The OMEGA system from SAIC is a powerful tool for weather forecasting and atmospheric dispersion modeling. The output from OMEGA is a large, multi-dimensional dataset which is difficult to interpret and visualize.
Delta HPC and SAIC worked together to develop powerful visualization capabilities, which greatly simplify interpretation and understanding of OMEGA predictions. Some of the challenges included the massive size of OMEGA output, the constantly changing adaptive mesh algorithms used by OMEGA, and the large scale geographic areas typically covered.
In just a few weeks, Delta HPC engineers developed a custom application for displaying OMEGA data in the CAVE. This application handled the size and complexity of OMEGA data, provided an easy to use interface, and presented the information in a way that both experts and ordinary people can understand.
These visualization tools were applied to data from hurricane Floyd (1999), and to the smoke dispersion from the Indonesian fires of 2002. These results were publicly shown at the SuperComputing 2002 conference in Baltimore.
OMEGA, the Operational Multiscale Environment model with Grid Adaptivity, was designed as an operational tool to support weather and hazardous atmospheric dispersion forecasting. Utilizing an adaptive unstructured grid technique developed by the aerospace community, and emphasizing the air surface interactions and boundary layer physics that drives the near-surface weather and dispersion problem, OMEGA provides a significant advance in the state-of-the-art.
OMEGA started as a regional model with static grid adaptation designed for vector/serial architectures. The first enhancement of the model was to invoke autotasking to perform loop-level parallelization of the code. This was followed by the implementation of dynamic adaptation, first in the vector/serial version of the model and later in the domain-decomposition parallel version. More recently we have started to perform global simulations using OMEGA with static adaptation in the vector/serial version and are currently completing the OMEGA capability cube with the addition of dynamic adaptation to the global model in the vector/serial and, finally, the parallel version.
OMEGA includes the embedded Atmospheric Dispersion Model (ADM). ADM consists of both Eulerian (grid based) and Lagrangian (grid free) transport schemes. In addition, the Lagrangian transport scheme includes both particle and Gaussian puff algorithms. XGRID allows the graphical analysis of both modes.
In September, 1997, illegal biomass burning in Sumatra and Kalimantan, Indonesia set fire to peat underlying the region. The resulting peat fires burned for week
s, emitting a great deal of smoke resulting in operational and health effects.
On Friday, September 26, 1997, twin disasters, both associated with the reduced visibility caused by the smoke, occurred. At 1:50 pm, Garuda flight GA-152, an Airbus A-300, descending for landing at Medan collided with terrain 32 km west of Polonia Airport. Less than 12 hours later, the Vikraman cargo ship and the Mount 1 supertanker collided in the Strait of Malacca about 11 p.m. near the central Malaysian town of Port Dickson, about 50 miles south of Kuala Lumpur.
On Monday, September 29, 1997, maritime insurance through the Strait of Malacca was lifted. On Tuesday, September 30, SAIC was contacted about simulating the conditions leading up to these events.
OMEGA can adapt dynamically to changing conditions in the forecast. In this simulation, we show the initial grid for a simulation of Hurricane Floyd. The simulation and the grid are updated each hour of adaptation for 65 hours. It is important to note that the grid does not move, but rather is altered by a chain of refinement and coarsening steps.
This dynamic adaptive grid maximizes computational efficiency – a critical factor for phenomena as complex as a hurricane. It resolves important processes, and does not require a priori knowledge.
Initial efforts used existing visualization tools, such as Cave5D. The OMEGA data presents unique challenges – it uses a prismatic grid structure, rather than a Cartesian grid. The grid and grid size typically changes for each time step in the simulation. The results are time-varying, and contain multiple types of information (temperature, pressure, wind direction and velocity, clouds, particulate levels, etc.)
Delta HPC’s engineers chose to develop a visualization application built around OMEGA’s data structure. This approach adapted to all of the unique requirements and characteristics of working with OMEGA. The application fully utilized the visualization power of the CAVE, and can support other environments.
Existing tools and infrastructure were used where possible – user interface tools, application frameworks, navigation and manipulation methodology, etc. As a result of combining existing frameworks with OMEGA specific data, the first prototype was displaying time-varying OMEGA data in the CAVE in just a few days.
Further refinement over the period of a few weeks provided a flexible application that allows users to fly around and visualize data from different angles, incorporate GIS (Geographic Information Systems) data, show realistic clouds and smoke plumes, and display different types of data.
For Further Information