Scientists are creating a digital twin of our planet that would be simulated in unprecedented detail to analyse the climate system in a way that has never been possible before
Switch on the news in any television channel, and you would be hit with shocking images of heatwaves across the US; wildfires in Europe; devastating floods and landslides in India with roads, houses, cars being swept away by torrential downpours and swelling rivers. In different parts of the world, El Nino is causing drought and deluge – disrupting economies and ruining livelihoods. Rising sea surface temperatures are triggering cyclones packing far greater power than ever imagined. According to the International Monetary Fund, climate change could shave 11%-14% off global economic output by 2050 compared with growth levels without climate change, and could push an additional 100 million people into poverty by 2030 due to food insecurity per World Bank. This has set off an urgent need to forecast climate with a greater accuracy that could better prepare us for risk mitigation.
Scientists around the world are collaborating to create a digital twin of our planet. The project codenamed Earth2, has an ambitious goal – to create Earth Virtualization Engines (EVE). It is a virtual replica of the physical Earth that can be used for monitoring, analysis, and simulation. EVE is a proposed digital infrastructure that would use high-performance computing (HPC) and artificial intelligence (AI) to create virtual Earths. These virtual Earths would be simulated in unprecedented detail, and would allow scientists to study the climate system in a way that has never been possible before.
A keynote speech delivered by NVIDIA CEO Jensen Huang, at the Berlin Summit for EVE initiative in early July, focused on the role of AI and accelerated computing in advancing climate science. Huang outlined three miracles that need to happen to create Earth Virtualization Engines: (i) simulating climate at high resolution, (ii) emulating climate physics using AI, and (iii) virtualising massive climate data. The EVE initiative aims to provide kilometre-scale climate information for sustainable management of the planet.
High Performance Computing
High Performance Computing (HPC) power plays a pivotal role in simulating the Earth in unparalleled details, allowing scientists to understand and forecast climate change with an amazing level of accuracy. The impetus for Earth Virtualization Engine (EVE) is the recognition that we are better together. The goal is to identify the information needs arising from the need to manage our planet sustainably and equitably, and the opportunities available to address these needs. The vision for EVE is a coordinated global effort to leverage technology and human ingenuity to meet the challenges of climate change.
HPC has recently reached a milestone that allows for global kilometre-scale climate simulations. These simulations, previously used for more fundamental process-based studies of weather systems and ocean circulation, can now be used to study Earth’s climate on a global scale.
HPC models are currently being developed in a few labs worldwide. Performance benchmarks show that these models can achieve a simulation throughput of 1 simulated year per day (SYPD), a benchmark for climate studies, using 1 petaflop (PF) computing capacity on nearly 1 km of meshes. The world’s most powerful computer, Fugaku, can sustain 16 PF while using just under 30 MW of power. If this capacity were dedicated to climate studies, it could produce small ensembles of 30-year simulations on a 2.5 km grid on a monthly basis using just half of the machine.
However, growth in computational throughput is currently limited by power consumption and memory bandwidth. While practical computing capacity can be expected to continue to grow, it’s hard to imagine global simulations being able to access more than a factor of four increase compared to Fugaku in the coming decade. On the model algorithmic side, dedicated engineering efforts show the potential for similar performance gains.
Taken together, a tenfold increase in throughput from a single facility is an ambitious, but not unrealistic, expectation for the coming decade. A facility with this capacity could generate thousands of simulated years at a 1.5 km to 2.5 km horizontal grid-spacing per year of operation.This would enable a significant exploration of the possible future climates.
Huang, in his speech, cast the spotlight on NVIDIA’sGH200 Grace Hopper Superchip and Modulus framework for accelerating climate research and making climate data more accessible. The Grace Hopper Superchip is an accelerated CPU designed from the ground up for giant-scale AI and high-performance computing applications. It delivers up to 10x higher performance for applications running terabytes of data. The Modulus framework is an open-source framework for designing, training, and fine-tuning physics-informed machine learning models.
He also discussed FourCastNet (short for ‘Fourier Forecasting Neural Network’, a global data-driven weather forecasting model), an Earth system emulator and predictor that can learn physics from real-world data. The model is capable of predicting extreme weather events with unprecedented speed and accuracy. It uses Adaptive Fourier Neural Operators to achieve high-resolution weather forecasting.
It can generate many possible weather trajectories when tethered to “checkpoints” created by a climate simulation in seconds to minutes. This enables rapid interactive exploration of massive ensembles of possible trajectories at high fidelity, and provides massive data compression. The longer the distance between checkpoints, the larger the compression achieved. FourCastNet today can tether between checkpoints spaced a month apart, achieving 700x data compression.
Nevertheless, despite the technological advances, there remains significant challenges to be overcome before the goals of EVE can be achieved.These include:
- Significant investment in HPC and AI: EVE will require a significant investment in HPC and AI infrastructure in order to run the complex simulations that it will be used to generate.
- Development of new methods and tools: EVE will require the development of new methods and tools for simulating the climate system and for analysing the data that it generates.
- Collaboration between scientists from different disciplines: EVE will require collaboration between scientists from different disciplines in order to take full advantage of its potential.
- Engagement with users from a variety of sectors: EVE will need to engage with users from a variety of sectors in order to ensure that its results are used to inform decision-making.
Know more about our Top Ranked PGDM in Management, among the Best Management Diploma in Kolkata and West Bengal, with Digital-Ready PGDM with Super-specialization in Business Analytics, PGDM with Super-specialization in Banking and Finance, and PGDM with Super-specialization in Marketing.