Efficient, graphics hardware-accelerated, atmospheric rendering and multi-field weather visualization

Kirk Jason Riley, Purdue University

Abstract

The evaluation of weather data from measurement and simulation remains an essential process for the prediction of weather. Therefore., weather visualization is critical to the efficient processing of the vast quantity of measured and simulated meteorological data. Atmospheric rendering is a notoriously difficult problem because the number of paths light may take through the atmosphere grows exponentially with the number of scattering events. This work demonstrates that efficient three-dimensional multi-field weather visualization and atmospheric rendering is possible through the application of modern graphics hardware. Although weather processes are three-dimensional in nature, they are often simplified to two-dimensional slabs, particularly at the synoptic scale. This results in critical information loss that otherwise could be used to better diagnose the meteorological event. Currently, meteorologists require field spotters to evaluate the full three dimensional event that is the formation of a severe storm. In order to achieve the same benefits afforded the weather spotter in prediction, to create more compelling training examples, and to improve model evaluation, a realistic weather visualization system is developed. Atmospheric rendering has more applications than weather visualization. Thus, the hardware-accelerated algorithms developed for approximating light scattering in the sky and interacting media provide a solution for efficient rendering of advanced optical effects from atmospheric particles in computer graphics. Additionally, because weather simulation and measurement require a wide variety of grid structures, a data model and rendering architecture are developed to accommodate the wide domain of data sources. This work provides a system for the simultaneous visualization of non-visual volumetric quantities, with the visual quantities, to provide a complete package for the evaluation of meteorological data. In order to maintain rendering efficiency, and thus, efficient exploration of the data, this implementation is designed for modern graphics hardware, taking full advantage of the flexible parallel-processing capabilities it provides.

Degree

Ph.D.

Advisors

Ebert, Purdue University.

Subject Area

Electrical engineering

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS