Summers are for science: A look at what Brown researchers were up to this summer

From the depths of the ocean to the surface of Mars, this summer proved a productive one for Brown researchers. Professors, postdoctoral students and undergraduate students made notable strides in the fields of artificial intelligence, geology, physics and more, with years of creativity and diligent work culminating in published research for many.

Here’s a look at some of the studies that were published this summer.

Mars

There was once liquid water on Mars. 

That is no secret, especially not for geologists, said James Head, professor emeritus of geological sciences and earth, environmental, and planetary sciences. What is novel, however, is the discovery of how recently there was water on Mars. Liquid water was present on the planet less than a million years ago — a blip for geologists. 

In a study published in Science in June, Head and other researchers, including Ashley Palumbo PhD’20, found through climate modeling that the tilt of Mars caused the presence of liquid water on the surface. Liquid water is likely responsible for creating gullies, networks of channels running down the sides of craters on Mars. 

The gullies still move today, as shown by time-lapse photographs of the surface, Head told The Herald. But that happens due to carbon dioxide, which changes the path of the gully as it evaporates. Head’s study proposes liquid water to be the main cause of these gullies, which often take the form of elaborate fans or channels that carbon dioxide evaporation cannot completely explain on its own.

Understanding the timeline of liquid water’s presence on Mars is essential for modeling its climate over billions of years and modeling it backward in time, a process called inverse modeling, Head explained. Geologists are still trying to understand whether Mars started out wet and rainy, and how it got to the dry, freezing state it is today, he added. The new study can help refine climate models for Mars, contributing to more accurate climate models for Earth. 

“We have 4.5 billion years exposed on Mars,” Head said. “So that really gives us an opportunity to understand the basic themes of planetary evolution, and then AI can help us predict what's going on in the future of the Earth as well.”

Sea ice 

Head is not the only researcher at the University who has been studying the movement of ice and water: So has Daniel Watkins, a postdoctoral research associate at the School of Engineering, although in a much faster and local setting than Mars.

In a study published in Geophysical Research Letters in July, Watkins, Assistant Professor of Engineering Monica Martinez Wilhelmus and researchers from NASA and Oregon State University showed that the deformation and motion of sea ice as it drifts are influenced by the changes in depth and topography of the sea floor. 

“Even though it may look peaceful in photographs, sea ice is constantly moving,” Watkins wrote in an email to The Herald. “It pulls apart, forming openings called leads, and then crushes back together.” 

During a trip to the Arctic in 2019, Watkins helped deploy a set of over 100 buoys placed directly on top of sea ice to measure their drift patterns via satellite, assisted by imaging from other satellites and shape detection technology. The research team measured both the ice’s drifting speed and its deformation, tracking its interactions with tidal activity, wind patterns and the topography of the ocean floor. 

When ocean depth changes, the study found, so do sea ice dynamics. That’s because the mountains and valleys of the seafloor affect tidal currents, which are stronger in shallow waters  — as do “sharp undersea features” like the cliff at the edge of a continental shelf. 

The study is important for climate models, according to Watkins. While the effects of wind patterns on ice have been well studied — and are therefore better represented in climate models — the effects of ocean dynamics on ice have received less attention. The study may allow for more accurate climate models that better represent ocean currents and ice dynamics. 

“As sea ice gets thinner, we expect that it will be more dynamic, and therefore more likely to follow ocean currents,” Watkins wrote. “Our study provides a challenging test case for simulations of the interaction between atmosphere, ice and ocean.”

Artificial intelligence 

While the current whirlwind of artificial intelligence technologies is improving rapidly, AI still makes mistakes.

Deep neural networks — artificial intelligence systems inspired by the human brain — are trained to recognize concepts and make decisions. But once a tool is trained, researchers don’t always know how the system makes decisions or the origin of its mistakes, said Thomas Serre, professor of computer science and cognitive, linguistic and psychological sciences. 

Concept Recursive Activation FacTorization for Explainability, a project by researchers at the Robert J. and Nancy D. Carney Institute for Brain Science and researchers in France, helps explain how visual AI systems go about solving problems, and thus how they make mistakes. CRAFT was presented in June at the Institute of Electrical and Electronics Engineers/Computer Vision Foundation Computer Vision and Pattern Recognition Conference in Vancouver.

By revealing what artificial intelligence looks at to make a determination, CRAFT can help explain mistakes. For example, CRAFT revealed that images of the large tench fish were often not identified by the fins on the fish or the head of the fish. Instead, “tench” was identified by a different head — most often of an older white man, holding the fish. 

The system, trained on that detail of images, might then begin to mistake images of older white men as “tench.” 

“That's what in computer vision or AI we call shortcuts,” Serre said. “Because of what we call biases in the data set — the fact that in many, if not most, images, the fish appears next to a fisherman — the neural network ends up using the presence of the fisherman as a predictor of the class label ‘tench’.” 

CRAFT has given insight into how differently humans and AI identify pictures. While humans might use the body or snout to identify a snake, AI might use the grass or sticks around the snake. 

Computer scientists can then “harmonize” AI to act like human identification, allowing greater accuracy in tools and insight into human visual processes, Serre said.

“We just don't want a deep neural network that just solves a task,” he explained. “We want it to solve it in a way that's consistent with humans.”

Parasites 

Over the past ten years, in what began as a project for graduate students in 2013, researchers at Brown have been observing how parasites change the phenotype expression in their hosts by studying parasitic control in amphipods, a small form of crustacean. 

Recent sequencing techniques have allowed the researchers to confirm changes in gene expression, proving what has previously been observed about infected amphipods. Their research was compiled and published in Molecular Ecology in August. 

When an amphipod is infected with a trematode — a parasitic worm — the crustacean turns bright orange and becomes sluggish, risking exposure to predators when it usually runs from light, study author, Chair of Ecology, Evolution and Organismal Biology and Professor of Natural History David Rand, who chairs the Department of Ecology, Evolution, and Organismal Biology, told The Herald. 

Shawn Williams PhD’21, Stephen Rong PhD’21, John Burley GS, Kim Neil PhD’20, Adam Spierer PhD’20, Wilson McKerrow PhD’18, postdoctoral researcher Yevgeniy Raynes, Amanda Lyons ’20, David Morgan MS’22, Bianca Brown MS’21 PhD’21, Nicholas Skvir PhD’22, Naima Emory Okami ’20 — along with researchers from William & Mary University, the University of Southern Mississippi, the U.S. Forest Service and the Marine Biological Laboratory are all listed as authors as well.

“This is blind evolution,” Rand said. “If there’s more color and less mobility, birds can pick them off and eat them. And then it will get into the gut of the bird and then the worm can continue its reproductive cycle.” 

The parasite also suppresses the immune response of the amphipod, delaying the rejection of the worm from the body. 

The modified gene expression caused by parasites is not unique to amphipods or birds. In mosquitos with the parasite that causes malaria in humans, there is some evidence that infected mosquitos are more likely to bite a human, Rand said. 

Understanding the mechanisms of manipulation might allow better interventions to reduce the transmission of malaria, he added, and contribute to the scientific community’s understanding of how biological systems operate.

Fluid dynamics

In science and engineering, spheres often serve as approximations for more complicated objects. With an eventual eye toward understanding how objects like microplastics, insects or machines might move at the interface between flowing water and the surface, a group of Brown researchers started with spheres. 

While drag force — the resistance force of a fluid — is already well understood for spheres fully underwater, the interaction between an object at an air-water surface is less understood by physicists, according to Robert Hunt, a postdoctoral research associate at the School of Engineering. 

Hunt co-authored a study alongside Assistant Professor of Engineering Daniel Harris, Professor of Engineering Yuri Bazilevs, Eli Silver ’21 and researchers from the University of Illinois Urbana-Champaign, published in Physical Review Fluids in August. It found that spheres partially submerged underwater experience more drag force than fully submerged spheres. 

“The water ‘piles up’ along the upstream face of the sphere, effectively applying more pressure … and pushing the sphere downstream,” Hunt wrote in an email to The Herald. Drag is greatest right before the sphere is fully submerged underwater, quickly decreasing once piled up water can flow over the top of the sphere. 

“When you think of things associated with water, a lot of them interact with the surface,” he wrote. “You may have heard about a ‘spherical cow,’ a meme pointing out the absurdity of approximations made in science and engineering. As absurd as it can sound, a sphere is often a great approximation for a much more complicated object.”