AsianScientist (Apr. 9, 2021) – Two years ago, the Intergovernmental Panel for Climate Change issued a stark warning: limit global warming to 1.5ºC by 2030 or risk cataclysmic consequences.
Should we surpass this limit to 2ºC, large swaths of the world will become uninhabitable—and not just for humans. Though seemingly insignificant, a half-degree difference could expose a third of the world’s population to extreme heat waves and wipe out nearly all coral reefs.
While Singapore has so far been spared the worst of climate change’s effects, it’s only a matter of time. With an increase of 0.25ºC per decade, the island is heating up twice as fast as other low-latitude regions. By the end of the century, daily temperatures of 35–37ºC could become the norm in an already-humid setting.
Under such conditions, our body’s ability to regulate temperature is compromised, resulting in potentially life-threatening illnesses like heat exhaustion and even heat stroke. As extreme heat reduces quality of life, it also threatens Singapore’s overall productivity and attractiveness as an investment destination.
But Singapore isn’t the only city that’s getting warmer. Despite covering only three percent of the Earth’s surface, cities emit more than 60 percent of total greenhouse gases and consume about 75 percent of global energy. With less than ten years left for the world to collectively beat the heat limits warned by the IPCC, scientists are harnessing supercomputers to design cooler, more liveable cities and turn the tide of climate change.
Understanding urban heat
Singapore’s gleaming cityscape may be instantly recognizable, but its rapid urbanization also comes at a cost. As shade-providing greenery is replaced by infrastructure that absorb or produce heat, the island’s urban centers are getting warmer compared to surrounding rural areas.
This phenomenon, called the urban heat island (UHI) effect, is part of a vicious cycle. Faced with a warming city, Singapore’s residents are resorting to air conditioning and private transport—two activities that, in turn, generate heat.
Greenery, infrastructure, air-conditioner use and transport are just a few of the many factors that influence urban heat in cities. Given the complex interrelationships between these elements and the veritable deluge of data, supercomputers have become essential tool for understanding urban heat and fighting climate change as a whole.
“Scientists use models to understand these interactions and project future changes,” noted Professor Dale Barker, director of the Centre for Climate Research Singapore (CCRS) of the Meteorological Service Singapore under the National Environment Agency.
“Weather and climate modeling is numerically intensive work, for which scientists have relied on computers for over half a century,” he added.
At the CCRS, for instance, Barker shared that they leverage an in-house Cray XC40 supercomputer called Athena with a peak capacity of 212 teraFLOPS, 6,336 cores and 3.6 GB RAM of memory per core.
Through Athena, CCRS produces numerical weather forecasts of temperature, wind, rainfall and other meteorological variables through their SINGV system. To deliver their next set of regional climate predictions, CCRS will also be partnering with the National Supercomputing Center (NSCC) Singapore to tap into up to 3.2 petaFLOPS of capacity.
Virtual cities, real impact
As Singapore rapidly digitalizes, addressing urban heat through computational means has become quite literally a hot topic in recent years. Rising up to the challenge is a team from the Agency for Science, Technology and Research (A*STAR) and the Housing & Development Board (HDB).
“The IEM integrates the urban planning and design process with environmental simulation,” explained Dr. Poh Hee Joo, a senior scientist at A*STAR’s Institute of High Performance Computing and one of the project’s lead researchers. “It allows users to predict the interrelationships and combined impact of solar, wind, temperature, noise and other environmental factors in an urban setting.”
Not only can the IEM model the combined effects of the environmental factors, but it can also simulate their individual impacts. Effects of urban elements like buildings and roads as well as natural features such as vegetation are captured on the platform. This makes the IEM the first truly integrated tool of its kind, as other commercially available modelers typically assess only a single environmental factor at a time.
To develop the IEM, Poh and his colleagues first painstakingly transformed three-dimensional (3D) geometric data of Singapore into a highly realistic simulation of the city. At the same time, 43 solar-powered sensor nodes were deployed in Punggol and Singapore’s eastern half to collect environmental data on wind, temperature and solar irradiation.
Data from these sensors were then transmitted wirelessly to the research team, allowing them to validate the models’ results. By allowing for easy visualization of the effect of various factors on the urban environment, users could first refine and test their designs computationally on the platform, reducing the risk of costly physical trial-and-error.
Pushing the boundaries of supercomputing, the team performed the first-ever 3D air flow simulation that depicts all of Singapore’s buildings on the IEM. This simulation, performed at a ten-meter horizontal resolution, was completed in only five days using 6,000 processors of NSCC’s ASPIRE 1 supercomputer.
According to the researchers, having an island-wide wind map could help identify wind corridors, allowing urban developers to plan and design cooler buildings that maximize wind flow, promote natural ventilation as well as reduce light from the sun.
“Therefore, the IEM will reduce the overall carbon footprint for new housing estates to achieve greater sustainability and liveability,” concluded Poh.
Poh and his team aren’t the only ones tackling the urban heat challenge. As quite literally indicated by its name, the project Cooling Singapore, led by Professor Gerhard Schmitt of the Singapore-ETH Center (SEC), is dedicated to identifying cooling strategies and designing climate-responsive guidelines for urban environments.
Since 2017, the project has brought together multi-disciplinary teams from partners like the SEC, National University of Singapore, TUM CREATE and Singapore-MIT Alliance for Research and Technology.
While the first phase of Cooling Singapore concentrated on selecting suitable tools to assess UHIs, the recently-kickstarted second phase will now put these tools into action by building an operational digital twin of Singapore. The platform, also known as Digital Urban Climate Twin or DUCT, will allow researchers and decision makers to explore various scenarios that contribute to urban heat.
Using the platform, users can simulate the UHI effect by calculating the impact of urban elements like buildings and roads on air temperature, then comparing
these results with a scenario where all urban areas are replaced with vegetation.
“You can also modify DUCT to run simulations for specific scenarios, like district cooling, [the presence] of solar panels or electric vehicles,” explained Dr. Heiko Aydt, project leader at Cooling Singapore, in a webinar organized by NSCC in September 2020.
According to Aydt, DUCT is meant to be a decentralized, modular platform with multiple components—a federation of models. These include micro- to macro-scale models of the climate from the meteorological service as well as computational models of UHI factors like land surface, traffic and building energy.
The platform will encompass outputs from A*STAR’s IEM, which is used by the Cooling Singapore team as a microscale urban climate model. DUCT also provides risk and impact models of the urban climate on the economy, environment and health. All this advanced modeling entails crunching millions of data points at the same time—a job supercomputers are certainly well-suited for.
Though Cooling Singapore’s second phase commenced only in September 2020, they’ve already implemented a few prototypes of the DUCT with in-house resources at the SEC.
“In this case, we used an internal computing cluster with a rather modest 72 CPUs and a bit short of 200 gigabytes of RAM,” shared Aydt.
Eventually, the research team hopes to deploy DUCT across several organizations using multiple machines. For this, the Cooling Singapore team is looking to soon tap onto NSCC’s computational resources to make DUCT’s simulations available on demand to researchers and policymakers alike.
“We want to have a solution that is intuitive and useful for the end-users,” concluded Aydt. “We plan to work closely with the NSCC and others to ensure that high performance computing resources can be tapped on in a seamless way.”
The supercomputer cooling conundrum
While it’s clear that supercomputers can help address Singapore’s cooling woes, ironically, they’re also part of the problem. With extreme data crunching comes extreme heat, and supercomputers can consume as much power as a small city. Though you’d think that most of the available power goes towards computing, a large portion actually goes into cooling the facilities and racks.
“I remember when I first saw Japan’s K computer— once the world’s fastest supercomputer—and the gigantic cooling plant right next to it,” recalled Schmitt, speaking in the same NSCC-organized webinar. “I knew then that these supercomputer centers would create significant cooling loads for their surroundings.”
With more researchers turning to supercomputers to study climate change and its impacts, the annual electricity bills of their organizations have also skyrocketed to millions of dollars each year.
“This is equivalent to tens of thousands of tons of carbon emissions per year, if fossil fuels are used in electricity production,” explained Barker.
For all the might of supercomputers, ensuring energy-efficiency is a major hurdle that needs to be overcome.
“It’s a challenge not only from a financial perspective, but also because climate science needs to lead the way in sustainable green computing,” he added.
Accordingly, a few countries have tapped onto renewable energy to solve this supercomputer conundrum. A good example is Iceland, which has since become an international home for climate and weather computing due to its abundant supply of renewable energy and its naturally cold, temperate climate. Combined, these two factors reduce the amount of energy required to cool supercomputers in Iceland.
Beyond alternative energy sources, newer technologies are also being developed to reduce the overall electricity bill. Some examples include GPU-based processors, artificial intelligence-based systems for managing load and even recycling waste heat to warm offices.
“Green computing in terms of FLOPS per watt is also increasingly part of the criteria for assessing bids when tendering for new climate supercomputers,” said Barker.
The demand for supercomputing resources in climate science research is unlikely to ease in the upcoming decades.
“More complexity, more local detail and even more scenarios mean that the supply will likely never satisfy the climate scientists’ insatiable desire for computing resources,” explained Barker.
Still, supercomputers could play a crucial role in keeping Singapore cool—so long as we find an efficient way to keep them cool as well.
This article was first published in the January 2021 print version of Supercomputing Asia.
Click here to subscribe to Asian Scientist Magazine in print.
Copyright: Asian Scientist Magazine.
Disclaimer: This article does not necessarily reflect the views of AsianScientist or its staff.