It’s National Science Week. This year’s theme is Deep Blue: innovations for the future of our oceans, and we are showcasing some of the exciting research happening at the College in this field.
Ever wondered just how accurate our weather forecasts are? What if they could be refined to offer even more insights on the world around us?
A team of researchers from The Australian National University (ANU) are looking at ways to improve Australia’s weather and climate predictions using the latest advanced computing techniques.
National agencies such as Australia’s Bureau of Meteorology (BoM) use operational codes to make long-range weather forecasts. However, the climate and weather models that we use are somewhat dated, and the current codes consume huge amounts of computer time.
Associate Professor Peter Strazdins, from the ANU College of Engineering and Computer Science, is leading a project developing techniques to port and adapt these legacy operational codes to modern supercomputers.
“I have always found the computer simulation of weather phenomena fascinating. I spent part of my study leave in 2010 at BoM, investigating the performance of their newly-adopted atmosphere code,” said Associate Professor Strazdins.
“Operational weather codes are extremely complex and require vast amounts of calculations to produce accurate results. Their codes were mostly developed decades ago, before modern-day supercomputers came into existence.”
Today’s supercomputers are truly leaps ahead of those past. They now contain processing nodes augmented with powerful accelerators, such as NVIDIA graphics processing units (GPUs).
The idea to apply this technology to weather science stemmed from knowledge sharing conversations between the industry, government, and university sectors.
“In 2015, I met some colleagues from NVIDIA and asked what kind of applications they would like more research on. Their answer was weather and climate. Shortly after, I met BoM Supercomputer Programme Director, Tim Pugh, and asked whether he’d be interested in seeing BoM's operational codes ported to GPUs. The answer was a resounding ‘yes’!
“It took some time, but my ANU Honours student Fan Yu achieved a superb port of HYSPLIT, a particle-tracking application. This is one of the most widely-used codes in the world. It would have been in heavy use predicting smoke dispersion in the recent Australian bushfire season,” said Associate Professor Strazdins.
BoM have been evaluating the HYSPLIT work from the ANU for over two years now, and are preparing it for deployment in the 2021 stage of the international Volcanic Ash Modelling Project.
The Australian-based agency are also seeking to integrate this work into the main HYSPLIT code base in the United States of America, which would speed up pollutant-tracking simulations worldwide.
While the project has mostly focused on atmosphere applications to date, it’s clear the techniques could be used to improve many other areas of weather and environmental science.
“Over the last year, we have also been working on a code called access-regrid. This code can resize a 3D grid of data, representing the atmosphere or ocean. This allows the data produced from a high-resolution simulation – like a weather forecast – to be used to create initial data for long-range simulations like a climate forecast,” said Associate Professor Strazdins.
“Performance improvements to the access-regrid code will be of great benefit to our weather outlooks. It will enable data to be more efficiently transferred between different kinds of simulations.”
From the skies to our oceans, Associate Professor Strazdins believes computer scientists can play an important role contributing to our understanding of our natural environment.
“Accurate and timely weather and climate forecasts are particularly important for a successful maritime industry. More generally, high resolution computer simulations will be useful in many areas of Australia’s Blue Economy. Computer scientists can greatly improve the speed of calculation in simulations by porting the codes to contemporary supercomputers. And this can permit greater accuracy.”
To discover events this August celebrating science and technology, and more stories on Deep Blue, see: How we’re celebrating National Science Week.