The south-west of Western Australia is a region of significant cropland farming and viticulture, worth $3,000-5,000 million annually. It is also a globally-recognised biodiversity hotspot and home to some of Australia’s most iconic forests. Everything the south-west is notable for is sensitive to changes in rainfall and temperature.
Global climate models (GCMs) are used to simulate past, current and future climate across the entire globe, at a spatial resolution of 100–250 km. But to better understand how climate change impacts manifest at local scales Dr Jatin Kala, an atmospheric scientist at Murdoch University, has been using Pawsey supercomputing and data storage to develop climate models at a 5 km resolution. These can be used to support local decision making in the south-west of Western Australia.
The south-west of Western Australia, a biodiversity hotspot and a region of significant agricultural production, is known to be vulnerable to climate change, with global models showing significant trends towards a warmer and drier climate across the entire south-west. However, GCMs have a limited ability to assess future climate changes at a more detailed local scale. To understand exactly how these changes will impact on the local environment, industries and communities in this region, and to plan appropriate response strategies, a more detailed regional climate projection is needed.
Dr Jatin Kala, an atmospheric scientist at Murdoch University explains: “The south-west of Western Australia is only represented by three or four points in the GCM, which is like saying the climate impacts on Perth and Busselton are exactly the same. But our climate and weather impacts are obviously more locally variable than that.”
Regionally-relevant information is needed to assess how climate change will impact on local physical, ecological and socio-economic systems, and to inform policy and planning responses. The challenge is to create higher resolution projections of climate change, to inform adaption planning and impact assessment studies across agriculture, forestry, conservation and urban planning.
Kala’s research team is developing regional climate models, which can dynamically downscale GCMs to a resolution of 5 km, and incorporate the finer-scale influences of topography, land cover, and mesoscale weather systems like sea breezes and winter cold fronts. This requires the use of both high-performance computing and data storage resources as the models become exponentially more detailed and complex.
Pawsey has been involved in climate modelling research since before it was even Pawsey. Kala’s early research was carried out on an iVEC computer with a total of only 192 individual processors, and his climate simulations were always limited by the computational power and data storage available. But using Pawsey’s current facilities, his regional climate projection models run using several hundred cores on Magnus (which has a total of 35,712 cores) and produce hundreds of terabytes of data.
“We need the supercomputing facilities to run the models, we need the data storage, and then to analyse all that data, we need to run even more code. This sort of research just isn’t possible without Pawsey,” says Kala.
“We can now do higher resolution projections over longer periods of time. We can evaluate how well our models simulate detailed historical weather records across a region over the past 30 years, and then run them forwards in time with greater accuracy. It is making our regional climate models much more useful.”
Kala’s research group is continuously refining and improving high-resolution climate projections for the south-west of Western Australia. These models are being used to explore climate change in the region, and specific issues like the impact on cereal crop production, local viticulture, or bushfire behaviour.
His datasets are currently available to other researchers on request, but together with other climate researchers using NCI, Pawsey’s partner institution in Canberra, Kala is working towards making regional climate model outputs covering all of Australia more generally available to the community.
With modern supercomputing, providing state-of-the-art climate prediction modelling at a finer resolution for local impact assessment is becoming a reality.