lloydt@watson.ibm.com
The visualization and analysis of large scientific data represents a very challenging task, especially in the earth, space and environmental sciences. The myriad of earth science data, for example, from observation (in situ and remotely sensed) and computation (simulations and empirical models) are complex and very large in volume. These data are multidimensional (typically two or three spatial dimensions, perhaps one or more non-spatial dimensions, e.g., energy), dynamic (time-varying in data and dimensionality) and consist of many parameters. There is enormous variation in the instrumentation used to observe the Earth that have consequences in the data geometries, sampling and error characteristics. Such variation is often compounded by inconsistencies in the data gathering process, especially for instruments that perform long-term monitoring of the Earth. These measurements each relate some aspect of the physical phenomena under observation. Typically they must be combined in order to glean some knowledge of the data. Furthermore, they are often used in conjunction with simulations to verify theory or as initial or boundary conditions for empiricial models. A long term goal of such on-going activities as well as planned data acquisition and computational efforts is to view the Earth as an integrated system. This would merge and define the interactions between the near-space environment, the atmosphere, the oceans, the land (both surface and subsurface), etc. An additional aspect of such work is the evaluation of the environmental effects of anthropomorphic activities. Greater cognizance of data characteristics and handling of the diversity of earth, space and environmental data can lead to effective solutions to their visualization and analysis. In addition, traditional visualization techniques have been quite limited in their utility in these disciplines. To illustrate the complexity of this challenge and viable solutions, consider some example problems using real data and real science. The first example concerns the analysis of spacecraft observations of global ozone, which are useful in understanding ozone depletion. An approach called correlative visualization, which is a set of methods to examine disparate data simultaneously, is applied to these observations as well as atmospheric dynamics data. For earth, space and environmental sciences applications, cartography must be introduced, which are methods of creating maps of the Earth.
Cartography is an ancient art and science of methods to project -- mathematically transform all or part of the surface of a sphere (e.g., the earth) onto a two-dimensional, flat surface or plane. The process of map projection introduces distortions of the data and/or its geometry. The choice of a specific projection method in visualization is very important for the proper communications of information. It is very much dependent on the visualization task (e.g., exploration, analysis, presentation, decision support, etc.). Too often, a very popular projection, such as Mercator, or a simple rectilinear projection is employed without knowledge of the resultant distortion of the visualized data. A brief survey of common projection methods is provided.
Observations made by the Total Ozone Mapping Spectrometer (TOMS) aboard NASA's Nimbus-7 spacecraft have been critical to the study of stratospheric ozone. Direct analysis of the TOMS data yields important information on the morphology of the annual austral depletion region. However, when these data are visually correlated with other relevant atmospheric data (e.g., objective analyses of temperature, geopotential heights, winds), information about the underlying diurnal atmospheric dynamics of the stratospheric polar vortex and potential contributions from the upper troposphere can be gleaned. This includes the formation and breakup of the depletion region each Antarctic spring. These data require care in their presentation so that artifacts due to the visualization process are not introduced and erroneously interpreted as features in the data. The provided form of these data is ill-suited for the study of such phenomena that occur continuously over a nominally spherical surface (i.e., it tears the data). In addition, they are not uniformly available for the entire earth or at least spatial regimes being examined. Each of the data sets being examined are generally not geographically coregistered and are defined on differing geometric structures. These characteristics require non-traditional techniques for visual correlation based upon registration of multiple data sets of disparate structure with cartographic warping of regular and irregular geometries. Such an approach does not introduce interpolation and its artifacts into the registration or realization process. It is also independent of the choice of realization technique, and hence, provides a framework for experimenting with different visualization strategies. As a result, the fidelity of the original data sets is preserved in a coordinate system suitable for three-dimensional, dynamic presentation and examination of upper atmospheric phenomena.