the transitions between the AOIs and multiple exploration metrics, we can quantificationally evaluate the proposed approach. The evaluation results verify the effectiveness of the approach and offer important references for applying the approach to analyze more datasets in different fields. Furthermore, an integrated visualization platform of observational data based on the Common Data Model has been proposed. The platform provides numerous basic visualization components and a unified interface for accessing all types of geoscience observation data to facilitate the design of new visualization techniques and systems.3. A 3D trajectory exploration approach to the observational data generated from a public transportation system. This approach is based on viewpoint selection algorithms and Google Earth, which consists of a workflow, two trajectory visualization techniques and three viewpoint selection algorithms. To be imported into Google Earth, the workflow utilizes KML as the description language of the generated visualizations and viewpoints. Combined with the environment information derived from the satellite images of Google Earth, analysts can conveniently perform exploration and analysis tasks. We improve the existing trajectory wall, and investigate two novel trajectory visualization techniques specialized in displaying the single route and multiple routes. To quickly explore detail, overview and dynamic variation of trajectory data, and reduce the impacts of the 3D visualization on the cognition effects of users, three types of viewpoints and the corresponding generation algorithms are investigated. The user can quickly explore and understand the visualization by sequentially visiting each viewpoint.