RESEARCH HIGHLIGHTS

Since being discovered in the 1920s and 1930s, the ozone layer has become synonymous with gauging our planet’s health. Protecting the ozone layer has become a common refrain everywhere from elementary school classrooms to industrial board rooms, culminating in the Montreal Protocol and its amendments and adjustments, a sweeping international ozone protection treaty signed in 1987.

This protocol preventing ozone depletion is considered one of the most successful environmental treaties of all time, with broad international adoption and support. However, there are still many unknowns about which natural and artificial processes influence ozone depletion.

Chief among them is the role polar stratospheric clouds (PSCs) play in this complex chemical process. As their name implies, these clouds form in polar regions high up in the Earth’s atmosphere, namely the stratosphere, and impact the chemistry of trace gases—small amounts of gases that are naturally part of the Earth’s atmosphere apart from the main constituents: oxygen, nitrogen, and argon. While these trace gases account for less than one percent of the gas in the atmosphere, some trace gases play a major role in ozone depletion.

For decades, researchers have refined their experimental toolkits to investigate changes of the ozone layer. Since the 1990s, a research group at the University of Wuppertal has played an active role in designing an instrument to be used by high-altitude planes and satellites. The first, the Cryogenic Infrared Spectrometers and Telescopes for the Atmosphere (CRISTA), was launched on two separate space shuttle flights in 1994 and 1997 and allowed researchers to measure small-scale chemical structures and dynamics within atmospheric trace gases located in the Earth’s stratosphere.

After the successful space missions, the Wuppertal researchers partnered with Forschungszentrum Jülich to mount the next-generation CRISTA device, CRISTA-NF, on a high-altitude research aircraft, M-55 Geophysica. While these experiments are capable of measuring trace gas distribution at high altitudes, many questions about the chemical composition of the cloud particles and their role in ozone depletion remain unanswered.

To that end, the team has partnered with researchers in the Climate SimLab at the Jülich Supercomputing Centre (JSC) to using high-performance computing (HPC) to model the impact of the cloud’s chemical composition on the CRISTA-NF measurements. The research collective recently published a paper that combines its experimental and computational approaches in Atmospheric Measurement Techniques.

Clouds form in simulated space

When studying complex mixtures of gases and small particles in the atmosphere, researchers use multiple instruments and techniques. CRISTA-NF measures electromagnetic radiation emitted by the gases and cloud particles along the “line of sight” and provides the researchers a spectrum in the mid-infrared range with characteristic peaks for each gas and the cloud particles. From the location of the peaks and their shape in the spectra, the scientists can directly tell which major trace gases and particles are present. However, to identify weaker signatures and to extract more properties, such as the trace gas concentration, cloud particle number density, and particle size, advanced simulation and retrieval techniques are required.

“Temperature changes with height, of course, but trace gas concentrations like water vapour also may suddenly change,” said Dr. Sabine Grießbach, JSC researcher and meteorologist. “Scientists look at a spectrum and can see, ‘this line belongs to CO₂, this line belongs to water vapour, this one belongs to HNO₃,’ et cetera. So I may already be able to tell from just looking at the spectrum that there might be this type of cloud or some aerosol, but to know the exact values—the concentration of the gases, the composition of the clouds and their particle sizes—we need to do a retrieval. A retrieval often means inverse modelling and this is what makes this a supercomputing-caliber problem.”

The spectra measured by CRISTA-NF in the Arctic at M-55 Geophysica flight altitude and below can only serve as static snapshots of conditions in a certain place at a certain time. These snapshots are not representative for an entire Arctic winter nor the Antarctic winter. Further, even advanced instruments are incapable of accurately measuring particles’ sizes, and while the differences may only be micrometres from one another, these minuscule changes can play a large role in how particles impact ozone chemistry.

When the Wuppertal team noticed an anomalous shape of a certain peak in experimental measurements, it reached out to Grießbach to leverage JSC resources to develop a pipeline between data gathered from CRISTA-NF and simulation.

In this case, inverse modelling means researchers start with an assumption about the state of the atmosphere, then simulate the spectrum and compare it with the measured spectrum. Based on the results of the comparison, the researchers start an iterative process until the difference between simulation and measurement is sufficiently small. However, if clouds are present, scattering on the particles has to be considered in the simulations, which makes the already computationally expensive process even more expensive. 

It gets even more complicated, though. There are three “types” of PSC particles:  ice, nitric acid trihydrate (NAT), and a chemical phenomenon born of nitric acid and water vapour binding to the sulphuric acid released from natural sources such as volcanoes and certain industrial processes, called supercooled ternary solution (STS).  The three types have different scattering properties, which adds a further complexity to the simulations. The researchers decided to perform simulations covering the large variability in terms of particle size and concentration for the three PSC types and mixtures between the three. Based on this comprehensive simulation data set the researchers then developed a computationally cheap method to derive the polar stratospheric cloud composition and the particle size from the many hours of CRISTA-NF measurements. Knowing composition and particle size helps scientists understand how these different PSC types ultimately impact ozone depletion.

In the mid-1970s, scientists identified seemingly inert gases that could be used for everything from aerosol spray cans to Styrofoam production to refrigeration technologies. These gases, made up of only carbon, chlorine, and fluorine, were aptly named Chlorofluorocarbons (CFCs). While CFCs were harmless near the surface of the Earth, their interactions with PSC particles in the stratosphere are harmful to the ozone layer. In short, PSC particles provide reaction surfaces for the breakup of CFCs, releasing chlorine gas that actively destroys the ozone layer. 

As the measurements exhibited an unexpected shape of the NAT signature, a special focus was put on this particle type.  Researchers identified NAT particles as playing a particularly significant role in ozone depletion.

While NAT particles bind nitric acid they can grow to relative large particles sizes. Larger particles sediment, or fall down, more quickly, and remove nitric acid. This impacts ozone chemistry, as nitrate is an important reaction partner for ozone-destroying chlorine by helping convert it to less harmful “reservoir gases”. 

For the Wuppertal and Jülich researchers, understanding how NAT particles interact with polar chlorine chemistry in the atmosphere is essential to more fully understanding how the ozone layer is depleted and for developing potential remediation strategies.

In fact, in the last 20 years, researchers had identified a peculiar feature in its spectrum data—the data seemed to indicate that the NAT particles had actually “shifted” in the spectrum, something that should not happen when looking at spectroscopy data. The Wuppertal/Jülich team set out to identify why this would have happened, and through its iterative approach, discovered that different sized NAT particles registered differently on the spectrum.

This kind of discovery helps researchers better understand how these complex interactions influence ozone depletion, and ultimately can provide important information in discussions related to geoengineering centered around seeding or injecting the atmosphere with different chemically reactive compounds to cool the planet, among other research areas.

Rising higher

From the experimental side of the research, the next frontier lies in space. The team plans to use its new method to help analyze a decade worth of global experimental data gathered from instruments mounted on satellites, such as Envisat MIPAS.

In order to meet this additional challenge, the team will also be moving into the next computational frontier—JSC’s JUWELS Booster module, which came online in late 2020. JSC is spearheading a novel course for HPC technology, developing a modular supercomputer over the coming decade that will consist of ultra-high-speed connections between modules consisting of different architectures. The initial JUWELS module is capable of roughly 12 petaflops, or 12 quadrillion calculations per second. The second module, the so-called “Booster,” will use GPU power to raise the total system performance to 85 petaflops (for more information on the JUWELS Booster, click here).

The team used the Jülich-developed JURASSIC code to run its simulations. As part of JSC’s Climate Science SimLab and co-developer of JURASSIC, Grießbach will play an active role in ensuring that JURASSIC is ported efficiently to make the best use possible of JUWELS’ enhanced capabilities. “For us, in the simplest sense, we in the SimLabs are supposed to help other users from the community who, of course, use a large variety of different codes,” Grießbach said. “JURASSIC is our own code, one that we know very well. While preparing the code and setup for the next scientific study we learn how to make efficient use of the new architecture—which part of a code works well on the GPU partition, and which part of the code should run on the CPU partition?”

-Eric Gedenk

This article originally appeared in the Autumn 2020 issue of InSiDE.